Datasets:

chemNLP

/

ncbi-text

like 0

But it was the proteins' seemingly bizarre movements that first caught the eye of William Earnshaw (now at the University of Edinburgh in the UK) and his colleagues in 1987. Previous work had demonstrated that sister chromatids can separate even when researchers cut the microtubules that form the spindle, suggesting that the centromere might house motors that propel the chromosomes. But cell biologists knew little about the centromere's architecture. In 1985, Earnshaw's lab snared the first three of the structure's components, which they dubbed the centromere proteins, or CENPs (). Using antibodies against the chromosomes' protein scaffold, Carol Cooke then identified a pair of proteins that cluster on the centromere. The Edinburgh team called this pair the inner centromere proteins, or INCENPs (). Earnshaw and colleagues could see INCENPs clinging to the arms and centromere in mitotic chromosomes. They took a closer look, treating cells so that the chromosomes were spread-eagled into the textbook “X” shape. They saw that some INCENPs adhered to the centromeres internal to the CENPs at the last points of contact between sister chromatids. Biochemical tests indicated that the INCENPs clung tightly to the chromosome's protein scaffold, so the researchers got a jolt when they tracked the proteins through mitosis. At the beginning of anaphase the INCENPs appeared to jump ship. Instead of following the chromosomes as they slid apart to the poles, the proteins festooned the microtubules in the middle of the mitotic spindle. And some snuggled up to the cell membrane at the point where the cleavage furrow later squeezes the cell in two. “Nobody had seen a chromosomal protein change its position that dramatically,” says Earnshaw. He and his colleagues had identified the first passenger protein (the two INCENPs turned out to be splice variants of the same molecule). INCENP concentrated at the site of presumptive cleavage furrow formation before myosin appeared there (), making it tempting to speculate that it was being dropped at the site after hitching a ride on the chromosomes as a “passenger.” But earlier functions were still under consideration. Earnshaw had found that cells injected with anticentromere antibodies failed to align chromosomes during metaphase and, although they kept going through mitosis, they did so with extremely defective spindles (). “This led me to think that the chromosomes must normally ‘give’ something to the spindle in metaphase that helped stabilize it in anaphase,” he says. The candidate for that something was INCENP. Later work showed that INCENP partners with three other passenger proteins—Aurora-B, survivin, and borealin—to form the chromosomal passenger kinase complex. This complex targets many proteins in the cell including histone H3, which acquires phosphate tags as the chromatin condenses at the onset of mitosis. The passenger complex also helps to fasten microtubules to the centromeres and to choreograph the separation of the two daughter cells. Just last year, the labs of Earnshaw and Hironori Funabiki (Rockefeller University, New York, NY) uncovered borealin (), also known as Dasra (), and showed that it's crucial for correctly attaching microtubules to the centromeres, stabilizing the spindle, and completing cell division. When Earnshaw and colleagues watched INCENP flitting about in 1987, they were seeing the protein complex on the job.

At the end of mitosis, after chromosome segregation, eukaryotic cells must inactivate the cyclin B–dependent kinases that lead them into and through mitosis. This inactivation is necessary for spindle disassembly, cytokinesis, and entry into a new round of DNA replication in the subsequent cell cycle. Critical to this process is cyclin B proteolysis triggered by the anaphase-promoting complex/cyclosome (). Inactivation of mitotic Cdks in budding yeast is driven by activation of a complex signal transduction cascade, called the mitotic exit network (MEN), which is required for mitotic exit and cytokinesis. The MEN comprises several factors, including a small G protein of the Ras family (Tem1), its activator (Lte1), several protein kinases and associated factors (namely Cdc5, Cdc15, Mob1/Dbf2, Dbf20, and Cla4), and a scaffold protein (Nud1). The latter acts as a platform for many MEN components at the microtubule organizing center or spindle pole body (SPB; ; ). A similarly organized pathway, the septation initiation network, drives cytokinesis in fission yeast (), and homologues of several MEN and septation initiation network factors can be found in multicellular eukaryotes. The ultimate effector of MEN signaling is the Cdc14 protein phosphatase, which on one side can directly reverse Cdk phosphorylation events () and on the other promotes inactivation of cyclin B–dependent kinases by triggering anaphase-promoting complex/cyclosome–dependent cyclin proteolysis and accumulation of their specific inhibitor Sic1 (for review see ). Though completed by the MEN in telophase, Cdc14 activation is already initiated during anaphase by the action of the Cdc14 early anaphase release (FEAR) pathway, which includes the polo kinase Cdc5 and the separase Esp1 (). To ensure balanced chromosome partitioning, inactivation of mitotic Cdks must not be initiated before telophase, i.e., before sister chromatid segregation is complete. This issue is vital for organisms like budding yeast, which define the cleavage plane early in the cell cycle and before bipolar spindle formation. In fact, in , the constriction between mother and daughter cells (bud neck) appears at the G1–S transition concomitantly with bud emergence and dictates where cytokinesis will later take place. The spindle positioning checkpoint is responsible for delaying cytokinesis until the spindle enters the bud. The Bub2/Bfa1 GTPase-activating protein (GAP) plays a key role in this process by keeping Tem1 inactive until the spindle is properly oriented, thus inhibiting MEN activation (for review see ). Bub2/Bfa1 is found on both SPBs soon after SPB duplication but only on the SPB directed into the bud, along with Tem1, from the onset of anaphase until the end of mitosis. This observation, along with the finding that the Tem1 activator Lte1 is localized specifically in the bud, led to the proposition that MEN activation is triggered by an encounter between Tem1 sitting on the bud-directed SPB and Lte1, thus coupling properly oriented spindle elongation with mitotic exit (; ). However, Lte1 is required for mitotic exit only at low temperatures () and is dispensable for inappropriate mitotic exit of mutants with spindle positioning defects (). Thus, other mechanisms, such as inactivation of the GAP Bub2/Bfa1, must ensure the timely activation of Tem1. Inhibitory phosphorylation of Bfa1 by the Polo kinase Cdc5, although not essential, clearly contributes to this task (). Unlike Bub2/Bfa1, Tem1 is also present on the mother-bound SPB from the time of bipolar spindle formation to telophase. This suggests that disappearance of Bub2/Bfa1 from that SPB at the onset of anaphase might be important for proper MEN activation. In agreement with this hypothesis, activation of the spindle position checkpoint by either microtubule depolymerization or mutations impairing spindle orientation preserves Bub2/Bfa1 on both SPBs (; ). In this paper, we investigate the role of Bub2/Bfa1 localization at SPBs in controlling mitotic exit. Our data indicate that Bub2/Bfa1 disappearance from the mother-bound SPB at the onset of anaphase, along with Cdc5 function, helps prompt Tem1 activation in telophase. This asymmetric disappearance of Bub2/Bfa1 requires Bub2 GAP activity and depends on a functional septin ring, consistent with the notion that passage of the daughter-oriented SPB through the bud neck signals mitotic exit (). Altogether, our data highlight a new molecular mechanism coupling MEN activation with passage of the nucleus through the bud neck. Because activation of the spindle position checkpoint leads to the persistence of Bub2/Bfa1 on both SPBs (; ), symmetric distribution of the GAP could contribute to keeping Tem1 inactive. In fact, Tem1 is also present on both SPBs during unperturbed anaphase (). We therefore asked whether Bub2's disappearance from the SPB remaining in the mother cell could be important for proper mitotic exit. We previously showed that a modified version of Bub2 with nine myc epitopes at the COOH terminus (Bub2-myc9) localizes symmetrically on both SPBs throughout the cell cycle (), unlike the GFP-tagged counterpart (). This difference was attributed to the assay we used at that time (), which relied on a chromosome-spreading technique. We therefore reinvestigated Bub2 subcellular localization by indirect immunofluorescence on fixed cells of strains expressing either Bub2-myc9 or the Bub2-HA3 and Bub2-HA6 variants, carrying three and six HA epitopes, respectively, at the COOH terminus. We found that Bub2-HA3 () localized asymmetrically on the SPB moving into the bud in 93% (±1.3%, = 289) of the cells undergoing anaphase, similar to Bub2-HA6 (not depicted), whereas Bub2-myc9 was present on both SPBs in 88.3% (±7.9, = 408) of the cells in the same stage of the cell cycle (). Therefore, symmetric localization is a peculiarity of Bub2-myc9, rather than an artifact attributable to the immunostaining procedure. Because Bub2 forms a complex with Bfa1 and either protein is necessary for proper localization of the other at SPBs (), we analyzed the localization of a fully functional Bfa1 variant tagged with six HA epitopes (Bfa1-HA6) in cells expressing Bub2-myc9 as the only Bub2 source. As previously shown (), Bfa1-HA6 was asymmetrically localized on the bud-directed SPB in 91.8% (±4.1%, = 319) of wild-type anaphase cells (), whereas it was found on both SPBs in 58.2% (±10.6%, = 446) of anaphase cells (), indicating that Bub2-myc9's persistence on the mother cell SPB prevents Bfa1's disappearance from the same SPB in many anaphase cells (). Similarly, a Tem1-HA3–tagged protein was found symmetrically localized on both SPBs in 83.8% (±0.8%, = 251) of anaphase cells expressing Bub2-myc9 (), whereas it was present on both SPBs in only 27.2% (±1.0%, = 174) of wild-type anaphase cells (). Symmetric localization of Bub2/Bfa1 did not cause any obvious cell cycle defect in otherwise wild-type cells (unpublished data). However, because Tem1 activation is likely regulated by redundant mechanisms, we looked for synthetic ef- fects between Bub2-myc9 and mutations affecting the MEN. As shown in , although Bub2-myc9 was perfectly tolerated by mutants defective in MEN proteins acting downstream of Tem1, such as , , and , it was lethal at the permissive temperature for , , , , and mutants, suggesting that its presence is toxic when Tem1 activation is impaired. In fact, Nud1 is thought to act as an anchor for Tem1 and other MEN components at SPBs, whereas Cdc5 promotes Tem1 activation at different levels (; ). Surprisingly, was not toxic for cells, presumably because Lte1 is dispensable for mitotic exit at 25°C (; ; ; ; ; ), whereas it caused a synthetic growth defect to cells that are defective in mitotic exit and in Lte1 activation (; ; ). Noticeably, deletion of was not lethal for any of the aforementioned mutants, indicating that the synthetic lethality of with the , , and conditional alleles is not attributable to 's loss of function. Accordingly, unlike deletion, expression of as the only Bub2 source in the cells did not impair checkpoint activation () and did not cause lethality to cells lacking the kinesin Cin8 (). Moreover, the allele, whose product localizes asymmetrically on SPBs during anaphase, was perfectly tolerated by , , and mutants (). Thus, Bub2-myc9 behaves like a gain-of-function variant down-regulating Tem1, and its deleterious effects might be linked to symmetric SPB localization of Bub2/Bfa1. To uncover the defects caused by Bub2-myc9 in MEN mutants, we chose to make a conditionally lethal strain. Because Cdc5 is an unstable protein, we generated a strain carrying a wild-type copy of under the control of the Met-repressible promoter. As shown in , cells were viable in medium lacking Met, where is expressed, but they were unviable in Met-containing medium at 25°C (permissive temperature for ). This behavior did not change when one extra copy of was integrated into the genome of cells (), indicating that is a dominant allele. To analyze the terminal phenotype caused by the combination, cell cultures of wild-type, , and strains, exponentially growing in the absence of Met, were synchronized in G1 with α factor and released at 25°C into fresh medium containing Met to shut off expression. The pheromone was added back 120 min after release to prevent cells from entering a second cell cycle. FACS profiles of DNA contents () and analysis of nuclear division and spindle elongation () showed that cells progressed normally through the cell cycle with kinetics similar to wild-type cells. Conversely, cells accumulated with 2C DNA content, two divided nuclei, and long anaphase spindles. In addition, unlike cells, they failed to bring about Clb2 proteolysis, Sic1 reaccumulation, and inactivation of the Clb2/Cdk1 kinase (), indicating that mitotic exit was compromised. In some experiments, spindles could eventually disassemble at 210–240 min after release from the G1 arrest, although cells could neither undergo cytokinesis nor rebud and re-replicate. Overexpression of from the galactose-inducible promoter had similar effects in cells as the presence of Bub2-myc9. In fact, cells expressing were unviable in galactose-containing medium at the permissive temperature and accumulated in telophase with 2C DNA content, two divided nuclei, and long anaphase spindles (Fig. S1, available at ), whereas overexpression of the same construct in wild-type cells caused no growth defect and at most a 15-min delay in mitotic exit compared with the isogenic untransformed strain. Thus, high levels of wild-type Bub2 recapitulate the effects of Bub2-myc9 in cells, consistent with the notion that is a gain-of-function allele. Because Bub2 acts in a complex with Bfa1, which is required for Bub2 localization at SPBs (), deletion of could be expected to bypass the mitotic exit delay of cells. Indeed, cells released from a G1 block in the presence of Met could exit mitosis, disassemble spindles, and reaccumulate mononucleate cells with kinetics similar to those of cells under the same conditions (), indicating that Bfa1 is required for Bub2-myc9 to exert its inhibitory function on mitotic exit in cells. Whereas the MEN is absolutely required for Cdc14 activation, mutants defective in the FEAR pathway only delay mitotic exit (for review see ). We therefore asked whether Bub2-myc9 could further compromise mitotic exit in separase mutant cells at the nonpermissive temperature. Separase inactivation at 37°C prevents sister chromatid separation () and mildly delays mitotic exit, allowing the undivided nuclei to enter a new round of DNA replication and accumulate with DNA contents higher than 2C (; ; ; ). Expression of Bub2-HA6, which was asymmetrically localized on SPBs, did not affect cell cycle progression of cells (Fig. S2, available at ). Conversely, expression of Bub2-myc9, which was present on both SPBs in most cells, prevented them from undergoing mitotic exit and entry into a new round of DNA replication (Fig. S2). Thus, the constitutive presence of Bub2/Bfa1 at both SPBs might inhibit mitotic exit when the FEAR pathway is compromised by separase inactivation. In addition, because the mitotic exit of cells has been attributed to the peculiar migration of the undivided nuclei and embedded SPBs into the bud (), which would lead to Tem1 exposure to Lte1 (), our data argue that the encounter between Lte1 and Tem1 in the bud is not sufficient to promote mitotic exit when Bub2-myc9 is present in the cells, and therefore the Bub2–Bfa1 complex is symmetrically localized on SPBs. Because expression of Bub2-myc9 prevented mitotic exit when either the MEN or the FEAR pathway was partially impaired, we directly tested to determine whether MEN hyperactivation by different means could bypass the mitotic exit defect of cells. As shown in , high levels of the Tem1 activator Lte1 can counteract the deleterious effects of Bub2-myc9 because galactose induction of a fusion could suppress the synthetic lethality, tipping the balance in favor of Tem1 activation. In addition, the allele (), encoding a constitutively active variant of the downstream MEN target, restored viability of cells (), consistent with their failure to activate the MEN. Conversely, deletion of either or and , whose gene products counteract MEN activation (; ), did not suppress (unpublished data). Remarkably, suppressed the lethal effects of the combination not only when expressed from either the promoter or from its attenuated version () but even when just one extra copy was expressed from its own promoter, suggesting that Tem1 becomes limiting in strains (). It is important to notice that overexpressed Tem1 did not disrupt the symmetric localization of Bub2-myc9 (unpublished data), suggesting that suppression is not attributable to Bub2 titration from SPBs. We investigated whether the toxic effects of Bub2-myc9 on mitotic exit were attributable to increased GAP activity on Tem1 by using bacterially expressed and purified 6×His-Tem1, maltose binding protein (MBP)–Bfa1, and GST-Bub2 fusions in a previously described in vitro GAP assay (). The rate of GTP hydrolysis and dissociation together was measured using Tem1 bound to γ-[P]GTP, whereas the rate of GTP dissociation alone was measured using Tem1 bound to the nonhydrolyzable GTP analogue γ-[S]GTP. As reported previously (), Tem1 showed on its own GTPase activity and, to a lower extent, GTP release (compare with Fig. S3, available at ). The presence of Bfa1 stabilized Tem1 in the GTP-bound form, whereas Bub2 had little or no effect on its own ( and Fig. S3). As previously shown (), Bub2 stimulated Tem1 GTPase activity in the presence of Bfa1 () but not GTP dissociation (Fig. S3). We then compared the GAP activity of purified GST-Bub2, GST-Bub2-HA3, and GST-Bub2-myc9. Surprisingly, Bub2-myc9 could not stimulate the GAP activity of GTP-bound Tem1, whereas Bub2-HA3 was as active as untagged Bub2 (). Assaying the GTPase activity of Tem1 in the presence of Bfa1 and increasing amounts of Bub2 or Bub2-myc9 confirmed that Bub2-myc9 is unable to stimulate Tem1 GTPase activity at any tested concentration (). These findings, along with the observation that Bub2-myc9 is proficient in activating the spindle positioning checkpoint (), argue that Bub2 GAP activity is dispensable for MEN inhibition. However, the GAP activity of Bub2 might be required to control SPB localization of the Bub2–Bfa1 complex. To investigate this possibility, we generated a mutant Bub2 variant, Bub2R85A, where an alanine residue replaces arginine 85, which appears to be the catalytic arginine in the GAP domain according to sequence comparison with other GAPs (; ). As shown in , the R85A substitution completely abolished the in vitro GAP activity of both untagged and HA-tagged Bub2, whereas it could not further affect the already impaired activity of Bub2-myc9. We then replaced the endogenous gene with the allele in a haploid yeast strain and analyzed the localization of the corresponding protein by in situ immunofluorescence. Unlike Bub2-HA3 and similar to Bub2-myc9, Bub2R85A-HA3 remained on both SPBs after the onset of anaphase in 75% of the cells (), indicating that Bub2 GAP activity is required to promote Bub2 release from the mother-bound SPB at the metaphase–anaphase transition. Remarkably, the R85A substitution completely knocked out the checkpoint function of Bub2, Bub2-HA3, and Bub2-myc9. In fact, like cells, cells were hypersensitive to the microtubule depolymerizing compound benomyl (Fig. S4 A, available at ) and re-replicated their DNA in the presence of nocodazole (Fig. S4 C), indicating that checkpoint response to spindle disruption is completely abolished. In addition, like cells, they were unable to activate the spindle position checkpoint, as judged by their ability to rebud in the presence of misoriented spindles (Fig. S4 B), caused by deletion of () and (; ). Therefore, because both the addition of the myc epitopes and the R85A substitution impair Bub2 GAP activity, the checkpoint defects observed in the R85A mutants should involve some Bub2 features other than GAP. Interestingly, the R85A substitution did not affect either the interaction of Bub2-myc9 and -HA3 with Tem1-GFP (Fig. S4 D) or the ability of Bub2-myc9 to pull down Bfa1-HA6 and vice versa (). However, unlike Bub2-myc9, Bub2R85A-myc9 failed to recruit Bfa1 at SPBs at any cell cycle stage (), in spite of its presence on both SPBs throughout the cell cycle (unpublished data), thus providing an explanation for its inability to engage the checkpoint. Altogether, our data suggest that Bub2 GAP activity is not directly involved in Tem1 inactivation but is rather required to regulate Bub2 and Bfa1 asymmetrical localization at anaphase. To gain insights into the mechanisms that regulate Bub2/Bfa1 disappearance from the mother-bound SPB at the onset of anaphase and their connections with mitotic exit control, we analyzed the distribution of Bub2-HA3 in different mutants. Because both interaction of cytoplasmic microtubules with the bud cortex (; ) and passage of the daughter-oriented SPB through the bud neck () have been proposed to signal mitotic exit, we selected mutants on the basis of their possible defects in the following processes: bud neck formation/localization, SPB regulation/localization, microtubule dynamics, and MEN activation. Because some of the analyzed mutations affect spindle positioning and therefore cause Bub2-HA3 to be maintained on both SPBs, the percentage of asymmetric versus symmetric Bub2-HA3 localization was scored only in cells undergoing properly oriented anaphase, as determined by DAPI staining of nuclei. ); the MEN components Cdc5 and -14; and the B-type cyclins Clb3 and -4, which have been shown to be localized asymmetrically on SPBs (). Conversely, lack of the plus-end microtubule binding protein Bim1 increased the fraction of anaphase cells with Bub2-HA3 on both SPBs from 10 to 25% (), suggesting that Bim1 contributes to the signal disappearance of Bub2 from the mother-bound SPB in anaphase. Strikingly, we found that protein kinases localized at the bud neck, namely Swe1, Gin4, and Hsl1, participate in promoting Bub2-HA3 disappearance from the mother-bound SPB. In fact, the fraction of anaphase cells with symmetrically localized Bub2-HA3 increased from 10 to 43, 18, and 28% in , , and mutants, respectively (). Deletion of further increased the percentage of and cells with Bub2-HA3 on both SPBs (), pointing to Swe1 as an important determinant for Bub2 asymmetry during anaphase. Consistent with a role for the bud neck in regulating Bub2/Bfa1 localization, Swe1-lacking cells where the septin ring was disrupted by either the septin mutation or overexpression of the dominant-negative allele () led to symmetric localization of Bub2-HA3 at SPBs in 64 and 70% of anaphase cells, respectively. This result can be partly explained by the failure of these mutants to properly activate or localize at the bud neck the kinases Gin4, Hsl1, and Swe1, whose recruitment to the bud neck depends on septins (; ; Longtine et al., 2000). Therefore, bud neck components are necessary to signal Bub2 elimination from the mother cell SPB when the spindle is properly oriented. To assess the effects of Bub2 retention at both SPBs on mitotic exit, we analyzed kinetics of spindle disassembly in cells overexpressing . As shown in , spindle disassembly, and therefore mitotic exit, was markedly delayed in these cells compared with wild type. It should be noted that, by 5 h, spindles had been correctly oriented in virtually all anaphase cells, suggesting that the lack of spindle disassembly in ∼50% of the cells was not attributable to spindle misorientation but rather to the persistence of Bub2 on both SPBs because it was abolished by deletion. Similar to what we found with Bub2-myc9, both expression of and the mutation caused growth defects in cells at the permissive temperature, and these synthetic effects could be rescued by deleting or (), further supporting the notion that symmetrically localized Bub2 is deleterious for mitotic exit when Cdc5 is not fully functional. Eukaryotic cells that divide asymmetrically must prevent mitotic exit and cytokinesis when the spindle is misoriented with respect to the cell division axis. This represents a major issue for budding yeast, where assembly of the bud neck at the G1–S transition defines the site of cell division, compared with other organisms that establish the cleavage plane only after bipolar spindle formation. Therefore, it is not surprising that the GTPase Tem1 is finely regulated, as it triggers mitotic exit and cytokinesis in budding yeast through MEN activation. Tem1 is kept inactive throughout most of the cell cycle by the Bub2–Bfa1 complex, which normally localizes with Tem1 at the bud-directed spindle pole from the anaphase onset to the end of mitosis. Regulation of the MEN on the mother-bound SPB might have an important role in controlling mitotic exit because activation of the spindle position checkpoint, which prevents mitotic exit, preserves symmetric Bub2/Bfa1 localization at SPBs (; ). On the other hand, unlike Bub2/Bfa1, Tem1 is also found on the mother-bound SPB when anaphase takes place properly (). Our characterization of the gain-of-function allele that allows localization of Bub2/Bfa1 on both SPBs throughout the cell cycle provides new evidence that Bub2/Bfa1 removal from the spindle pole staying in the mother cell contributes to triggering mitotic exit. The role of Bub2/Bfa1 disappearance from the mother-bound SPB in mitotic exit seems to overlap with other ways of activating Tem1, as it becomes apparent only when Tem1 itself or the scaffold spindle pole component Nud1 or the polo kinase Cdc5 function are crippled. In addition, we found that Bub2-myc9 is also lethal to cells overexpressing (unpublished data), which we recently revealed to be involved in Tem1 inactivation (). Nud1 activates the MEN by recruiting Tem1 and other MEN components to SPBs (), whereas Cdc5 triggers mitotic exit through different mechanisms (for review see ). On the other hand, the observation that temperature sensitivity of the and mutants used in this study can be partially rescued by deletion of () suggests that they are impaired in Tem1 activation. Consistently, higher dosage suppresses lethality, indicating that the latter is likely attributable to either Tem1 sequestration or a failure to properly activate it. Although direct inhibitory phosphorylation seems to be a major function of Cdc5 in promoting mitotic exit, a mutant version of Bfa1 no longer phosphorylatable by Cdc5 (Bfa1-11A; ) was not synthetically lethal with Bub2-myc9 (), indicating that Cdc5 likely regulates Bfa1 function by additional means besides direct phosphorylation. Another task that Cdc5 might carry out to promote mitotic exit is phosphorylation of the Tem1 activator Lte1 (). However, deletion does not cause synthetic effects when combined with the allele (). Therefore, Cdc5 might contribute to MEN activation and/or Bub2/Bfa1 inhibition through redundant mechanisms. One of them is likely related to the involvement of Cdc5 in the FEAR pathway (; ). Accordingly, when we inactivated the latter by the separase mutant allele, Bub2-myc9 prevented mitotic exit (Fig. S2). Coupling between mitotic exit and nuclear partitioning has been proposed to be triggered by exposure of Tem1, carried by the daughter-oriented SPB, to Lte1, which is constrained in the bud (; ; ). Although this mechanism might be important for Tem1 activation in late anaphase at low temperatures, our data suggest that it might not be sufficient to drive mitotic exit when Bub2/Bfa1 remains symmetrically localized on SPBs. In fact, mitotic exit and entry into a new round of DNA replication do not take place in mutant cells when Bub2/Bfa1 is present on both SPBs, in spite of the encounter between Tem1 and Lte1. Accordingly, Lte1 was found to be dispensable for the unscheduled mitotic exit of mutants defective in the spindle position checkpoint (). We therefore propose that disappearance of Bub2/Bfa1 from the mother-bound SPB contributes to couple mitotic exit with properly oriented chromosome partitioning. Interestingly, a role for the mother cell in controlling the MEN has recently been highlighted by the finding that the Kin4 kinase, involved in the spindle position checkpoint, is specifically localized in the mother cell (; ). The Bub2–Bfa1 complex is proposed to prevent mitotic exit by stimulating Tem1 GTPase activity both in budding and in fission yeast (; for review see ). According to this hypothesis, knocking down the GAP activity of the complex should allow Tem1 activation even in conditions triggering a checkpoint response, i.e., microtubule defects or spindle misorientation. Because Bub2 but not Bfa1 carries a conserved GAP domain, we directly tested this hypothesis by substituting the putative catalytic arginine (R85) with alanine (; ). Indeed, unlike wild-type Bub2, Bub2R85A completely lacked in vitro GAP activity and caused checkpoint defects similar to deletion. Surprisingly, we also found that Bub2-myc9 had no detectable in vitro GAP activity, although it could normally support the checkpoint, suggesting that Bub2 likely contributes to Tem1 inhibition by means other than stimulating its GTPase activity. It is important to emphasize that we and others () have shown that Tem1 on its own has a high rate of GTP hydrolysis, as well as guanosine 5′-diphosphate (GDP) release, unlike other Ras-like G proteins. In agreement with Tem1's ability to switch by itself between GTP- and GDP-bound forms, Lte1 mitotic exit function does not seem to be related to its putative guanine nucleotide exchange factor activity on Tem1 (). Rather, it could be linked to its ability to stimulate Tem1 recruitment to the daughter-directed SPB after anaphase (). Thus, among several possible models, we favor the idea that Bfa1 alone is responsible for Tem1 inhibition in cells (). Indeed, Bfa1 has been shown to be able to inhibit Tem1 and mitotic exit independently of Bub2 (; ), perhaps by preventing its cycling between GTP and GDP binding () and/or by inhibiting its binding to Cdc15 (). If Bfa1 alone can account for Tem1 inhibition in cells, the different abilities of Bub2R85A and Bub2-myc9 in activating the spindle position checkpoint could be explained by their different abilities to recruit Bfa1 at SPBs. In fact, whereas Bub2-myc9 is more effective than wild-type Bub2 at keeping Bfa1 at both SPBs during anaphase, Bub2R85A fails to bring Bfa1 to either SPB throughout the cell cycle. Of course, such a model does not rule out the possibility that Bub2 GAP activity helps inhibit Tem1 upon checkpoint response in wild-type cells. In any case, our data indicate that Bub2 GAP activity promotes the disappearance of the Bub2–Bfa1 complex from the mother-bound SPB at the onset of anaphase () because both Bub2R85A and Bub2-myc9 are maintained on both SPBs from S phase to telophase. Upon spindle position checkpoint activation, Bub2 would be required to maintain Bfa1 at SPBs (), whereas its GAP activity could render the system more dynamic and help release Tem1 from Bfa1, along with Cdc5-dependent Bfa1 phosphorylation. Lte1-dependent recruitment of Tem1 on the bud-directed SPB after anaphase would also contribute to MEN activation (). Whether the in vivo target of Bub2 GAP activity in promoting its own disappearance from the mother-bound SPB is Tem1 or other proteins remains to be established. One possibility is that bud neck G proteins get exposed to Bub2 only when the daughter-directed SPB crosses the bud neck, thus signaling Bub2/Bfa1 disappearance from the mother-bound SPB. The finding that bud neck components are required for Bub2 asymmetric localization at SPBs (see the following paragraph) supports this hypothesis. Inappropriate mitotic exit of spindle positioning–defective mutants often correlates with interaction of the spindle with the bud neck (). In addition, during the unperturbed cell cycle, mitotic exit is tightly linked to the passage of one SPB through the bud neck (). Our data clearly indicate a relationship between Bub2/Bfa1 disappearance from the mother-bound SPB and the function of bud neck components (i.e., PAK kinases; septins; and the protein kinases Hsl1, Gin4, and Swe1), thus providing a molecular basis for the aforementioned results. In fact, impairment of bud neck kinases allows Bub2, and presumably its partner Bfa1, to persist on both SPBs even when the spindle is properly oriented during anaphase. Localization of bud neck components takes place in a hierarchical manner, with PAK kinases contributing to assembly of the septin ring (), which is in turn essential for recruiting Gin4 and Hsl1 to the bud neck (; Longtine et al., 2000), where they are required for Swe1 localization (Longtine et al., 2000). This suggests that Swe1 might promote Bub2/Bfa1 disappearance from the mother-bound SPB more directly than upstream components. However, other bud neck components beside Swe1 are likely implicated in this process because the fraction of anaphase cells with symmetrically localized Bub2 further increases upon septin ring disruption by a mutation or overexpression. We therefore propose that passage of the daughter-directed SPB through the bud neck signals the removal of Bub2/Bfa1 from the mother-bound SPB, thus setting Tem1 free of inhibition at this spindle pole. This, together with the Lte1-mediated recruitment of additional Tem1 at the daughter-directed SPB and the Cdc5-dependent inhibition of Bfa1, perhaps taking place at the same SPB, would trigger mitotic exit (). How the signal is transmitted from the SPB passing through the bud neck to the mother-bound SPB is unclear at the moment, but the plus-end microtubule binding protein Bim1 might be implicated in the signaling, as its lack partially disrupts the asymmetric localization of Bub2. Uncovering the molecular details of this process will be an important challenge for the future and will shed light on the mechanisms coupling mitotic exit and spindle positioning in yeast as well as in other eukaryotic organisms. All yeast strains (Table S1, available at ) were derivatives of or were backcrossed at least three times to W303 (, , , , , and ). Cells were grown in YEP medium (1% yeast extract, 2% bactopeptone, and 50 mg/l adenine) supplemented with 2% glucose (YEPD), 2% raffinose (YEPR), or 2% raffinose and 1% galactose (YEPRG). Unless otherwise stated, α factor, nocodazole, and benomyl were used at 2, 15, and 12.5 μg/ml, respectively. Synchronization experiments were performed at 25°C. For galactose induction of synchronized cells, galactose was added half an hour before release from α factor. cells were grown in synthetic medium lacking Met, whereas shutoff of the promoter was done by resuspending cells in YEPD medium supplemented with 2 mM Met. Bacterial cells were grown in LD broth (1% bactotryptone, 0.5% yeast extract, and 0.5% NaCl, pH 7.25) supplemented with 50 μg/ml ampicillin and 34 μg/ml chloramphenicol. Standard techniques were used for genetic manipulations (; ). To generate the plasmid (pSP82), a NdeI–XhoI PCR fragment containing open reading frame was cloned in the NdeI–XhoI sites of a pRS305 vector carrying the promoter (pSP81). pSP82 integration was directed at the locus by EcoRI digestion. Single integration of the plasmid was assessed by Southern analysis. To clone under the promoter (plasmid pSP67) an HpaI–SphI PCR product containing the coding region was cloned into [BamHI]–SphI of a –bearing YIplac211 vector. pSP67 integration was directed to the locus by StuI digestion. The copy number of the integrated plasmid was verified by Southern analysis. To clone under the promoter (plasmid pSP233), a BamHI–SalI PCR product was ligated into BamHI–SalI of a -bearing YCp vector (). To generate a -containing YCp vector (pSP237), a SmaI PCR fragment bearing the coding region was ligated into SmaI of pFL39. To generate HA-tagged alleles, , , and were tagged immediately before the stop codon by one-step gene tagging (). The pRS315 plasmid () was a gift from E. Schiebel (Center for Molecular Biology Heidelberg, Heidelberg, Germany). The allele was produced by site-directed mutagenesis (QuikChange site-directed mutagenesis kit; Stratagene) on pSP279, a YIplac211 vector carrying 355 bp of 5′ noncoding and 663 bp of coding region of . Integration of the mutagenized plasmid (pSP285) was directed to the locus by BamHI digestion, thus generating a full-length mutant allele plus a truncated gene. Copy number of the integrated plasmid was verified by Southern analysis. To express in , wild-type or R85A mutant , either untagged or tagged with three HA or nine myc epitopes, were amplified by PCR from genomic DNA and subcloned in the EcoRI site of pGEX6P-2rbs (a gift from A. Musacchioi, Instituto Europeo di Oncologia, Milan, Italy) to generate pSP359 (), pSP295 (-), pSP296 (), pSP358 (), pSP304 (), and pSP312 (). A BamHI–PstI PCR fragment containing the open reading frame was cloned into BamHI–PstI of pPROEX HTa (Invitrogen) to generate pSP276. The construct () was a gift from M. Geymonat (National Institute for Medical Research, London, UK). BL21 carrying pLysE plasmid (Novagen) and ×, , GST, GST-, GST-, GST-, GST-, GST-, and GST- expression plasmids were grown in LD broth containing ampicillin and chloramphenicol at 37°C for 3 h, transferred to 14°C for 1 h, and induced with 0.1 mM isopropyl-1-thio-β--galactopyranoside for 15 h. Cells expressing MBP-Bfa1, Tem1-6×His, and different GST-Bub2 fusion proteins were resuspended, respectively, in the following cold lysis buffers: 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 2 mM DTT, and 1 mM 4-(2-Aminoethyl)-bezenesulfonylfluoride; 50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 2 mM MgCl, and 10 mM imidazole supplemented with a cocktail of protease inhibitors (Complete; Boehringer); and 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, and 2 mM DTT supplemented with a cocktail of protease inhibitors. Cells were incubated with 1 mg/ml lysozyme in ice for 30 min, placed at 37°C for 5 min, and sonicated at 4°C. The extract was then clarified by centrifugation at 15,000 rpm for 30 min at 4°C. Tem1-6×His fusion protein was purified by affinity chromatography with Ni-NTA columns (QIAGEN). The MBP-Bfa1 fusion protein was purified using amylose resin (New England Biolabs, Inc.), whereas the different GST-Bub2 fusion proteins were purified with glutathione-Sepharose (GE Healthcare). After elution, the fusion proteins were dialyzed against 50 mM Tris-HCl, pH 7.5, and 200 mM NaCl and stored at –80°C. For quantification, purified proteins were analyzed by Coomassie staining and by Western blot with anti-GST polyclonal antibodies (Santa Cruz Biotechnology, Inc.), anti-MBP Mab antibodies (New England Biolabs, Inc.), and 6×His Mab (CLONTECH Laboratories, Inc.). GTPase assays were performed according to . In brief, 1 μg of Tem1-6×His was incubated in 25 μl of loading buffer (20 mM Tris-HCl, pH 7.5, 25 mM NaCl, 5 mM MgCl, and 0.1 mM DTT) containing 0.1 MBq of γ-[P]GTP or 0.03 MBq of γ-[S]GTP in the absence or presence of MBP-Bfa1 for 10 min at 30°C. The reaction was then put on ice, and 10 μl of reaction were added to 50 μl of reaction buffer (20 mM Tris-HCl, pH 7.5, 2 mM GTP, and 0.6 μg/μl BSA) containing GST-Bub2. The mixture was incubated at 30°C, and for each time point 10 μl of the reaction was diluted in 990 μl of cold washing buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 5 mM MgCl). The samples were filtered through nitrocellulose filters, washed with 12 ml of cold washing buffer, and air dried, and the filter-bound radioactivity nucleotide was determined by scintillation counting. Each assay was repeated at least two times, and reproducible results were obtained. Immunoprecipitations were performed as described in . Bub2-myc9 was immunoprecipitated from 1 mg of total extract by protein A–Sepharose beads cross-linked to anti-myc antibodies. Bub2-HA3 and Bfa1-HA6 were immunoprecipitated from 1 mg of total extract by protein A–Sepharose beads cross-linked to anti-HA antibodies. For Western blot analysis, protein extracts were prepared according to . Proteins transferred to Protran membranes (Schleicher & Schuell) were probed with 9E10 mAb for myc-tagged Bub2, with 12CA5 mAb for HA-tagged Bub2 and Bfa1, and with polyclonal antibodies against GFP for GFP-tagged Tem1 (Invitrogen), Swi6, Clb2, and Sic1. Secondary antibodies were purchased from GE Healthcare, and proteins were detected by an enhanced chemiluminescence system according to the manufacturer. Histone H1 kinase activity was measured as previously described (). Flow cytometric DNA quantitation was determined according to on a FACScan (Becton Dickinson). In situ immunofluorescence was performed according to . Immunostaining of α-tubulin was performed with the YOL34 monoclonal antibody (Serotec) followed by indirect immunofluorescence using rhodamine-conjugated anti–rat antibody (1:100; Pierce Chemical Co.). Immunostaining of myc-tagged proteins was done with the 9E10 mAb, whereas that of HA-tagged proteins was done with the 16B12 mAb (Babco), followed by indirect immunofluorescence using CY3-conjugated goat anti–mouse antibody (1:1,000; GE Healthcare). Cytokinesis defects were assessed upon cell wall digestion with zymolase. Digital images were acquired on a fluorescent microscope (Eclipse E600; Nikon) equipped with a charge-coupled device camera (DC350F; Leica) at 20°C with an oil 100× 1.3 NA PlanFluor objective (Nikon), using FW4000 software (Leica). Fig. S1 shows that overexpression of in cells causes effects similar to Bub2-myc9. Fig. S2 shows that symmetrically localized Bub2 prevents exit from mitosis in cells. Fig. S3 demonstrates that Bub2 does not stimulate GTP dissociation from Tem1. Fig. S4 shows that the GAP-defective mutant fails to activate the spindle position checkpoint. Online supplemental material is available at .

Multinucleated cells are found in a variety of organisms and are integral to processes as diverse as the early development of the fruit fly, bone remodeling, placenta formation, and cancer metastasis. However, information on how multinucleated cells regulate nuclear division is limited to relatively few systems. Fusion experiments with mammalian cells demonstrated that multiple nuclei in a shared cytoplasm synchronize their nuclear division cycles (). These experiments suggested that separate nuclei within a common cytoplasmic environment will alter their normal cell cycle kinetics and align temporally with the other nuclei present. Synchronous mitosis has also been observed in binucleate yeast cells (). Furthermore, in naturally multinucleated cells, mitoses often occur either synchronously, as in the slime mold (), or parasynchronously, with a linear wave of nuclear division spreading across a cell, as in the filamentous fungus (). One hypothesis for the molecular basis for synchronous mitoses in multinucleated cells is that cell cycle regulators shuttle between nuclear and cytoplasmic compartments, which would allow them to diffuse and coordinate multiple nuclei in the same cytoplasm. Thus, the dynamic localization of the cell cycle machinery may promote continual communication between individual nuclei and the cytoplasm and could explain synchronous nuclear division cycles, although in no instance has this hypothesis been tested in multinucleated cells. In some cases, nuclei appear to behave independently of one another despite sharing a common cytoplasm. In newly fertilized binucleate sea urchin embryos, the two pronuclei have been shown to initiate nuclear envelope breakdown and mitosis at different times if one nucleus is paused as a result of DNA damage (). Signaling for cell cycle progression was also nuclear limited in experiments involving binucleate yeast cells in which a single nucleus received DNA damage (). Similarly, in binucleated cultured marsupial cells, anaphase entry is asynchronous if chromosome attachment is delayed in one spindle (). Thus, at least for the case of responding to DNA damage or unattached chromosomes, multinucleated cells can undergo asynchronous mitosis. Furthermore, the filamentous fungi and are also thought to have asynchronous mitoses, but in neither case has the phenomenon been studied in depth (; unpublished data). Thus, it seems to be possible for the nucleus itself to control the decision to enter mitosis, although the molecular basis for such nuclear autonomy in any system remains unknown. In this study, we present evidence for and initial molecular analysis of asynchronous nuclear division cycles in the multinucleated filamentous fungus . This fungus shares a common ancestor with before the duplication of the budding yeast genome (; ). Over 90% of the genome contains syntenic homologues in budding yeast, and, in particular, all cell cycle control proteins described for budding yeast are present with homology ranging from 20 to 95% identity (). Despite this conservation of cell cycle gene sets, unlike budding yeast, cells grow exclusively as hyphae containing many nuclei in linear arrays within the same cytoplasm. We predict that these hyphal cells undergo a closed mitosis like many fungi, including , and there is evidence that the hyphae are compartmentalized with partial septa that contain pores large enough for nuclei to pass between compartments (). We have generated nuclear pedigrees with time-lapse video microscopy showing that nuclei in cells divide asynchronously in the same cytoplasm. We analyzed this apparent nuclear autonomous behavior and evaluated how periodic cell cycle proteins act in such a system in which nuclei are not in sync with each other. Our experiments showed that both G1 and mitotic cyclins are present across all cell cycle stages. We present evidence that the continuous cytoplasm in these cells permits proteins to diffuse so that nuclei continually receive proteins from the cytosol, which are expressed from neighboring nuclei. These data have led us to hypothesize that differences in cell architecture and spatial organization may have selected for alternative modes of cell cycle control during the evolution of and . When combined, our experiments suggest that cyclin-dependent kinase (CDK) inhibitors rather than cycles of accumulation and complete destruction of the cyclins provide the primary source of oscillation in the cell cycle machinery of . Growth and nuclear dynamics were followed in the growing tip region of cells using in vivo fluorescence and time-lapse video microscopy. Nuclei were readily visualized using a GFP-tagged histone (H4) protein, . As can be observed in the accompanying video and has been reported previously, these nuclei oscillated rapidly and tumble through the cytoplasm in three dimensions, frequently exchanging places through by-passing one another (Video 1, available at ; ). Multiple mitoses were easily observed in the growing tip compartment, and, remarkably, in most cases, nuclei divided asynchronously, independently of neighboring nuclei in close physical proximity (, A and B; and Video 1). The distance between nuclei varied between <1.0 μm, when nuclei pass each other, and 7 μm, with a mean of 3.6 ± 0.1 μm. Nuclear division occurred irrespective of the distance from the hyphal tip, and no spatial patterns of mitosis could be discerned from the time-lapse data. One way for cells to establish asynchrony would be for some nuclei to be arrested in G and act as “insulators” or barriers between actively dividing nuclei. To investigate this possibility, we generated a series of nuclear pedigrees in which nuclei in living cells were followed through several generations of division and were also charted as to their relative positions in the cell (). In all pedigrees examined, most nuclei divided, yielding daughter nuclei that also divided ( = 11 independent pedigrees). The mean time between divisions of a nucleus, which represents the total nuclear division cycle length, was 112 min but varied greatly from 46 to 250 min ( = 41 mitoses; ). Interestingly, variability in cell cycle length was observed within a single lineage starting from one nucleus. Daughter nuclei produced from the same mitosis do not necessarily divide again in the same time interval (). Frequently, daughter nuclei by-passed neighboring nuclei soon after mitosis was finished (X on pedigree in ); however, asynchrony was still observed in the absence of such by-passing events. Furthermore, sudden increases in Hhf1-GFP intensity, which may correlate with S phase, generally occurred at different times in sister nuclei produced from the same mitosis. Thus, most nuclei appear to be capable of division but with variable timing, and, surprisingly, they act independently of neighboring nuclei. The time-lapse data lead to the prediction that adjacent nuclei are in different phases of their division cycle at the same time. To determine the cell cycle stages of neighboring nuclei, we visualized tubulin () in fixed cells and spindle pole bodies (SPBs; fungal equivalent of the centrosome; ) in living cells. Overall, 62 ± 4.8% of nuclei had a single SPB (noted as 1 in figures), 18 ± 2.6% had duplicated, adjacent SPBs (noted as 2 in figures; scored based on either two distinct dots or a brighter and expanded area of fluorescent signal compared with 1), and 20 ± 2.7% showed separated SPBs and a spindle (noted as Meta for metaphase and Ana for anaphase in figures; , A and B; > 1,000). Similar proportions were observed by evaluating microtubules in fixed cells with a tubulin antibody () and observing SPBs with (). We found that only 35% of nuclei ( > 1,000) were in the same spindle stage as their neighbors as judged by SPB morphology, and these were nearly all nuclei with only a single SPB. Focusing specifically on nuclei with mitotic spindles and separated SPBs, >80% were beside a neighbor with only a single or just duplicated SPB, showing that even actively dividing nuclei do not seem to influence neighbors in different cell cycle stages. Four-dimensional time-lapse microscopy of hyphae expressing and revealed that one nucleus frequently underwent SPB duplication, whereas neighboring nuclei remained with a single SPB (Fig. ). Nuclei spent an extended period with duplicated but not separated SPBs (from 20 min to >1 h, with a mean time of 47 ± 6 min; = 11 nuclei). Thus, nuclei in appear to transit the different stages of the cell cycle regardless of the state of their neighbors, and many nuclei pause for extended time periods before or during bipolar spindle assembly. The asynchrony described in and could come about in either of two ways: each nucleus may independently modulate the lengths of the division cycle phases as suggested by the pedigrees, or all nuclei could undergo a standard cell cycle, but because the nuclei move around, each nucleus is out of phase with respect to its neighbors. To address these possibilities, we used the microtubule poison nocodazole to arrest nuclei in mitosis and synchronize this normally asynchronous system. If nuclei have a standard cycle, synchrony should persist after release from the arrest. All cytoplasmic microtubules were depolymerized by nocodazole, producing tubulin staining limited to SPBs that were used to determine cell cycle stage. After a 4-h incubation in the drug, >70% of the nuclei appeared in synchronous arrest with adjacent, bright SPBs, whereas control cells treated with DMSO still displayed asynchrony similar to that of wild-type cells ( = 200 nuclei for each condition; , Fig. S1 B, and not depicted). Cells were released from arrest into fresh media, and nuclear division cycle progression was followed. Initially after release (15-, 30-, and 45-min time points), the extensive network of cytoplasmic microtubules was repolymerized before recovery of nuclear microtubules. By 1 h after release, the cytoplasmic networks of microtubules were largely rebuilt, and the majority of nuclei was in synchrony and had extended spindles indicative of anaphase/telophase. However, by 2 h after the release from nocodazole, only 30% of the nuclei were in the same spindle stages as their immediate neighbors, much like the starting cell population, indicating that asynchrony was quickly restored (). These data suggest that each nucleus seems to have an autonomous clock that functions independently of nuclei in their vicinity. Are there clues in the genome to explain the molecular basis for nuclear asynchrony? The genome shows striking synteny patterns and similarity in gene set to despite their different lifestyles and growth form. Whole genome comparisons between the two organisms strongly suggest that they diverged before the duplication of the budding yeast genome, and, thus, the genome lacks many gene pairs that are retained in budding yeast after this duplication (). genes are named based on their syntenic homologue in , and in the cases where a syntenic gene pair exists in yeast, the single homologue in is called by both names separated by a slash. We examined homologues of the core cell cycle machinery in and found that the entire gene set is present at syntenic positions, with amino acid identity ranging from 31 to 86% (). Thus, and share similar transcription factors, cyclins, the CDK, inhibitors, activators, and degradation machinery involved in cell cycle control. How might this conserved network of cell cycle regulators be constructed to direct asynchronous mitosis in a multinucleated cell? Cyclin proteins are integral to the cell cycle clock in eukaryotic cells and, when complexed to a CDK, drive cell cycle progression (). Periodic transcription and degradation of cyclins is believed to be a central mode of oscillation that is critical for orderly cell cycle progression. Upstream sequences of Ag and Ag showed the conservation of transcriptional regulatory elements, including some Swi4/6 cell cycle box-binding factor and Mbp1–Swi6 cell cycle box-binding factor transcription factor–binding sites in the Ag promoter and Fkh1p, Fkh2p, and Mcm1p sites in the Ag promoter (consensus sequences summarized in ). Furthermore, protein sequences have degradation motifs, including possible PEST sequences in AgCln1/2p and the destruction (D)-box and KEN-box motifs in AgClb1/2p (; ; ; ; ). The AgClb1/2p sequence has one additional NH-terminal D box, one of two nuclear export sequences (NESs) that is compared with ScClb2p, and one NLS (). Based on the homology and similar domain composition of the cyclins compared with those of budding yeast, we would predict that individual nuclei are expressing and degrading distinct cyclins depending on their cell cycle stage. Given that protein translation occurs in a common cytoplasm, however, it is unclear how independent transcriptional programs and oscillating protein levels are achieved by neighboring nuclei that are out of phase with each other. In an effort to determine whether such periodic expression patterns could be locally established in a syncytial cell, the cyclins were localized. The predicted G1 cyclin and the B-type cyclin were epitope tagged at their endogenous loci with 13 copies of the c-myc epitope to evaluate cyclin protein distribution in the cell. These epitope-tagged strains displayed normal growth and wild-type levels of nuclear asynchrony, and proteins of the proper size were recognized on an anti-myc Western blot ( and not depicted). Cyclin protein localization in the cell was determined using indirect immunofluorescence. The myc antibody did not show any signal in untagged wild-type strains (). As would be predicted, AgCln1/2p was concentrated in nuclei with a single SPB and was diffusely present in the cytoplasm (, C and D; left). Surprisingly, however, this G1 cyclin homologue is also readily visible in nuclei with mitotic spindles. Additionally, AgCln1/2p was observed at many hyphal tips, which are sites of polarized growth (, left). Similarly, the B-type cyclin homologue AgClb1/2p was concentrated in nuclei with two SPBs and metaphase spindles as expected but, remarkably, was also clearly present in nuclei with a single SPB and in anaphase nuclei (, C and D; right). Both types of cyclins seem to be present throughout all cell cycle stages (). AgClb1/2p was found in 91% of nuclei in a mycelium, and AgCln1/2p was observed in 74% of nuclei ( > 200 nuclei). Thus, the proportion of nuclei with a given cyclin is greater than the proportion of nuclei in the corresponding spindle stage (on average only 20% nuclei are in mitosis, yet >90% have mitotic cyclin). Furthermore, nuclei lacking either type of cyclin signal were not in any specific cell cycle stage as determined by spindle appearance, suggesting that these unstained nuclei are results of technical limitations in resolving all proteins by immunofluorescence rather than a large oscillation in cyclin levels in specific stages of the cell cycle (Fig. S2, available at ). There was some variability in the intensity of signals between nuclei, but this also did not correlate with cell cycle stage, and we predict this is caused by variability inherent in immunofluorescence in fungal cells where there is irregularity in digestion of the cell wall. Thus, the cyclins in do not display complete degradation in sync with cell cycle progression but, instead, are found in most nuclei regardless of their cell cycle stage, suggesting that many nuclei have different types of cyclins present at the same time. Synchrony in multinucleated systems is hypothesized to depend upon the free diffusion of mitotic regulators in the cytoplasm. Could asynchrony be established simply by limiting diffusion and concentrating cell cycle factors in nuclei? If so, the nuclear limited localization of mitotic cyclins in cells could compartmentalize the cell and prevent communication between individual nuclei leading to asynchrony. It is unclear, however, if nascent cyclin proteins expressed from a single nucleus and translated in the common cytoplasm are free to diffuse and enter neighboring nuclei of different cell cycle stages. To evaluate the basis of the ubiquitous nuclear cyclin localization, we generated heterokaryon cells in which only a subset of nuclei express a tagged version of the cyclin. We then asked whether the nuclei that do not themselves contain an epitope-tagged cyclin gene have detectable levels of tagged cyclin proteins expressed from neighboring nuclei. 13myc was expressed from an autonomous replicating sequence plasmid that also contained lac operator repeats (lacO) and was put into an strain that had GFP-lacI-NLS integrated in the genome (; ). Autonomous replicating sequence–containing plasmids are maintained by in a heterokaryon state such that only a subset of nuclei ever contain a plasmid under limited selection pressure, and plasmids are readily lost from nuclei after the removal of all selection (). Nuclei containing the plasmid and, thus, also the tagged version of were visible by a bright “dot” of GFP-lacI-NLS where it concentrates as a result of the lacO repeats present on the plasmid (, left). When all selection for the plasmid was removed, the bright dots of GFP signal were lost from many nuclei so that only a fraction (28%; = 300 nuclei) of nuclei retained a plasmid (5 h after release from selection; , right). We then asked whether the AgClb1/2-13myc protein was visible in nuclei that did not appear to contain a plasmid. Tagged cyclin proteins were detected in nearly all nuclei regardless of whether they contained the plasmid expressing the epitope-tagged gene. In hyphae in which very few GFP dots were visible, AgClb1/2-13myc protein was still apparent in many nuclei. Thus, nuclei that clearly lacked the plasmid containing the tagged gene had tagged protein product. This product most likely derived from neighboring nuclei that still possessed a plasmid or, in some cases, may still contain a reservoir of stable protein. This experiment suggests that mitotic cyclins are freely diffusing, taken up by, and maintained in neighboring nuclei. To further address how mitotic cyclin nuclear localization contributes to asynchrony, we shifted a proportion of the cyclin protein out of the nucleus with the addition of a NES to the AgClb1/2-13myc protein (). If nuclear sequestration of newly made cyclin protein is important for nuclear autonomy, we would predict that the displacement of more mitotic cyclin into the cytoplasm, where it could potentially more freely diffuse, would lead to an increase in synchrony of nuclear division. The additional NES led to an increase in the cytoplasmic levels of AgClb1/2-13myc protein as visible by immunofluorescence but still left sufficient protein in the nucleus to promote normal growth and nuclear division. There was no increase in cell cycle synchrony observed despite the increase in cytoplasmic levels of cyclin proteins between cells with an inactive NES or active NES fusion protein (). Interestingly, however, these cells did show a higher proportion of nuclei with duplicated SPBs or mitotic spindles, suggesting that cyclin protein concentration gradients were altered by this displacement, leading to an increase in the frequency of mitosis (). Thus, nuclear asynchrony or autonomy seems to be established independently of the origins, levels, and localization of mitotic cyclins. Cyclin proteins appear to be present in all cell cycle stages, which raises the question of how nuclei cycle with accuracy. To further examine the possible persistence of a pool of mitotic cyclin protein across all cell cycle stages, B-type cyclins were evaluated during a nocodazole arrest and release experiment. This allowed AgClb1/2p localization and protein levels to be followed through a synchronous mitotic exit, the time period when mitotic cyclins are predicted to be degraded. Mitotic cyclins were readily observed both by immunofluorescence and by Western blots on whole cell extracts even in cells where the majority (73%) of nuclei had extended spindles and separated DNA indicative of late anaphase/telophase (). This suggests that B-type cyclin degradation does not correlate with mitotic exit and that any regulative degradation, if present, must occur on only a small fraction of the total protein. Thus, mitotic cyclins seem to persist across the mitosis–G1 transition and potentially may be inactivated for G1 by direct inhibition rather than degradation. When was expressed in yeast cells either from its own promoter or from a GAL promoter, yeast cells showed a delay in telophase, further suggesting that the protein may be somewhat intrinsically stable compared with the yeast cyclins (Fig. S3, available at ). These localization and protein level data suggest that cyclin levels do not dramatically fluctuate in a cell cycle–dependent manner in these syncytial cells. The AgClb1/2p homologue, which is an essential gene in , does contain two NH-terminal D-box sequences that, in other systems, have been shown to lead to anaphase-promoting complex-mediated degradation by the proteosome (). In , yeast, and cells, the expression of cyclins lacking the D boxes leads to cell cycle arrest as a result of an inability to exit mitosis (). If some amount of cyclin destruction is required in vegetatively growing cells, D-box mutants should display a clear cell cycle block or delay. The expression of mutants in which either D-box was deleted individually and ) or both were deleted () led to no growth defect or alterations in nuclear division based on the proportion of nuclei in different cell cycle stages and levels of asynchrony (, C [top] and D; > 200 cells scored for each strain). To determine whether these mutant strains were delayed in exiting mitosis, cells were synchronized with nocodazole and released to observe a synchronous mitosis in the population. (As with , in the first hour after release from nocodazole, cells rebuild cytoplasmic microtubules before restoring nuclear microtubules). When compared with cells with only wild-type , both the single and double D-box mutants showed similar rates of return to asynchrony and similar proportions of nuclei in mitosis after 120 min, suggesting no significant delay in mitotic exit (, t = 120 min). Surprisingly, protein levels of AgClb1/2-db1p, -db2p, and -dbdp were comparable with wild-type AgClb1/2p (). We predict that the D-box sequences may be required in a different growth stage but are dispensable for normal nuclear division during vegetative growth. Normal mitotic progression does not appear to require sharp degradation of the mitotic cyclin pools, yet nuclear division is apparently accurate and faithful in . What prevents premature entry into mitosis given the presence of mitotic phase–promoting cyclin early in the cell cycle? One possibility is that mitotic CDK inhibitors such as AgSwe1p and AgSic1p are acting in a nuclear-autonomous fashion and repress any CDK–Clb1/2p complexes present in G1 nuclei. Although mutants lacking showed minimal mitotic defects (unpublished data), 35% of cells stopped growth with 10–20 nuclei, aberrant multipolar spindles, and irregular nuclear distribution and density (, A [middle] and B). A proportion of cells (55%) formed a developed mycelium, but, out of these, the majority ultimately arrested growth with up to 100 densely packed nuclei (, bottom). About 10% of mutant spores produced viable and sporulating mycelia, potentially as a result of the acquisition of a suppressing mutation or the production of multinucleated spores during sporulation that contain a wild-type nucleus. Many cells in an mutant population show defects in nuclear division, suggesting that AgSic1p may help limit the premature activity of mitotic CDK complexes. If AgSic1p is a source of AgCDK/Clb1/2p oscillation, the protein may vary in localization and/or abundance across different stages of the cell cycle. We localized AgSic1p-13myc protein by immunofluorescence and evaluated its behavior during different nuclear division cycle stages to examine whether this predicted CDK inhibitor may oscillate in any way. We found that AgSic1p undergoes notable changes to its sub-nuclear localization, correlating with different stages of nuclear division (). In nuclei with a single SPB, AgSic1p was present throughout the nucleoplasm. Interestingly, the AgSic1p was not uniformly distributed; rather, there were foci of intense signal in addition to a more diffuse signal across the nucleus. In nuclei with duplicated SPBs, the nucleoplasm signal weakened, and some protein was also observed on short spindles. When metaphase spindles were present, AgSic1p was barely detectable in the nucleoplasm, and, occasionally, a low level of AgSic1p was apparent on spindles. In anaphase, as soon as any separation was apparent between the stained DNA, AgSic1p reappeared on elongating spindles as an intense signal. AgSic1p was observed on spindles throughout late anaphase/telophase as well as in the nucleoplasm of the daughter nuclei. These dynamics would suggest that AgSic1p's subnuclear localization is regulated to alter either its own activity or its contact with interacting proteins through different stages of the nuclear cycle. In this study, we provide evidence for nuclear-autonomous division in multinucleated fungal cells. Nuclear pedigrees and observations of spindle morphology in neighboring nuclei demonstrate that nuclei divide asynchronously in these cells and display independence in a common cytoplasm. Cyclin proteins appear to be highly enriched in nuclei but able to diffuse in the cytoplasm without disturbing asynchrony. Remarkably, the mitotic cyclin protein is present in all stages of the nuclear division cycle, suggesting that oscillation of CDK activity is not primarily generated by periodic expression and complete degradation of cyclins (). From our experiments, we propose that the syncytial nature of these cells has favored the evolution of a cell cycle oscillator built primarily upon the action of CDK inhibitors rather than control of cyclin protein abundance. These data raise two unique problems for these cells: how does cell cycle progression occur accurately given the apparent lack of coordinated oscillation in cyclin proteins, and how do nuclei behave independently in the same cytoplasm? Current models of cell cycle control are rooted in the principle of a biochemical oscillator built through the intertwined regulation of synthesis and destruction of many proteins (). In yeast, ∼800 genes display periodic expression over the cell cycle, including cyclins, CDK inhibitors, degradation machinery, and transcription factors, and many of these proteins similarly have periodic destruction patterns (). Extensive experimental data combined with recent mathematical models have shown that this choreographed synthesis and destruction program enables the cell cycle to function as a bistable system that alternates between two stable but mutually exclusive and irreversible states, G1 and S/G2/mitosis phases (; ). In yeast, the basis of this bistability is thought to originate from essentially two redundant, although distinct, biochemical oscillators: one rooted in a negative feedback loop that triggers mitotic cyclin (ScClb2p) degradation and the other, which is a “relaxation oscillator” that ensures alternating states of high Clb2p/low Sic1p and low Clb2p/high Sic1 activity (). Only one of the two oscillators needs to be present for yeast to divide. Importantly, however, both oscillators require fluctuating protein levels of either Clb2p or Sic1p to properly function and maintain the bistable nature of the system (). It appears as though the cell cycle oscillators in may not require temporally regulated synthesis and complete degradation of at least some key cell cycle control proteins. Conceivably, there is highly localized degradation of a small but functionally significant fraction of cyclin proteins that has escaped our detection in these studies. However, lack of a dominant phenotype in cells expressing cyclin D-box mutants suggests that regulated proteolysis of cyclins is not critical for regulating mitosis during normal vegetative growth. The continuous cytoplasmic streaming of a syncytial cell complicates the establishment of an oscillating protein gradient through degradation alone because new protein is always supplied by the cytoplasm. Thus, multinucleated cells like may have evolved to favor the modulation of CDK activity by inhibitors to generate bistability in the cell cycle circuit. Although at least the AgCln1/2p and AgClb1/2p cyclins are consistently present in nuclei throughout the cell cycle, we predict that they cannot be simultaneously active. In these cells, the feedback loops that run the cell cycle oscillator would lead to changes in activity instead of changes in protein synthesis and degradation. In fact, a bistable cell cycle oscillator has been shown to function with constant levels of mitotic cyclins in egg extracts simply through modulating the activity of Cdc2 (CDK; ). Our data suggest that Sic1p acts in cells to ensure alternating states of CDK activity even in the presence of relatively constant AgClb1/2p levels. Nuclear-autonomous regulation of a single factor such as the inhibitor AgSic1p could be sufficient to provide the necessary oscillation in CDK activity for DNA synthesis, spindle construction, and chromosome segregation without great changes in cyclin levels. recently presented evidence in that oscillation in CDK activity could be generated without the degradation of mitotic cyclins simply by the overexpression of ScSic1p (). In these cells, ScClb2p was modified to be present in G1 phase, as is observed in the normal cycle, but was held inactive by an overabundance of Sic1p. Furthermore, ScSic1p becomes essential in yeast cells overexpressing ScClb2p (). Clearly, AgSic1p makes a vital contribution to accurate division in these multinucleated cells and may prevent G1 nuclei with AgClb1/2p from entering mitosis before the completion of S phase. The AgSic1p homologue shows relatively low identity to the yeast homologue, suggesting it could have diverged to be more stable and active and may act as a more potent CDK inhibitor. Additionally, AgSic1p subnuclear localization dynamics present the possibility that it is tightly regulated in the nuclear cycle. In the middle of mitosis, when CDK/Clb activity is at a peak, AgSic1p levels diminish in the nucleus potentially as a result of either nuclear autonomous degradation or transient nuclear export. The rapid accumulation of AgSic1 shortly thereafter in early anaphase, however, argues that nuclear import rather than new synthesis would replenish the nuclear protein pool. Thus, we hypothesize that multinucleated cells such as may generate oscillation in CDK activity through posttranslational regulation of inhibitor and cyclin protein activity rather than through tightly controlled abundance. It is conceivable that other cyclins in the system, such as AgClb3/4p or AgClb5/6p, may fluctuate, but this will be tested in future experiments. How does each nucleus cycle independently? One simple possibility is that asynchrony or nuclear autonomy could be generated as a result of the nature of the cell cycle circuits in , which, as discussed above, may be subtly different than those used in uninucleated cells. Interestingly, recent modeling and experimental work by in egg extracts in which a positive feedback loop in the Cdc2/anaphase-promoting complex circuit is blocked (producing damped oscillations in CDK activity) led to asynchrony among nuclei in the extracts. In , the core circuit regulating CDK oscillations may be wired in such a way that transitions between division stages are gradual (which happens when positive feedback loops are repressed in the model), making nuclei divide out of sync. Alternatively, cell cycle stage identity is given by periodically transcribed genes that may produce proteins that are translated in a common cytoplasm but, unlike the cyclins, could be directed back to their transcriptional “mother” nucleus. This factor could be a direct regulator of Sic1p that ensures temporal regulation of the localization and, potentially, activity of the CDK inhibitor. What are possible mechanisms to limit the presence of such an “identity” factor to a specific stage nucleus? Potentially, mRNAs could be targeted to a domain of the ER adjacent to nuclear pores to facilitate rapid uptake into nuclei after translation, or the newly translated proteins could be immediately complexed with nuclear import factors. Conceivably, such a protein could actually be translated within the nuclei, thereby avoiding posttranslational diffusion within the cytoplasm (). Alternatively, perhaps such a factor is constitutively transcribed but would be selectively blocked in certain cell cycle stages by small regulatory RNA molecules that exist only within the nucleus. The nuclear pore complex itself could be remodeled in a cell cycle–dependent manner such that certain proteins are only allowed to enter nuclei of a particular cell cycle phase (; ). Finally, this could be a factor that is associated with some cell cycle event, such as DNA replication, and could be activated in a nuclear-autonomous manner only upon completion of the event in an individual nucleus. Thus, there are many ways to conceive of how the nuclei could maintain independent cell cycle stage identity within the same space. Why would a multinucleated cell evolve mechanisms to ensure asynchronous nuclear division? What advantage does a nuclear-autonomous cell cycle bring to a multinucleated cell? Such cells may limit mitosis so that it does not occur synchronously to prevent a sudden doubling of the nucleocytoplasmic ratio. A nuclear-autonomous cell cycle is a possible way to maintain the balance of nuclei in the cell, ensuring that large fluctuations in the nuclear number do not occur without concurrent growth. A nuclear-autonomous cycle also enables nuclei in a specific subcellular location to divide, such as near a branch point or at a site rich in nutrients, without having to duplicate or transport more distant nuclei. Thus, asynchronous mitosis enables conservative and more spatially directed control of mitosis. When combined, these experiments demonstrate that the basic eukaryotic cell cycle control network may have evolved in diverse ways to accommodate the unique geometry of a multinucleated cell. Given the evolutionary relationship between budding yeast and , it is clear that dramatic differences in cell behavior can be achieved through modulating a very similar set of proteins. media, culturing, and transformation protocols are described in and , and strains are listed in . Nocodazole stocks were 3 mg/ml in DMSO and used at a final concentration of 15 μg/μl (Sigma-Aldrich). Standard methods were used for culturing and plasmid manipulations (). Plasmids generated and used in this study are listed in Table S1 (available at ). PCR was performed using standard methods with Taq polymerase from Roche, oligonucleotides were synthesized at MWG, and all restriction enzymes came from New England Biolabs, Inc. or Roche. Oligonucleotide primers are listed in Table S2. The wild-type strain was transformed with the pHHF4-GFP to make the AgHHF1 strain (). This strain was always grown under selective conditions with 200 μg/ml G418 (Calbiochem/Merck) to maintain the plasmid during time-lapse video microscopy for pedigree analyses. was tagged at the COOH terminus with GFP at its genomic locus using PCR-mediated gene targeting with oligonucleotides SPC42-GFPS1 and SPC42-GFPS2 and the pGUG template to generate ASG38, which was verified using oligonucleotides SPC42 I and GG2. The 13-myc-GEN cassette for was generated by digesting pFA6-13myc with BglII and ligating the gel-purified 3.2-kb fragment to the 1.7-kb fragment generated from digesting pGEN3 with EcoRV–BamHI. To COOH-terminally tag the cyclin genes and AgSic1p, this 13-myc cassette was amplified with oligonucleotides Clb2-MycF and Clb2-MycR for , Cln2-MycF and Cln2-MycR for , and Sic1-MycF and Sic1-MycR for . The resulting PCR products were cotransformed in yeast with the plasmid pAG7578 (for or pAG5016 (for ) to produce pCLB2-13myc and pCLN2-13myc. was tagged at its endogenous locus by direct transformation with the PCR fragment. pCLB2-13myc was digested with BglI and NcoI, and pCLN2-13myc was digested with ApaI before the transformation of cells to tag the endogenous cyclin genes, creating ASG43 and ASG60, which were verified with oligonucleotides Clb1-Myc-I or Cln2-Myc-I and MycI. To generate the ASG46 strain, the plasmid pHHF1 was digested with BglII and SacI. Blunt ends were generated using the 5′–3′ exonuclease activity of Vent polymerase. The 1,747-bp fragment was subcloned into pAIC opened at the ScaI site (). The new plasmid pHPH001 carried a COOH-terminal fusion of GFP (S65T) to the ORF of histone H4 with flanking homologies to the locus. HHF1-GFP was under the control of the promoter and was terminated by . 5 μg pHPH001 was isolated from digested with EcoRI–HindIII, and transformed into the partially deleted locus of the strain (). Transformants were obtained on minimal medium lacking adenine (ASG46). The -tagging cassette for was generated by digesting pRedStar2 (provided by M. Knop, European Molecular Biology Laboratory, Heidelberg, Germany; ) with BamHI and SphI and ligating the purified 3-kb fragment to the 1.5-kb fragment generated by digesting pGEN3 with BamHI and SphI. This was used as a template to amplify using oligonucleotides with homology to the COOH terminus of Ag, SPC42RedS1, and SPC42RedGS2. The resulting PCR product was cotransformed into yeast cells together with pAG19228 to generate pAG-SPC42RFP, which was then digested with XbaI and MscI to transform strain ASG46 and tag the endogenous gene with to produce ASG50, which was verified with oligonucleotides Red I and SPC42 I. deletion mutants were made using pairs of gene name S1/S2 oligonucleotides for each locus, which contained 45-bp homology upstream and downstream of the ORFs. Wild-type cells were transformed with PCR products amplified off the p template with the S1/S2 oligonucleotide pairs. The primary transformation produces heterokaryon cells, which have a mixture of wild-type and transformed nuclei; thus, even mutations in essential genes produce viable transformants. Transformants were verified with oligonucleotide gene name G1/G4/I, and three independent transformants were characterized for each mutant. To evaluate phenotypes of lethal mutants, the heterokaryon was sporulated to produce uninucleated spores, which were then germinated under selective conditions. The db mutants were made using an overlap PCR approach, which deleted amino acids 27–36 (db1) and/or amino acids 84–96 (db2) in AgClb1/2. To generate pAKH47 (db1), pCLB2-13myc was used as a template for a PCR reaction with clb2dbA and clb2dbB to produce a 643-bp product and a second reaction with clb2dbC and clb2dbD to create a 537-bp product. Oligonucleotide clb2dbA was homologous to a region 500 bp upstream of the D box, and clb2dbD was homologous to a region 600 bp downstream of the D box. Oligonucleotide clb2dbB contained homology to immediately upstream of the D box but lacked the D-box sequences between nt 81 and 108. Additionally, the 3′ end of clb2dbB has homology to the region immediately downstream of the D box, which is also present in the 5′ end of oligonucleotide clb2dbC. This overlapping homology was then used in a second PCR reaction involving the two products of A + B and C + D and additional clb2dbA and clb2dbD primers to create a 1.1-kb fragment containing an in-frame deletion of the D box (db1) and part of the AgCLB1/2 ORF. To generate a full-length db mutant, this 1.1-kb fragment was cotransformed into yeast strains with pCLB2-13myc, which had been digested with NheI located just upstream of the D box. The gap-repaired plasmid containing the deleted D box was confirmed by sequencing and was called pclb2Δdb1-13myc. The GFP-lacI-NLS integration cassette for was generated with pLKL55Y, which was provided by C. Pearson and K. Bloom (University of North Carolina, Chapel Hill, NC). pLKL55Y was digested with Sap1 and HindIII to generate a 2-kb fragment, and blunt ends were generated with Vent polymerase and ligated into the Sca1 site of the pAIC plasmid. This produced two plasmids, pAKH35 and pAKH36, which were verified by sequencing and differ only in the orientation of lacI-NLS-GFP. pAKH35 and pAKH36 were digested with HindIII–Not1 and transformed into the partially deleted locus of the strain () to generate AKHAg26 and AKHAg27, which contain the lacI-NLS-GFP integrated at the ADE2 locus. These strains were verified by PCR and showed diffuse, nuclear GFP signal. For construction of the forced localization cassettes of the mitotic cyclin , the pAKH20 and pAKH21 cassettes were amplified with the oligonucleotides Clb-NES-P1 and Clb-Myc-P2, each giving rise to a 2,800-bp fragment. These PCR fragments were cotransformed into yeast cells together with pAG7578 to generate pAKH25 and pAKH26. pAKH25 and pAKH26 were digested with Stu1, NcoI, and BsrG1, and the resulting DNA was transformed into cells, resulting in the strains AKHAg19 and AKHAg22. These new strains were verified using oligonucleotide CLB1/2-I1 and G3. cells were processed for immunofluorescence as described for yeast cells () with slight modifications. Young mycelium containing ∼75–100 nuclei were fixed for 1.5 h in 3.7% formaldehyde (Fluka) and digested in 1.0 mg/ml zymolyase + 1% β-meracptoethanol for 30–45 min before antibody incubation. Anti-myc and tubulin stainings were performed sequentially so as to limit cross reactivity, beginning with mouse anti-myc (Santa Cruz Biotechnology, Inc.) at 1:100, AlexaFluor488 anti–mouse (Invitrogen) at 1:200, rat antitubulin (YOL34; Serotec) at 1:50, and AlexaFluor568 anti–rat (Invitrogen) at 1:200 with Hoechst (Invitrogen) dye to visualize nuclei at 5 μg/ml. Rabbit anti-GFP (Invitrogen) was used to detect GFP-lacI-NLS at a 1:100 dilution, and AlexaFluor488 anti–rabbit was used at a 1:200 dilution (Invitrogen). Antibody dilutions and washes were made with PBS + 1.0 mg/ml BSA. The microscope used for all fixed cell images (immunofluorescence and Hoechst stainings with cells mounted in standard fluorescent mounting medium containing 1 mg/ml p-phenylenediamine in 90% glycerol) was essentially as described in and consisted of an imaging microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) with plan Neofluar 100× Ph3 NA 1.3 objectives. It was equipped with a 75 WXBO and a 100 WHBO illumination source controlled by a shutter and filter wheel system (MAC2000; Ludl Electronics). The camera was a back-illuminated cooled CCD camera (TE/CCD-1000PB; Princeton Instruments). The following filter sets for different fluorophores were used: No. 10 for AlexaFluor488 and No. 20 for rhodamine/AlexaFluor568 (Carl Zeiss MicroImaging, Inc.); and No. 41018 for GFP (Chroma Technology Corp.). The excitation intensity was controlled with different neutral density filters (Chroma Technology Corp.). MetaMorph 4.6r9 software (Universal Imaging Corp.) controlled the microscope, camera, and Ludl controller and was used for processing images. The microscope used for time-lapse images was a similarly equipped microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) with a plan Neofluar 63× NA 1.25 or 100× (same as above) objective. The 100× objective was driven by a PIFOC P721.10 objective nanopositioner (Piezo device) to move quickly through multiple planes in the Z axis (Physik Instrumente). A Micromax back-illuminated CCD camera (512BFT; Princeton Instruments) was used for rapid acquisition and transfer of images. Stacks of tubulin images were processed using the “no-neighbors” deconvolution within MetaMorph, and, like all other fixed specimens, the Z-stacks were compressed using stack arithmetic, brightness and contrast were adjusted, and images were overlayed using color align. Time-lapse conditions used XBO fluorescence at 5–10% intensity and exposures of 100 ms that were acquired every 30 s. For HHF1-GFP videos, a 63× objective was used, and a single-phase image with transmitted light was taken followed by 5 × 0.5–μm images in the Z axis with the EGFP filter and fluorescent light. For videos, a single Hhf1-GFP image was acquired followed by 16 × 0.5–μm images in the Z axis with the rhodamine filter to capture the dynamics of Spc42-RFP using the 100× objective. Focusing was automatically performed using the find focus function in MetaMorph software, and for videos, a Piezo motor directed the objective to allow more rapid movement through the Z axis than with the motorized stage. Images were acquired at a constant ambient temperature of 25–27°C. Figs. S1 shows still images from a time-lapse video observing SPBs and nuclei. Fig. S2 shows the categorization of nuclei that lack detectable cyclin signal according to nuclear cycle stage. Fig. S3 shows that causes a transient delay in the budding yeast cell cycle. Table S1 presents the plasmids that were used in this study. Table S2 presents the oligonucleotide primers used in this study. Video 1 shows the dynamics and divisions of nuclei in a growing hyphal tip. Online supplemental material is available at .

Cytokinesis is the physical cell division process that follows the duplication and spatial segregation of the genetic material (). In animal cells, a complex microtubule structure termed the central spindle or spindle midzone forms between the retreating chromosomes in anaphase and organizes many proteins essential for the process of cytokinesis (). One of the critical components of this structure is the microtubule-associated protein (MAP) PRC1 (protein-regulating cytokinesis 1; ; ; ). PRC1 not only promotes the formation of stable microtubule bundles () but also acts as a scaffold for the kinesin-4 family chromokinesin KIF4 (; ). In PRC1-depleted cells, KIF4 fails to localize to the central spindle, and, rather than forming a dense array, the central spindle microtubules spread out toward the cell cortex (). Like PRC1, KIF4 is required for cytokinesis (; ; ) and for chromosome condensation and segregation (); however, the molecular details of these different functions are not understood. More is known about the molecular functions of the kinesin-6 family motor proteins MKlp1 and MKlp2, which also have essential functions in cytokinesis (; ; ; ). MKlp1 forms a heterotrimeric complex with the Rho GTPase–activating protein Cyk-4 and the Rho guanosine diphosphate–GTP exchange factor ECT2 (; ; ). This complex is needed to control the activation state of Rho at the cell cortex during the formation and ingression of the cleavage furrow (; ). In vertebrates, MKlp2 is a docking partner of polo-like kinase 1 (Plk1) at the central spindle () and is also required for the transport of Aurora B kinase to the central spindle in anaphase cells (). In the absence of MKlp2, Aurora B is unable to phosphorylate its targets at the central spindle, including MKlp1, and cytokinesis fails. Whether or not PRC1 has a general function in regulating these mitotic kinesins in addition to KIF4 is not known. Therefore, we have further investigated the function of the central spindle MAP PRC1 and uncovered links to multiple mitotic kinesins with reported functions in cytokinesis, including KIF4, MKlp1, and MKlp2. In addition, we identify a novel regulatory pathway in human cells involving citron kinase and the kinesin-3 family motor protein KIF14. To investigate the function of PRC1 during cytokinesis in human cells, it was immunoaffinity purified from anaphase cell extracts, and the obtained complexes were analyzed by mass spectrometry (). This approach revealed that PRC1 complexes contain KIF4 and MKlp1 as expected as well as MKlp2 and KIF14 (). These proteins specifically precipitate together with PRC1 because the antibody used for the immune precipitation recognizes a doublet band at the size expected for the two PRC1 splice variants expressed in HeLa cells; this is depleted in cells treated with small interfering RNA (siRNA) duplexes targeting PRC1 (). The presence of KIF14 was unexpected because it was previously reported to function in chromosome segregation but not cytokinesis in a high through-put RNA interference screen (). Analysis of KIF14 complexes with specific antibodies confirmed that PRC1 was present (). Furthermore, KIF14 was found to localize to the central spindle and midbody () and, like KIF4, to depend on the presence of PRC1 for this localization (; ). These observations suggest that KIF14 might function in cytokinesis. To test this idea, KIF14 was specifically depleted from cells using siRNA duplexes (). After 48 h of transfection, Western blotting for KIF14 showed that it was efficiently depleted (), central spindle staining was absent from anaphase cells (), and >60% of KIF14-depleted cells were binucleated (). No obvious defects in micro-tubule structures () or chromosome segregation were observed (unpublished data), and, consistent with this, KIF14-depleted cells had two equally formed nuclei rather than unequal or micronuclei (). KIF4 is thought to be important for both chromosome structure and cytokinesis, and this dual function is reflected in its localization to both chromosomes and the central spindle (; ). If KIF14 had a similar dual function, it, too, would be expected to display a similar pattern of localization. Like PRC1, KIF14 was present on the spindle midzone throughout mitosis but was highly enriched on the midbody during cytokinesis, whereas PRC1 remained spread along the microtubule bundles and apparently did not concentrate at the midbody (). In metaphase cells, the bulk of KIF14 was cytoplasmic, and although a weak spindle microtubule staining was observed, it did not obviously localize to the chromosomes (). To further support the specific nature of this localization, mapping experiments were performed to identify the central spindle–targeting region of KIF14 (). The unique amino-terminal extension of KIF14 adjacent to the motor domain was found to localize to the central spindle and midbody (). None of the other regions of KIF14 displayed the same localization, although motor domain–containing fragments weakly associated with the spindle during metaphase (). At no time was any association with the chromosomes observed. Because KIF14 and PRC1 are present in a complex and have similar localizations, the role of PRC1 in KIF14 targeting was then investigated. A biochemical analysis of KIF14 using immune precipitation of tagged constructs revealed that amino acids 1–356, the extension amino terminal to the motor domain, also form the minimal region tested that was capable of binding to PRC1 (). In contrast, the forkhead-associated (FHA) domain-containing stalk and tail region of the motor was unable to bind PRC1 and also failed to target to the central spindle (). Therefore, the PRC1-binding site of KIF14 maps to the region containing the central spindle-targeting signal. This is different from the kinesin-6 family motors MKlp1 and MKlp2, which target to the central spindle via their carboxy-terminal stalk and tail domains (). Interestingly, in neither case does it appear that the kinesin motor domain is needed for specifying central spindle localization. Together, these findings indicate that KIF14, like PRC1, is required for cytokinesis but raise questions about the previously reported function in chromosome segregation (). To establish why cytokinesis depends on KIF14, the localization of a panel of central spindle and cleavage furrow proteins was examined in KIF14-depleted cells (). In KIF14-depleted cells, citron kinase was lost from the central spindle and cleavage furrow region (). This was a specific effect because the central spindle proteins MKlp1, MKlp2, KIF4, ECT2, and the cleavage furrow proteins anillin, actin, and RhoA all displayed the expected normal localizations in both control and KIF14-depleted cells (). To test whether this was caused by a direct interaction between KIF14 and citron kinase, different central spindle proteins were immune precipitated from anaphase cell extracts using specific antibodies (). Consistent with the results presented in , KIF14, KIF4, MKlp1, and MKlp2 could be detected in PRC1 immune precipitates (). Furthermore, PRC1 could be detected in KIF14 immune precipitates (). Citron kinase was only found in PRC1 and KIF14 immune precipitates (), suggesting that it binds to KIF14 but not other mitotic kinesins. Finally, astrin, another mitotic spindle MAP, did not precipitate with PRC1, citron kinase, or any of the motor proteins tested (), supporting the idea that mitotic kinesins make specific interactions with PRC1. Fragments of KIF14 were then tested for their ability to interact with citron kinase, to provide more evidence for the direct interaction of these proteins, and to map the interaction site on KIF14 (). For this purpose, cells were transfected with GFP-tagged KIF14 and FLAG-tagged wild-type and kinase-dead forms of citron kinase. GFP-tagged KIF14 was only detected in immune precipitates with the kinase-dead form of citron kinase but not the wild-type kinase or the control protein INCENP (inner centromere protein; ). This was despite the fact that kinase-dead citron kinase was typically expressed at a lower level than the wild-type protein (; compare the first two lanes), possibly because of its toxic effects on cell growth and division (). A similar result was seen for endogenous KIF14, which was also only found to efficiently coprecipitate with the kinase-dead form of citron kinase (). This interaction was not mediated by the amino-terminal extension involved in binding to PRC1 or the kinesin motor and FHA domains but required amino acids 901–1,649 forming the stalk and tail region (). Therefore, KIF14 makes interactions with at least two other components of the central spindle components. First, amino acids 1–356 interact with PRC1, and this is needed for central spindle targeting of the motor protein. Second, the stalk and tail region encompassing amino acids 901–1,649, although unable to target to the central spindle alone, form the binding site for citron kinase. Interestingly, this interaction appears to depend on the activation state of citron kinase, although the significance of this is unclear at present. The inactive mutant of citron kinase has a dominant-negative effect on cytokinesis (), and the results presented here suggest this may, in part, be caused by sequestration or improper regulation of KIF14. Citron kinase has been reported to be required downstream of Rho for the final steps of cytokinesis in (; ). However, in mammalian cells, the situation is a little more confusing because it is unclear if citron kinase is an essential component of the cytokinesis machinery (; ). Thus, this issue was reinvestigated in HeLa cells using siRNA duplexes to deplete citron kinase (). Western blotting showed that citron kinase was efficiently depleted from HeLa cells using siRNA without altering the levels of KIF14 or α-tubulin (). After 50 h of citron kinase depletion, a 15-fold increase in binucleated cells was found compared with the control (), but no obvious defects on chromosome segregation or cleavage furrow ingression were observed (, telophase figures; and not depicted). KIF14 targeting to the central spindle and midbody was significantly reduced in these cells (). Formation of the MKlp1-containing midbody matrix structure is one of the latest stages of cytokinesis before abscission (, ). Interestingly, the binucleated cells formed after the depletion of citron kinase or KIF14 contain MKlp1-positive midbody remnant structures (, arrows). This is different from cells depleted for MKlp2 or PRC1, which fail to form a midbody (; ; ). These observations suggest that citron kinase is required for the localization of KIF14 and that citron kinase and KIF14-depleted human cells fail to complete cytokinesis at a late stage after or in parallel with formation of the midbody. In this study, we have identified a network of interactions between KIF14, citron kinase, the central spindle MAP, and the microtubule-bundling protein PRC1. Although it has recently been reported that KIF14 is needed for chromosome segregation (), our data indicate that it is, in fact, required for the process of cytokinesis. This is consistent with observations made on the KIF14 homologue KLP38B (). Mutations in the gene result in the formation of polyploid cells without causing defects in other aspects of mitosis such as chromosome segregation (). Although it is unknown whether KLP38B interacts with , the homologue of citron kinase, mutants and insect S2 cells in culture depleted of using RNA interference have severe defects in cell division similar to those we reported in human cells (; ; ; ). The presence of an FHA domain is one of the defining features of the kinesin-3 family motor proteins (; ). FHA domains are phosphopeptide-binding motifs (), implying that KIF14 either makes phospho-dependent interactions with other proteins or itself. Neither the interaction of KIF14 with PRC1 nor citron kinase appears to depend on the FHA domain, and, to date, we have been unable to identify any other interaction partners of KIF14. Observations on KIF1A, another kinesin-3 family motor protein in which the FHA domain negatively regulates the kinesin motor activity (), support the idea that this could also be a regulatory feature in KIF14. The observations that KIF14 interacts more strongly with the kinase-dead form of citron kinase perhaps indicate that it is a substrate for this kinase. However, preliminary observations indicate that this may not be the case (unpublished data), and this issue requires further investigation. An alternative possibility is that citron kinase interacts with KIF14, depending on its activation state. It has been reported that citron kinase acts downstream of Rho and depends on Rho for its activity (; , ; ; ), so this could provide a regulatory pathway modulating the interaction of citron kinase with KIF14. Citron kinase is reported to phosphorylate myosin light chain and, thus, exert a regulatory effect on actomyosin-mediated contractility of the cleavage furrow (; ). How this relates to KIF14 function is unclear because KIF14-depleted cells do not appear to have defects in the localization of central spindle and cleavage furrow components other than citron kinase (). One explanation for this could be that human citron kinase is only required very late in cytokinesis, perhaps to control the final stages of abscission in cooperation with anillin as in insect cells (; ). These findings may also have a wider relevance because KIF14 is encoded on chromosome 1 within the minimal region at 1q31-1q32 amplified in a wide variety of different tumors (). Of the 14 genes encoded in this region, only KIF14 was overexpressed in tumors and tumor cell lines, indicating it may be an important factor in oncogenesis (). Defective regulation of cytokinesis has been proposed to be a possible route to the generation of aneuploid cells and, hence, tumorigenesis (). Our observations that KIF14 is needed for efficient cytokinesis, together with the finding that it is overexpressed in some human tumors, support this idea and indicate that KIF14 and other components of this pathway such as citron kinase may be potentially interesting targets for therapeutic intervention. We have previously shown that MKlp2 is necessary for the relocation of the Aurora B kinase from the centromeres to the central spindle at the metaphase to anaphase transition () and is also an important docking partner for Plk1 at the central spindle (). Our observations on KIF14 and citron kinase suggest that spatial control of protein kinases is a general principle of mitotic regulation (). Based on the data presented here and published observations on the insect homologue of citron kinase (; ), we propose a model for KIF14 and citron kinase function (). In this model, the heterotrimeric centralspindlin complex of MKlp1, Cyk-4, and ECT2 locally regulates the activation state of Rho family GTP-binding proteins at the cleavage furrow during furrow ingression (; ; ). KIF14 is required for the localization of citron kinase, whereas Rho controls the activation state of citron kinase and, hence, the phosphorylation of myosin light chain (; ; ). PRC1 can be seen as a master regulator or central spindle matrix protein that is required not only to organize antiparallel microtubule bundles () but also to localize the many motor proteins associated with this structure. This is supported by our observations that endogenous PRC1 interacts with MKlp1, MKlp2, and KIF14 ( and ) as well as KIF4 (). Together, these pathways, each defined by a different kinesin motor protein, contribute to the temporal regulation of cleavage furrow ingression and the final stages of cytokinesis. Whether citron kinase has substrates in addition to myosin light chain and the role these might play in controlling cytokinesis are the subjects of further investigation. These findings add another layer to the complex regulatory network of MAPs, kinesin motor proteins, and mitotic kinases necessary for the process of animal cell division. Antibodies were obtained as follows: mouse monoclonals to α-tubulin (clone DM1a; Sigma-Aldrich); Aurora B (clone AIM-1; Becton Dickinson); mouse monoclonal to citron kinase (Becton Dickinson); mouse monoclonal to Rho A (clone 55; Becton Dickinson); rabbit polyclonals to MKlp1 SC-867 (0.2 mg/ml; Santa Cruz Biotechnology, Inc.); the MKlp1 motor domain (); the KIF4 amino acids 738–1,232; KIF14 (BL358; Bethyl Laboratories Inc.); ECT2 amino acids 1–388; INCENP (); astrin amino acids 1,014–1,193; anillin amino acids 417–656; and affinity-purified sheep polyclonal to MKlp2 (; ). PRC1 antibodies were raised in rabbits against the full-length hexahistidine-tagged protein purified from insect cells and were affinity purified against maltose-binding protein–tagged full-length PRC1 expressed in bacteria. Antibodies are directed to human proteins unless indicated otherwise. Secondary antibodies conjugated to HRP, CY2, and CY3 were obtained from Jackson ImmunoResearch Laboratories. KIF14 was amplified from a human testis cDNA library (CLONTECH Laboratories, Inc.) using pfu polymerase (Stratagene). PRC1 constructs have been described previously (). Mouse citron kinase constructs were obtained from F. Di Cunto (University of Torino, Torino, Italy) and have been described previously (). All constructs were confirmed by DNA sequencing (Medigenomix). Mammalian expression constructs were made in pEGFP-C2 (CLONTECH Laboratories, Inc.) or pcDNA3.1+ (Invitrogen) modified to encode either myc- or triple FLAG-epitope tags. HeLa S3 cells were grown and arrested with 1.6 μg/ml aphidicolin for 19 h, released for 6 h in fresh growth medium, and arrested for 14 h with 100 ng/ml nocodazole. Mitotic cells obtained by shake-off were plated in fresh growth medium and released for 70 min. Cell pellets were lysed in lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 40 mM β-glycerophosphate, 10 mM NaF, 1% [vol/vol] IGEPAL, 0.1% [wt/vol] deoxycholate, 100 μM ATP, 100 μM MgCl, 2 mM Pefabloc, and complete protease inhibitor cocktail [Roche Diagnostics]). For immune precipitations, 2 μg of affinity-purified antibody, 20 μl protein G– or A–Sepharose beads, and 8 mg of extract in a total volume of 500 μl were incubated for 2 h at 4°C. Beads were washed twice with 1 ml lysis buffer and twice with 1 ml wash buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 40 mM β-glycerophosphate, 10 mM NaF, 0.1% [vol/vol] IGEPAL, and 1 mM Pefabloc). HEK293T cells plated on 15-cm–diameter dishes were transfected using 8 μg of the required plasmid DNA and 24 μl Fugene-6 (Roche Diagnostics) according to the manufacturer's instructions. After 40 h, cells were washed three times in ice-cold PBS and were lysed in immunoprecipitation (IP) buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% [vol/vol] IGEPAL, 2 mM Pefabloc, and complete protease inhibitor cocktail). Beads were washed twice with 1 ml IP buffer and twice with 1 ml of wash buffer. Bound proteins were eluted in 50 μl of sample buffer and were analyzed by SDS-PAGE followed by mass spectrometry () or Western blotting. HeLa S3 and U2OS cells were cultured at 37°C and 5% CO in DME containing 10% FCS. Plasmid transfection and RNA interference were performed as described previously (). Proteins were targeted with the following sequences: MKlp2 with 5′-AACCACCTATGTAATCTCATG-3′; MKlp1 with 5′-AAGCAGTCTTCCAGGTCATCT-3′; PRC1 with 5′-AAGGCTTCTAGGCGTGAGGAG-3′; KIF14 with 5′-TTCCCGATCTCATTCAGTTTT-3′; and citron kinase with 5′-ATGGAAGGCACTATTTCTCAA-3′. The GL2 and lamin A controls were described previously (). For Western blotting, cells from three wells of six-well plates were washed in 2 ml PBS and lysed in 70–80 μl of 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 0.1% (wt/vol) Triton X-100. For each lane of a minigel, 10 μg of the protein lysate was loaded. Cells to be imaged were fixed for 12 min in PTEMF (20 Pipes-KOH, pH 6.8, 0.2% [vol/vol] Triton X-100, 10 mM EGTA, 2 mM MgCl) and blocked with 2% (wt/vol) BSA in PBS. Cells to be stained with phalloidin to detect actin were fixed for 20 min with 3% (wt/vol) PFA, quenched for 10 min with 50 mM ammonium chloride, and permeabilized with 0.1% (vol/vol) Triton X-100 for 5 min. For RhoA staining, cells were fixed and permeabilized by incubation in 10% (wt/vol) tricholoracetic acid for 15 min. All solutions were made in PBS, and antibody staining was performed for 60 min using a 1,000-fold dilution of antiserum or purified antibody at a final concentration of 1 μg/ml. Coverslips were mounted in 10% (wt/vol) Moviol 4–88, 1 μg/ml DAPI, and 25% (wt/vol) glycerol in PBS. Images were collected using a microscope (Axioskop-2; Carl Zeiss MicroImaging, Inc.) equipped with a 63× NA 1.4 plan Apochromat oil immersion objective and standard filter sets (Carl Zeiss MicroImaging, Inc.), a 1,300 × 1,030 pixel cooled CCD camera (CCD-1300-Y; Princeton Instruments), and Metavue software (Visitron Systems). Images were cropped in Adobe Photoshop 7.0 and were sized and placed using Adobe Illustrator 10.0 (Adobe Systems).

In eukaryotic organisms, transcription is spatially separated from translation by a nuclear envelope. Consequently, gene expression requires nuclear export of mature mRNA. Although the distribution of individual mRNA export factors has been studied, as has that of several nuclear mRNAs, the use of bimolecular fluorescence complementation (BiFC) analysis makes it possible to study the in vivo formation of complexes between different export factors that evidence indicates are functionally associated with RNA. We have used this approach to study the distribution, dynamic behavior, and relationship of Y14–nuclear export factor 1 (NXF1) complexes to RNA synthesis. The assay relies on the reconstitution of fluorescent YFP by the association of two nonfluorescent YFP half-molecules, each linked to one of two proteins, whose interactions are of interest (). Evidence indicates that many or all of the complexes visualized are associated with RNA. Thus, monitoring the interaction of Y14 and NXF1 by BiFC indirectly allows the observation of potentially export-competent mRNA. Y14 is known to bind mRNA as part of the exon–exon junction complex (EJC) at a late stage of splicing () and remains bound to mRNA until translation in the cytoplasm (). Bound to the EJC, NXF1 (also called TAP) promotes export of the mature mRNA (for reviews see ; ). We show that coexpression of the two modified proteins, YC-Y14 and YN-NXF1, carrying the COOH- and NH-terminal parts of YFP, respectively, allows observation of a characteristic BiFC pattern in cell nuclei. Unexpectedly, BiFC fluorescence accumulated in speckle-associated patches, suggesting an active role for speckles in mRNA processing, although they are otherwise considered mainly as storage sites for splicing and export factors (). Findings also provided insight into the idea that the nuclear retention of RNA is one way in which nature regulates gene expression. Concordantly, it had been found that only a small fraction of all transcribed RNA is exported to the cytoplasm, although most of nuclear polymerase II–derived RNA is maturely spliced and polyadenylated (; ; ). Studies using BiFC to visualize Y14–NXF1 export complexes provide new evidence relating to the nuclear retention of mRNA in vivo. Upon cotransfection of YC-Y14 and YN-NXF1, MCF7 cells emitted YFP fluorescence depending on BiFC maturation for 2 h at 30°C (). Fluorescence was observed in >90% of the cells. The signal was characterized by its nuclear localization and the composition of patchy accumulations embedded in a diffuse background. In nucleoli, the signal level was very low. Immunostaining of the YC epitope (the COOH-terminal area of YFP) essentially colocalized with the BiFC pattern (). Y14 tagged by full-length YFP displayed a similar pattern, except that it also stained nucleoli (, YFP-Y14). In contrast, patchy accumulations were less obvious with YFP-tagged NXF1, where focal accumulations aligned at the nuclear periphery appeared as a characteristic expression pattern (, YFP-NXF1). When the fusion partners YC-NXF1 and YN-Y14 were cross exchanged, YFP fluorescence was not reconstituted, even though transfection efficiency as revealed by YC immunostaining was still high (unpublished data). Furthermore, BiFC as obtained with YC-Y14 and YN-NXF1 did not occur if NXF1 was mutated by deletion of the NH-terminal 60 or 371 amino acids (). Deletion of the first 60 amino acids of NXF1, which are known to interact with the EJC protein E1B-AP5 (), is supposed to prevent BIFC formation caused by disturbed interaction with Y14, whereas the deletion of the NH-terminal 371 amino acids totally compromises interaction with RNA (; ; Liker et al., 2000). With both mutations, BiFC occurred in <10% of the cells, even though transfection efficiency was >90%, as determined by YC immunodetection, and the nuclear distribution, which was monitored by full YFP tags (YFP-NXF1 Δ1–371 and YFP-NXF1 Δ1–60), was not seemingly influenced by the mutation (). In the few cases of BiFC, fluorescence was very weak and different from the characteristic patchy pattern of BiFC with full-length NXF1. In particular, BiFC with YN-NXF1 Δ1–371 only occurred in cells with shrunken nuclei and abnormally condensed chromatin. The aforementioned experiments support the hypothesis that functional interaction of the proteins YC-Y14 and YN-NXF1 is a prerequisite for the reconstitution of fluorescent YFP. As such, BiFC should depend on active mRNA transcription. Accordingly, BiFC was compromised upon RNA polymerase II (pol II) inhibition with either 5,6-dichloro-1-β--ribofuranosylbenzimidazole (DRB) or α-amanitin at 37°C, even though expression of the YC epitope still appeared unperturbed (; α-amanitin not depicted). In contrast, pol II inhibition after BiFC maturation at 30°C did not abolish the signal, which remained stable within 5% for a period of at least 6 h (). Immunoprecipitation of YC-Y14 in nuclear extracts of BiFC-matured cells confirmed the expected interactions. Targeted against the YC epitope of YC-Y14, immunoprecipitation coprecipitated YN-NXF1, NXF1, Y14, and radiolabeled RNA (). Signal patches of BiFC from YC-Y14 and YN-NXF1 resembled speckles in respect to size, shape, and number. Speckles are nuclear bodies enriched in splicing and export factors (; ; ). Colocalization of the BiFC patches with speckles was investigated by immunodetection of SC35, a marker protein for speckles (). All BiFC patches colocalized with SC35 domains, and vice versa. However, the BiFC patches were larger than their corresponding SC35 domain, surmounting the speckle perimeter by up to 500 nm. In ∼10% of the cells, the perispeckle signal was even higher than the SC35-positive center. Export capability of the BiFC complexes was investigated in digitonin-permeabilized cells. Permeabilization allows observation of unidirectional nuclear export () because soluble factors necessary for reimport become depleted. Upon permeabilization, the whole nuclear level of BiFC fluorescence decreased by two thirds within 5 min () and remained stable for at least 5 min further. However, the remaining YFP fluorescence was immediately extinguished upon administration of 5 mM ATP. To further verify the role of ATP in releasing the Y14–NXF1 complexes, cells were pretreated with sodium azide to deplete cellular levels of ATP. Sodium azide treatment reduced overall BiFC intensity. In particular, the diffuse background appeared extracted and a diffuse rim emerged at the nuclear periphery. Upon digitonin permeabilization, the signal level decreased by 50%. Again, the administration of 5 mM ATP caused the complete loss of fluorescence (). To measure immobile fractions and the recovery speed of BiFC complexes in speckles and the nonspeckle nucleoplasm, FRAP was performed within 2-μm-diam circular regions of interest (). Immobile fractions measured by FRAP were confirmed by fluorescence loss in photobleaching (FLIP; ). Because temperature dependence was of special concern in this study, FRAP was also measured for YFP-Y14 to observe a possible shift from 37 to 30°C and RT (). Besides a marginal trend of decreased immobile fraction in the cold, temperature had no significant effect within the error of individual measurements. For the BiFC signal, immobile fractions were much higher in speckles (46%) than in the nonspeckle nucleoplasm (21%). In contrast, recovery time of the mobile fraction was similar in both compartments (1.6 s). Sodium azide treatment significantly increased the immobile fractions (74 and 55%) and also decreased the recovery times. In particular, fluorescence recovery was much slower in speckles compared with the nonspeckle nucleoplasm (). Performed in a timescale of 35 s, FRAP measured the mobility of fluorochromes already established at the time of bleaching. Bleaching all fluorescence of the complete nucleus, in contrast, allowed observation of de novo BiFC formation (). It took roughly 2 h for a steady state of restoration, which resembled the initial state. Separate analysis of the time course of recovery within speckles and the nonspeckle nucleoplasm revealed that speckles reached only 80% of their initial value. Recently, in vivo analysis of nuclear mRNA greatly contributed to our current understanding of RNA dynamics. However, measuring the dynamics of processed endogenous mRNA in vivo has been challenging. Studies that visualize mRNA export factors individually cannot distinguish functionally associated reporters from unbound reporters. We used the BiFC approach to visualize the interaction of Y14 and NXF1. Furthermore, evidence indicates that these complexes are associated with nuclear RNA. Thus, monitoring the interaction of target molecules instead of labeling single molecules by fluorescent protein tags or by FISH probes allows discrimination of false positive backgrounds from noninteracting entities. Compared with fluorescence resonance energy transfer, which is another method used to visualize protein–protein interaction in situ, the signal-to-noise ratio is higher in the BiFC experiments because the signal is measured directly. In contrast, fluorescence resonance energy transfer analysis requires cross talk compensation between donor and acceptor channels, which intrinsically increases statistical noise. As NXF1 binds to processed mRNA and mediates the interactions of the export complex with the nuclear pore complex, it is considered a mRNA export receptor (; ; ). The combination of the splicing-associated protein Y14 with NXF1 was thought to label spliced mRNA with putative export capacity. BiFC from YC-Y14 and YN-NXF1 reproducibly demonstrated a specific nuclear distribution of the interacting proteins, revealing spatial association with the speckle compartment in particular. The BiFC approach is particularly well suited for in vivo observations. A peculiar requirement is the maturation at a low temperature (30°C). Because cells still grow and divide at this temperature, mRNA processing and export go on at physiologically acceptable rates. Accordingly, even though we observed a trend in reduction of the immobile fraction of YFP-Y14 in speckle-associated patches, temperature dependence of immobile fractions and recovery times upon photobleaching were not significant. Thus, we consider the shift to 30°C of marginal influence on the analysis of RNA maintenance by the BiFC assay. Several experiments were performed to demonstrate that the BiFC assay labels functional interaction of the two proteins with RNA. First, the cross exchanged pair YN-Y14–YC-NXF1 did not reconstitute YFP, indicating that BiFC depended on the sterical configuration of the linked proteins, Y14 and NXF1, and that the YC and YN parts did not interact in solution to reconstitute fluorescence. Second, the characteristic BiFC pattern was not observed if the NH-terminal 60 or 371 amino acids of NXF1 were deleted, even though topological expression of the proteins was not affected. The NH-terminal part of NXF1 is responsible for RNA binding and interaction with the EJC (; ; ). According to , deletion of the first 371 amino acids blocks nuclear export and impairs viability of the cells. This may explain the bad condition of many cells transfected with YN-NXF1 Δ1–371. BiFC fluorescence, with YN-NXF1 Δ1–371 exclusively seen in shrunken nuclei, may have evolved because of aggregation of the BiFC chimeras. The first 60 amino acids are involved in interaction of NXF1 with the EJC protein E1B-AP5 (). Because NXF1 Δ1–60 still binds to RNA (, ), interaction of YN-NXF1 with YC-Y14 may occur with low efficiency, explaining the few cases where BiFC fluorescence could be monitored, although at very low intensity. Third, immunoprecipitation of YC-Y14 from nuclear extracts not only coprecipitated YN-NXF1 but also Y14, NXF1, and radiolabeled RNA, demonstrating that all expected interactions of YC-Y14 did occur. In particular, this experiment shows directly that the YC tag does not prohibit interaction with NXF1. Fourth, BiFC from YC-Y14 and YN-NXF1 requires pol II activity. Inhibition of pol II by DRB or α-amanitin before BiFC maturation prohibited fluorescence reconstitution, whereas already established fluorescence was not affected by pol II inhibition. All together, these experiments suggest that BiFC of YFP from YC-Y14 and YN-NXF1 monitors endogenous mRNA at a terminal state of maturation. Because BiFC-YFP represents the interaction of YC-Y14 and YN-NXF1, the characteristic BiFC pattern should be part of both YFP-Y14 and YFP-NXF1 distributions. This was the case, as both constructs showed nuclear location with accumulation in specklelike patches. Differences from the characteristic BiFC pattern may be attributable to entities that are not functionally associated with RNA, such as the nucleolar localization of YFP-Y14. The focal accumulations of YFP-NXF1 at the nuclear edge may represent cytoplasmic entities that become docked to nuclear pores before import (). The BiFC complexes accumulated within speckles and the perispeckle region, suggesting a functional role of the speckle compartment in mRNA processing, including maturation and/or transport. Furthermore, the pronounced enrichment of BiFC complexes in the perispeckle space indicates that speckles associate with a specific nuclear environment. The accumulation of RNA in speckles had already been observed for both microinjected and endogenous RNA that were labeled by FISH with either an oligo d(T) probe or sequence-specific probes (; ; ; ). Because BiFC has a maturation time that is >2 h (), the observed accumulation in speckles derives from complexes of rather long nuclear residence time. The poor mobility of BiFC complexes was experimentally revealed by FRAP and FLIP and was further confirmed in cell permeabilization assays. Upon cell permeabilization, roughly two-thirds of the BiFC signal was eluted. The remaining fluorescence of an apparently immobile fraction completely disappeared upon administration of ATP. One might argue that ATP destroys the BiFC complexes. However, YN-NXF1 still coprecipitates with YC-Y14 upon ATP administration, indicating that the effect of ATP is the release of the complexes. ATP-dependent mobilization of splicing- and export-related factors were described for the alternate splicing factor () and the EJC protein SRm160 (). ATP-dependent export was also observed for in vitro–transcribed rab11 mRNA (). In cells treated with sodium azide, the diffuse background of BiFC complexes strongly decreased. The speckle-associated accumulations became more prominent, and a rim emerged at the nuclear periphery. This reorganization may be caused by the extraction of a soluble part of BiFC complexes or by the reduced release of bound complexes. Accordingly, a lower fraction of fluorescence was eluted upon digitonin permeabilization. The peripheral rim was broad and not sharply delimited, unlike the focal accumulations of NXF1. It may represent a zone of weak binding for mRNA, which only becomes visible when the diffuse background of presumptively mobile BiFC complexes disappeared. In vivo analysis of mobility by FRAP was kept simple. Half-life recovery and immobile fractions were read out directly from the recovery plots of bleach-corrected raw data. Neither curve fitting to decipher the complexity of presumptively superposed events nor modeling of diffusion constants wer performed. Thus, the data presented are valuable for direct comparison only. Recovery of YFP-Y14 was faster, compared with BiFC-YFP. Because the YFP-Y14 signal does not represent a pure population of complexed molecules, a fraction of unbound molecules will account for this apparent faster mobility. One might also argue that BiFC formation, by itself, impedes mobility of the complexes. Measured by FRAP or FLIP, roughly half of the BiFC-forming complexes in the speckle-associated patches were immobile, and the immobile fraction was lower in the nonspeckle nucleoplasm, suggesting a functional association of BiFC complexes with speckles. Further evidence for functional association with speckles compared with haphazard immobilization caused by the artificial BiFC tag is provided by the efficient release of immobile BiFC complexes upon ATP administration in permeabilized cells. Upon incubation of cells in sodium azide, most of the BiFC signal became immobile. According to the ATP-dependent release in permeabilized cells, ATP depletion may have impaired the release of bound complexes rather than decreased the nuclear export rate. FRAP of diffusible molecules is a fast process. In our study, an approximately steady state was reached within a period of 35 s. In contrast, bleaching the entire pool of fluorescence in a cell allows the observation of fluorescence turnover instead of diffusion. Upon total bleach of a nucleus, the initial BiFC pattern recovered within a period of ∼2 h, demonstrating its turnover even within BiFC patches. Previous studies reported that speckles contain poly(A) RNA (; ). As shown by and , at least part of this speckle RNA comes from transcription of protein-coding genes. Although possibly all actively transcribed genes associate with the surface of speckles (; ), some genes' interaction with speckles is particularly intimate, as their transcription products pervade the associated speckle (“typeI genes;” ). Transcripts can even move over and accumulate in the entire speckle compartment of the nucleus (). Enrichment of messengers in speckles, however, appears to depend on the particular conditions of either the cell state or the transcript. Thus, observable accumulation throughout the entire speckle compartment of the induced gene p21 transcripts occurred in <20% of the induced cells (). Comparing the distribution of collagen 1α1 and 1α2 transcripts in cultured fibroblast nuclei, found significant different distribution patterns. The collagen 1α1 product pervades the entire associated speckle. Because transcripts of a splice mutation of collagen 1α1 that never leave the nucleus also pervade the entire associated speckle (), one may argue that speckle-associated mRNA is retained before export. This also explains why their prevalence remains practically stable upon transcripition inhibition (; ). Long residence time is also a characteristic feature of the observed BiFC complexes because of the duration of fluorescence maturation. There may be complexes that exit the nucleus quicker. These, however, would not be visible by the assay, as the complex is supposed to decay in the cytoplasm. Although we cannot exclude the possibility that the BiFC complexes are artificially bound within speckles because of their tag, they may equally represent a fraction of normally retained RNA. In any case, the assay reveals a possible involvement of the speckle-associated nuclear space in retention of processed mRNA. BiFC of YFP from YC-Y14 and YN-NXF1 indicates that mammalian cell nuclei harbor NXF1-associated mRNA with very long residence times, implying that association of RNA with NXF1 is insufficient for immediate export. Because the BiFC signal can be released by ATP, an ATP-dependent mechanism may be involved in the export prohibition of NXF1-associated RNA. Full-length NXF1 and Y14 sequences were PCR amplified and cloned into peYFP-C1 (CLONTECH Laboratories) using EcoRI–BamHI (New England BioLabs, Inc.) restriction sites. Subsequently, the eYFP was substituted by the 1–154 or the 155–238 fragment of eYFP. These fragments were PCR amplified and inserted using AgeI–XhoI or AgeI–BspEI (New England BioLabs, Inc.) restriction enzymes to generate fusion proteins with the NH- or COOH-terminal parts of YFP (YN-NXF1, YN-Y14, YC-NXF1, and YC-Y14, respectively). NXF1 deletion mutants (YN-NXF1 Δ1–60 and YN-NXF1 Δ1–371) were linked to full-length YFP or to the NH-terminal part of YFP using EcoRI–BamHI. For P labeling of RNA (), cells were washed with prewarmed, phosphate-free DME (Invitrogen) and incubated with 0.4 mCi P (GE Healthcare) per 200 ml of phosphate-free DME supplemented with 50 mM Hepes (Invitrogen) for 4 h at 30°C. Cells were fixed with 4% formaldehyde in PBS with 2% sucrose for 15 min on ice, permeabilized in 0.2% Triton X-100 for 3 min on ice, preincubated with 4% goat serum for 10 min at RT, and incubated with primary antibody (mouse α-SC35; Sigma-Aldrich; rabbit α-GFP; Abcam pvc) for 30 min at 37°C. The GFP antibody cross reacts with the YC epitope. Samples were washed and incubated with Cy5-labeled secondary antibodies (Dianova) for 30 min at 37°C. Coverslips were mounted with Vectashield containing DAPI (Linaris) for direct observation. Pol II transcription was inhibited by incubation in 50 μg/ml DRB or α-amanitin in DME for 6 h. For energy starvation, cells were incubated for 20 min at 30°C in PBS containing 10 mM sodium azide, 50 mM of deoxy-glucose, 1 mM MgCl, and 0.5 mM CaCl (). For selective permeabilization of the cell membrane, unfixed cells were incubated for 5 min at 30°C with 40 μg/ml digitonin (Calbiochem) in transcription buffer (), following the protocol used by . In experiments with ATP supply after permeabilization, digitonin was replaced by transcription buffer containing 5 mM ATP. Transfected cells were trypsinized, suspended in prewarmed DME, and washed with ice-cold PBS. For the cytoplasmic preparation, the cell membrane was permeabilized for 5 min on ice with digitonin buffer supplemented with complete EDTA-free protease inhibitor cocktail (Roche), 0.1 mM PMSF (Sigma-Aldrich), and 250 mM KCl. Cells were spun down at 800 for 3 min, the supernatant was taken off, and cells were respun at 15,000 and saved. The pellet containing the nuclear fraction was resuspended in buffer containing 0.1% NP-40, incubated on ice for 10 min, and spun at 15,000 for 3 min, and the supernatant was saved. For immunoprecipitation, 200-μl protein A–Sepharose beads (GE Healthcare) were washed in NP-40 buffer and blocked with 100 μg/ml of tRNA and 1 mg/ml BSA. Beads were incubated with antibody (α-GFP; Abcam pvc) at a ratio of 20:1 for 1 h, in a total volume of 500 μl. Cell extracts were incubated with either antibody-coated beads or beads alone overnight. Beads were washed three times in 1 ml of buffer and resuspended in 300 μl of buffer. All steps were performed at 4°C. Proteins were isolated from the beads and denatured by boiling in Laemmli buffer (2% SDS, 20% glycerol, 20 mM Tris-Cl, pH 8, 2 mM EDTA, 1 mM DTT, and 0.1 mg/ml Bromophenol blue) for 5 min. After the immunoprecipitation, 1 ml TRIzol (Invitrogen) was added, mixed, and incubated at RT for 5 min. 200 μl chloroform was added, mixed, and incubated at RT for 15 min. Phases were separated by centrifugation at 15,000 for 5 min. The upper phase was saved, 10 μg of glycogen was added, and RNA was precipitated with 500 μl isopropanol. The pellet was washed with 70% ethanol, resuspendend in 5 μl of nuclease-free water, and spotted onto a nitrocellulose membrane. Radioactivity was exposed on BioMax MS film (Kodak). Fluorescence images were acquired by confocal laser scanning microscopy (model TCS SP2; Leica) with a 63×, NA 1.32, oil objective at pinhole size 1 Airy and a nominal voxel size of 58 × 58 × 284 nm. DAPI, YFP, and Cy5 were excited by 405-, 514-, and 633-nm laser light and emission was detected at 410–500, 520–590, and 640–700 nm, respectively. Cross talk was minimized by serial acquisition of the fluorescence color channels. The digital images were analyzed with either LSM software (Leica) or analySIS pro 3.2 (Soft Imaging System GmBH). All micrographs in have been deconvoluted (blind three dimensional) with analySIS pro 3.2. Frequency distributions of intensities were collected either over a regions of interest or through binary masks, which were configured by global thresholding, morphologic filtering, and interactive editing. Two features of interest were compared, which were the BiFC signal in speckles (corresponding to the SC35 domains), including a perispeckle compartment and the nuclear area, except for speckles and nucleoli. Note that the “nonspeckle nucleoplasm” is heterogeneous, as it contains chromatin, nuclear bodies, and parts of the speckle compartment that were not identified as speckles. Four postbleach images were taken every 107 ms, followed by four images every 214 ms and 30 images every 1.07 s. The period of recovery measurement was adjusted to reach saturation (up to 100 s). After subtraction of the mean of image black level, the recovery data were corrected for FLIP effects and postbleach loss of fluorescence during recovery observation and transformed into relative values according to I = I/I × T/T, with I being the initial intensity and I the intensity at the time t in the bleached spot, and T and T being the total nuclear intensity before bleach and at time t, respectively. Immobile fractions and the period to reach half saturation (half-life) were directly read out from plotted FRAP curves. For FLIP analysis, approximately one half of a cell nucleus was bleached 100 times every 1.6 s at maximum laser intensity and beam expander 1 while the unbleached part of the nucleus was not illuminated. Two dimensional image acquisition before and after bleaching was conducted with pinhole size 6 Airy. For data analysis, the black level was subtracted and the postbleach image corrected for observation-induced bleaching. The factor for bleach correction was calculated from the fluorescence decay, I/I, of an adjacent unbleached cell observed under the same conditions. A complete loss of fluorescence across the entire cell nucleus was achieved with 25 repeated scans at maximum laser intensity. Fluorescence recovery was then observed over 2 h, and images of the same confocal plane were recorded 4 times every 30 s, followed by 8 times every 1 min, 10 times every 5 min, and 6 times every 10 min. The images were black level subtracted and the time-dependent increase in intensity was measured as mean intensity through masks designed for speckles and the nonspeckle nucleoplasm.

The ER provides an environment that facilitates the folding and assembly of newly synthesized secretory and transmembrane proteins and actively participates in the quality control of these proteins. By these means, correctly folded proteins are allowed to exit the ER to reach their final destination, whereas incompletely folded or misfolded molecules are retained in the ER. Quality control in the ER is achieved by two independent mechanisms, the productive folding mechanism, which uses various molecular chaperones and folding enzymes localized in the ER (ER chaperones), and the ER-associated degradation (ERAD) mechanism, by which unfolded or misfolded proteins are retrotranslocated from the ER to the cytoplasm to be degraded by the ubiquitin-dependent proteasome system (; ; ; ). Under a variety of conditions collectively termed ER stress, however, quality control becomes inefficient, resulting in the accumulation of unfolded proteins in the ER. Essentially, all eukaryotic cells are equipped with a system to cope with protein unfolding or misfolding in the ER. Interestingly, the strategy taken is well conserved from yeast to humans: components of both the productive folding and ERAD mechanisms are induced in response to ER stress by a transcriptional program termed the unfolded protein response (UPR), leading to accelerated refolding and degradation of unfolded proteins, respectively (; ; ; ). Nonetheless, an as-yet-unanswered critical question is how eukaryotic cells maintain the homeostasis of the ER by balancing the refolding and degradation mechanisms to counteract ER stress or, in other words, how proteins to be refolded are discriminated from those to be degraded. Transmembrane proteins are capable of transmitting signals across the membrane, and eukaryotic cells indeed use some of these molecules as sensors and transducers of ER stress for the UPR. Among these, Ire1p/Ire-1/IRE1 is an ER membrane–bound endoribonuclease conserved from yeast to humans. IRE1 initiates unconventional (frame switch) splicing in response to ER stress and has the substrates yeast HAC1 mRNA and metazoan XBP1 mRNA encoding the UPR-specific transcription factors Hac1p and XBP1, respectively. As a result of frame-switch splicing, the DNA binding domain and activation domain of these substrates are joined to produce a transcription factor capable of activating transcription efficiently (; ). A second transmembrane UPR signal transducer is pek-1/PERK, an ER membrane–bound protein kinase conserved in metazoan cells. PERK activated in response to ER stress phosphorylates the α subunit of eukaryotic initiation factor 2, resulting in a general attenuation of translation (). The third UPR transducer is ATF6, an ER membrane–bound transcription factor whose involvement in the UPR has to date been clarified for mammalian cells only. Upon ER stress, ATF6 is subjected to regulated intramembrane proteolysis, leading to liberation of the cytosolic transcription factor domain, which translocates into the nucleus to activate transcription (). We are currently investigating the roles of the IRE1–XBP1 and ATF6 pathways in the homeostasis of the ER and have proposed a time-dependent phase-transition model based on differential properties of the two signaling pathways (; see Discussion). Various components of ERAD have also been identified (; ). Among these, ER degradation enhancing α-mannosidase–like protein (EDEM) is a key component, as it targets misfolded glycoproteins to the proteasome, presumably by directly recognizing a signal that is present in misfolded glycoproteins, such as Man8GlcNAc2 structure (). Transcription of EDEM is ER stress inducible, and its induction depends on the IRE1–XBP1 pathway (). The mammalian genome contains two further EDEM-like molecules, designated EDEM2 and -3, which are also involved in the ERAD of misfolded glycoproteins and are ER stress inducible in a manner dependent on the IRE1–XBP1 pathway (; ; Molinari, M., personal communication). Recently, two groups independently identified an interesting human protein that shows weak homology to yeast Der1p, a protein involved in yeast ERAD (; ), and designated it Derlin-1 (Der1p-like protein; ; ; ). This protein is involved in a process similar to the ERAD used by human cytomegalovirus to escape from the immune system (). In this process, the US11 protein of human cytomegalovirus causes retrotranslocation of major histocompatibility complex class I heavy chain (HC) from the ER to the cytosol for degradation by the proteasome, in which the cytosolic ATPase p97 plays an essential role in extracting class I HC through the ER membrane (). Given that Derlin-1 interacts with functional US11 but not with nonfunctional US11 and that it recruits p97 to the ER membrane, Derlin-1 is considered to provide the missing link between events on the lumenal side of the ER (recognition of class I HC by US11) and those on the cytosolic side (extraction catalyzed by p97; ; ). The mammalian genome contains two additional Der1p-homologous proteins, designated Derlin-2 and -3. It was recently shown that Derlin-2 forms a complex not only with p97 but also with mammalian homologues of yeast Hrd1p and -3p, proteins involved in yeast ERAD (), similar to Derlin-1 (; ). However, it is not known whether Derlin-2 or -3 is required for ERAD. We demonstrate that Derlin-2 and -3, proteins regulated by the UPR, are functional components of ERAD for misfolded glycoproteins and that Derlin-2 at least is a target of the IRE1–XBP1 pathway, similar to EDEM. We also discuss the role of the mammalian UPR in distinguishing proteins to be refolded from those to be degraded. We conducted a microarray analysis to identify novel ER stress–inducible genes that might be involved in determining the fate of proteins that are unfolded or misfolded in the ER. Total RNA was isolated from HeLa cells that had been treated for 8 h with or without tunicamycin, an inhibitor of protein -glycosylation known to cause ER stress (), and the difference in expression levels was determined using the DNA chip cDNA Microarray kit human I carrying 12,814 genes. Results showed ∼300 genes that were induced more than twofold, among which we focused on one, CGI-101, because of its 2.9-fold induction by tunicamycin treatment (),the ability of hydrophobic stretches in its translational product to span the ER membrane several times (), and, most important, the weak homology it exhibited to yeast Der1p (). Although the role of Der1p in ERAD remains poorly understood, these results implied that CGI-101 may be a gene that connects the UPR with ERAD. Our database search also indicated that the human genome contains a gene somewhat similar to CGI-101 (designated ; available from GenBank/EMBL/DDBJ under accession no. ), the transcription of which we subsequently found to be induced in response to ER stress (see ). During the course of our subsequent analyses, proteins encoded by CGI-101 and turned out to be identical to Derlin-2 and -3, respectively (see Introduction). We decided to follow this nomenclature hereafter and continued to investigate their functions, which remain largely uncharacterized. A database search indicated the presence of two variants of Derlin-3, probably resulting from alternative splicing. We refer to these short and long forms as Derlin-3 transcriptional variant (tv) 1 and 2, respectively; tv2 has 30 more amino acids at the COOH terminus than does tv1 (). Derlin-2 and -3 show ∼75% identity, whereas that between Derlin-1 and -2 or Derlin-1 and -3 is ∼30%. Transcript structures for Derlin-1, -2, and -3 deposited in the data bank are schematically shown in . Northern blot hybridization revealed that both a short form and a long form of Derlin-2 mRNA are expressed in both human and mouse and that in both species the short form is more abundant than the long form (see and ). Similarly, short and long Derlin-3 mRNAs also exist, with the short form being more abundant in most human tissues; however, neither short nor long Derlin-3 mRNA was detected in mouse embryonic fibroblasts (MEFs; unpublished data). As shown in (B and C), Derlin-1 mRNA was expressed ubiquitously. Expression of Derlin-2 mRNA also appeared ubiquitous; the apparent discrepancy in the level of Derlin-2 mRNA in skeletal muscle between (lane 6) and (lane 3) is likely attributable to a difference in the amount of mRNA loaded (compare the levels of β-actin mRNA). In contrast, expression of Derlin-3 mRNA was restricted to several tissues. Both Derlin-2 and -3 mRNAs were detected abundantly in placenta and pancreas (), whereas Derlin-3 mRNA was much more abundant than Derlin-2 mRNA in spleen and small intestine (). The biological significance of these findings remains a subject for future investigation. We conclude that the summed expression of Derlin-2 and -3 mRNAs is comparable with that of Derlin-1 mRNA. We asked whether the transcriptional induction of Derlin-2 in response to ER stress depends on the IRE1–XBP1 pathway, similar to that observed for EDEM. Northern blot hybridiza- tion was performed for total RNA isolated from IRE1α+/+, IRE1α−/−, XBP1+/+, and XBP1−/− MEFs that had been treated with 10 μg/ml tunicamycin. Consistent with our previous results (), mRNA encoding binding protein (BiP), a major ER chaperone, was similarly induced in response to tunicamycin treatment in both wild-type cells and cells lacking IRE1α or XBP1, whereas induction of EDEM mRNA observed in wild-type cells was abolished in cells lacking IRE1α or XBP1 (). We found that Derlin-2 mRNA was induced in response to tunicamycin treatment, albeit less efficiently than BiP mRNA (, lanes 1–5 and 11–15), consistent with the results of microarray analysis shown in . We also found that Derlin-1 mRNA was ER stress inducible (, lanes 1–5 and 11–15). Importantly, induction of Derlin-1 or -2 observed in wild-type cells was greatly attenuated in cells lacking IRE1α or XBP1 (, lanes 6–10 and 16–20). These results indicated that Derlin-1 and -2 are targets of the IRE1–XBP1 pathway of the UPR. Derlin-3 mRNA was induced in tunicamycin-treated human cells, such as 293T, and HeLa cells, and the extent of induction was greater than that of Derlin-2 mRNA (). However, the dependence of Derlin-3 mRNA induction on the IRE1–XBP1 pathway could not be examined because Derlin-3 mRNA was not detected in MEFs (unpublished data). Nonetheless, Derlin-3 is considered to be a target of the IRE1–XBP1 pathway because the induction time course of Derlin-3 mRNA was similar to that of Derlin-2 and EDEM mRNA but clearly slower than that of BiP mRNA (), reflecting the differences in DNA binding specificity and activation mechanisms between ATF6 and XBP1 (see Discussion). These results suggested that, similar to Derlin-1 and EDEM, Derlin-2 and -3 might be components of ERAD. To determine the localization of Derlin-2 and -3, their NH-terminal c-myc epitope–tagged versions were transfected into HeLa cells. An indirect immunofluorescence analysis of transfected cells revealed that Derlin-2 or -3 was colocalized with Sec61β (), a component of the translocon in the ER (). It should be noted that overexpression of Derlin-1 created dot-like structures around the ER as reported previously (; ), whereas that of Derlin-2 or -3 did not do so, although their expression levels appeared comparable when detected with anti-myc antibody ( and see ). Derlin-1 is considered to be a protein that spans the ER membrane four times, with both its NH and COOH termini facing the cytosol (; ), and Derlin-2 and -3 contain four hydrophobic stretches at positions similar to that of Derlin-1 (). To determine actual topology, Derlin-2 or -3 tagged with the c-myc epitope at the respective COOH terminus was transfected into human embryonic kidney (HEK) 293 cells and microsomes isolated from transfected cells were digested with trypsin. Calnexin, which is a type I transmembrane protein in the ER, and calreticulin, which is a soluble lumenal protein, were used as controls. As shown in , the COOH termini of Derlin-2 and -3 were not protected against trypsin digestion, as was also the case for Derlin-1 and calnexin but not for calreticulin and the NH terminus of calnexin, which remained intact. Similar protease-sensitivity experiments conducted on Derlin-2 and -3 tagged with the c-myc epitope at the respective NH terminus gave rise to complicated results (unpublished data). The mode of insertion into the ER membrane may not have been uniform when NH-terminally tagged Derlin-2 or -3 was overexpressed, even though they were functional (see the following section). This notion was supported by the finding from the pulse-chase analysis that NH-terminally tagged proteins were much more unstable than COOH-terminally tagged proteins, albeit with more abundant expression (). Formal conclusion that endogenous Derlin-2 and -3 are four-transmembrane proteins in the ER with both their NH and COOH termini facing the cytosol thus requires analysis of their topology by raising antibodies against their NH and COOH termini. We next examined whether Derlin-2 and -3 are indeed involved in ERAD using two approaches, overexpression and knockdown experiments. Because a previous study indicated that Derlin-2 does not participate in US11-mediated degradation of class I HC, a transmembrane protein in the ER (), we chose as substrate a soluble lumenal protein misfolded in the ER, the null Hong Kong (NHK) mutant of α1-proteinase inhibitor (α1-PI, also called α1-antitrypsin; ). NHK glycosylated and misfolded in the lumen of the ER is recognized by EDEM for destruction by the proteasome in the cytosol (). Pulse-chase experiments followed by immunoprecipitation analysis with antibody against α1-PI showed that NHK was degraded with a half-life of <3 h (, lanes 1–3). Importantly, overexpression of Derlin-2 or -3 accelerated the degradation of NHK to a half-life of ∼1.5 h (, lanes 4–9). This acceleration was comparable with that observed with overexpression of EDEM (). Interestingly, both Derlin-2 and -3 were coimmunoprecipitated with NHK when cells were lysed with buffer containing 1% NP-40 (), and this association was lost when they were lysed with buffer containing 1% SDS (not depicted). Thus, as in the case of Derlin-1, Derlin-2 and -3 interact with the substrate to be degraded. We then asked whether overexpression of Derlin-2 and -3 also affects degradation of nonglycoproteins misfolded in the ER. A previously modified NHK designated NHK(QQQ), in which all asparagines of three potential -glycosylation sites are mutated to glutamine, was shown to degrade quite rapidly (Hosokawa, N., personal communication) with a half-life of ∼1 h, which is even shorter than that of NHK (, lanes 1–3). Results showed that overexpression of Derlin-2 or -3 did not significantly accelerate the degradation of NHK(QQQ) (, lanes 4–9), indicating that Derlin-2 and -3 induced by the UPR assist in the degradation of misfolded glycoproteins but not in that of misfolded nonglycoproteins. Unexpectedly, both Derlin-2 and -3 coimmunoprecipitated with NHK(QQQ) (), indicating that the binding activity of Derlin-2 and -3 is not simply correlated with their degradation activity. In this connection, we also found that EDEM located upstream of Derlin-2 and -3 in the degradation process (see the following section) can bind to NHK and NHK(QQQ) with similar efficiency (unpublished data). We then examined the effect of reducing Derlin-2 and -3 levels on the degradation of NHK. Double-stranded oligonucleotides corresponding to two regions each of Derlin-2 or -3 were inserted into the mammalian expression vector pSUPER to produce short hairpin RNAs (shRNAs) that act as short interfering RNA–like molecules in transfected cells (). The knockdown efficiencies of these constructs were determined by Northern blot hybridization. As shown in , they effectively suppressed expression of Derlin-2 or -3 without affecting that of other Derlins significantly. We therefore transfected HEK293 cells with these pSUPER derivatives together with the plasmid for expression of NHK. Pulse-chase experiments revealed that degradation of NHK became inefficient in cells with reduced levels of Derlin-2 or -3 (), indicating that they are functional components of ERAD for misfolded glycoprotein. We obtained no indication from reporter assays that the UPR was activated in these knockdown cells (unpublished data). Because a reduction in the expression of Derlin-2 or -3 alone effectively blocked the degradation of NHK, we explored the possibility that Derlin-2 and -3 are heterooligomerized. Results of cotransfection experiments clearly showed coimmunoprecipitatation of HA-tagged Derlin-3 with myc-tagged Derlin-2 and that of myc-tagged Derlin-2 with HA-tagged Derlin-3 in a fairly stoichiometric manner (, lanes 6 and 8, respectively). On the other hand, Derlin-1 was found to be poorly associated with Derlin-3 (, lanes 2 and 4), suggesting that Derlin-2 and -3 may constitute a group distinct from Derlin-1, as suggested from the distances in the phylogenic tree (). As a reduction in the expression of Derlin-2, Derlin-3, or both had little effect on the synthesis and secretion of the wild-type α1-PI (), we concluded that they are required for the ERAD of a soluble misfolded glycoprotein but not for the maturation of a correctly foldable protein. We next determined whether Derlin-2 and -3 are able to associate with known components of ERAD, such as p97, by pulse labeling HEK293 cells expressing COOH-terminal c-myc–tagged versions of Derlin-2 or -3 by transfection, followed by immunoprecipitation. Analysis with anti-p97 antibody revealed that every Derlin coimmunoprecipitated with p97 (, a, lanes 7–10). In contrast, analysis with anti-myc antibody did not detect significant p97 coimmunoprecipitation with any of the Derlins (, a, lanes 2–5). However, when FLAG-tagged p97 was overexpressed by cotransfection, each Derlin coimmunoprecipitated with p97 (, b, lanes 7–10) and FLAG-p97 coimmunoprecipitated with each Derlin (, b, lanes 2–5). Further, coimmunoprecipitation of each Derlin with p97 was observed when each Derlin was tagged with the c-myc epitope at the respective NH terminus (, c, lanes 7–10). These results are consistent with those of previous studies demonstrating the association of Derlin-1 and -2 with p97 (, ; , ). We then examined whether Derlins are also able to associate with EDEM. A similar immunoprecipitation analysis was performed in HEK293 cells expressing HA-tagged EDEM and one of the three Derlins tagged with the c-myc epitope at the respective NH terminus by transfection. Immunoprecipitation with both anti-myc and anti-HA antibodies revealed an association between EDEM and Derlin-2 and -3 (, a, lanes 2–4 and 7–9) but not Derlin-1 (, a, compare lanes 5 and 10 with lanes 1 and 6, respectively), indicating a functional difference between Derlin-1 and a group consisting of Derlin-2 and -3. This notion was further confirmed by the next series of immunoprecipitation analyses. Immunoprecipitation with anti-p97 antibody revealed no association between EDEM and p97, even though HA-tagged EDEM was overexpressed in HEK293 cells by transfection (, b, compare lane 6 with 5). Importantly, however, immunoprecipitation with anti-p97 antibody did reveal that there was an association between EDEM and p97 when Derlin-2 was overexpressed simultaneously with EDEM (, b, compare lane 7 with 6). In contrast, cooverexpression of Derlin-1 did not mediate the association of p97 with EDEM (, b, compare lane 8 with 6). As the level of Derlin-1 coimmunoprecipitated with p97 was lower than that of Derlin-2 in this experiment (, b, compare lane 8 with 7), we increased the amount of Derlin-1–expressing plasmid DNA in the next experiment. As a result, the level of Derlin-1 coimmunoprecipitated with p97 became comparable with that of Derlin-2 (, c, compare lane 4 with 3). Nevertheless, only Derlin-2 mediated the association of EDEM with p97 (, c, compare lane 3 with 4). We further examined whether Derlin-2 mediates association of NHK with p97. Immunoprecipitation analysis was performed in HEK293 cells expressing NHK and Derlin-2 or -1 tagged with the c-myc epitope at the respective NH terminus by transfection. Immunoprecipitation with anti-myc antibody revealed coimmunoprecipitation of NHK with Derlin-2 but not with Derlin-1 (, compare lanes 6 and 7 with 5). Immunoprecipitation with anti-p97 antibody showed no direct interaction between p97 and NHK (, lane 2). Importantly, however, p97 became associated with NHK when Derlin-2 was overexpressed simultaneously (, lane 3) and the NHK–Derlin-2 complex recovered with anti-p97 antibody was quantitatively similar to that recovered with anti-myc antibody (, lanes 3 and 6, respectively), strongly indicating that Derlin-2 associated with p97 was the same molecule as that associated with NHK. In contrast, Derlin-1 failed to mediate the association of NHK with p97 (, lane 4). These results prompted us to examine the effect of a reduction in the level of Derlin-1 on the degradation of NHK. It was recently reported that a sequence corresponding to a part of Derlin-1 successfully decreased the level of Derlin-1 without affecting that of Derlin-2 when incorporated into the shRNA vector (). We confirmed this observation by immunoblotting in our system (, lane 3). Similarly, our shRNA vector for Derlin-2 succeeded in decreasing the level of Derlin-2 without affecting that of Derlin-1 (, lane 4). Interestingly, knockdown of Derlin-1 had little effect on the degradation of NHK, whereas that of Derlin-3 reproducibly blocked it (). On the basis of these results, we concluded that Derlin-2 and -3, but not Derlin-1, function downstream of EDEM and are required for the degradation of NHK, a misfolded glycoprotein substrate of ERAD, although we cannot formally rule out the possibility that the residual amount of Derlin-1 in Derlin-1 knocked down cells (, lane 3) is sufficient for the degradation of NHK. Proteins misfolded in the ER are degraded by the proteasome in the cytosol via a process called ERAD, in which misfolded proteins are recognized inside the ER lumen, targeted to a channel, extracted from the ER membrane, and delivered to the proteasome (; ). Misfolded glycoproteins are recognized by presentation of an N-linked oligosaccharide processing intermediates, such as Man8GlcNAc2 (), which acts as a motif that triggers binding to EDEM (). In this way, glycoproteins that are unable to fold properly are transferred from calnexin, a glucoprotein-specific molecular chaperone, to EDEM (; ). Subsequently, the cytosolic multifunctional protein p97 acts to extract the glycoprotein from the ER membrane (; ; ; ; ). However, the events that link these two steps have remained undefined. The recent identification of Derlin-1 has brought new insights into the molecular mechanism of the US11-mediated retrotranslocation of class I HC, providing a missing link between events in the ER and those in the cytosol (; ). Further, the findings that Derlin-1 binds to misfolded and ubiquitinated proteins and that RNA interference of Derlin-1 in evokes ER stress suggest a more generalized role of Derlin-1 in ERAD (). Recent studies demonstrated that Derlin-1 is associated with ubiquitin ligases such as HRD1 and gp78 via binding to p97 and vasolin-containing protein (VCP)–interacting membrane protein and that Derlin-2 also forms such a multiprotein complex, consisting of p97, VCP-interacting membrane protein, HRD1, and HRD3/SEL1L (; ). However, it remains to be determined whether Derlin-2 is required for ERAD. We show that the properties of Derlin-2 and -3 are exactly those expected for proteins able to provide the missing link between EDEM and p97 in the process of degrading glycoproteins misfolded in the ER. Derlin-2 and -3 are transmembrane proteins that span the ER membrane multiple times ( and ) and are required for the ERAD of misfolded glycoprotein ( and ). They are associated with the degradation substrate, p97, and EDEM ( and ). p97 and EDEM form a complex only in the presence of Derlin-2 (), and Derlin-2 mediates the association of p97 with the degradation substrate (). In contrast, Derlin-1 does not mediate the association of p97 with either EDEM or the degradation substrate (). This specificity might be attributable to differences in amino acid sequences between Derlin-1 and -2 () or, as Derlin-1 and -2 or -3 appear to be associated with different cellular proteins (), to differences in accessory proteins. Interestingly, MEFs express only Derlin-2 mRNA, whereas human cells express both Derlin-2 and -3 mRNAs ( and and not depicted). In addition, Derlin-2 and -3 mRNAs are distributed differently in various human tissues (). The finding that the knockdown of Derlin-2 or -3 alone effectively blocked the degradation of a misfolded glycoprotein () suggested that they may form heterooligomers to function when expressed simultaneously, and this was indeed shown to be the case (). Similarly, they may form homooligomers when expressed singularly. It is of great interest whether Derlin-2 and -3 themselves can form a channel through which misfolded glycoproteins move from the ER lumen to the cytosol, as has been proposed for Derlin-1 (; ). We are also aware of a recent study showing that the proteasome binds directly to Sec61 (), supporting the idea that misfolded proteins are retrotranslocated from the ER to the cytosol through the translocon as originally proposed (). Thus, it appears that we are a far way from any conclusive determination on this issue. The UPR consists of transcriptional control only in yeast, and the Ire1p-mediated program covers transcriptional induction of not only ER chaperones but also components of ERAD in response to ER stress (; ; ). On this basis, the UPR plays little role in determining the fate of unfolded or misfolded proteins in yeast ER. The activated refolding system and degradation system may deal with unfolded proteins in a competitive manner. In contrast, metazoan cells have developed the ability to attenuate translation via activation by PERK in response to ER stress, thereby decreasing the burden on the ER (). Mammals have further evolved two signaling pathways, namely the ATF6 and IRE1–XBP1 pathways, for transcriptional induction of ER chaperones and components of ERAD in response to ER stress (). The transcription factor ATF6 is activated by a posttranslational mechanism, whereas the transcription factor XBP1 is activated by a posttranscriptional mechanism, causing the active form of ATF6 to be detected in cell extracts earlier than the active form of XBP1. In addition, XBP1 has broader target specificity than ATF6. Thus, the transcription of genes unaffected by the active form of ATF6 can be later induced by the active form of XBP1, allowing a time-dependent decision (). Accumulating evidence indicates that the ATF6 pathway mainly controls the expression of ER chaperones, whereas the IRE1–XBP1 pathway regulates the expression of not only ER chaperones but also components of ERAD such as EDEM, EDEM2, and HRD1 (; ; ; ; ). Our present analysis adds Derlin-1, -2, and probably -3 to the list of targets of the IRE1–XBP1 pathway ( and ). We have also found that the misfolded glycoprotein NHK was degraded more slowly in IRE1α−/− cells than in IRE1α+/+ cells (), whereas the misfolded nonglycoprotein NHK(QQQ) was degraded at similar rates in both cell types (unpublished data). These results strongly support our time-dependent phase-transition model for determining the fate of proteins that are unfolded or misfolded in the mammalian ER, at least as far as glycoproteins are concerned, in which the ATF6-mediated unidirectional phase (refolding only) is shifted to the XBP1-mediated bidirectional phase (refolding plus degradation) depending on the quality, quantity, or both of unfolded or misfolded proteins in the ER (). The availability of multiple pathways for transcriptional control confers diversity for mammalian cells to adjust to the accumulation of unfolded proteins in the ER. Total RNA was isolated from HeLa cells that had been untreated or treated with 2 μg/ml tunicamycin for 8 h by the acid guanidium/phenol/chloroform method using ISOGEN (Nippon Gene) and was further purified using RNeasy Midi (QIAGEN). 20-μg aliquots of total RNA from tunicamycin-treated and untreated cells were labeled with cyanine 3– and cyanine 5–2′-deoxycytidine 5′-triphosphate, respectively, using a Direct-Label cDNA synthesis kit (Agilent Technologies) in three of six independent analyses. In the other three analyses, the labeling dyes were swapped to avoid false induction resulting from biased labeling. Labeled cDNAs were mixed and hybridized with a human 1 cDNA microarray kit (Agilent Technologies), on which 12,814 human cDNA clones were spotted. Fluorescence intensities of cyanine 3 and 5 were determined for each spot using a GenePix 4000B and GenePix Pro 4.0 software (Axon Instruments, Inc.) and normalized so that the sum of cyanine 3 intensities of all spots was equal to that of cyanine 5 intensities of all spots. Fold induction caused by tunicamycin treatment was defined as the ratio of normalized intensity for treated cells to that for untreated cells. Recombinant DNA techniques were performed according to standard procedures (). Human Derlin-1 and -2 partial cDNAs containing the respective complete open reading frame were obtained by reverse transcription–coupled polymerase chain reaction using total RNA isolated from HEK293T cells and the following primers: 5′-ATCTTGGCTACCTGTGGGTCGAAGATGTCG-3′ and 5′-AACGCAGTTGTTAAGTGCACCCAGCACTGG-3′ for Derlin-1 and 5′-GGAAGATGGCGTACCAGAGC-3′ and 5′-GCTGCTTTAACCTCCAAGCC-3′ for Derlin-2. Mouse Derlin-2 partial cDNA containing the complete open reading frame was obtained similarly using total RNA isolated from MEF and the primers 5′-GGAAAGATGGCGTACCAGAGC-3′ and 5′-TCTCGTAATTGGCACCGCTG-3′. cDNAs for human Derlin-3, mouse Derlin-1, and mouse Derlin-3 were purchased as IMAGE cDNA clones (IMAGE ID 30338943, 2655640, and 3582647, respectively) from Open Biosystems. pCMV-Myc (CLONTECH Laboratories, Inc.) and pcDNA3.1/Myc-His(+)A (Invitrogen) were used to express each Derlin tagged with the c-myc epitope at the NH and COOH terminus, respectively. pCMV-SPORT2-EDEM-HA to express HA-tagged EDEM as well as pREP9–α1-PI and pREP9-NHK to express the wild-type α1-PI and its NHK mutant were described previously (; ). Plasmids to express NHK(QQQ) and FLAG-tagged p97 were gifts of N. Hosokawa (Kyoto University, Kyoto, Japan) and M. Tagaya (Tokyo University of Pharmacy and Life Science, Tokyo, Japan) respectively. HEK293, HEK293T, and HeLa cells, as well as IRE1α+/+, IRE1α−/− (), XBP1+/+, and XBP1−/− (gifts of L. Glimcher, Harvard University, Cambridge, MA) MEFs, were cultured at 37°C in a humidified 5% CO/95% air atmosphere in Dulbecco's modified Eagle's medium (glucose at 4.5 g/liter) supplemented with 10% fetal bovine serum, 2 mM glutamine, and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin). Transfection was performed using Fugene6 (Roche) and Lipofectamine 2000 (Invitrogen) transfection reagents according to the manufacturers' instructions. Total RNA was isolated from cultured cells using ISOGEN. 5 μg each of total RNA were subjected to 1.2% agarose gel electrophoresis containing 2.2 M formaldehyde and transferred to a nylon membrane. RNA isolated from human tissues was separated by electrophoresis and blotted on a membrane (Human MTN blot and Human 12-lane MTN blot; CLONTECH Laboratories, Inc.). Digoxigenin (DIG)-labeled cDNA probes were prepared using polymerase chain reaction according to the manufacturer's instruction (Roche) and hybridized with RNA blotted on a membrane in an ultrasensitive hybridization buffer (ULTRAhyb; Ambion). Subsequent reaction with anti-DIG antibody (Roche) and treatment with chemiluminescent detection reagent CDP-star (GE Healthcare) were performed according to the manufacturers' specification. Chemiluminescence was visualized using an LAS-1000plus LuminoImage analyzer (FujiFilm). HeLa cells grown directly on slide glass in a 24-well plate were transfected with plasmid to express each Derlin tagged with the c-myc epitope at the respective NH terminus. 36 h later, cells were fixed with 4% formaldehyde at room temperature for 10 min, permeabilized with 0.2% Triton-X 100 in PBS, reacted with mouse anti-myc antibody (Santa Cruz Biotechnology, Inc.) and rabbit anti-Sec61β antibody (Upstate Biotechnology) for 1 h, and incubated with FITC-conjugated anti–rabbit IgG antibody (MP Biomedicals) or rhodamine-conjugated anti–mouse IgG antibody (Cappel). Fluorescence was visualized using a laser-scanning microscope (Eclipse E800; Nikon). HEK293 cells were transfected with plasmid to express each Derlin tagged with the c-myc epitope at the respective COOH terminus. Transfected cells were homogenized in sucrose buffer (50 mM Tris-Cl, pH 8.0, containing 0.32 M sucrose, 1 mM dithiothreitol, and 1 mM EDTA). Postnuclear supernatant obtained after 1,000 centrifugation was treated with various concentrations of trypsin at 4°C for 15 min. Digestion was terminated by adding SDS sample buffer and boiling for 15 min. Immunoblotting was performed using antibodies against myc (Santa Cruz Biotechnology, Inc.), calnexin (StressGen Biotechnologies), and calreticulin (Affinity BioReagents, Inc.) and by reaction with Western blotting luminol reagent (Santa Cruz Biotechnology, Inc.). Chemiluminescence was visualized using a LAS-1000plus LuminoImage analyzer. Antibodies against Derlin-1 and -2 were purchased from MBL International Corporation, and anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody was obtained from Trevigen. Cells were incubated for 30 min in methionine- and cysteine-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 2 mM glutamine and 10% dialyzed fetal bovine serum. Cells were then pulse labeled with 4.1 Mbq/dish EXPRESS protein labeling mixture (NEN Life Science Products) and chased in fresh complete medium. Cells were lysed in buffer A (50 mM Tris-Cl, pH 8.0, containing 1% NP-40, 150 mM NaCl, and protease inhibitors). Immunoprecipitation was performed with anti–α1-PI antibody (DakoCytomation), anti-myc antibody (Santa Cruz Biotechnology, Inc.), anti-HA antibody (Santa Cruz Biotechnology, Inc.), or anti-VCP antibody (Affinity BioReagents, Inc.) and protein A– or G–coupled Sepharose beads (GE Healthcare). Beads were washed in high ionic buffer (50 mM Tris-Cl, pH 8.0, containing 1% NP-40 and 150 mM NaCl) twice, washed in low ionic buffer (10 mM Tris-Cl buffer, pH 7.4), and boiled in Laemmli's sample buffer. Immunoprecipitates were subjected to SDS-PAGE, and radioactive bands were analyzed using a FLA-3000G FluoroImage analyzer (FujiFilm). Cells were also lysed in buffer A plus 1% SDS, boiled for 5 min, diluted with four volumes of buffer A, and centrifuged at 13,000 . The resulting supernatant was used for subsequent immunoprecipitation. Double-stranded oligonucleotides corresponding to five regions each of human Derlin-2 and -3 were inserted into the shRNA expression vector pSUPER (OligoEngine) between the BglII and HindIII sites. HEK293 cells were transfected with 4 μg of pSUPER or pSUPER derivatives for Northern blot hybridization. Two regions each were found to decrease Derlin-2 and -3 mRNA levels effectively, i.e., sh-Derlin-2, 441–459; sh-Derlin-2, 737–755; sh-Derlin-3, 2160–2178; and sh-Derlin-3, 3088–3106, where A of the initiation codon ATG is set as 1. The sequence to knockdown Derlin-1 was obtained from a recent study ().

SP-C (surfactant protein C) is a transmembrane protein that is synthesized as a 191– or 197–amino acid proprotein by type II epithelial cells of the lung. Processing of the proprotein in the distal secretory pathway liberates a 35–amino acid peptide that is secreted into the airspaces as a component of pulmonary surfactant. Mutations in the gene encoding SP-C () have been associated with the development of sporadic and familial interstitial lung disease (ILD). In all cases to date, the mutation was present on only one allele consistent with a dominant-negative effect. The index mutation, a heterozygous base substitution of A for G at the first base of intron 4 (c.460 + 1 G→A), led to the internal deletion of 37 amino acids from the ER luminal domain, generating a truncated proprotein (SP-C; ). Two separate mutations associated with ILD, SP-C and SP-C, were detected in three kindreds (; ; ). The age of onset and penetrance of ILD varied markedly in all three kindreds. Studies in transiently transfected cells suggested that the c.460 + 1 G→A mutation led to misfolding of the mutant proprotein, retention of SP-C in the ER, activation of the unfolded protein response (UPR), and apoptosis (; ; ). SP-C was also associated with cytotoxicity and lung dysmorphogenesis when expressed in type II cells of transgenic mice (). The UPR is activated by conditions that perturb ER homeostasis, including the accumulation of misfolded proteins (). This response encompasses translational and transcriptional changes within the cell to alleviate the stress and to promote restoration of ER homeostasis. A model for the time-dependent induction of the UPR has been proposed, suggesting that translational repression via PERK activation/eIF2α phosphorylation occurs first followed by the cleavage of ATF6, activation of IRE1/XBP-1, and expression of ATF6 and XBP-1 target genes (). If ER homeostasis cannot be restored by these pathways or by the induction of adaptive responses, apoptosis may occur as a means of avoiding the untoward effects of cell necrosis. ER stress–induced apoptosis has been associated with induction of the transcription factor C/EBP homologous protein, activation of c-Jun amino-terminal kinase via IRE1, and activation of the ER stress–specific caspases 4 (human; ) and 12 (mouse; ; ; ; for review see ). Although the effects of acute ER stress, which is imposed by xenotoxic agents such as thapsigargin and tunicamycin, are well established, little is known about the molecular pathways involved in adaptation to chronic ER stress imposed by a misfolded protein. The variability in the age of onset and penetrance of disease in the SP-C and SP-C pedigrees suggests that both genetic and environmental factors may influence the manifestation of lung disease. Based on the results of the aforementioned studies in human patients and transiently transfected cells, experiments were designed to test the hypotheses that chronic ER stress imposed by misfolded SP-C promotes adaptation and cell survival and adaptation increases susceptibility to environmental stress. Clonal cell lines stably expressing SP-C or SP-C were generated to identify cytoprotective pathways that are associated with adaptation to the constitutive expression of misfolded SP-C and to assess the cytotoxic effects of environmental stress on adapted cells. To determine the molecular mechanisms underlying SP-C-induced cytotoxicity, HEK293 cell lines stably expressing SP-C or SP-C were generated. Multiple clonal lines were obtained for each construct, and two lines were chosen for subsequent experimentation based on equivalent expression of SP-C mRNA, which was initially assessed by RT-PCR () and subsequently confirmed by microarray analysis (). These cell lines were morphologically indistinguishable by light microscopy () or electron microscopy (not depicted) and exhibited similar doubling rates (not depicted). Basal SP-C protein levels were assessed by Western blot analysis of cell lysates with an antibody directed against the NH-terminal peptide of the proprotein (proSP-C), a region which is unaffected by the Δexon4 mutation. Despite equivalent mRNA levels, expression of the SP-C protein was barely detectable compared with SP-C, which is consistent with rapid proteasome-dependent turnover of the mutant proprotein (). SP-C protein was previously shown to be rapidly degraded in a proteasome-dependent manner and failed to be exported from the ER when transiently expressed in HEK293 cells (). To determine the subcellular localization of the SP-C variants in stably transfected cell lines, cells were stained with antibodies directed against proSP-C and LAMP-1 and analyzed by confocal microscopy. A bright, punctuate staining pattern was observed for proSP-C in SP-C cells, whereas cells expressing SP-C showed faint, diffuse staining that was detected only in the presence of proteasome inhibitor (). SP-C traffics to the lamellar body in type II cells but is redirected to the lysosome in cells that lack lamellar bodies (for reviews see ; ). Colocalization of SP-C with the lysosomal marker LAMP-1 demonstrated efficient export of wild-type protein from the ER (). In contrast, SP-C exhibited a faint and diffuse staining pattern that colocalized with an anti-KDEL antibody, an ER marker (). These data in clonal cell lines, coupled with previous results in transiently transfected cells (), suggest that SP-C is correctly folded and exported from the ER, whereas SP-C fails quality control and is rapidly degraded by the proteasome. Molecular pathways induced by the chronic expression of SP-C were identified by transcriptional profiling of SP-C and SP-C clonal cell lines (the complete dataset can be found at ; GenBank/EMBL/DDBJ accession no. ). Unexpectedly, known components of the UPR/ER-associated degradation (ERAD) pathways were not increased in cells that constitutively expressed SP-C, which is in contrast to results in transiently transfected cells (). However, several genes associated with anti- and pro-apoptosis pathways were differentially expressed in the SP-C clonal cell line, including BAX and Bcl-2 (, Fig. S1, and Table S1; available at ). Interestingly, two transcripts linked to the NF-κB pathway, interleukin-1 receptor-associated 1 (IRAK1) and the γ subunit of the Iκ-B kinase complex (IKBKG), were increased in the SP-C cells, which is consistent with alterations in the NF-κB signaling pathway. Analyses of the 5′ flanking sequences of apoptosis-associated genes revealed that ∼44% (15/34) contained putative NF-κB binding sites, suggesting that they may be direct targets of NF-κB. To determine whether NF-κB activity was increased in SP-C cells, an NF-κB luciferase reporter construct was transiently transfected into the clonal cell lines. SP-C cells exhibited an 8.6-fold increase in basal NF-κB activity compared with cells stably transfected with empty vector or SP-C (). Cotransfection of a stabilized, nonphosphorylatable form of I-κBα (super repressor [SR]) with the NF-κB reporter resulted in a dose-dependent decrease in luciferase activity in SP-C cells that was completely suppressed to levels detected in untreated SP-C cells (). IκBα protein in SP-C cells was decreased compared with SP-C cells, which is consistent with increased NF-κB activity (). Treatment of SP-C cells with SN50 peptide to inhibit the nuclear translocation of NF-κB resulted in cellular retraction and a significant increase in cell death, as indicated by propidium iodide (PI) staining () and MTS reduction assay (). Treatment of SP-C cells with a control peptide, SN50, at equivalent doses and duration had no effect (). Cells expressing SP-C were unaffected by SN50 or SN50 treatment (not depicted). Treatment of SP-C cells with SN50 decreased basal NF-κB activity as measured by a NF-κB reporter () and inhibited nuclear translocation of the p65 subunit of NF-κB (). Altogether, these results suggest that NF-κB, mediated in part by the p65 subunit, plays an important cytoprotective role in adaptation to the constitutive expression of SP-C. The increased expression of pro-apoptotic transcripts in SP-C cells () suggested that adapted cells may be more susceptible to secondary stress. Infection with respiratory syncytial virus (RSV) preceded the onset of ILD in several patients carrying the SP-C mutation. To determine whether the expression of SP-C increased susceptibility to RSV, the clonal cells lines were infected at a multiplicity of infection (MOI) of 10 pfu/cell, and cell viability was assessed 24 h after infection. Although cell viability was minimally affected in SP-C cells, as indicated by the low level of PI staining ( [b and f]), the infection of cells expressing SP-C resulted in significant cell death ( [d and h]). Western blot analysis demonstrated that SP-C protein accumulated in a manner directly correlated with the viral titer, whereas SP-C protein levels were unaffected (). Several viruses, including adenovirus, influenza, and hepatitis B and C viruses, have been reported to induce ER stress/UPR pathways (; ; ; ). To determine whether RSV activated the UPR in cells stably expressing SP-C, an UPR element luciferase reporter construct was transiently transfected into cells expressing SP-C or SP-C before infection with RSV (). Baseline levels of luciferase activity were similar between the cell lines, which is consistent with the microarray data, indicating that the UPR was not activated in cells stably expressing SP-C. However, luciferase activity was increased in both cell lines after RSV infection, and the effect was exacerbated in SP-C cells (2.5-fold increase in SP-C vs. 4.4-fold increase in SP-C; ). Furthermore, RSV-induced cell death appeared to be independent of NF-κB, as NF-κB activity was sustained in both RSV-infected SP-C and SP-C cells (). Consistent with this finding, UPR activity was not induced in SP-C cells when NF-κB activity was inhibited with SN50 (). Collectively, these data indicate that RSV infection was associated with accumulation of the mutant proprotein, pronounced activation of the UPR, and increased cell death independent of NF-κB activity. Accumulation of mutant SP-C proprotein in RSV-infected SP-C cells suggested that proteasome function was inhibited. To assess the impact of RSV infection on proteasome activity, a proteasome reporter construct, pZsProSensor-1, was transiently transfected into the clonal cell lines. Cells were infected with 1–10 MOI of RSV and analyzed 24 h later by fluorescence microscopy and FACS analysis. Fluorescence in untreated SP-C cells (mean fluorescence intensity [MFI] = 305) was higher than that observed in untreated SP-C cells (MFI = 156), indicating a modest inhibition of basal proteasome activity in the presence of the mutant proprotein (, a [e] vs. b [e]). Infection of SP-C cells with RSV resulted in a dose-dependent increase in fluorescence detected by microscopy and FACS ( [e–h]). In contrast, fluorescence was not altered in SP-C cells infected with RSV ( [e–h]). Transfection efficiency was similar between the two cell lines, and the proteasome reporter was activated in SP-C cells treated with MG-132 (). These results demonstrate that basal proteasome function is decreased in SP-C cells and that this perturbation is exacerbated during RSV infection. To determine whether decreased proteasome function induced ER stress and activation of apoptosis independently of RSV infection, the clonal cell lines were treated with the proteasome inhibitor MG-132. The SP-C cell line was exquisitely sensitive to proteasome inhibition. The majority of cells retracted and detached from the plate after treatment with 250 nM MG-132 for 18 h, and a more pronounced effect was detected at the 500 nM of concentration ( [g and h]). SP-C cells were minimally affected by similar treatment ( [a–d]). Cell viability was decreased ∼10% in SP-C cells and 50% in SP-C cells treated with 500 nM MG-132 (). To determine whether MG-132–induced cytotoxicity was associated with SP-C accumulation, immunoblot analysis was performed with a proSP-C antibody after MG-132 treatment. There was a dose-dependent increase in SP-C protein, whereas levels of the SP-C protein were unaffected by MG-132 treatment (). Accumulation of unfolded proteins in the ER results in the activation of the IRE1–XBP-1 pathway (for review see ). Active XBP-1 was not detected in SP-C cells after MG-132 treatment; in contrast, robust activation of XBP-1 was detected in SP-C cells after proteasome inhibition (). Collectively, these results demonstrate that proteasome inhibition resulted in dose-dependent cytotoxicity associated with the accumulation of SP-C and the activation of XBP-1. To determine whether the accumulation of SP-C after proteasome inhibition was associated with apoptosis, cell lines were treated with MG-132 and stained with the fluorescent DNA-intercalating dye H33342 to assess nuclear architecture. Cells expressing SP-C showed a diffuse, uniform staining of the nucleus typical of viable, healthy cells (, left panel and inset). Although the majority of cells expressing SP-C detached from the plate after MG-132 treatment and were excluded from this analysis, the cells that remained attached exhibited punctuate staining and nuclear condensation that is consistent with apoptosis (, right panel and inset). Furthermore, treatment of SP-C–expressing HEK293 cells with MG-132 for 18 h resulted in the activation of caspase-3 (). Caspase-3 activation was not detected in SP-C cells treated with MG-132, indicating that this event was specific for cells expressing mutant SP-C (). As previously reported, the expression of SP-C in the distal lung epithelium of transgenic mice resulted in cytotoxicity associated with lung dysmorphogenesis (). Immunohistochemistry was performed on fetal lung sections from a transgene-positive animal (, TG#2) and a wild-type littermate control (, WT) with an antibody that detects the cleaved isoform of caspase-3. Robust staining for active caspase-3 was detected in sloughed epithelial cells in the distal airway and isolated intact epithelial cells in the transgenic animal, whereas lung epithelial cells in the wild-type animal were negative (). Together, these results demonstrate that the accumulation of SP-C, induced by proteasome inhibition in vitro or overexpression in vivo, was associated with the activation of apoptotic pathways. In contrast, RSV infection resulted in specific, extensive death of SP-C–expressing cells () without evidence of caspase-3 activation (). The SP-C mutation was chosen for these experiments because of its link to ILD in human patients (, ), the severity of this mutation on the structure of the proprotein, and its association with cytotoxicity and lung dysmorphogenesis when expressed in transgenic mice (). Results of experiments performed in this study using stably transfected cells indicated that the SP-C cell line adapted to the constitutive expression of misfolded SP-C and that NF-κB played a pivotal role in the adaptive response. Infection of cells expressing SP-C with RSV resulted in enhanced cytotoxicity associated with decreased proteasome function, accumulation of the mutant proprotein, and cell death. Collectively, these results suggest that adaptation to chronic ER stress imposed by misfolded SP-C may promote resistance to ILD, whereas environmental insults, such as viral infection, may trigger the onset of disease in patients with mutations in . Although the effects of acute ER stress, imposed by xenotoxic agents such as thapsigargin and tunicamycin, are well established, little is known about the molecular pathways involved in adaptation to chronic ER stress imposed by a misfolded protein. The lack of a cytotoxic response coupled with the absence of ER stress induction in SP-C clonal cells suggested that these cells successfully adapted to the constitutive expression of misfolded SP-C. Several lines of evidence support this hypothesis. First, the level of SP-C mRNA in the SP-C cell line was comparable with that in the SP-C cell line, thus excluding low expression of the mutant protein as a reason for the survival of SP-C cells. Second, in contrast to the wild-type protein, SP-C protein was barely detectable, indicating that the misfolded proprotein was recognized as terminally misfolded and rapidly degraded via the ERAD pathway. Furthermore, chronic expression of SP-C was associated with modestly reduced proteasome function that is consistent with near saturation of this degradative pathway. Third, components of the ER stress/UPR pathways were not transcriptionally up-regulated in the clonal SP-C cell line, which is in marked contrast to results in transiently transfected HEK293 cells (; ). Fourth, chronic expression of SP-C was associated with differential expression of both pro- and anti-apoptotic genes, approximately half of which contained putative NF-κB–binding sites in their promoters. Lastly, the inhibition of NF-κB, which is known to promote cell survival in response to a variety of stresses, resulted in the death of SP-C cells. Collectively, these results are consistent with an NF-κB–dependent adaptive response to chronic ER stress imposed by the constitutive expression of SP-C. Although the induction of other molecular pathways involved in cell survival may occur, the activation of NF-κB appears to be critical for survival in HEK293 cells that constitutively express SP-C. The role of NF-κB in the regulation of cytokine-stimulated proinflammatory gene expression is well established (). In this study, up-regulation of proinflammatory cytokines was not detected by microarray analyses; however, it is certainly possible that inflammation associated with lung injury and repair contributes to pathogenesis in vivo. NF-κB activity is also increased in response to the accumulation of transmembrane proteins in the ER, including the E3/19k protein of adenovirus (), middle human hepatitis B surface protein (), expression of major histocompatibility complex class 1 in the absence of β2-microglobin protein (), and p450 (). In addition, transient expression of the misfolded Z variant of α1-antitrypsin in CHO and HEK293 cells resulted in NF-κB activation associated with decreased levels of IκBα and IκBβ (). Activation of NF-κB in response to the accumulation of newly synthesized membrane or misfolded secretory proteins is referred to as the ER overload response. Constitutive activation of NF-κB promotes survival of a wide range of cells, including B cells, hepatic cells, and cancer cells, and is a major target for cancer therapy (). NF-κB promotes cell survival, in part, by activating anti-apoptotic genes, including members of the inhibitor of apoptosis family, TRAF1 and TRAF2, and the Bcl2 homologues Bfl-1/A1 and Bcl-X (). Recent studies using tunicamycin and thapsigargin as stressor agents demonstrated an association between ER stress–induced PERK activation, eIF2α phosphorylation, and NF-κB activation (; ), effectively linking the UPR and the ER overload response. Although a definitive role for NF-κB activation in response to ER stress has not been established, it is thought to promote cell survival in this context. Transient transfection of HEK293 cells with SP-C resulted in specific up-regulation of BiP, XBP-1, and HedJ1, which is consistent with the induction of UPR and ERAD to promote clearance of the misfolded protein and the alleviation of ER stress (; ). In contrast, the constitutive expression of SP-C resulted in the differential expression of apoptosis-related genes with an apparent balance shifted toward NF-κB–dependent anti-apoptotic responses. Identification of environmental triggers that tip the balance toward apoptosis/cell death are clearly important for understanding disease pathogenesis in humans with mutations. ILD is often associated with infection and/or inflammation (; ). The marked variability in severity and age of onset of lung disease in the SP-C pedigrees suggested that environmental factors might be involved in triggering the onset of ILD (; ). Consistent with this hypothesis, five individuals in the SP- C kindreds were diagnosed with viral infection before the manifestation of lung disease. Importantly, cells stably expressing SP-C were much more susceptible to viral-induced cell death, suggesting that infection may be an environmental trigger for ILD. Although RSV infection in vivo is associated with epithelial cell damage and cellular desquamation, these cytopathic effects are primarily mediated by an augmented immune response initiated by the infected host cells rather than viral-induced cell death. In support of this hypothesis, mice depleted of CD4 and CD8 cells exhibited persistent viral replication without illness (). Furthermore, RSV infection of cultured airway epithelial cells (MOI of ∼20; i.e., twice the dose used in this study) did not result in cytopathology (). Consistent with these findings, HEK293 cells constitutively expressing SP-C exhibited relatively little cell death in response to RSV infection (). In marked contrast, RSV infection of cells expressing SP-C resulted in extensive cell death associated with proteasome inhibition and accumulation of mutant proprotein. In vivo, injury of alveolar epithelial cells is an early event in progressive pulmonary fibrosis. Impaired reepithelialization leads to sustained proliferation/activation of interstitial fibroblasts and myofibroblasts, resulting in the accumulation of extracellular matrix and, ultimately, fibrosis (; ). The increased susceptibility of some individuals to severe RSV pneumonia may further contribute to variability in the severity and onset of the disease (). Although RSV infection was specifically cytotoxic for SP-C–expressing cells, the molecular pathway leading to cell death is not clear. Cell death did not appear to be apoptosis dependent, as caspase-3 cleavage was undetectable in the RSV-infected cells (). RSV has been shown to inhibit apoptosis of cultured epithelial cells via the activation of EGF receptor–extracellular signal-regulated kinase and PI3K–AKT signaling pathways, thereby promoting self-replication (; ). In this study, RSV-induced cytotoxicity was associated with proteasome dysfunction, accumulation of the mutant proprotein SP-C, and pronounced activation of the UPR. Induction of ER stress and proteasome inhibition in RSV-infected SP-C cells may be related, in part, to forced synthesis of viral membrane proteins, including the F, G, and SH proteins. UPR activation was higher in SP-C cells than in SP-C cells, suggesting that the production of RSV membrane proteins superimposed on the expression of misfolded SP-C protein may overwhelm the degradative capacity of the proteasome, leading to accumulation of cytotoxic forms of SP-C and subsequent cell death. Therefore, it is possible that proteasome dysfunction and accumulation of SP-C drives the cell down a nonapoptotic cell death pathway in the presence of RSV. In contrast, accumulation of SP-C in transgenic mice and MG-132–treated cells (i.e., in the absence of RSV infection) resulted in caspase-3 activation, which is typical of apoptosis (; ). Based on the findings of this study, the following model is proposed. SP-C protein is correctly folded and exported from the ER, whereas SP-C is terminally misfolded and degraded by a proteasome-dependent pathway. Constitutive expression of terminally misfolded SP-C results in chronic ER stress and an NF-κB–dependent cytoprotective response. Superimposition of a secondary stress, such as RSV infection, on the constitutive expression of SP-C leads to proteasome dysfunction, accumulation of SP-C, and cytotoxicity. Accumulation of SP-C and cytotoxicity may contribute to the pathogenesis of ILD associated with mutations in . Recombinant TNFα protein was obtained from PeproTech. The MTS reduction assay kit and Dual-Luciferase Reporter Assay System were purchased from Promega; MTS reduction assays were performed in 96-well plates according to the manufacturer's protocol. MG-132, SN50, and control peptide (SN50) were obtained from EMD Biosciences. Full-length human wild-type SP-C (SP-C) cDNA and SP-C, generated as previously described (), were subcloned into pTRE2-Hyg (BD Biosciences) to generate SP-C/pTRE2-Hyg and SP-C/pTRE2-Hyg. To generate stably transfected cell lines, HEK293 Tet-Off cells (CLONTECH Laboratories, Inc.) were transiently transfected with SP-C/pTRE2-Hyg or SP-C/pTRE2-Hyg using LipofectAMINE 2000 (Invitrogen) in the presence of 250 μg/ml hygromycin B as the selection agent. Doxycycline was present in the media during the selection process to obtain regulatable lines (i.e., expression of SP-C or SP-C was dependent on doxycycline withdrawal). However, all lines that were initially regulatable eventually reverted to constitutive expression of the transgene. Therefore, SP-C or SP-C cells for this study were cultured in the absence of doxycycline and G418 (selection agent for the tTA cassette). Clonal colonies were isolated, amplified, and screened for integration of the transgene by RT-PCR using SP-C–specific primers and Western analyses using an antibody directed against the NH-terminal peptide of proSP-C (). Total RNA for RT-PCR and real-time PCR analysis was isolated using the acidified guanidinium method (), treated with DNase I (DNA free; Ambion), and reverse transcribed into cDNA using SuperScript II Reverse Transcriptase (Invitrogen). Forward and reverse primer sequences for human XBP-1 are 5′-GGACTTAAGACAGCGCTTGG-3′ and 5′-TGAGAGGTGCTTCCTCGATT-3′. PCR reactions for XBP-1 were performed for 35 cycles, and products were separated on 4% agarose gels. Verification of microarray data for selected genes was performed by real-time PCR analysis with a SmartCycler (Cepheid) using reaction conditions as previously described (). RNA samples for microarray analysis were prepared as previously described () and hybridized to the GeneChip Human Genome U133 Set (HG-U133A and HG-U133B; Affymetrix, Inc.) according to the manufacturer's protocol. Affymetrix Microarray Suite 5.0 was used to scan and quantitate the gene chips under default scan settings. Normalization was performed using the Robust Multichip Average model (,). Data were further analyzed using Significance Analysis Of Microarrays () and Genespring 7.2 (Silicon Genetics). Detection of differential expression was performed using random permutation and Welch's approximate test for mutant and control groups at P ≤ 0.01, false discovery rate ≤ 10%, a minimal of twofold changes in absolute ratio, and a minimum of two present calls by Affymetrix algorithm in three samples with the relative higher expression. Gene ontology analysis was performed using the database for annotation, visualization, and integrated discovery (). Potential protein–protein interactions were identified using PathwayAssist (Ariadne Genomics). NF-κB–binding sites (GGGACTTTCC) were scanned in −2-kb promoter regions of all differentially expressed genes from microarray analysis using MatInspector (Genomatix), allowing a maximum of one mismatch. HEK293 stable cell lines were propagated as previously described (). Cell lysate preparation, protein standardization, and Western analysis were performed as previously described (). Nuclear extracts were isolated as previously described (). Antibody sources are as follows: anti–proSP-C (), anti–caspase-3 (Cell Signaling), antihistone H2B (Imgenex), anti-p65 subunit of NF-κB (Santa Cruz Biotechnology, Inc.), and anti-actin (). Cells for confocal microscopy were prepared as previously described (). Primary antibodies included a polyclonal anti–proSP-C (), a monoclonal anti-LAMP1/CD107A (Research Diagnostics, Inc.), and a monoclonal anti-KDEL (StressGen Biotechnologies). Secondary antibodies included anti–rabbit FITC-conjugated and anti–mouse Texas red–conjugated secondary antibodies (Jackson ImmunoResearch Laboratories). Vectashield Hardset (Vector Laboratories) was used as the mounting medium. Fluorescence was visualized on a confocal microscope (LSM510; Carl Zeiss MicroImaging, Inc.) using FITC/Texas red filters and a 40× NA 1.3 objective. HEK293 cells were detached from the culture plate 24 h after RSV infection/transient transfection by treatment with trypsin/EDTA, washed once with PBS, and resuspended in FACS buffer consisting of PBS with 0.1% FBS and 0.05% sodium azide. Cell-associated fluorescence was measured on a FACScalibur flow cytometer using CellQuest software (BD Biosciences). For each sample, 20,000 events were acquired. Immunohistochemistry was performed on mouse fetal lung samples that were previously described () using an antibody that detects the cleaved form of caspase-3 (R&D Systems) at a 1:5,000 dilution. Localization of antigen–antibody complexes was performed as previously described (). The A2 strain of RSV was amplified, purified, and quantitated in Hep2 as previously described (). Clonal cell lines were infected with RSV using a previously established protocol for adenoviral infection (). For studies in which proteasome activity was assessed after RSV infection, the proteasome sensor plasmid pZsProSensor-1 (CLONTECH Laboratories, Inc.) was transfected into the cells using LipofectAMINE 2000 24 h before RSV infection. The pZsProSensor-1 plasmid encodes a destabilized GFP consisting of amino acids 422–461 of the degradation domain of mouse ornithine decarboxylase protein fused to the COOH terminus of the naturally occurring reef coral species protein (ZsGreen). This fusion protein is rapidly degraded by the proteasome in an ubiquitin-independent manner. The NF-κB–dependent ELAM-1 promoter-driven firefly luciferase plasmid (pELAM-luc) was obtained from M.J. Fenton (University of Maryland, Baltimore, MA). The p5xATF6GL3 plasmid (also known as the UPR element luciferase reporter) was obtained from R. Prywes (Columbia University, New York, NY). The IκB SR expression plasmid was obtained from R. Hay (University of St. Andrews, St. Andrews, Scotland). The pRL-TK plasmid was purchased from Promega. 400 ng of firefly luciferase construct was cotransfected with 50 ng pRL-TK using LipofectAMINE 2000 (Invitrogen). Cells were harvested 48 h after transfection, and luciferase activity was quantified using the Dual-Luciferase Assay system in a Berthold multitube luminometer. Data are plotted as the ratio of firefly/renilla activity to correct for transfection efficiency among samples. Data were analyzed with InStat version 3.0 (GraphPad Software). Values are presented as means ± SD. Multiple comparisons were made by analysis of variance between groups using the Tukey-Kramer multiple comparisons test, and paired samples were analyzed by tests. In both cases, statistical significance was defined as P < 0.05 or less. Fig. S1 is a model of functional associations between differentially expressed genes in SP-C cells associated with apoptosis using PathwayAssist software. Table SI lists descriptive information of apoptosis-associated genes from the model in Fig. S1. Online supplemental material is available at .

Mammalian skin epithelium is a self-renewing tissue that constitutes the barrier between an organism and its environment. To provide the organism with this essential function, epidermis must balance proliferation and differentiation (; ). Its innermost basal layer adheres to an underlying basement membrane rich in ECM. This layer contains proliferative keratinocytes that are typified by their expression of genes encoding integrins and growth factor receptors, particularly EGF receptor (EGFR; also referred to as ErbB1), as well as the structural keratins 5 and 14 (K5 and K14; ; ). As basal cells move upward, they repress basally expressed genes and switch to expressing a set of differentiation-associated proteins, including keratins K1 and K10. As keratinocytes continue their trek, they further adjust their transcriptional program to culminate in the production of dead, flattened squames that are sloughed from the skin surface as new cells moving outward replace them. Epidermal homeostasis is under tight transcriptional regulation (). Sequence motifs for the binding of the AP-2 family of transcription factors are found in most epidermal promoters and enhancers irrespective of terminal differentiation status (; ; ; Zeng et al., 1997; ; ; ; ; ). Of the five known murine AP-2 proteins, four are differentially expressed in the skin. Of these, is most highly expressed (; ), making it an attractive candidate transcription factor for regulating epidermal-specific transcription. Although a role for AP-2 factors in epidermal gene expression seems likely, a clear picture as to how they may be involved has not yet emerged. Do AP-2 family members promote or repress proliferation and/or differentiation? Are these effects dependent on the particular family member expressed or the relative differentiation stage of the keratinocytes? Often, studies have led to seemingly opposing conclusions. In cultured keratinocytes, for example, AP-2α seems to repress the promoter activity of the basal cell keratin gene (), but in vivo, AP-2 factors are expressed throughout the epidermis and K5 is restricted to basal cells. In hyperproliferative skin, factors are coexpressed with suprabasally (). Data on the role of AP-2 proteins in other epithelial cells offer little assistance in resolving these issues, where both active and repressive roles for AP-2 proteins have been described (; ; ). In mammary carcinoma cell lines, for instance, 5′ regulatory sequences for the growth-promoting and the ErbB subfamily of EGFR genes seem to be positively regulated by AP-2α (), whereas overexpression of AP-2α appears to be growth and proliferation inhibitory (). Similarly, in breast cancer tissue, enhanced expression of is a frequent occurrence, and yet diminished expression has often been cited as a poor prognostic marker for breast cancer survival (; ). These tantalizing but often contrasting results underscore the importance of resolving the possible link between AP-2α and epithelial growth. A major difficulty in evaluating how AP-2 family members orchestrate transcriptional regulation in skin epidermis stems from the disparate results obtained from functional studies across different vertebrate species. In frog embryos, injection of antisense oligonucleotides leads to the loss of epidermal character and the gain of neural gene expression (). In contrast, the embryonic epidermis of mice lacking AP-2α seems to develop normally, although early perinatal lethality has precluded analyses of postnatal mouse skin (; ; ). The physiological relevance of other AP-2 family members in skin also remains unknown, as –null mice do not display a skin phenotype, and targeting of results in lethality before epidermal and follicle development (; ). In this study, we use conditional gene targeting of to explore the functional significance of AP-2α in postnatal skin development. We show that AP-2α functions in the epidermis by repressing gene expression as cells exit the basal layer and commit to terminally differentiate. We show that nuclear AP-2α is present normally in some basal and many suprabasal epidermal cells and that it is essential for governing the EGFR-mediated control of epidermal cell proliferation. Other AP-2 family members do not appear to compensate in the suprabasal differentiating epidermal layers, where expression fails to switch off in the absence of AP-2α. Upon growth factor signaling and EGFR activation, –null epidermis displays hyperproliferation and formation of papilloma-like invaginations accompanied by abnormal suprabasal elevation of activated Akt. Finally, we show that, mechanistically, the promoter possesses AP-2–binding sites that are occupied by AP-2α and that transcriptionally temper receptor gene expression. In vitro, loss of AP-2α elevates gene transcription, and regulatory circuitries for the phosphoinositol-3 kinase (PI3K), Akt, and MAPK fail to function properly. These findings have major implications for understanding why reductions in expression have been associated with tumorigenesis and cancer. Recently, mice were genetically engineered so that essential coding sequences of the gene were flanked by loxP sites ( ; ). To elucidate the role of AP-2α in postnatal skin epidermis, we bred these animals to mice harboring a recombinase transgene, which efficiently expresses Cre throughout epidermis by embryonic day (E) 15.5 (). line was phenotypically indistinguishable from wild type (WT). lines were produced at expected Mendelian ratios of genotypes according to PCR analyses of genomic DNA (). Microarray analyses revealed the presence of , , and mRNAs in E18.5 skin epidermis, whereas is primarily in dermis (unpublished data; for expression analyses, see ; ; ). We corroborated and extended these results by conducting real-time PCR on mRNAs isolated from WT epidermis that was purified from skin by dispase treatment. Expression of and mRNAs was particularly high (). showed that the signal for 's Mfloxed coding exon was abolished, verifying the efficacy of the targeting event. Importantly, the absence of AP-2α did not appreciably affect the expression of any of the other family members, indicating a failure to compensate at the level of gene expression (). Immunoblot analyses confirmed the expression of and in WT skin epidermis and verified that the targeting event resulted in the loss of AP-2α protein production and also underscored the specificity of the AP-2α antibody (). By immunofluorescence, anti-AP2α labeled basal and suprabasal nuclei of WT cells, whereas anti-AP2γ preferentially labeled basal nuclei ( and not depicted). skin, and anti-AP2γ staining was unchanged. mice as conditional knockout (KO [cKO]). In agreement with and expected from previous reports of the full KO of in mice, the skin surface of newborn cKO pups appeared similar to that of their WT littermates (not depicted; ; ; ). Upon histological inspection, however, cKO skin displayed a thickened interfollicular epidermis (). Despite the complete absence of AP-2α, the thickening was not uniform across the postnatal day (P) 0–3 epidermis and began to wane altogether by P6. These aberrations had not been noted previously in the straight KO animals that died at birth. Whether this is attributable to the relatively mild nature of the defects or to strain-related differences in the mice used for the two studies () was not addressed. Our conditional targeting strategy enabled us to examine, for the first time, the consequences of AP-2α loss to postnatal skin development. As the animals aged, cKO adult mice progressively lost their hair in select regions of their coats. The loss of hair was most pronounced on the ventral thoracic surface, where a large area of skin always became bald (). Histological analyses revealed that in contrast to the typically thin epidermis of adult WT animals, the cKO animals displayed hyperthickened, papilloma-like undulations throughout these expansive areas (). As judged by light and electron microscopy, these undulations displayed epidermal rather than hair follicle morphology. In addition to thoracic skin epithelium, dorsal ear epidermis was also markedly affected, typified by both hyperthickening and parakeratosis in cKO skin (). To examine the consequences of ablation on epidermal differentiation, we first used immunofluorescence microscopy (). In regions where the –null epidermis appeared morphologically normal, differentiation markers were expressed in patterns indistinguishable from WT skin. These included basal keratins K5 and K14, the spinous layer keratins K1 and K10, and granular layer proteins filaggrin and loricrin ( and not depicted; for review see ). In hyperthickened regions, the classical changes in gene expression associated with the hyperproliferative state were observed (), including sustained expression of keratins K5 and K14 in the suprabasal layers and induction of the outer root sheath keratins K6, K16, and K17 in the suprabasal layers of the epidermis (; and not depicted). These abnormalities were accompanied by an approximately two- to threefold increase in proliferation in the basal layer as judged by BrdU incorporation () and by labeling with antibodies against the proliferating nuclear antigen Ki67 (not depicted). The biochemical abnormalities noted in hyperthickened neonatal regions were enhanced in the adult papilloma-like invaginations (). In addition, the papilloma-like regions displayed some suprabasal proliferating cells as judged by staining for Ki67, which is nuclear in cycling cells (, top; arrows). These data are suggestive of a perturbation in the mechanism that normally restricts dividing cells to the basal layer. In contrast, we observed no obvious change in apoptosis, as indicated by either histology or antibody staining against the activated form of caspase 3. The caspase 3–positive cells (, bottom; arrows) in catagen-phase hair follicles undergoing cyclic apoptosis provided a nice internal control for these stainings. Because the alterations we observed in differentiation patterns are classical features of hyperproliferation, they could represent a reflection of an abnormality in the proliferative machinery of the epidermis. Alternatively, if these changes were a primary consequence of a loss of AP-2α, it is possible that the alterations in differentiation could perturb the epider-mal barrier, which is a process known to result in indirect hyperproliferation (). To begin to understand the underlying basis for the phenotype, we first performed a dye penetration assay on newborn WT and –null pups (). With the exception of the tail and umbilicus, which were severed in both WT and null pups, toluidine blue dye did not penetrate the skin (). When coupled with the mosaic hyperproliferative defects seen in the –null skin, these results indicated that there must be other molecular events that are dependent on AP-2α and that render the epidermis susceptible to an imbalance in proliferation and differentiation. Members of the EGF and insulin growth factor (IGF) families are particularly important in regulating proliferation in the epidermis (for review see ). Of these, and genes were particularly interesting in that they both harbor AP-2–binding sites in their promoters. Additionally, constitutive expression of in transgenic mice results in a transient epidermal hyperproliferation that wanes in the adult and can reemerge after wounding and/or mechanical stress (; ), whereas overexpression of constitutively active members of the receptor family can lead to papilloma undulations throughout the skin (). Finally, EGF injection in sheep skin results in concomitant hair loss and epidermal thickening (), whereas EGFR down-regulation has been associated with hair follicle formation (). Collectively, although all prior gene studies have argued for a positive rather than a negative effect of AP-2α on EGF signaling pathway genes (; ; ; ), the significant physiological parallels prompted us to focus on the status of genes involved in EGF signaling in our cKO animals. We first performed semiquantitative RT-PCR on mRNAs from purified hyperproliferative neonatal cKO and WT epidermis. By this criteria, mRNA levels were significantly elevated (). Immunoblot analysis showed that the increase in mRNA expression was reflected at the protein level (). Immunofluorescence microscopy furtherrevealed sustained expression of EGFR in the suprabasal layers of –null epidermis (). This difference was evident by P0, making it an early consequence of the targeting event. Furthermore, the increase in epidermal thickening was associated with the phosphorylated (i.e., activated) form of the receptor. The epidermis of adult mice is typically thin, and its proliferative activity is low. Correspondingly, it was not surprising to see a single layer of anti-EGFR labeling with very little phospho-EGFR labeling in WT adult skin (). In contrast, the lesional regions of –null epidermis displayed elevated levels of not only total but also autophosphorylated EGFR. The null–associated up-regulation in total EGFR was confirmed by immunoblot analyses, as was the elevated levels of activated EGFR in lesional regions of the cKO mice (). A priori, the observed elevation in EGFR and/or signaling could be caused either directly by the loss of AP-2α or indirectly by the enhanced proliferation that occurs concomitantly with the absence of AP-2α. To distinguish between these possibilities, we first examined P6 –null and WT back skin (). Despite the similar morphologies and thickness of WT and –null epidermis at this age (), EGFR was still elevated in the absence of AP-2α. Interestingly, however, antibodies against the phosphorylated form of EGFR showed no difference between WT and cKO skin, suggesting that the EGFR was not active. Further consistent with the lack of EGFR signaling and hyperproliferation in these regions was the lack of anti-K6 staining in the epidermis. null epidermis appeared largely normal (). By immunofluorescence analysis, however, the EGFR levels were clearly elevated. In contrast to hyperproliferative cKO thoracic skin, the levels of tyrosine-phosphorylated EGFR remained low. This was confirmed by immunoblot analyses (). Based upon these data, two important findings emerged. First, AP-2α loss resulted in increased expression. Second, the hyperproliferative phenotype correlated with the activation of these EGFR tyrosine kinases, a process which is typically dependent on ligand. TGFα is both a ligand for EGFR and a potent growth factor of epidermal keratinocytes in vitro (). In nonlesional adult back skin, mRNAs were extremely low in both WT and cKO mice (). In contrast, in neonatal (P2) epidermis, where EGFR activity and proliferation were detected in both WT and cKO mice, mRNAs were readily detectable (). The levels were slightly higher in cKO relative to WT P2 epidermis, and this and/or other EGF signaling molecules could enhance the effects of elevated EGFR in neonatal cKO mice. Altogether, the patterns of TGFα, EGFR, and EGFR activation were consistent with the phenotypic effects observed. To further explore these parallels, we took advantage of the previous observation that topical application of 12--tetradecanoylphorbol-13-acetate (TPA) causes elevation of TGFα, leading to activation of EGFR in adult skin (). If, as we surmise, the epidermal hyperproliferation seen in lesional –null skin is at least in part caused by regional activation of elevated EGFRs, nonlesional –null back skin might be expected to show elevated thickening upon TPA treatment. At the doses of TPA administered in this study, TPA treatment had only a modest effect on the back skin of control adult mice. In contrast, these TPA treatments caused hyperthickening of the –null back skin accompanied by activation of EGFRs (). These morphological aberrations were accompanied by an increase in proliferation and biochemical alterations that are typical of a hyperproliferative state (not depicted). Even after 4 mo of treatment, however, no signs of tumor formation or progression were observed, indicating that AP-2α loss alone was not sufficient to cause skin tumorigenesis. To further probe the underlying mechanisms responsible for the up-regulation of in the absence of AP-2α, we cultured primary keratinocytes from neonatal WT and cKO mouse skins. To evaluate the data, it was first essential to examine the expression of and in vitro to see whether the preferential suprabasal expression of was recapitulated in culture, as we had observed in epidermis (). To examine this, we induced terminal differentiation by elevating the calcium levels in the medium and then performed immunoblot analyses on protein extracts from low- and high-calcium exposed cells. As shown in , within 24 h of the calcium shift, the expression of the spinous layer markers and were elevated, although appreciable involucrin mRNA was present even in the low-calcium state. Within the same time frame, AP-2α was largely unaffected by calcium treatment relative to internal protein loading standards. In contrast, AP-2γ was markedly decreased upon calcium-induced differentiation. Moreover, the levels of AP-2γ were not appreciably changed in the –null cells (). Together, these data were consistent with our prior in vivo data presented in . Next, we conducted immunoblot studies in which we could examine EGFR levels under conditions where we could more rigorously control for microenvironment and cell numbers. As shown in , EGFR levels were elevated in the –null keratinocytes relative to their WT counterpart. Under enriched culture conditions, both cell populations displayed phosphorylated EGFR (not depicted). These changesin –null keratinocytes were also reflected at the mRNA level (). Real-time PCR showed a fourfold increase in mRNAs in KO versus WT cells. mRNA levels were only modestly elevated in the –null state (not depicted). If AP-2 acts to repress gene transcription, it should bind to the endogenous promoter. Previous studies have shown that recombinant AP-2α can bind to an AP-2 consensus motif in the human promoter (; ). In the mouse promoter, multiple AP-2 consensus binding sites are situated within a kilobase upstream from the transcription initiation site (, ovals). Two of these sites are conserved across mammalian species (, ovals highlighted in red). To evaluate whether endogenous AP-2α binds directly to one or more of these sites in epidermal cells, we conducted chromatin immunoprecipitation (ChIP) assays using the monospecific anti–AP-2α antibody. Anti–AP-2α antibodies specifically immunoprecipitated chromatin–protein complexes that contained an ∼500-bp DNA fragment encompassing the conserved putative AP-2α–binding sites (). In contrast, ChIP analysis of WT skin did not display PCR bands with primers corresponding to either the promoter or downstream regions that did not contain AP-2–binding motifs (not depicted). These data provided the first ChIP data illustrating the binding of AP-2α to an endogenous gene that contains AP-2 consensus binding motifs. To test whether gene transcription is affected by the loss of AP-2α, we transfected Ca grown keratinocytes with a luciferase reporter gene driven by a 1.1-kb promoter fragment harboring the AP-2α–binding sites. By 48 h, luciferase activity was markedly elevated in –null versus WT cells (). Point mutations in the AP-2α motifs raised promoter activity levels in WT but did not affect –null cells. Finally, we infected WT and KO keratinocytes with a retroviral expression vector and repeated the reporter assay. Exogenous expression of resulted in a potent repression of luciferase activity driven by the WT promoter but not its mutant counterpart (). These data underscore the AP-2–specific nature of the manipulations and the repressive effects on transcription. These findings further suggest that AP-2α functions in suppressing gene expression as epidermal cells commit to terminally differentiate. To evaluate how downstream signaling events may be affected when EGFR levels are elevated, we first examined the proliferative potential of primary –null and WT epidermal keratinocytes. Both populations grew equally well when plated at high density on a fibroblast feeder layer and in rich growth media (unpublished data). However, when subjected to more stringent conditions such as sparse plating with no feeder layer for long term growth (>5 d) or low serum/growth factor– supplemented medium, only KO keratinocyte cultures grew well (representative examples in ). Moreover, the growth advantage of the KO keratinocytes was largely eliminated by the addition of AG1478, an EGFR-specific protein kinase inhibitor, to the culture medium, suggesting that increased EGFR signaling might be responsible for the enhanced proliferative potential observed in KO keratinocytes. EGFR stimulation typically leads to downstream activation of the MAPK family members Erk1/2, but it can also lead to activation of Akt kinases. Both are required for epidermal growth and differentiation (; ). To determine whether the loss of AP-2α and elevated EGFR signaling results in an elevated activation of MAPK (Erk1/2) and/or Akt pathways, we stimulated 24-h serum-starved cells with EGF. Within a minute after stimulation, the two populations activated Erk1/2 with similar kinetics (). Relative to total Erk1/2, the phosphorylated (i.e., activated) Erk1/2 signals were always higher in the KO keratinocytes. This was visualized over a range of EGF concentrations (). Although –null keratinocytes activated more Erk1/2 in response to EGFR ligands, the most significant difference observed was in EGF-mediated Akt activation. Again, the first signs of activation appeared within a minute after exposure to EGF. However, relative to equivalent levels of total Akt, phosphorylated (i.e., activated) Akt was nearly an order of magnitude higher in KO cells across a range of EGF concentrations. Remarkably, as little as 0.5 ng/ml EGF was sufficient to elicit these differences in Erk1/2 and Akt activation. Curiously, the differential effects on Akt appeared to be specific for EGFR signaling, whereas the differential activation of Erk1/2 was also seen with IGF-1, another potent stimulator of keratinocyte growth and survival (). The effects were obliterated by treatment with the PI3K-specific inhibitor , indicating that Akt activation in EGF-treated keratinocytes was mediated through PI3K activation (). To test the physiological relevance of these findings, we examined the status of activated Akt in our mice. In nonlesional regions of –null skin where EGFRs were elevated but not activated, phospho-Akt levels remained low and comparable with WT skin ( and not depicted). However, in either TPA-treated or thoracic cKO skin, activated Akt was markedly elevated. Even in WT skin, TPA treatment resulted in mildly up-regulated levels of activated Akt, which is consistent with the modest epidermal thickening we noted previously (). Based upon these criteria, the mechanisms uncovered in our in vitro analyses appeared to be operative in vivo. AP-2–binding sites have been found in a myriad of genes differentially expressed in both basal and suprabasal compartments of mammalian epidermis (; for review see ). However, to date, their functional significance has remained elusive. In this study, we have shown that AP-2α plays a role in the switch between epidermal cell proliferation and differentiation, and we have provided the first experimental evidence to integrate this transcription factor into a key signaling pathway in the skin. The finding is especially important given that some studies have highlighted AP-2α as a tumor suppressor (; ; ), whereas others have suggested a growth-promoting effect of the protein (; ). Our study clarifies how AP-2α governs epidermal proliferation and provides important new insights into why inverse correlations are often found between AP-2α and EGFR family members in human epithelial cancers. Our studies place AP-2α in the EGFR signal transduction cascade. Using in vivo and in vitro approaches, we found that gene expression was elevated upon genetic ablation of or by mutation of the AP-2–binding sites in the promoter. Correspondingly, gene expression was down-regulated when AP-2α was added back to the KO cells through retroviral transgene expression. Although AP-2α is likely to have many other target genes in epidermis, several of the phenotypic abnormalities in the conditionally null mice can be explained by EGFR misregulation. EGFRs have long been known to play an important role in regulating the development of the epidermis and its appendages (; ; ; ). In mammals and birds, overexpression or injection of EGF can arrest epidermal appendage development and promote epidermal thickening, and, conversely, activation of EGFRs is typically diminished in areas of follicle formation (; ; ). Postnatally, EGFRs are predominantly expressed in the basal epidermal layer and down-regulated as cells commit to terminally differentiate (). Once activated, EGFR has the ability to activate Ras–MAPK signaling, which is often linked to proliferation, as well as PI3K–Akt signaling, which is more typically associated with cell survival (; ). Relevant to the data presented here, tumorigenesis can arise from superactivated Ras–MAPK signaling in transgenic mouse skin but only in conjunction with EGFR signaling to activate PI3–Akt (). A key feature of EGFR signaling is that its activation is dependent on ligand stimulation. Made and secreted by salivary glands, EGF can enter the bloodstream and reach epidermal tissue, whereas TGFα is the major autocrine growth factor of the epidermis. TGFα is down-regulated postnatally but can be up-regulated upon injury or TPA treatment (). That postnatal skin is limiting for EGFR ligands is graphically illustrated by the papilloma-like undulations that develop in skin of mice engineered to overexpress an EGFR relative (ErbB2) that lacks the requirement for ligand stimulation (; ). When placed in the context of these prior studies, we can account for many of the features of the seemingly complex phenotype of our cKO mice. The hyperproliferation that occurs in cKO newborn epidermis despite elevated EGFR levels suggests that the ligand pool for these receptors must be quite large at this age. The levels appear to be saturating for WT but not cKO EGFR levels, and this is reflected by the observed enhancement of phosphorylation and activation of the EGFRs in newborn cKO skin. Soon after birth, hyperproliferation and phosphorylation of EGFRs returns to normal in cKO despite sustained elevated EGFR levels in cKO skin. After embryogenesis, the ligand pool is known to be attenuated along with other growth-promoting signals, and our results indicate that these pool levels have dipped below saturation even for WT EGFRs. Finally, the markedly hyperproliferative epidermis and phosphorylation of EGFRs in lesional but not nonlesional regions of adult cKO thoracic and ear skin and TPA-treated cKO back skin is reflective of the elevated ligand pools that are known to arise upon mechanical irritation and TPA treatment. Additionally, when animals lick their wounds, the EGF-rich saliva could further contribute to the robust proliferative response in the ventral chin area. An additional point worthy of mention is that the –null papilloma-like undulations were not as severe as the spontaneous skin tumors that arise from chemical or transgenic mutations in epidermally expressed Ha-Ras. We surmise that the underlying reason for this may be rooted in the normal feedback regulatory loops that are operative in dampening the deleterious effects of overly active EGFR signaling. Indeed, for sustained epidermal hyperproliferation and tumorigenesis to occur on a background of Ha-Ras mutations, sustained EGFR signaling must still be permitted to activate and sustain elevated PI3K–Akt signaling (). Our data shed additional insight onto this signaling circuitry by demonstrating that the EGFR response to Akt activation is different than that of IGF receptor in epidermal cells and that in situations in which EGFR signaling is hyperactive, Akt activation can be very high and suprabasal proliferation can occur. Although Akt activation is typically associated with cell survival, several examples that link PI3K–Akt activation to proliferation may be relevant to the findings we report. Sustained PI3K–Akt activation also occurs in liver tumors, where it posttranslationally silences the C/EBPα transcriptional repressor that controls hepatocyte proliferation (). Intriguingly, C/EBPα has also been implicated in epidermal differentiation and is repressed in some skin cancers (). Although beyond the scope of this study, such a mechanism could explain why null–mediated hyperactivation of the EGFR–PI3K–Akt pathway leads to an increase in proliferation and epidermal thickening in vivo. After 15 yr since the original implication of AP-2 in transcriptional regulation in the epidermis (; ), the functional importance of these proteins in epidermis is now emerging. Recent knockdown experiments underscore a role for AP-2α in embryonic skin development (), and our studies now reveal an essential role for mammalian AP-2α in orchestrating the balance between epidermal proliferation and differentiation. This process can be largely explained by AP-2α's repressive effects on gene transcription, but given the repertoire of epidermally expressed genes with AP-2–binding sites, there are likely to be many additional key genes controlled by AP-2α. Why have these other target genes not surfaced in our analyses? The answer seems, at least in part, to reside in functional redundancy among AP-2 family members, four of which are expressed in the epidermis. This is substantiated by the fact that both in vivo and in vitro, the consequences of AP-2α ablation appeared to be more dramatic in the differentiating cells where expression is down-regulated. Additionally, appears to be more highly expressed in those basal cells that are not actively cycling, leading us to speculate that AP-2's role may be most critical at the juncture between proliferation and differentiation in the epidermis. Evaluating the degree of AP-2 functional redundancy in the epidermis must await functional studies on the other members of the AP-2 family that are expressed in skin. Cell type and differentiation-specific cofactors are likely to further impact the complexities of when and how AP-2 proteins act in the epidermis. Studies in other systems have already indicated that interactive partners for AP-2 proteins can influence whether AP-2 proteins act as transcriptional repressors or activators (). Differences in AP-2 recognition motifs are also likely to influence target gene specificity, and the differential expression of AP-2 family members and possibly putative AP-2 cofactors could further magnify differences in the spatial and temporal behavior of putative AP-2 target genes within a tissue. Such differences are also likely to contribute to our understanding of why the loss of AP-2β results in massive apoptosis in the kidney () and why the loss of AP-2α in neural crest impaired craniofacial development and pigmentation (). These issues are also likely to underlie the seemingly opposing findings that the addition of recombinant AP-2α to a nuclear extract from a human squamous cell carcinoma line in vitro led to an increase in transcription (), whereas AP-2α in mouse epidermis and in cultured epidermal keratinocytes had an inhibitory effect on promoter activity (our study). The finding that AP-2 proteins positively regulate cell proliferation in some cells and inhibit growth in others may help in the future to explain the seemingly disparate results obtained concerning the roles for AP-2 family members in human cancers (; ; ; ; ). AP-2α floxed mice ( ) and transgenic mice were generated as described previously (; ). Genotyping was conducted by PCR of tail skin DNAs. For light or fluorescence microscopy, tissues were embedded in optimal cutting temperature compound and frozen on dry ice. For semi-thin sections and transmission EM, tissues were fixed and processed as previously described (). For indirect immunofluorescence, 10-μm sections were permeabilized with 0.1% Triton X-100 and washed. For mouse monoclonal antibodies, we used the MOM kit (Vector Laboratories); for antibodies from other species, we used 2.5% normal donkey serum, 2.5% normal goat serum, 1% BSA, 2% gelatin, 0.1% Triton X-100 as block, and antibody diluent. Primary antibodies used are as follows: mouse: AP-2α (1:5; T. Williams) and phospho-EGFR (1:200; Upstate Biotechnology); rabbit: K6 (1:200; Fuchs laboratory), K1 (1:250; Fuchs laboratory), Ki67 (1:1,000; Novocastra laboratories), active caspase 3 (1:500; R&D Systems), K17 (1:1,000; a gift from P. Coulombe, The Johns Hopkins University School of Medicine, Baltimore, MD), EGFR (1:100; Upstate Biotechnology), and phospho-Akt (1:250, Cell Signaling); guinea pig: K5 (1:200; Fuchs laboratory); rat: BrdU (1:150, AbCam) and β4 integrin (1:250, BD Biosciences). For phospho-EGFR staining, sections were fixed in cold 100% methanol, primary antibody incubations were performed as above, and the signal was amplified using the ABC kit (Vector Laboratories) and visualized using the tyramide signal amplification Plus Fluorescence detection kit (PerkinElmer). In situ hybridizations of K5 transcripts were performed as described previously (). Probe synthesis was performed according to the manufacturer's instructions (Roche). The histology, immunofluorescent, and in situ hybridization images were taken by a mot plus microscope (Axioskop 2; Carl Zeiss MicroImaging, Inc.). The objectives used were 20× NA 0.5 plan Neofluar ∞/0.17 and 40× NA 1.3 oil plan Neofluar ∞/0.17 (Carl Zeiss MicroImaging, Inc.). The images were taken at room temperature in antifade as an imaging medium for immunofluorescent images and 80% glycerol for hematoxylin and eosin (H&E) images. The fluorochromes used were FITC, Texas red/rhodamine red X, and DAPI. A slider camera (SPOT RT; Diagnostic Instruments) and MetaMorph 6 (Molecular Devices) software were used to acquire the pictures. Adobe Photoshop 6.0 software was used for contrast and brightness adjustment. The immunoblot and PCR gel images were acquired with an Alphaimager (Innotech), and Quantity One software (Bio-Rad Laboratories) was used for contrast and brightness adjustment. Enzymatic separation of epidermis from skins and primary newborn mouse epidermal cell cultures were performed as described previously (). To induce keratinocyte differentiation in culture, the calcium concentration of the media was raised from 0.05 to 1.5 mM for 24 h, after which protein and RNA were extracted. For growth comparisons, 2.5 × 10 of freshly isolated cells were plated with feeder layers in 24-well dishes. For stringent growth conditions, serum was adjusted to 2%, and/or growth factor supplements and the addition of fibroblast feeder cells were omitted. For growth comparisons with the presence of AG1478, 200 nM of the inhibitor (Sigma-Aldrich) was added to the serum daily after plating. Keratinocyte numbers were determined with a coulter counter. Nucleotides −1,180–29 (transcription initiation site = 0) of the mouse promoter were cloned upstream of the firefly luciferase gene in the pGL2 basic plasmid (Promega). Cells were transferred into 12-well dishes and grown to 30–40% confluency before Fugene 6 (Roche) reagent-assisted transfections of 2 ng cytomegalo virus– luciferase DNA (control) and 100 ng of either WT or mutant promoter firefly luciferase constructs (pEGFRpr-Luc or pMut-EGFRpr-Luc) or empty vector (p-Luc; Promega). 48 h after transfection, luciferase assays were performed as described previously (; ). Transfection efficiency was 2–3%. A retroviral vector containing the full-length mouse AP-2α cDNA was engineered and infected for AP-2α rescue studies (70% efficiency). Mice were anesthetized, and their backs were shaved. Each mouse received 2.5 μg TPA (dissolved in acetone) on the right half of the back twice per week for 4 mo. As a control, the left half was treated with acetone. In vivo ChIP was performed as described previously (). The presence of AP-2α sites was confirmed by rVista analysis of 5′ upstream sequences as defined by the ECR Browser and Ensemble software (European Bioinformatics Institute and Sanger Institute). AP-2 sites were chosen for ChIP analysis based on the conservation and alignment between mouse and at least one other species, including human, canine, and rat, and clustering of sites when applicable. As a control, PCR was also performed using primers that recognize other sites within the same promoter or downstream portions of the same gene to demonstrate the specificity of the pull-down. SDS-PAGE and immunoblot analyses were used to identify the AP-2α, AP-2γ, EGFR, MAPK (Erk1/2), and Akt proteins. AP-2α and AP-2γ proteins were identified with monoclonal antibodies against AP-2α and AP-2γ (Santa Cruz Biotechnology, Inc.), respectively. The active (i.e., phosphorylated) forms of the EGFR, MAPK, and Akt proteins were identified with phosphospecific antibodies against EGFR (Tyr1173; Upstate Biotechnology), MAPK (Thr183 and Tyr185 in Erk2; Sigma-Aldrich), and Akt (Ser473; Cell Signaling), respectively. The total levels of these proteins were identified with panel antibodies against EGFR (Upstate Biotechnology), MAPK (Cell Signaling), and Akt (Cell Signaling), respectively.

Fibroblasts synthesize, organize, and maintain connective tissues during development and in response to injury and fibrotic disease (; ; ). Cells cultured in three-dimensional (3D) collagen matrices have been used to study fibroblast–matrix interactions in a tissue-like environment. Fibroblast morphology in the 3D environment ranges from dendritic to stellate to bipolar, depending on matrix stiffness and tension (; ), which is similar to cells in tissues (; ) and quite distinct from the flattened morphology of fibroblasts on two-dimensional (2D) tissue culture surfaces. Cells can exert mechanical force on their surroundings (; ; ; ), and fibroblasts in 3D collagen matrices use this force to contract the matrix (; ; ; ; ; ). The mechanism by which fibroblasts regulate the contraction of 3D collagen matrices has been shown to vary according to growth factor stimulus, mechanical environment, and the differentiation state of the cells. The physiological agonists PDGF and lysophosphatidic acid (LPA) both stimulate floating matrix contraction, even though these agonists have opposite effects on the movement of cellular dendritic extensions within the matrices. PDGF increases their protrusion; LPA causes their retraction (). Studies with C3 exotransferase showed that the small G protein Rho is required for floating matrix contraction by either PDGF or LPA (), but only PDGF-stimulated, and not LPA-stimulated, contraction was inhibited by blocking the Rho effector Rho kinase (; ). Therefore, PDGF and LPA regulate floating collagen matrix contraction, in part, by different signaling mechanisms, and it has remained an open question as to whether there is a point of convergence. p21-activated kinases (PAKs) were first identified as Rac- and Cdc42-interacting proteins () and are now known to be important in the regulation of cytoskeletal organization and cell migration (; ). Early on, PAK1 was recognized as a downstream effector for PDGF (; ), but more recently was shown to also be important for LPA-mediated signaling (; ). In this study, we show that the PDGF and LPA signaling pathways that regulate matrix contraction converge on PAK1 and its downstream effector cofilin and that contraction depends on cellular ruffling activity, rather than on protrusion and retraction of cellular dendritic extensions. We also show that, depending on the agonist, different Rho effectors are required to cooperate with PAK1 to regulate matrix contraction, Rho kinase in the case of PDGF and mDia1 in the case of LPA. We used small interfering RNA (siRNA) to knock down PAK1 expression in human fibroblasts. shows an example of immunoblot analysis performed on cell lysates prepared from cells after a 36-h transfection with PAK1-specific double-stranded siRNA. Levels of PAK1 in the PAK1 siRNA, but not in mock-transfected cells, were reduced by almost 95% without affecting levels of PAK2. shows the morphology of PAK1-silenced versus mock-transfected cells by fluorescence visualization of actin. Compared with control cells, knocking down PAK1 had no detectable effect on cell spreading or response to PDGF and LPA in 2D culture. Treatment with PDGF caused the appearance of small lamellipodia along the cell margins, and treatment with LPA increased formation of actin stress fibers. Transfection studies with modified PAK1 constructs have demonstrated a role for PAK1 in cell motility (Sells et al., 1999). Consistent with this finding, human fibroblast migration was decreased by knocking down PAK1. shows the typical appearance of cultures that were scrape wounded and incubated in medium containing PDGF. During the initial 3–6 h of culture, both mock-transfected and PAK1 knockdown cells extended lamellipodia into the scrape region. By 24 h of incubation, however, PAK1 knockdown cells had migrated substantially further into the wound region compared with the controls. shows the results quantitatively with the difference in cell migration evident by 12 h. Fibroblasts within 3D collagen matrices protrude a dendritic network of extensions that expands in response to PDGF stimulation and retracts in response to LPA stimulation (). (1 h) and (4 h) show representative photomicrographs of the network under basal (BSA), expanded (PDGF), and retracted (LPA) conditions. Fibroblast dendritic extensions have microtubule cores ( and , green) with ruffling, actin-rich tips (red). Retraction of the extensions in response to LPA stimulation occurred within 1 h in PAK1-silenced or mock-transfected cells, after which most fibroblasts either were round or had short extensions. Expansion of the dendritic network in response to PDGF was reduced in PAK1 knockdown cells. shows morphometric analysis of a representative experiment. After 4 h in basal- and PDGF-containing medium, the projected cell area, branch length, and number of branches were all lower in PAK1-silenced cells. Therefore, PAK1 played a role in regulation of fibroblast dendritic extensions, but formation of the extensions was PAK1 independent. The actin-rich tips of fibroblast extensions (protruded or retracted) showed prominent ruffling activity. In general, ruffling was more evident in PDGF- and LPA-stimulated cells than in basal (BSA) medium, and B and B show that cell ruffling induced by PDGF or LPA was markedly reduced in PAK1-silenced cells, compared with mock-transfected cells. Therefore, PAK1 appeared to be necessary for cell ruffling, regardless of whether the cells were stimulated by PDGF or LPA. Collagen matrix contraction experiments also were performed using PAK1-silenced and mock-transfected cells. Contraction was quantified by measuring the diameter of collagen matrices before (∼12 mm) and after contraction and then calculating the difference. shows that after 4 h in the presence of LPA or PDGF, contraction was ∼6–7 mm for control cells and ∼2 mm for PAK1 knockdown cells. The latter value is comparable to the level of basal contraction without agonist stimulation. These findings demonstrated that PAK1 was required for both PDGF- and LPA-stimulated matrix contraction. Current studies suggested PAK1 as a potential point of convergence in PDGF and LPA regulation of cell ruffling and collagen matrix contraction. Activation of cell ruffling and PAK in response to PDGF stimulation depends on PI3 kinase (; Sells et al., 2000). In previous papers, the role of PI3 kinase in contraction of floating collagen matrices was tested, but the findings were inconsistent (; ; ). In preliminary experiments, we established that 20 μM inhibited PDGF-stimulated PI3 kinase activity based on measurement of Akt phosphorylation (unpublished data). demonstrates that this concentration of the PI3 kinase inhibitor (LY) blocked PDGF-stimulated, but not LPA-stimulated, matrix contraction. In addition, shows that blocking PI3 kinase inhibited PDGF-stimulated, but not LPA-stimulated, cell ruffling. Therefore, PI3 kinase was required for both cell ruffling and contraction stimulated by PDGF, whereas the link between LPA and PAK1 appeared to be PI3 kinase independent, as has been reported (). LPA receptors couple to multiple G protein signaling pathways (; ), and Gα has been implicated in floating collagen matrix contraction (). shows that overnight treatment with pertussis toxin inhibited LPA-stimulated, but not PDGF-stimulated, matrix contraction. In addition, shows that pertussis toxin treatment inhibited cell ruffling stimulated by LPA. These findings provided evidence for a link between PAK1-dependent cell ruffling and contraction stimulated by LPA, which was distinct from PDGF. The actin dynamics required for fibroblast ruffling can be controlled at the level of actin-depolymerizing factor (ADF)/cofilin by the PAK1 effector LIM kinase (; ; ; ). shows that with fibroblasts in 3D collagen matrices, PDGF and LPA stimulated cofilin1 phosphorylation and that stimulation was inhibited in PAK1-silenced cells. Contraction experiments were performed after knocking down cofilin1 using siRNA, which, as shown in , could be reduced by >70%. demonstrates that in cofilin1-silenced cells, collagen matrix contraction was inhibited. Also, cells ceased their ruffling activity (unpublished data). Therefore, it could be concluded that cofilin1 was a downstream effector for PAK1 in LPA- and PDGF-stimulated collagen matrix contraction. Together, the experiments identified PAK1 as a downstream convergence point for the regulation of both cell ruffling and collagen matrix contraction stimulated by PDGF and LPA. As already mentioned, PDGF- and LPA-dependent floating collagen matrix contraction requires activity of the small G protein Rho (), but only PDGF-stimulated contraction was dependent on the Rho effector Rho kinase (; ). Consistent with this observation, shows that blocking Rho kinase (Y) inhibited PDGF-dependent, but not LPA-dependent, contraction of floating collagen matrices. Moreover, blocking Rho kinase did not cause a decrease in cell ruffling (unpublished data). Therefore, rather than functioning in the same signaling pathway as PAK1, it seemed likely that PAK1 and Rho acted in parallel cooperative fashion. Along with Rho kinase, mDia1 has been implicated as a Rho effector for regulation of actin cytoskeletal dynamics and force generation (; ). Therefore, we tested the possibility that mDia1 might be the Rho effector required for LPA-stimulated matrix contraction. This was accomplished by knocking down mDia1expression. shows that levels of mDia1 in mDia1-specific siRNA, but not in mock-transfected cells, were reduced by >95%. mDia1-silenced cells were able to spread on collagen-coated coverslips and form vinculin-containing focal adhesions, although the cells were rounder and had reduced actin stress fibers compared with controls (unpublished data), as has been reported (). shows the results of collagen matrix contraction studies with mDia1-silenced and mock-transfected cells. Silencing mDia1 selectively inhibited LPA-stimulated contraction without affecting PDGF-stimulated contraction. However, knocking down mDia1 had no effect on cell ruffling (unpublished data). Therefore, different Rho effectors cooperated with PAK1 in fibroblast contraction of floating collagen matrices depending on the agonist used to stimulate contraction, which was Rho kinase in the case of PDGF and mDia1 in the case of LPA. Control experiments were also performed to confirm activation of mDia1 by LPA. This was accomplished by taking advantage of the observation that LPA-stimulated formation of stable (nocodazole-resistant) microtubules depends on mDia1 (; ). Serum-starved mDia1-silenced and mock-transfected cells were agonist stimulated and tested for the development of nocodazole- resistant microtubules (; ). shows that after agonist stimulation by PDGF or LPA, a subpopulation of microtubules in mock-transfected cells became nocodazole resistant, but that in mDia1-silenced cells LPA was no longer able to stimulate formation of stable microtubules. summarizes the results of our studies. We found that PDGF and LPA regulate floating collagen matrix contraction through signaling pathways that converge on PAK1 and its downstream effector, cofilin, and that contraction depends on cellular ruffling activity. Moreover, different Rho effectors were observed to cooperate with PAK1 in regulating contraction, Rho kinase in the case of PDGF and mDia1 in the case of LPA. Previous work (for review see ) had demonstrated that the physiological agonists LPA and PDGF stimulated fibroblast contraction of floating collagen matrices, but the relationship between the signaling mechanisms involved had not been elucidated. These agonists have opposite effects on the overall movement of fibroblast dendritic extensions: retraction in response to LPA and protrusion in response to PDGF (). Moreover, blocking Rho kinase or myosin II activity was shown to inhibit PDGF-dependent, but not LPA-dependent, matrix contraction (; ). PAK1 is a downstream effector for both PDGF- (; ) and LPA-mediated signaling (; ). Therefore, we analyzed the possible role of PAK1 in matrix contraction, using siRNA to silence PAK1 expression, and learned that knocking down PAK1 resulted in inhibition of contraction stimulated by either agonist. Microinjection and expression studies had implicated PAK1 in ruffling and motility of cells on 2D surfaces (; , , ). Although PDGF and LPA have opposite effects on the overall movement of fibroblast dendritic extensions, both agonists were found to stimulate membrane ruffling activity. However, ruffling by cells in 3D matrices was inhibited in PAK1-silenced cells. In addition, blocking PI3 kinase selectively inhibited PDGF-stimulated cell ruffling and matrix contraction, whereas blocking Gα selectively inhibited LPA-stimulated ruffling and contraction. As previously stated, PAK1-regulation of the actin dynamics required for fibroblast ruffling can be controlled at the level of ADF/cofilin phosphorylation by the PAK1 effector LIM kinase (; ; ; ). We found that LPA and PDGF both stimulated cofilin1 phosphorylation, and that stimulation was blocked in PAK1-silenced cells. Moreover, silencing cofilin1 with siRNA blocked LPA- and PDGF-dependent matrix contraction, as well as membrane ruffling. Therefore, we suggest that matrix contraction requires cellular ruffling activity stimulated by PAK1 and cofilin1. Although LIM kinase is the likely intermediate between PAK1 and cofilin1, we have not succeeded in developing conditions using siRNA to effectively silence LIM kinase expression and directly test its role. On collagen-coated coverslips, PAK1-silenced fibroblasts showed markedly decreased migration, but cells spread normally and increased formation of lamellipodia after PDGF stimulation. That PAK1-silenced cells formed normal lamellipodia in 2D culture, but lacked ruffles in 3D matrices, suggests that lamellipodia and ruffles may be regulated independently; PAK-independent mechanisms of cell ruffling and lamellipodia formation have been previously described (; ). Differences in fibroblast adhesion to 3D matrices versus 2D coverslips may also be important. Besides LIM kinase, regulation of ADF/cofilin and fibroblast ruffling can be controlled by integrin interactions and testicular protein kinase 1 (; ). Cells interacting with 3D matrices have fewer stress fibers, smaller focal adhesions, and decreased activation of focal adhesion kinase compared with cells in 2D culture (; ; Wozniak et al., 2004). It is possible, therefore, that in 3D matrices fibroblasts become completely dependent on the PAK1 pathway, whereas human fibroblasts on 2D collagen-coated surfaces can regulate cell ruffling and lamellipodia formation by multiple mechanisms. Also noteworthy is the discovery that the dendritic extensions of PAK1-silenced fibroblasts in 3D matrices showed decreased expansion in response to PDGF. This decrease may have resulted from an increase in microtubule catastrophe in the absence of PAK1 (, ) because microtubules were shown to be required for formation of the fibroblast dendritic network (). Based on molecular architecture, PAKs can be categorized into two subgroups with three members each (; ; ). The group I PAKs, including PAK1, share a high degree of amino acid homology and influence diverse cellular processes, such as cellular morphology, migration, and gene regulation. Our results identify PAK1 as the major isoform involved in regulation of fibroblast ruffling in 3D collagen matrices and collagen matrix contraction, although expression of PAK2 is at least five times higher than PAK1 in human fibroblast based on Western blotting results. Structurally, the major differences between PAK1 and -2 are in the NH-terminal regulatory domain in which PAK1 has five proline-rich Src homology 3 (SH3)–binding motifs, compared with two SH motifs in PAK2 (). It has been reported that phosphorylation of the threonine 212 residue in one of PAK1's unique SH3 motifs is important for regulation of neuronal growth cone dynamics (). Whether phosphorylation at this site is also important in regulation of fibroblast contraction of collagen matrices has yet to be determined. Although shows PDGF and LPA converging separately on PAK1, Rac is likely to be immediately upstream of PAK1. Preliminary studies showed that dominant-negative expression of N17Rac1, but not of N17Cdc42, completely abolished the formation of dendritic extensions (unpublished data), which was a phenotype more pronounced than observed by knocking down PAK1. Given the potential for indirect effects by overexpression of dominant-negative Ras family mutants (), coupled with the difficulty of transfecting genes into early passage human fibroblasts, our initial results cannot yet be clearly interpreted. During collagen matrix contraction, individual collagen fibrils are translocated toward the cell surface (; ; ). Such translocation of collagen fibrils, when it involves fibroblasts on 2D coverslips, requires a mechanism of contractile force generation (). In the case of cell migration, PAK1-dependent cell ruffling is usually thought to be important for cells to reach forward (; ), whereas the small G protein Rho has been implicated in generation of contractile force required for cell translocation (; ). This study builds on our previous finding that Rho was required for floating matrix contraction () to show that different Rho effectors are involved in contraction dependent on the agonist (i.e., Rho kinase in the case of PDGF and mDia1 in the case of LPA). Because blocking Rho kinase or silencing mDia1 using siRNA inhibited contraction selectively, and neither treatment blocked cell ruffling, we propose that the Rho effectors act in parallel to and cooperatively with the PAK1 signaling pathway and play a role in the force generation required for matrix contraction. At this time we can only speculate as to why LPA requires mDia1 rather than Rho kinase for floating matrix contraction. Certainly, LPA activates Rho kinase in these cells because blocking Rho kinase has been shown to inhibit retraction of dendritic extensions without preventing matrix contraction (; ). Rho and Rho kinase have been implicated in myosin II–dependent force generation (; ). The role of mDia1 in force generation is less clear, however (; ). Significantly, mDia1 has been implicated in the stabilization of both microtubules () and actin filaments (), and microtubule dynamics (; ) as well as microfilament depolymerization () have the potential to generate force independently of actomyosin. DME and trypsin/EDTA solution were obtained from Invitrogen. BSA (fatty acid free) and LPA were obtained from Sigma-Aldrich. Vitrogen 100 collagen was obtained from Cohesion Technologies, Inc. PDGF (BB isotype) was obtained from Upstate Biotechnology. , pertussis toxin, and were obtained from Calbiochem-Novabiochem. Total PAK- (C-19), PAK1- (N-20), cofilin1-, phospho-cofilin1– (mSer 3), and mDia1-specific antibodies were obtained from Santa Cruz Biotechnology, Inc. rhodamine-B-isothiocyanate–conjugated phalloidin, oligofectamine, and Opti-MEM I were obtained from Invitrogen. Human foreskins were obtained from anonymous donors and provided by the University of Texas Southwestern Medical Center. Fibroblasts from human foreskin specimens (<10th passage) were maintained in 75-cm tissue culture flasks (Falcon) in DME supplemented with 10% FBS. Fibroblasts were harvested from monolayer culture with 0.25% trypsin/EDTA for 4 min at 37°C, followed by 10% FBS in DME. All incubations with cells were performed at 37°C in a humidified incubator with 5% CO. For experiments with collagen matrices, cells in neutralized solutions of 1.5 mg/ml of collagen were prewarmed to 37°C for 3–4 min, and 0.2-ml aliquots were placed in 24-well culture plates (Corning). Unless otherwise specified, cell density was 2 × 10 cells/matrix for matrix contraction and immunoblotting experiments and 2 × 10 cells/matrix for observing cell morphology. Each aliquot of collagen matrix occupied an area outlined by a 12-mm diam circular score within a well. After polymerization for 60 min, matrices were gently released from the underlying culture dishes with a spatula and allowed to float in 0.5 ml of basal medium (DME containing 5 ml/ml BSA). Growth factors and inhibitors were added at the times indicated in the figure legend. For experiments with collagen-coated surfaces, harvested cells (2 × 10) were incubated for the times indicated in the figure legend on 12-mm glass coverslips. The coverslips were coated for 20 min with 50 μg/ml collagen and then rinsed with Dulbecco's PBS (1 mM CaCl, 0.5 mM MgCl, 150 mM NaCl, 3 mM KCl, 1 mM KHPO, and 6 mM NaHPO, pH 7.2). Subsequently, the cultures were incubated with 1 ml DME containing 5 mg/ml BSA and growth factors or inhibitors as indicated in the figure legend. To knock down PAK1, cofilin1, and mDia1 expression in human fibroblasts, the following primer pairs for siRNA were designed and obtained from the University of Texas Southwestern Medical Center siRNA core facility. PAK1 siRNA: 5′-AACACACAAUUCAUGUCGGTT-3′ and 5′-AACCGACAUGAAUUGUGUGTT-3′. Cofilin1 siRNA: 5′-GCGGUGCUCUUCUGCCUGAUU-3′and 5′-UCAGGCAGAAGAGCACCGCUU-3′. mDia1 siRNA: 5′-AUUCUUCUGCAUCAUAUGGTT-3′ and 5′-CCAUAUGAUGCAGAAGAAUTT-3′. For annealing, 20 μM of each single-strand 21-nt RNA was incubated in annealing buffer (100 mM potassium acetate, 30 mM Hepes-KOH, pH 7.4, and 2 mM magnesium acetate) for 2 min at 95°C, followed by 2 h at 37°C. To accomplish high efficiency transfection, fibroblast cultures (60–70% confluent) were rinsed with antibiotic-free DME and treated with trypsin-EDTA for 1 min to elicit cell rounding, but not detachment. Subsequently, antibiotic-free 10% FBS/DME was added at a ratio of 4:1 to quench the trypsinization. After cells were rinsed with antibiotic-free DME, they were incubated with Opti-MEM I containing 700 nM siRNA- (PAK1 and mDia1) or 500nM siRNA-(cofilin1) annealed oligonucleotides. After 12 (PAK1 and mDia1) or 36 h (cofilin1), the transfection medium was removed and replaced with 10% FBS/DME containing antibiotics for an additional 24 h, at which time cells were subcultured. Mock-transfected cells were treated with only the sense direction oligonucleotide at double the concentration. To measure 2D migration, mock- and PAK1-silenced fibroblasts were incubated overnight on collagen-coated coverslips with DME containing 5 mg/ml BSA and 0.1% FBS. The cell cultures were scrape wounded with a pipette tip and then incubated in DME containing 5 mg/ml BSA and 50 ng/ml PDGF. For immunostaining, collagen matrix samples were fixed for 10 min with 3% paraformaldehyde in PBS (3 mM KCl, 1 mM KHPO, 150 mM NaCl, and 6 mM NaHPO, pH 7.2.) at room temperature, blocked with 2% BSA/1% glycine in PBS for 30 min, and permeabilized for 15 min with 0.5% Triton X-100 in PBS. Samples were then incubated for 1 h at 37°C with mouse anti–β-tubulin (1:100 dilution in 1% BSA/PBS) followed by 45 min at 37°C with FITC-conjugated goat anti–mouse IgG. For actin staining, samples were incubated with Alexa Fluor 594–conjugated phalloidin (1:200 dilution in 1% BSA/PBS) for 30 min at 37°C. Samples were mounted on glass slides with Fluoromount G. (Southern Biotechnology Associates, Inc.) Images were collected using a fluorescent microscope (Eclipse 400; Nikon) using 10×/0.45, 20×/0.75, and 40×/0.75 infinity corrected objectives (Plan Apo; Nikon). Images were collected at room temperature using a camera (SenSys; Photometrics) and MetaView acquisition software (Universal Imaging Corp.). Subsequent image processing was performed using Photoshop 5.5 or 7.0 (Adobe) in accordance with the image acquisition and manipulation instructions.

Amputation or tissue removal can lead to the regeneration of lost structures in some vertebrate species, such as the salamanders (e.g., the newt and the axolotl; ; ; ). For example, adult newts can rebuild entire limbs, tails, and jaws through an epimorphic regeneration process that leads to the restoration of complete and functional tissue architecture (). Epimorphic limb regeneration proceeds by rapid wound closure and is critically dependent on the formation of a multipotent mesenchymal growth zone, the blastema, which gives rise to the newly formed limb (). Data show that mature tissues in the stump (e.g., bone, cartilage, and skeletal muscle) respond to amputation by disorganization, histolysis, and increased cellular proliferation. This process is generally referred to as the dedifferentiation step leading to the formation of blastema progenitors (). However, the resolution of our picture on the contributing tissues at the cellular level is low at present. It is unclear to what extent differentiated cells reverse mature phenotypes and to what extent undifferentiated cells, such as stem cells, residing within differentiated tissues become activated, followed by their incorporation into the blastema. The lack of molecular markers has also obstructed the prospective isolation of blastema progenitors. Skeletal muscle is an important contributor to blastema formation (). The skeletal muscle fiber is a syncytial (multinucleate) cell type, whose differentiation during embryonic development is characterized by the cellular fusion of somite-derived precursors (; ). An intriguing aspect of the regenerating salamander appendages is the reversal of differentiation. Both static analyses and dynamic in vivo tracing showed that skeletal muscle fibers break up, the syncytium becomes fragmented as a response to limb or tail removal, and muscle-derived mononucleate progeny significantly contribute to the blastema (; , ; ; ). Isolated salamander myotubes can also undergo a cellularization process by which the syncytium turns into mononucleate progeny after reimplantation into the regenerating limb (; ). Although adult mammals do not form a blastema after limb amputation, their skeletal muscle tissue regenerates after injury (). However, mammalian skeletal muscle regeneration does not involve cellularization of the syncytium. Instead, a stem cell population called satellite cells, which express markers such as Pax7, M-cadherin, and Myf5, reenters the cell cycle, proliferates, and incorporates into nascent or into preexisting myofibers during mammalian muscle regeneration (; ). Mammalian satellite cells reside between the basal lamina and the sarcolemma of the myofiber (). Earlier studies identified a cell population that is closely apposed to the myofiber in the adult newt limb as well. But in contrast to mammals, these cells were shown to be completely encapsulated by a basement membrane (; ), and it has remained unsettled whether adult newts possess a cellular population that is equal to mammalian satellite cells. In addition, it has not been established whether dedifferentiation of skeletal muscle leads to the activation of a stem cell population within the tissue and if such cells could contribute to the new limb. To start addressing these questions we combined histological analyses and in vitro culture of single newt myofibers, along with implantation and tracing of labeled myofiber-derived cells. We find that the salamander myofiber contains a satellite cell population. As we can distinguish between the process of cellularization of the syncytial myofiber on one hand and satellite cell activation on the other, the quantitative aspects of these two separate events can be examined. Satellite cell activation prevails in our model of skeletal muscle plasticity, leading to the production of a multipotent progeny population. Therefore, the data highlight the possibility of promoting blastema formation by the activation of cellular and molecular programs that also operate in mammals. To test whether newt skeletal muscle in the limb contains a satellite cell population, we used a monoclonal antibody against Pax7, which is a specific marker of skeletal muscle satellite cells. As shown in , similar to mammalian muscle, Pax7 cells are present in newt limb skeletal muscle. However, a basement membrane surrounds the Pax7 cells (). No Pax7 cells were detected outside the skeletal muscle tissue (unpublished data). To test whether Pax7 cells reenter the cell cycle after limb amputation, we immunostained limb sections with an antibody raised against phosphorylated histone 3 (H3P), which marks mitotic cells (). We found that Pax7 cells are largely quiescent in the uninjured limb, but become mitotic after limb removal (; and ). We saw Pax7 cells outside of skeletal muscle tissue 4 d after amputation, and detected Pax7 cells within the blastema upon formation (). These data show that quiescent satellite cell activation is a response to limb removal and the findings suggest that satellite cells leave their niche to incorporate into the blastema. To understand the cellular basis of the plasticity of skeletal muscle fibers, we established an ex vivo culture of living, intact single newt myofibers. We isolated and plated single myofibers that were viable and displayed characteristic morphology, such as Z band striation marking the boundaries of the sarcomeres (). This technique has previously been used to establish single myofiber culture from both mammalian and salamander species with no contamination from other tissues or cell types (; ). As indicated by the presence of Pax7 () and M-cadherin cells (), muscle fibers from the newt limb could be copurified with a satellite cell population after isolation and plating. Similar to the in vivo analyses, we found an additional basement membrane between the myofiber itself and the satellite cells, as indicated by the Pax7–collagen type IV double immunostaining (). Thus, newt single myofibers can be isolated containing the myofiber proper, along with the tightly associated satellite cells. The model in shows the location of newt satellite cells compared with their mammalian counterparts. We observed that on average 7.7% of the 2,167 nuclei in a representative sample of 55 single fibers were found to be in satellite cells, and 9 myofibers were devoid of satellite cells. Three-dimensional confocal microscopic analyses showed the complete absence of cells outside of the basal lamina, indicating that the myofiber cultures did not contain contaminating cells from elsewhere. All satellite cells were encased by basement membrane directly after attachment, and 99% of the cells in satellite cell positions were Pax7. On average, after 7 d in culture the myofibers started to produce proliferating progeny cells. shows a myofiber directly after attachment and with proliferating progeny after ∼15 d in culture. Video 1 (available at ) illustrates the budding of single cells from the myofiber, and shows the single frame sequence of one budding event taken from the time-lapse movie capture in Video 1. Budding of cells continued until the myofiber hypercontracted and detached from the substrate. Myofiber-derived cells migrated onto the surrounding substrate and proliferated. At this stage it was unclear whether the proliferating progeny cells were derived by cellularization of the myofiber itself and/or by activation of quiescent satellite cells. To distinguish between these two events, we injected a fluorescein-conjugated nuclear-localizing dextran (NLS-dextran) into the myofibers directly after the attachment of the myofiber to the substrate (). This lineage tracer cannot be transferred between cells and, therefore, should only label myonuclei. In agreement with this, none of the Pax7 satellite cells were labeled with NLS-dextran. Conversely, none of the NLS-dextran–labeled myonuclei were Pax7 (). Out of the 70 single myofibers that we observed, we were only able to detect two mononucleate cells at one occasion that appeared to contain NLS-dextran (), and these two cells did not proliferate. All other proliferating cells were NLS-dextran negative. Because the fluorescent NLS-dextran signal was easily detectable in all of the myonuclei and we analyzed the myofiber-derived progeny at 12-h intervals, we can exclude the possibility that the NLS-dextran signal was diluted because of rapid proliferation. These data show that satellite cell activation, rather than cellularization of the syncytium, resulted in a proliferating cell progeny population in our culture system. These proliferating satellite cells retained Pax7 expression and were also positive for MyoD for several generations (). In accordance with earlier observations on mammalian myofiber cultures (), Pax7 expression became heterogeneous in prolonged newt satellite cell progeny cultures (unpublished data). Because the blastema is a multipotent tissue, we tested whether newt satellite cells were able to adopt anything other than myogenic fates. When cultured in myogenic medium, satellite cell progeny readily formed myotubes, which expressed M-cadherin and myosin heavy chain (). Western blot analyses confirmed the up-regulation of myosin heavy chain and M-cadherin during myogenesis, which was concomitant with the increased number of myotubes and the decreased number of myoblasts in the culture (). Simultaneously, Pax7 levels dropped in the protein extracts (). When cells were exposed to adipogenic media, we detected that at least 30% of the cells contained lipid droplets and displayed adipocyte morphology. In contrast, the few myotubes that were visible in the adipogenic media did not contain lipid droplets (). Cells that were not cultured in adipogenic media were negative for Oil Red staining (). When satellite cell progeny were cultured in osteogenic media, we saw that 10% of the cells produced alkaline phosphatase–positive foci () and that the cells produced calcium deposits stained by Alizarin red (). In addition, clonal analysis also indicated that the progeny are multipotent, displaying myogenic (not depicted), adipogenic, and osteogenic potential (Fig. S2, available at ). These data show that skeletal muscle satellite cell progeny can adopt nonmyogenic fates and indicate that satellite cells could represent a multipotent blastema progenitor population. To test whether satellite cells are able to contribute to newly formed limb tissues, we injected labeled satellite cell progeny intramuscularly before amputation. Satellite cell progeny were labeled with BrdU before injection, during their in vitro expansion. As a control, we injected the contralateral limbs with PBS before amputation at the same axial level. As blastema formation and regeneration occurred we saw that a large number of the injected, BrdU-labeled cells appeared in clusters within the blastema at all analyzed stages of the regeneration. At the medium bud stage, BrdU-labeled cells were found within both the blastema () and, strikingly, the epidermis (). BrdU-labeled cells were not detected in the contralateral regenerate, which was injected with PBS before amputation (Fig. S1, available at ). There was no difference in the speed and morphology of regeneration between cell- and PBS-injected limbs. BrdU-labeled cells were also clearly visible in the late bud stage regenerate, although the intensity of the BrdU label varied more, compared with the medium bud stage regenerate (). The contralateral PBS-injected regenerate was also devoid of BrdU-labeled cells at this stage (Fig. S1). All four injected limbs developed cartilage at this stage, and BrdU-labeled cells were detected within newly formed cartilage tissue in all four cases (). These results show that implanted satellite cell progeny can give rise to new tissues during limb regeneration and indicate that metaplasia may occur during salamander limb regeneration. The ability to form a regeneration blastema, which leads to the epimorphic regeneration of complex body structures, is restricted to some amphibians and fish among vertebrates (). A conundrum of regenerative biology is why mammals, with a few exceptions, do not form a blastema or a blastema-like structure despite the fact that they can functionally repair some tissues, such as skeletal muscle () and liver (). Of particular interest is whether the generation of progenitor cells during epimorphic regeneration in salamander and during mammalian tissue repair proceeds by the activation of different or overlapping mechanisms. A unique feature of blastema formation in salamanders is the process of dedifferentiation of stump tissues that follows appendage removal. The possibility to induce blastema formation and regeneration in mammals through the activation of a comparable dedifferentiation program has been proposed (; ; ). This is especially valid for skeletal muscle tissue because dedifferentiating skeletal muscle is a significant source of blastema progenitors. Although the potential role of stem cells in blastema formation has been suggested (; ; ), no such cells have been previously identified in the newt limb. Hence, it is still not clear whether the term dedifferentiation solely refers to the reversal of the differentiated state of mature cells, to the activation of stem cells in the disorganizing tissues, or to a combination of these two definitions. If both processes coexist, the quantitative aspects of their relative contribution in vivo remain to be elucidated. Our data clearly show that satellite cells, which are comparable to mammalian skeletal muscle stem cells, exist in newt skeletal muscle as well. First, we found that newt satellite cells or their progeny express molecular markers, such as Pax7, M-cadherin, and MyoD, all of which are expressed by mammalian satellite cells or their progeny as well (). Second, when we isolated single myofibers a satellite cell population was copurified, despite the presence of an additional basal lamina between the satellite cell and sarcolemma. Third, similar to the mammalian myofiber cultures, we observed that satellite cell activation occurred that was characterized by cell cycle reentry and proliferation of the satellite cell progeny population. Finally, we showed that the satellite cell progeny population in newts is multipotent, which has also been observed in mammals (; ; ). Thus, the results indicate that newts do not represent an exception in the vertebrate phyla, and like other amphibians (; ) and mammals they also contain Pax7 stem cells in their skeletal muscle tissue. However, the additional basement membrane that separates newt satellite cells from the sarcolemma may reflect that newt satellite cells are in some respect evolutionary intermediates between interstitial stem cells and satellite cells, which were found to be separate populations in mammals (; ). Identification of further stem cell populations in newt skeletal muscle, along with functional studies, could address this issue. The satellite cell progeny population was able to adopt nonmyogenic fates in vitro and they incorporated into the regeneration blastema after intramuscular injection before amputation. We also noted a contribution to the epidermis and detected satellite cell progeny within newly formed cartilage tissue. The observed multipotentiality of satellite cell progeny does not directly address the question of whether activated satellite cells adopt divergent fates without in vitro expansion. However, the onset of tissue-specific molecular differentiation programs and the large number of satellite cell progeny within various tissues, which did not alter the speed and mode of regeneration, suggest that the integrated satellite cell progeny are functional. Furthermore, lineage shifting across germ layer boundaries has been shown to occur during salamander tail regeneration (). Clearly, additional experiments are required to assess the plasticity of satellite cells in vivo and to establish whether metaplasia characterizes salamander limb regeneration. Nevertheless, in light of the available observations, a plausible hypothesis is that skeletal muscle dedifferentiation results in a significant contribution by satellite cells to the blastema and to the regenerate. Pax7 cells are also found in the blastema of the regenerating axolotl tail () and tail regeneration in the tadpole also involves satellite cell activation (). These observations further suggest an important role of satellite cells in the regeneration of missing body parts in vertebrates. In a study similar to our own, showed that limb myofibers isolated from axolotl larvae undergo cellularization and fragmentation. The authors noted that only 3.5% of the myofibers contained the satellite type of cells and that these were not observed in their skeletal muscle fiber plasticity model. We saw that 86% of the isolated myofibers contained satellite cells and that only satellite cell progeny proliferated in our culture system, although we could not detect any sign of proliferating progeny that could have been derived by cellularization of the myofiber. At present, it is unclear whether the discrepancies between our observations and the model presented by reflect phylogenetic or ontogenetic differences, or are caused by dissimilarities in the experimental paradigms. However, both studies underpin the necessity to further assess the quantitative aspects and functional relevance of satellite cell activation that leads to multipotent progeny on one hand and cellularization and/or fragmentation of the syncytium on the other during limb regeneration. Our results show that epimorphic limb regeneration activates such programs, which lead to regeneration of muscle tissue in mammals after injury. Mammalian skeletal muscle responds to various challenges, such as stretching or mechanical damage, by activating a proliferation program in satellite cells that is followed by differentiation and fusion into myotubes and into myofibers. In this context, it is interesting to note the study by , which showed that amputation as such was not sufficient to produce blastema progenitors. Instead, a mechanical stimulus (minor clipping of the muscle fiber) was required for the generation of progeny from dedifferentiating axolotl tail muscle in vivo (). The exact identity of signals that link tissue injury to blastema formation needs to be elucidated, as it may reveal key aspects of blastema formation involving both myofiber fragmentation and concomitant stem cell activation. Formation of a blastema-like structure, although a rare event, is possible in mammals, as exemplified by the healing capacity of MRL mice and by the seasonal regeneration of deer antlers (; ). The question is how blastema formation is induced in mammals and how it can be promoted. We propose skeletal muscle satellite cells as a potential target in the promotion of mammalian blastema formation. The following primary antibodies were used: mouse monoclonal anti-Pax7 IgG (Developmental Studies Hybridoma Bank), mouse monoclonal anti–myosin heavy chain IgG (MF20; Developmental Studies Hybridoma Bank), mouse monoclonal anti–M-cadherin IgG (Clone 1B11 used for immunofluorescence; nanotools GmbH), rabbit polyclonal anti–collagen type IV antibody (Rockland Immunochemicals, Inc.), rabbit polyclonal anti–M-cadherin antibody (used for immunoblotting; Invitrogen), rabbit polyclonal anti-H3P antibody (Upstate Biotechnology), rat monoclonal anti-BrdU IgG (Trichem ApS), rabbit polyclonal anti-MyoD antibody (Santa Cruz Biotechnology, Inc.), anti-WE3 monoclonal IgG (Developmental Studies Hybridoma Bank), mouse monoclonal anti–collagen type II IgG (CHEMICON International, Inc.). For immunofluorescence studies, primary antibodies were detected with appropriate species-specific Alexa Fluor–conjugated secondary antibodies (Invitrogen). All experiments were performed according to European Community and local ethics committee guidelines. Adult red-spotted newts, , were supplied by Charles D. Sullivan Co., Inc. and maintained in a humidified room at 15–20°C. Animals were anesthetized by placing them in an aqueous solution of 0.1% ethyl 3-aminobenzoate methanesulfonate salt (Sigma-Aldrich) for 15 min. Forelimbs were amputated by cutting just proximal to the elbow or wrist, and the soft tissue was pushed up to expose the bone. The bone and soft tissue were trimmed to produce a flat amputation surface. Animals were left to recover overnight in an aqueous solution of 0.5% sulfamerazine (Sigma-Aldrich) before being placed back into a 25°C water environment. At specified time-points, the regenerating limbs were collected after anesthetization. Tissue samples were mounted on cork using Gum Tragacanth (Sigma-Aldrich), snap frozen in isopentane (VWR), and cooled to freezing point in liquid nitrogen. 5-μm-thin frozen sections were thawed at room temperature and immediately fixed in acetone/methanol (1:1) for 5 min at –20°C. Sections were blocked with 20% normal goat serum (DakoCytomation) diluted in PBS for 30 min at room temperature. Sections were incubated with a relevant primary antibody overnight and with secondary antibodies for 1 h at room temperature. Antibodies were diluted in blocking buffer and sections were mounted in mounting medium (DakoCytomation) containing 100 ng/ml DAPI (Sigma-Aldrich). Newts were anesthetized and decapitated. The skin was removed from the underside of the forelimbs, exposing the musculature. Excess fat and connective tissue was carefully removed from around the musculature. A group of muscles located between the elbow and wrist were isolated with forceps and carefully dissected away from the bone, handling only the tip of the muscle to prevent damage. Digestion with type I collagenase (Sigma-Aldrich) solution (0.2% wt/vol in DME; Invitrogen) supplemented with 1% Glutamax (Invitrogen) and 1% penicillin/streptomycin (Invitrogen) was performed in a water bath at 25°C for 3–4 h. All media used in this and subsequent cell cultures were diluted 24% with distilled water. After digestion, myofibers were disaggregated as previously described (). Single myofibers were placed in 35-mm Falcon culture dishes (BD Biosciences) coated with 1 mg/ml Matrigel (BD Biosciences) in DME supplemented with 13% FCS (Invitrogen), 1% Glutamax, 1% penicillin/streptomycin, and 1% insulin (Sigma-Aldrich) and cultured at 25°C. Myofiber cultures were fixed in 2% PFA at various time points and processed for immunofluorescence studies. The protocols for immunofluorescent staining of cells and newt single myofibers were followed as previously described (), with the exception that cells and myofibers were fixed with 2% PFA. A synthetic polypeptide containing the NLS of the polyomavirus large T antigen, CGYGVSRKRPRPGC, was synthesized by Thermo Electron Corporation. The peptide was covalently linked to fluorescein-conjugated dextran (70 kD; Invitrogen) via the COOH-terminal cysteine residue, using the heterobifunctional cross-linker sulfo-SMCC (Pierce Chemical Co.) as described previously (; ). Myofibers were injected with NLS-conjugated fluorescein-dextran directly after their attachment, using a Femtojet in combination with an Injectman (Eppendorf AG). Myofiber cultures were analyzed using both brightfield and fluorescence microscopy at 12-h intervals before fixation or passaging of the myofiber-derived cells. For myogenic differentiation, satellite cell progeny were grown to 90–100% confluency and incubated in DME supplemented with 0.5% horse serum (Invitrogen), 1% Glutamax, 1% penicillin/streptomycin, and 1% insulin. After 3 and 6 d in differentiation medium, cells were fixed with 2% PFA and processed for immunofluorescence studies. For immunoblotting, cells were lysed with RIPA buffer supplemented with a protease inhibitor cocktail (Roche). 2 μg of each cell lysate was separated on a 10% PAGE gel and transferred to nitrocellulose membrane. The membrane was blocked with 5% dry milk fat and 0.1% Tween 20 (Sigma-Aldrich) in PBS and subsequently probed with primary antibodies. Primary antibodies were recognized with species-specific streptavidin-conjugated secondary antibodies (GE Healthcare). Membranes were developed using an ECL detection kit (GE Healthcare). For adipogenic and osteogenic differentiation, cells were grown to 90–100% confluency and incubated in adipogenic and osteogenic media as described previously (). Cells in adipogenic medium were stained with Oil red (). Cells in osteogenic medium were stained with Alizarin red (), and alkaline phosphatase was detected using kit 85 (Sigma-Aldrich) according to the manufacturer's instructions. For clonal analyses, cells were cultured at a density of 0.5–1.0%, so that single cells were clearly discernible. Single cells were isolated with cloning cylinders (Sigma-Aldrich) and incubated for 30 s in trypsin-EDTA (0.05% trypsin and 0.53 mM EDTA; Invitrogen) at room temperature. Trypsinized single cells were transferred to one well of a 24-well culture plate that contained a 1:1 ratio of normal and conditioned proliferation media (13% FCS, 1% Glutamax, 1% insulin, and 1% penicillin/streptomycin). Proliferating clonal cells were maintained at a confluency of no more than 60% to avoid spontaneous differentiation before being subjected to differentiation studies. Satellite cell progeny were grown in the presence of 10 μM BrdU for 6 d before injection. 20,000 cells were suspended in 4 μl PBS diluted with 24% water. Animals were anesthetized and cells were injected using a Hamilton syringe intramuscularly in the upper forelimb halfway between the elbow and shoulder. 30 min after injection the limb was removed just above the elbow as described in Animals and procedures. Contralateral limbs were injected with PBS to serve as control. The regenerates were harvested at different time points and processed for immunohistochemistry. An LSM 510 Meta laser microscope with LSM 5 Image Browser software (both Carl Zeiss MicroImaging, Inc.) was used for confocal analyses. A microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) with Openlab 3.1.7 software (Improvision Ltd.) was used for brightfield and fluorescence microscopy analyses. Images were taken at room temperature and were further processed using Photoshop (Adobe) according to the JCB guidelines. Fig. S1 shows that the progeny of injected BrdU-labeled satellite cells are found in the regenerate, but not in the contralateral regenerate. Fig. S2 shows a multipotent satellite cell progeny clone. Video 1 shows the derivation of proliferating mononucleate cells from a 10–14-d-old newt myofiber in vitro. Online supplemental material is available at .

Postsynaptic scaffolds are thought to be crucial for receptor immobilization at both excitatory and inhibitory synapses (; ; ; ; ). Synaptic receptor–scaffold complexes interact with different cytoskeletal elements (; ; ). These interactions stabilize the complex and are thought to participate in the entry and exit of receptors and/or scaffold elements at postsynaptic sites (). At inhibitory synapses, gephyrin represents a core protein on the cytoplasmic side of the postsynaptic plasma membrane. Gephyrin harbors two oligomerization domains and is thought to generate a reversible postsynaptic scaffold for the immobilization of glycine receptors (GlyRs) and individual subtypes of γ-aminobutyric acid A receptors (GABARs; ; ). Structural analysis of both gephyrin's oligomerization and GlyRβ subunit binding sites revealed that dimeric gephyrin interacts with GlyRβ subunits and that subsequent multimerization is required to form a hexagonal gephyrin lattice (; ). Consequently, disassembly of the gephyrin scaffold must occur to enable dynamic changes at the postsynaptic specialization. Long distance intraneuronal transport of neurotransmitter receptors and associated proteins is typically mediated by microtubule-based motor complexes (). Notably, individual synapse-associated proteins, which locate at postsynaptic densities, are reported to act as adaptor proteins between neurotransmitter receptors and motor protein complexes (, ; ). For instance, the glutamate receptor–interacting protein 1 (GRIP1) functions as a transport adaptor that links intracellular α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptor GluR2 subunits to the kinesin superfamily motor KIF5 (). Moreover, GRIP1 binds to plasma membrane–inserted AMPA receptors at the postsynaptic specialization (; ), suggesting a dual role for receptor-associated proteins in transport and postsynaptic scaffold reactions. The use of a common set of proteins for both transport and plasma membrane anchoring may contribute to transport specificity and postsynaptic integration or to removal of receptors that underlie synapse formation and plasticity (). We investigated the dynamics of gephyrin and show that intracellular gephyrin forms a transport complex with inhibitory GlyR and the dynein motor. Our data suggest that this triple complex participates in receptor–scaffold dynamics at inhibitory synapses. To examine whether gephyrin particles are subjects of active transport in dendrites, time-lapse video microscopy was applied on cultured hippocampal neurons from different developmental stages expressing a previously described GFP–gephyrin fusion protein (). In mature neurons cultured for 12–14 d in vitro, gephyrin autofluorescent particles were recruited within dendrites in both the anterograde and retrograde direction ( and not depicted). Particle movement was observed in a discontinuous manner, with alternate mobility and immobility of particles over time. A quantitative evaluation of GFP–gephyrin transport packets revealed on average only 1.8 mobile particles per cell during image acquisition, a value that represents 2.2% of the total clusters. Control stainings with the synaptic marker SV2 () confirmed the maturity of the culture, as indicated by the high density of synaptic contacts (unpublished data). Notably, the number of mobile particles in our system is consistent with data published by , which show that 2% of gephyrin-positive immunogold particles locate in the neuronal cytoplasm, whereas the majority of gephyrin is associated with submembrane regions. GFP–gephyrin clusters were transported at mean velocities of 1.3 ± 0.1 μm/min, a value that resembles the transport characteristics of the postsynaptic density protein PSD-95 (; ). To test whether the recruitment of gephyrin particles represents active transport or diffusion processes, we applied time-lapse microscopy with image acquisition rates of 1 frame/s (unpublished data). As indicated by the absence of undirected movement, we concluded that active transport processes, but not diffusion, drive gephyrin particle recruitment in our system. Two gephyrin particle populations of different sizes were prominent within neurites. Notably, only the small fluorescent particles were mobile and frequently added to larger immobile particles ( and Video 1, available at ). In addition, small-sized particles left immobile clusters over time (). An evaluation of the relative size of GFP–gephyrin particles in the culture system revealed that mobile particles are on average ∼2.6-fold smaller compared with immobile particles (). Therefore, we hypothesized that mobile particles might represent gephyrin transport units participating in size increase and reduction of preexisting postsynaptic gephyrin scaffolds. To test this, we applied immunostaining against endogenous gephyrin and the vesicular inhibitory amino acid transporter (VIAAT), which is a marker labeling inhibitory presynaptic terminals. In this assay, large endogenous gephyrin aggregates exclusively colocalized with presynaptic boutons (), indicating that this particle population indeed represents gephyrin scaffolds at mature postsynaptic sites (). In contrast, no colocalization with synaptic marker was obtained for small-size particles (), which were frequently mobile in our system. Endogenous gephyrin clusters at nonsynaptic sites were on average 2.3-fold smaller than synaptic clusters (). For the simultaneous detection of both axon-terminal boutons and gephyrin in living neurons, we loaded GFP–gephyrin–expressing neurons with FM4-64 dye, which visualizes synaptic vesicle recycling and is therefore indicative of synaptic activity (). As seen in , gephyrin transport units emerged from active synaptic contacts (yellow) and merged with other FM4-64–positive terminal boutons over a time period of several minutes (). Notably, the mobile particle moved relatively fast between the individual synaptic contacts, but was delayed at active FM4-64–labeled presynaptic sites (Video 2). Together, these data suggest that intracellular transport processes recruit the gephyrin transport units underlying postsynaptic scaffold remodeling at inhibitory synapses. Because gephyrin is a direct binding partner of plasma membrane GlyRβ subunits (), we asked whether gephyrin and GlyR could be subjects of intracellular cotransport. Because 98% of neuronal gephyrin locates at plasma membrane regions (), we first examined whether gephyrin associates with intracellular vesicle fractions at all (). Sucrose gradient centrifugation on 160,000 vesicle-enriched pellets revealed that GlyR sedimentation peaked around 1.6 M, a molarity that is highly enriched with gephyrin immunoreactivity (). Consistent with an intracellular GlyR–gephyrin association, we also obtained in vitro binding of GlyR and gephyrin from cytoplasmic vesicle–rich fractions using either GlyR- or gephyrin-specific antibodies for precipitation (). Intracellular GlyR–gephyrin complexes should not colocalize with synaptic markers. To address whether one could identify small puncta of GlyR and gephyrin coimmunoreactivity that do not represent synaptic cluster formations, we immunostained cultured hippocampal neurons known to contain synaptic GlyR (; Fig. S1 A, available at ) against endogenous gephyrin, GlyR, and synaptophysin. Indeed, small GlyR–gephyrin–colocalized puncta, which likely represent en route transport particles, were frequently identified at nonsynaptic regions (), suggesting that these particles are the subjects of cotransport. A quantitative evaluation revealed ∼90% colocalization of GlyR and gephyrin clusters in this system, with 81% of these coclusters locating at synaptophysin-positive synapses (). This finding indicates that ∼20% of GlyR–gephyrin coclusters are nonsynaptic, a value that includes candidate particles in transit. To confirm cotransport of both particles in a living system, we expressed fluorescent fusion proteins in the cultured neurons monomeric red fluorescent protein (mRFP)–gephyrin and GFP–GlyRβ and performed double-channel time-lapse video microscopy. Both fusion proteins formed clusters in distal neurites, were recognized by GlyR- or gephyrin-specific antibodies, and displayed strong colocalization ( and Fig. S1 B). Indeed, mRFP–gephyrin transport units comigrated together with GFP–GlyRβ particles over time, indicating an intracellular transport complex of both proteins ( and Video 3). The intracellular recruitment and/or transport of individual synaptic proteins can be regulated in an activity-dependent manner. This includes the synaptic localization of profilin I and II (; ), which are both factors that interact with actin and gephyrin. Moreover, neuronal depolarization alters the phosphorylation state of MAP2 and, consequently, the stability of microtubules (), which represent tracks for intraneuronal transport. To test whether alterations of electrical parameters functionally influence GlyR–gephyrin cotransport, we depolarized cultures with physiological concentrations of KCl and/or applied the GlyR-specific antagonist strychnine during time-lapse analysis. Notably, the average velocity of GFP–gephyrin particle transport increased significantly either when cultures were depolarized or when GlyR function was blocked (). Also, the combined parameters (KCl + strychnine) resulted in an increase of average particle velocity, as compared with control conditions. In addition, strychnine-mediated GlyR blockade, but neither KCl-mediated depolarization nor bicuculline-mediated GABAR blockade, caused a significant increase in the average number of mobile GFP–gephyrin particles (). This observation was accompanied by a shift from anterograde to retrograde transport in the presence of strychnine (), suggesting that GlyR inactivity, but not KCl-mediated neuronal depolarization, causes retrograde GlyR–gephyrin removal from synaptic sites. In accordance with this finding are previous observations, in which chronic GlyR blockade through the antagonist strychnine over several days caused a recruitment of GlyR clusters toward neuronal somata in cultured neurons (; ). Together, these results provide functional evidence for an activity-dependent component regulating GlyR–gephyrin cotransport, as GlyR blockade not only alters GlyR recruitment but also the recruitment of comigrating gephyrin polypeptides. They further suggest a molecular motor system representing the driving force for neuronal recruitment of this complex. Gephyrin interacts with dynein light chains (DLCs; ); however, as these polypeptides represent components of both dynein and myosin motors, it could not be predicted whether gephyrin associates with one motor system or the other. Therefore, we performed coimmunoprecipitation with 400,000 pellets using antibodies specific for gephyrin and 74-kD dynein intermediate chains (DICs), the latter representing exclusive components of the dynein motor complex (). Endogenous gephyrin was specifically enriched in the precipitate, but was not detectable when unspecific immunoglobulins were used. In addition, the use of control IgG and two independent antibodies directed against DIC revealed specific coimmunoprecipitation of DIC with the gephyrin complex ( and not depicted), indicating that both gephyrin and DIC are components of the same complex. Triple labeling of endogenous proteins also confirmed dynein association and further revealed that sites of dynein–gephyrin or dynein–GlyR colocalization are occasionally found close to, but mainly not at, synaptic sites, suggesting that they represent molecules in transit (). In accordance with our quantitative evaluation in , this observation was obtained under experimental conditions with ∼80% of GlyR or gephyrin clusters in colocalization with either synaptophysin or the VIAAT marker, respectively. As a control for our time-lapse experiments, we also expressed GFP–gephyrin fusion proteins in cultured hippocampal neurons and analyzed colocalization with the dynein motor complex. Consistently, individual GFP–gephyrin particles are also found in colocalization with the 74-kD DIC in dendrites (Fig. S1 C). To test whether GlyRs, gephyrin, and dynein are associated in a triple complex formation, we performed immunoprecipitation using GlyR-specific antibodies. Immunodetection of the individual binding partners, as immobilized on the same membrane, revealed that GlyR was specifically precipitated and that both gephyrin and DIC, but not GluR2, were subjects of coprecipitation (). In accordance, triple immunodetection of GlyR, gephyrin, and dynein heavy chain (DHC) in cultured hippocampal neurons frequently displayed sites of triple colocalization () with ∼13.5% of GlyR and/or gephyrin puncta colocalizing with DHC, as compared with 5.7% colocalization with an unrelated control (; P < 0.001). Together, our data indicate the existence of a transport complex consisting of gephyrin functioning as an adaptor protein that couples vesicular GlyR with the dynein motor complex (). Dynein motor complexes mediate a variety of functions in neurons (; ). To show a direct and functional link between gephyrin transport and dynein-mediated reactions, we initially searched for a cell system with less complexity. Remarkably, gephyrin does not reach the plasma membrane compartment upon heterologous expression in human embryonic kidney 293 (HEK293) cells, but rather accumulates in intracellular aggregates (). We hypothesized that HEK293 cells may lack the appropriate anterograde transport system to recruit gephyrin toward the plasma membrane. In this case, dynein-mediated retrograde transport reactions could accumulate gephyrin polypeptides at the microtubule-organizing center (MTOC), which harbors the minus ends of microtubules. Indeed, double detection of heterologously expressed YFP–gephyrin and endogenous γ-tubulin, representing a marker for MTOCs (), revealed that intracellular gephyrin aggregates highly colocalized with MTOC structures (Fig. S2, available at ). In contrast, by overexpressing dynamitin, which represents a widely used blocker of dynein motor function (), gephyrin aggregates were no longer found at MTOCs (Fig. S2). Therefore, we conclude that gephyrin tends to aggregate in this cell type because of its multimerization domains (, ). Gephyrin is further transported in microtubule minus-end directions via the dynein motor complex, and the blockade of dynein-mediated transport relocalizes gephyrin aggregates within the cytoplasm of these nonneuronal cells. Next, we aimed to inhibit dynein function in neurons. Based on structural observations on the gephyrin polypeptide (, ), we expressed an NH-terminal truncated gephyrin polypeptide (amino acids 2–188) fused to GFP in mature cultured hippocampal neurons. This polypeptide harbors the trimerization motif, but lacks the dimerization motif of gephyrin and therefore represents a dominant-negative protein. Thus, this gephyrin deletion mutant interferes with the incorporation of endogenous gephyrin into a hexagonal scaffold formation (), thereby reducing gephyrin cluster stability over time. As a result, cells that highly express this deletion mutant are represented by loss of endogenous gephyrin clusters in neurites within 24 h of expression, with very few aggregates remaining in the cell somata (). Based on the hypothesis that postsynaptic gephyrin cluster formation and/or remodeling is a constant steady-state process ( and ) that involves dynein-mediated retrograde transport reactions, the GFP–gephyrin 2–188 mutant was coexpressed together with the dynein transport inhibitor dynamitin (). A control of dynamitin function in hippocampal neurons is shown in Fig. S3 (A and B, available at ). Notably, dynamitin overexpression prevented the dominant-negative–induced loss of endogenous gephyrin clusters, which remained at synaptic sites under these conditions (). These results are likely to depend on a severe slowdown of gephyrin scaffold turnover upon dynein inhibition, which prevents the incorporation of truncated polypeptides (encoded by GFP–gephyrin 2–188) into preexisting gephyrin scaffold formations and, thus, the loss of endogenous gephyrin clusters. In any case, we show for the first time a functional requirement of dynein for gephyrin recruitment in neurons. Dynein-dependent transport processes require microtubules as tracks for transport, and the depolymerization of microtubules interferes with dynein-dependent cargo recruitment in both neurons and other cell types (). To analyze whether the depolymerization of microtubules interferes with gephyrin transport, we also treated mature cultured hippocampal neurons with nocodazole, a microtubule-depolymerizing agent (). As indicated by immunocytochemistry using an α-tubulin–specific antibody, microtubules depolymerized within 15 min of nocodazole application (Fig. S4 A, available at ). By use of the assay, as shown in , dominant-negative–induced loss of gephyrin clusters was prevented in the absence of microtubules (Fig. S4 D), confirming that gephyrin recruitment is also a microtubule-dependent transport process. Inhibition of dynein transport by dynamitin causes the blockade of various dynein-dependent transport reactions. To more specifically interfere with gephyrin–dynein interactions, we generated a mRFP fusion protein (mRFP–gephyrin 181–243; ) harboring the DLC-binding motif of gephyrin. This gephyrin peptide was previously shown to specifically interact with DLCs (). Remarkably, overexpression in cultured hippocampal neurons revealed that this isolated gephyrin-derived polypeptide is able to functionally interact with the dynein motor complex, as detected by retrograde movement of mRFP fusion proteins over time (). This finding corroborates the aforementioned data indicating that dynein specifically binds and transports gephyrin. A control that overexpression of this fusion protein does not generally interfere with dynein transport is shown in Fig. S3 C, indicating that the formation of the Golgi complex (GM130 immunostaining of cis-Golgi), known to depend on functional dynein, is normal in cells that overexpress mRFP–gephyrin 181–243. We further coexpressed GFP–gephyrin and mRFP– gephyrin 181–243 in neurons to analyze whether the red fluorescent fusion protein is able to compete with binding to dynein and therefore affects retrograde transport of full-length GFP–gephyrin. As seen in , localization of full-length GFP–gephyrin clusters to distal dendrites was normal. Consistent with this, anterograde movement of full-length GFP–gephyrin particles was detectable in cells expressing both fusion proteins for ∼12 h (unpublished data). However, upon overexpression of mRFP–gephyrin 181–243, transport of full-length GFP–gephyrin particles in retrograde directions was undetectable ( [cells] = 34; [clusters] > 1,000). In contrast, mRFP fusion proteins consisting of gephyrin residues 181–243 in the same cells were highly mobile in retrograde directions (), suggesting that overexpression of the isolated binding motif competes with full-length, most likely multimerized, GFP–gephyrin for dynein interaction. Finally, upon expression of mRFP–gephyrin 181–243 for several days the anterograde transport route of GFP–gephyrin was also decreased and ultimately undetectable in neurons expressing both fluorescent proteins, thereby confirming our previous observations ( and ), which suggested that gephyrin is the subject of a highly regulated turnover process including both transport directions. As a result, blockade of one transport direction also affects the opposite directed transport on a longer time scale. However, more importantly, these results indicate that an isolated DLC-binding motif of gephyrin competes with full-length GFP–gephyrin particle transport, thereby corroborating aforementioned loss-of-function results using dynamitin and nocodazole, which show that gephyrin binds the dynein motor complex via DLC interactions and is functionally transported by this molecular motor system. In accordance to these observations, combined neuronal expression of GFP–GlyRβ and mRFP–gephyrin 181–243 led to similar results (Fig. S5, available at ). Consequently, gephyrin and/or GlyR–gephyrin remodeling processes at inhibitory postsynaptic sites essentially depend on active transport via dynein. Using a combination of biochemical, immunocytochemical, and time-lapse assays in neurons, we functionally investigated the dynamics of gephyrin, a scaffold component at postsynaptic specializations of inhibitory synapses. We demonstrate that gephyrin particles enter and leave active synapses in the range of minutes and can be subject of cotransport together with the inhibitory GlyR. Furthermore, the molecular motor dynein binds and colocalizes with GlyR–gephyrin transport units and functionally recruits them in a microtubule-dependent manner along neuronal processes. Our data postulate a GlyR–gephyrin–dynein transport complex that is involved in retrograde transport processes underlying postsynaptic remodeling. As revealed in a recent electron microscopy quantitative analysis (), 86.5% of gephyrin locates at synaptic and perisynaptic regions in neurons. These values are consistent with our quantitative evaluation as shown in . Another significant proportion of gephyrin (11.5%) is also found in association with the plasma membrane at extrasynaptic sites, whereas only 2% of gephyrin locates in the cytoplasm. These values demonstrate the tendency of this polypeptide to assemble at plasma membrane regions, a process that requires multimerization through a trimerization motif in the NH-terminal and a dimerization motif in the COOH-terminal part of the protein (, ). It has been hypothesized that gephyrin generates a reversible scaffold for postsynaptic receptor recruitment (; ) that underlies constant exchange of material to subsequently regulate the number of receptors available for synaptic transmission (). However, until now there has been no experimental evidence for a dynamic exchange of gephyrin at synapses. Analysis of gephyrin over time revealed that only a small percentage of particles are mobile within neurite projections. This value is consistent with small amounts of gephyrin locating in the neuronal cytoplasm () and suggests that intracellular gephyrin represents molecules in transit. Our quantitative analysis of GlyR–gephyrin–dynein colocalization is slightly higher than the mobile particles observed; however, it has to be considered that not all transport complexes attach to microtubules at a given time (), a conclusion that is consistent with our observation of discontinuous movements with alternate mobility and immobility over time. The velocity of gephyrin particles (∼1.3 μm/min) closely resembles the transport characteristics of PSD-95, a molecule that is also involved in scaffold formation at postsynaptic sites of excitatory synapses (; ). Notably, mobile gephyrin particles were added to or released from active synapses in the range of several minutes, suggesting that gephyrin transport is involved in fast modular assembly/remodeling of scaffold size and/or neurotransmitter receptor transport over time. In contrast to the prominent association of gephyrin and GlyR at plasma membranes, both immunostaining and immunoprecipitation using high-speed fractions indicated that intracellular gephyrin is also associated with GlyR in the cytosol, interactions likely to represent cotransported molecules in transit. Indeed, dual-channel time-lapse analysis revealed retrograde cotransport of gephyrin and GlyR fusion proteins, thereby functionally confirming the existence of intracellular GlyR–gephyrin transport complexes. Notably, upon addition of physiological KCl concentrations, neuronal depolarization, as well as blockade of GlyR-mediated inhibition, enhanced the transport velocities of GFP–gephyrin particles in our system, suggesting that feedback mechanisms might exist that cross talk between the neuronal surface membrane and the intracellular transport machinery. It is known that alterations in neuronal activity recruit other components to and from synapses (; ; ; ) and depolarization of neonatal hippocampal slices also increases phosphorylation of the microtubule-associated protein MAP2, thereby impairing its ability to stabilize microtubules (). Although the exact mechanisms are currently unknown, it is possible that activity-regulated stability of cytoskeletal elements might contribute to the velocity of cargo delivery in neurons. Because gephyrin directly binds the inhibitory GlyR and functionally associates with GABAR subtypes (; ; ), we also analyzed the transport characteristics of GFP–gephyrin in the presence of the GlyR or GABAR antagonists strychnine or bicuculline, respectively. Although this situation is not physiological, it showed that the blockade of GlyR, but not of GABAR, affects GFP–gephyrin transport. In fact, upon GlyR blockade, the number of mobile GFP–gephyrin particles increased by >100% with a distinct shift of transport in the retrograde direction. Because this strychnine-mediated shift of gephyrin transport was observed both with and without KCl, it seems likely that the effects mediated by KCl and strychnine are independent of each other. Different studies have previously reported that chronic strychnine-mediated blockade of GlyR causes an intracellular receptor accumulation at neuronal somata near the nuclear compartment (; ). Later it was hypothesized that strychnine triggers the disappearance of GlyR from synapses (). In confirmation and addition to this view, our data for the first time functionally demonstrate that retrograde GFP–gephyrin transport and/or gephyrin–GlyR cotransport is directly sensitive to GlyR blockade. Moreover, consistent with our dual-channel time-lapse experiments, they provide another functional con- nection of GlyR and gephyrin cotransport in neurons. Other transport complexes consist of neurotransmitter receptors and postsynaptic scaffold components (). For instance, mLin2/CASK and GRIP1 comigrate in a kinesin-dependent manner with -methyl--asparate (NMDA) or AMPA receptors, respectively (, ). Hence, there has been an interest in the question of which motor system would drive a gephyrin–GlyR complex within neurons. Although a yeast two-hybrid screen identified the motor components Dlc-1 and -2 as gephyrin-binding partners (), until now it was not clear whether these interactions point to actin- or microtubule-based transport reactions because DLCs are components of both myosin and dynein complexes (; ). Applying both immunoprecipitation from intracellular fractions and immunocytochemistry, we demonstrate that both intracellular gephyrin and GlyR bind and colocalize with either DICs or DHCs, which are exclusive components of the dynein motor complex. Dynein motors are known to transport a large variety of cargo molecules in neurons and other cell types (; ; ); therefore, it is consistent that only a small proportion of dynein puncta colocalize with gephyrin at a given time. Remarkably, colocalized puncta were mainly found at nonsynaptic sites, a finding that suggests dynein motor complexes do not directly reach the postsynaptic specialization. This result resembles observations made for NMDA receptor synaptic delivery (). A motor-cargo complex consisting of the microtubule-dependent kinesin KIF17 and the synaptic NMDA receptor subunit protein NR2B also colocalizes close to, but not at synaptic sites. Whether or not microtubule-based motors generally reach the postsynaptic site is currently unclear. However, actin filament–based motor systems, known to mediate short-distance transport (), might contribute to cargo recruitment at synapses (; ; ). Functional evidence that gephyrin recruitment depends on active transport through the dynein motor complex was given in different independent assays. First, active transport was confirmed by the lack of particle diffusion at high image acquisition rates. Second, loss-of-function experiments using dynamitin blockade of dynein in both HEK293 cells and neurons confirmed the association of gephyrin and dynein. In addition, by expressing a fluorescently labeled peptide that harbored the DLC-binding site of gephyrin (mRFP–gephyrin 181–243), we could show that this polypeptide is, on its own, highly mobile in neurons, indicating that it is able to associate with a molecular motor. Because short-term overexpression of this binding site was able to compete with retrograde GFP–gephyrin or GFP–GlyRβ transport in time-lapse experiments, but not with anterograde-dependent transport of clusters to distal dendrites, our data strongly indicate that GlyR– gephyrin complexes are functionally recruited via dynein. Notably, our time-lapse analysis revealed that gephyrin and GlyRs move not only retrogradely but also in anterograde directions over time. Thus, it will be a challenge to identify the anterograde transport system that is required for gephyrin and/or GlyR recruitment toward the synapse. A GlyR–gephyrin–dynein triple transport complex is likely to contribute to the regulation of synaptic receptor number and the regulation of gephyrin scaffold, which in turn provides the platform for trapping of diffusing plasma membrane receptors. In this respect it is also important to understand whether dynein-mediated retrograde recruitment of the complex mainly represents receptor recycling processes, degradation of receptors, or both. Based on our observations that depolarization and/or receptor blockade influences transport parameters of the gephyrin–GlyR complex, it appears that cross talk between neuronal activity mechanisms and the neuronal transport machinery might be an important platform for the modulation of synaptic strength at inhibitory synapses. The GFP–gephyrin fusion construct has been previously described (). To generate YFP–gephyrin, the vector pEYFP-C1 (BD Biosciences) was restricted with BglII and treated with Klenow polymerase (Roche) in the presence of deoxyribonucleotide triphosphates to generate the pEYFP-C2 vector. The gephyrin complementary DNA was subsequently subcloned into pEYFP-C2 as a HindIII–KpnI fragment. To generate mRFP–gephyrin, the mRFP1 coding sequence was subcloned as a NheI–SacI fragment into GFP–gephyrin, thereby replacing GFP by mRFP1. To generate GFP–gephyrin 2–188, a PCR product encoding amino acids 2–188 of gephyrin was cloned as a HindIII–SalI fragment into pEGFP-C1 (BD Biosciences). For the generation of GFP–GlyRβ, a BglII restriction site was introduced after the signal peptide of a GlyRβ construct in pRK5. The GFP coding sequence was introduced as a BglII–BglII PCR product into this locus. Furthermore, the GlyRβ 3′-UTR was introduced into the PstI site of pRK5. To generate mRFP–gephyrin 181–243, a PCR product encoding mRFP1 was cloned as a NheI–SacI fragment into GFP–gephyrin 181–243 (), thereby replacing GFP by mRFP1. GFP–dynamitin and Dynamitin–myc were obtained from R. Vallee, Columbia University, New York, NY. GFP–Shank 1 obtained from C. Sala, University of Milan, Milan, Italy. mRFP1 was obtained from R.Y. Tsien, University of California, San Diego, La Jolla, CA. The following antibodies were used for immunoprecipitation and Western blotting: mouse anti-gephyrin (1:1,000; Synaptic Systems GmbH), mouse anti-gephyrin (1:250; BD Biosciences), rabbit anti-gephyrin (1:4,000; Alexis), mouse anti-GlyR, clone mAb4a (1:250; Synaptic Systems GmbH), rabbit anti-GlyR (1:100; Sigma-Aldrich), mouse anti-dynein intermediate chain (1:1,000; CHEMICON International, Inc.), mouse anti-dynein intermediate chain (1:1000; Sigma-Aldrich), and mouse anti-GluR2 (1:1,000; CHEMICON International, Inc.). The following antibodies were used for immunofluorescence: mouse anti-gephyrin (1:100; Synaptic Systems GmbH), rabbit anti-gephyrin (1:100; Qbiogene), mouse anti-GlyR, clone mAb4a (1:100; Synaptic Systems GmbH), goat anti-DHC, clone S-19 (1:100; Santa Cruz Biotechnology, Inc.), mouse anti-dynein intermediate chain (1:100; CHEMICON International, Inc.), rabbit anti-VIAAT (1:200; obtained from B. Gasnier, Centre National de la Recherche Scientifique, Paris, France; ), mouse anti–α-tubulin (1:1,000; Sigma-Aldrich), mouse anti–γ-tubulin (1:100; Sigma-Aldrich), mouse anti-myc (1:100; Sigma-Aldrich), rabbit anti-myc (1:100; Sigma-Aldrich), mouse anti–synaptic vesicle (SV2; 1:100; Developmental Studies Hybridoma Bank), rabbit anti-synaptophysin (1:100; DakoCytomation), goat anti-synaptophysin (1:100; Santa Cruz Biotechnology, Inc.), and mouse anti-GM130 (1:100; Sigma-Aldrich). The following secondary antibodies were used: CY3-, CY2-, or CY5-conjugated donkey anti–goat, anti–mouse, or anti–rabbit (all 1:500; Dianova). Brains of five postnatal day (P) 10 juvenile rats were homogenized in buffered sucrose solution containing 320 mM sucrose, 2 mM DTT, 1 mM EDTA, 1 mM EGTA, and 4 mM Hepes-KOH, pH 7.4, supplemented with proteinase inhibitor cocktail (Roche), 2 mM ATP, and 5 mM MgCl. The homogenate was centrifuged at 1000 for 10 min (P1) and the resulting postnuclear supernatant further clarified at 10,000 for 10 min (P2). Supernatant from this medium speed centrifugation was processed by another 160,000 centrifugation step to collect small membrane organelles (P3). The pellet was resuspended in 1.5 ml of 2 M sucrose solution and subjected to a linear 0.3–2 M sucrose-density gradient centrifugation. The gradient was centrifuged at 160,000 in a rotor (model SW49Ti; Beckman Coulter) for 12 h at 4°C. Fractions of ∼750 μl were collected from the top of the gradient using a defractionator (Labconco). 20 μl of each fraction was used for SDS-PAGE and Western blotting. Rat brains of five P10 animals were dissected in ice-cold PBS and homogenized in IM-Ac buffer, containing 20 mM Hepes, 100 mM KCl, 5 mM EGTA, and 5 mM MgCl, pH 7.2. The buffer was supplemented with proteinase inhibitor cocktail (Roche), 5 mM DTT, and 2 mM MgATP. The homogenate was clarified by centrifugation at 1000 for 10 min and the postnuclear supernatant was used for the following steps. First, the supernatant was centrifuged at 10,000 for 10 min to pellet large membrane organelles (P2). The remaining supernatant was then centrifuged at 100,000 to collect small membrane organelles (P3). Finally, remaining organelles and large protein complexes were pelleted at 400,000 for 60 min (P4). After diaminopimelate cross-linking of antibodies to magnetic beads (Invitrogen), antigen from P3 or P4 fractions was immobilized, followed by extensive washing steps with either IM-Ac buffer or IM-Ac buffer containing 0.5% Triton X-100. Bound proteins were eluted by boiling in SDS-containing sample buffer and examined by Western blotting. Primary cultures of hippocampal neurons were prepared from mice or rats at P0 and P1, as previously described (; ). Cells cultured between 4 and 12 d in vitro were used for transfection with either 4 μg Lipofectamine 2000 (Invitrogen) in serum-free medium or by a calcium phosphate coprecipitation protocol (). For KCl experiments, hippocampal neurons were incubated in neurobasal/B27 medium containing 10 mM KCl for 1–3 min before fluorescent imaging. For FM-dye labeling of active synapses, cells were exposed to 15 μM FM4-64 for 1 min in 31.5 mM NaCl, 90 mM KCl, 5 mM Hepes, 1 mM MgCl, 2 mM CaCl, 30 mM glucose, and 50 μM DL-AP5. Imaging was performed in neurobasal/B27 medium containing 50 μM DL-AP5 and 10 μM of 6-cyano-7-nitroquinoxaline-2,3-dione. For GlyR or GABAR blockade, neurons (12 d in vitro) expressing GFP–gephyrin were analyzed by time-lapse imaging in 10 mM Hepes and 10 mM KCl containing either 500 nM strychnine or 10 μM bicuculline. Drugs were added immediately before imaging. HEK293 cells were cultured on glass coverslips. For heterologous expression, cells were microinjected. Plasmids were purified using a complementary DNA purification kit (QIAGEN). Cells were kept in 10 mM of prewarmed Hepes buffer, pH 7.4, during the procedure. 30 ng/μl concentrations of DNA were microinjected into nuclei () using a Transjector 5246 coupled with an Injectman system (both Eppendorf AG) at 70–100 hPa for 0.1–0.3 s. Visual control was obtained by the use of an inverted microscope equipped with a 63× long distance phase-contrast objective (Axiovert 35 and LD Plan Neofluar, respectively; Carl Zeiss MicroImaging, Inc.). For depolymerization of microtubules, the neurobasal/B27 medium of cultured hippocampal neurons was removed 3 h after transfection and stored separately. Cells were then incubated in fresh medium containing 10 μM nocodazole (Sigma-Aldrich) for 15 min, washed with PBS/10 mM glucose, and stored in the original neurobasal/B27 medium. After 24 h at 37°C/5% CO, cells were fixed and processed for immunocytochemistry. Fluorescence imaging was performed with an inverted laser scanning confocal microscope (model TCS-SP2; Leica) using a 63× objective. For simultaneous multichannel fluorescence, images were taken in a sequential channel recording mode. Images from time-lapse experiments were taken with a TCS-SP2 microscope or an inverted fluorescent microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) combined with a charge-coupled device camera (SPOT RT-SE; Sony). Images were taken at various intervals ranging from every 30 to every 60 s. Cells at the microscope stage were temperature controlled (37°C) and either CO controlled or kept in Hepes-buffered medium. For analysis of time-lapse data, dendrite length, size of puncta, and intensity histograms, Power Scan software TCS-NT (Leica), the analySIS software package 2.5 (Soft Imaging System GmbH), and the MetaVue 6.2r6 software (Universal Imaging Corp.) were used. Puncta were quantified using Scion Image 1.63 software (National Institutes of Health). The statistical significance of experiments was assessed with the test. Fig. S1 represents control figures for – . Fig. S2 shows the effect of dynamitin on YFP–gephyrin localization in HEK293 cells. Fig. S3 is a control figure for and . Fig. S4 shows that depolymerization of microtubules prevents the dominant-negative–induced loss of gephyrin clusters in neurons. Fig. S5 shows that mRFP–gephyrin 181–243 interferes with GFP–GlyRβ retrograde transport in neuronal dendrites. Video 1 shows the recruitment of GFP–gephyrin as shown in . Video 2 shows the recruitment of GFP–gephyrin toward and/or from FM4-64–positive presynaptic terminal boutons, as shown in . Video 3 shows the cotransport of GFP–GlyRβ and mRFP–gephyrin, as shown in . Online supplemental material is available at .

Synaptic transmission efficacy largely depends on the morphology of dendritic spines, which are highly specialized protrusions on the neuronal surface that receive most of the central excitatory input (; ). Dendritic spines are highly motile structures capable of undergoing rapid morphological plasticity in response to external signaling events (; ). Such architectural changes are achieved via dynamic modifications of the local actin cytoskeleton, the major cytoskeletal element present in dendritic spines (; ; ). Much information has been gathered describing the many actin-regulatory mechanisms that cells use to control diverse morphogenic events (). A group of proteins that have acquired a most noticeable position are the Rho GTPases, of which RhoA, Rac1, and cdc42 are the best-characterized members (; ). Rho GTPases act as molecular switches, existing in an active GTP-bound and an inactive guanosine diphosphate (GDP)–bound state (; ). Depending on this, these proteins trigger modifications in the actin polymerization state via specific downstream effectors. As they are ubiquitously distributed, the activity of Rho GTPases must be precisely monitored in space and time, allowing for the many different architectural actin-dependent modifications occurring at different cellular domains (). Thus, for example, the protrusive or quiescent status of axonal growth cones is determined by the dynamic state of the underlying actin cytoskeleton, which, in turn, is locally regulated by specific membrane-signaling events in a Rho GTPase–dependent manner (). In principle, a similar spatially and temporally restrictive control over membrane-signaling events could be involved in the expansion/retraction of dendritic spines upon synaptic stimulation. In support of this, in cultured hippocampal neurons, activation of AMPA and -methyl--asparate (NMDA) receptors (NMDARs) results in dendritic spine collapse in an actin-dependent manner (; ). Furthermore, manipulation of Rho GTPase activity via the expression of constitutively active or dominant-negative mutants affects dendritic spine number and shape (for review see Dillon and Goda, 2004). In optic tectal neurons, the stimulation of glutamate receptors (GluRs) lowers RhoA activity (). However, many questions remain open: how do these three locally restricted events—membrane receptor activation, modulation of RhoA activity, and actin dynamics—relate mechanistically? How does the activation of excitatory neurotransmitter receptors influence the activity of RhoA in spines? What are the molecular mechanisms controlling the underlying actin cytoskeleton? To address these issues, we performed a series of cell biological approaches in embryonic rat hippocampal neurons in culture during the time of ongoing excitatory synaptic activity. ext-link #text I n t h i s s t u d y , w e i l l u s t r a t e d t h a t R h o A i n t e r a c t s w i t h G l u R s a t t h e l e v e l o f t h e P S D a n d r e g u l a t e s d e n d r i t i c s p i n e a c t i n i n a R O C K – P I I a - d e p e n d e n t m a n n e r . S u c h R h o A - d e p e n d e n t d e n d r i t i c s p i n e F - a c t i n r e g u l a t i o n i s r a p i d l y ( w i t h i n m i n u t e s ) m o d u l a t e d i n r e s p o n s e t o d i r e c t ( A M P A a n d N M D A ) o r i n d i r e c t ( 5 5 m M K C l ) s t i m u l i . F u r t h e r m o r e , w e h a v e m a n a g e d t o d i f f e r e n t i a t e b e t w e e n i G l u R a n d m G l u R b e h a v i o r t o s u c h s t i m u l i a n d , m o r e o v e r , d i s s e c t e d t h e d i f f e r e n c e s b e t w e e n N M D A - a n d A M P A - t y p e n e u r o t r a n s m i t t e r r e c e p t o r s . Primary cultures of rat embryonic hippocampal neurons were prepared as described previously (). For biochemical analysis, 150,000 cells were plated per 30-mm plastic dish coated with 0.1 mg/ml poly--lysine. For morphological analysis, 100,000 cells were plated per 60-mm dish, each with six poly--lysine–coated glass coverslips, and for in vitro time-laps observations, the cells were plated on 40-mm poly--lysine–coated glass coverslips. All times referred to in the text and figure legends are after seeding (time of plating is 0 h). Y27632 was used at 24 nmol/ml (Calbiochem). Optionally, neurons were incubated with 1 μM MK-801 (Sigma-Aldrich) or 2 μM CNQX (Qbiogene) for 10 min before high potassium treatment (see below). The following antibodies were used: mouse monoclonal anti-RhoA (26C4; Santa Cruz Biotechnology, Inc.), mouse monoclonal anti-ROCK (clone 21; BD Transduction Laboratories), mouse monoclonal anti-PSD95 (Upstate Biotechnology), mouse monoclonal antisynaptophysin (Boehringer), rabbit polyclonal anti–α-synuclein (a gift from C. Sanchez, Centro Biologia Molecular, Madrid, Spain), rabbit polyclonal anti-mGluR1 (AB-1504; Chemicon), rabbit polyclonal anti-PIIa (a gift from W. Witke, European Molecular Biology Laboratory, Monterotondo, Italy), rabbit polyclonal anti-GluR type 1 (Upstate Biotechnology), rabbit polyclonal anti-NMDAR2a (AB1555P; Chemicon), rabbit polyclonal anti-ST (a gift from M. de Hoop, Aventis, Frankfurt, Germany), and goat polyclonal anti-Dia1 (V20; Santa Cruz Biotechnology, Inc.). Goat anti–rabbit and goat anti–mouse AlexaFluor350/488/568 (Invitrogen) and donkey anti–mouse, goat anti–rabbit, and donkey anti–goat HRP-conjugated antibodies (GE Healthcare) were used as secondary antibodies. Cells were fixed in PFA/SEM buffer (4% PFA, 0.12 M sucrose, 3 mM EGTA, and 2 mM MgCl in PBS), quenched with 50 mM ammonium chloride, and extracted with 0.1% Triton X-100. Specific protein detection was performed using previously mentioned antibodies (see previous section), and F-actin was labeled with TRITC-conjugated phalloidin (Sigma-Aldrich). Optionally, cells were labeled with a lipophilic tracer (DiI; Invitrogen) to visualize the plasma membrane. For this, fixed cells were incubated with a 67-μg/ml working solution of DiI prepared in 1× PBS for 30 min at room temperature before mounting. For visualization of active RhoA only, cells were incubated with 0.1% Triton before fixation with PFA/SEM to extract all soluble protein forms. Cells were observed using a microscope (DMIRE2; Leica) equipped with 40, 63, and 100× objectives (Leica) and a digital camera (Q550; Leica), and images were captured using the Qfluoro software (Leica). Optionally, samples were analyzed in a confocal scanning microscope (LSM 5110; Carl Zeiss MicroImaging, Inc.) on a platform (Axiovert 100 M; Carl Zeiss MicroImaging, Inc.). For in vitro time-lapse experiments, the plasma membrane of mature hippocampal neurons was labeled with a lipophilic tracer (DiI; Invitrogen). In brief, cells were incubated for 1 min at 37°C with a 67-μg/ml working solution of DiI prepared in equilibrated growth medium. The cells were incubated in HBSS for 10 min under culture conditions, and time-laps recording was performed. For this, the cells were treated as described previously (). In brief, cells were placed in a temperature-controlled FCS-2 long-term observation chamber (Bioptechs) and positioned on the stage of an inverted microscope equipped with 40, 63, and 100× objectives (Leica), and images were captured using the Qfluoro software (Leica). Pictures of chosen cells were taken in 1-min intervals. In some cases, the cells were incubated with a 2-μM solution of photosensitive nitroveratryloxycarbonyl-caged ROCK inhibitor (caged Y27632; Invitrogen) for 15 min. Caged Y27632 remains inactive until hydrolyzed under UV light (wavelength ≤ 360 nm). To locally activate the compound at a selected dendritic segment (dendrite 1), the microscope pinhole was closed to the minimum, and the selected region was stimulated by a 500-ms exposure to light of short wavelength (band pass filter 360/40 nm). Pictures of the DiI-labeled dendritic tree (absorption of 549 nm and emission of 565 nm) were taken in 1-min intervals. For control purposes, a second dendritic segment (dendrite 2) just outside of the pinhole was observed before and after local uncaging of Y27632 at dendrite 1, and pictures were taken at time points −1 min and +10 min of uncaging at dendrite 1. To verify the specificity of the effects observed after local uncaging of Y27632, we incubated DiI-labeled cells with 0.14% DMSO in HBSS (vector buffer used for caged Y27632) and stimulated selected dendritic segments (dendrite 3) under the same conditions as described for caged Y27632. The protocol used to purify synaptosomal fractions from adult rat brain is based on well-established methods used by Cohen et al. and . 6 g of adult rat brains were homogenized in 4 vol/g of buffer A (0.32 mM sucrose, 1 mM MgCl, 0.5 mM CaCl, 1 mM NaHCO, chymostatin, leupteptin, antipain, pepstatin, and 1 mM dithiothreitol) at 800 rpm/7 strokes in a Dounce glass homogenizer. After the addition of 10 vol/g of buffer A, the homogenate was centrifuged at 1,400 for 10 min to recover the supernatant S1 and the pellet P1. P1 was resuspended in 4 vol/g of buffer A, homogenized at 800 rpm/3 strokes, and recentrifuged at 700 for 10 min. The resulting supernatant was combined with S1 and centrifuged at 13,800 for 10 min. The obtained supernatant (S2) was separated from the pellet P2 and centrifuged at 100,000 for 1 h. The resulting supernatant (S3) constitutes the cytosolic fraction. P3 was resuspended in 24 ml/10g wet weight of buffer B (0.32 mM sucrose, 1 mM NaHCO, 1 mM EGTA, 1 mM dithiothreitol, chymostatin, leupeptin, antipain, and pepstatin) and homogenized to obtain the crude synaptosomal fraction. To obtain the pure synaptosomal fraction, the sample was loaded on a discontinuous sucrose gradient (1 and 1.4 M sucrose) and centrifuged for 65 min at 82,500 . The synaptosomal fraction was recovered from the interphase between 1 and 1.4 M sucrose. Protein amount was calculated, and a 4-mg/ml solution was prepared with buffer B. An equal volume of a solution composed of Triton X-100, 0.5 mM Hepes/KOH, and protease inhibitors was added and stirred for 15 min on ice. The sample was centrifuged at 28,000 for 40 min to obtain supernatant LS1. LS1 was centrifuged at 165,000 for 120 min to obtain pellet LP2. LP2 was then homogenized in 2 ml of buffer B and loaded onto a discontinuous sucrose density gradient composed of 1.0, 1.5, and 2.1 M sucrose and was centrifuged at 201,800 for 60 min. A PSD fraction (PSD I) was obtained from the interphase between sample and 1.0 M sucrose. Protein amounts of the cytosolic, crude syaptosomal fractions were calculated using spectrophotometric analysis. 1.2 ml of PSD I fraction was concentrated to a final volume of 70 μl using chloroform–methanol precipitation. Synaptosomal fractions (500-μl sample volume at 2 mg/ml) were optionally treated with 5- or 55-mM KCl solutions (see next section). Where noted, synaptosomal preparations (500-μl sample volume at 2 mg/ml) were treated with GDP (to 1.0 mM; Pierce Chemical Co.), GTPγS (to 0.1 mM; Pierce Chemical Co.), or ddHO (in the case of control experiments). Samples were incubated at 30°C for 30 min under constant agitation, and the reaction was stopped by placing the samples on ice and adding 32 μl of 1 M MgCl. Activation of NMDA and/or AMPA receptors was performed by incubating synaptosomal preparations with 10 μM NMDA (Sigma-Aldrich) or AMPA (Qbiogene) for 3 min at 37°C under gentle agitation. Optionally, synaptosomal preparations were incubated at 37°C with 1 μM MK-801 (Sigma-Aldrich), 2 μM CNQX (Qbiogene), 100 μM DHPG (Sigma-Aldrich), or 400 μM AIDA (Qbiogene) for 5 min before high potassium treatment (see next section) or incubation with AMPA or NMDA. Coverslips were placed into 55 mM KCl (high potassium) buffer (10 mM Hepes, 2.2 mM CaCl, 0.33 mM NaHPO, 0.44 mM KHPO, 4.2 mM NaHCO, 5.6 mM glucose, 77 mM NaCl, and 55 mM KCl) or 5 mM (low potassium) buffer (10 mM Hepes, pH 7.2, 2.2 mM CaCl, 0.33 mM NaHPO, 0.44 mM KHPO, 4.2 mM NaHCO, 5.6 mM glucose, 127 mM NaCl, and 5 mM KCl) and incubated for 3 min at 37°C and 5% CO in a humid chamber. Coverslips were then processed for immunofluorescence or replaced into the original growth medium for 22 h (termed 55 mM + wash) before fixation. For synaptosomal preparations, low or high potassium buffer was added to purified samples and incubated at 37°C for 3 min under gentle agitation. These samples were brought to 4°C on ice and used immediately for immunoprecipitation or RhoA activation assay (see below). Active RhoA was isolated from neuronal lysates and synaptosomal preparations using the EZ-Detect RhoA activation kit (Pierce Chemical Co.). In brief, cell lysates or synaptosomal preparations were incubated with recombinant GST-Rhotekin in the presence of the designated SwellGel Immobilized Glutathione Discs (Pierce Chemical Co.) at 4°C for 1 h (gentle rocking). The column was then centrifuged briefly at 7,200 and washed with washing buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl, 1% NP-40, 1 mM DTT, and 5% glycerol). The gel-bound active RhoA was eluted by adding 50 μl of sample buffer (125 mM Tris-HCl, pH 6.8, 2% glycerol, 4% SDS, and 0.05% bromophenol blue). The resulting elutes were then used immediately for Western blotting (see below). Synaptosomal preparations were precleared with prewashed protein G–Sepharose beads and were incubated with 3 μg anti-RhoA antibody for 1 h at 4°C. Subsequently, protein G–Sepharose beads were added, and samples were incubated overnight at 4°C under gentle rotation. Samples were then washed twice (20 min each) with immunoprecipitation buffer (1% Triton X-100, 100 mM NaCl, 2 mM EDTA, 10 mM Tris-HCl, 1 mM NaVO, pH 7.5, and protease inhibitors), twice (20 min each) with high salt buffer (same as immunoprecipitation buffer but with 500 mM NaCl and no Triton X-100), and once (20 min) with low salt buffer (same as immunoprecipitation buffer but no NaCl or Triton X-100). Beads were pelleted in between washes by centrifugation at 1,600 for 30 s. After the final wash, beads were pelleted down by high-speed centrifugation, and the supernatant was analyzed by Western blotting (see below). Approximately 40 μg of cytosolic and pure synaptosomal fraction and 22 μl of concentrated PSD I fraction were loaded on 12% SDS gels and separated by electrophoresis. Separated proteins were transferred to nitrocellulose filters. Filters were blocked by incubation in 3% BSA in PBS with 0.1% Tween-20. Filters were then incubated with the indicated primary antibodies and with the respective secondary HRP-conjugated antibodies (see Antibodies). Signal detection was performed using an ECL detection kit (GE Healthcare) before exposure to photosensitive films. The obtained autoradiogram was scanned at high resolution (3000 Pro Scanner; Epson) and exported for densitometry analysis (see next section) using National Institutes of Health (NIH) Image 1.63 software. Unless noted differently, dendritic spine number was calculated in individual mature hippocampal neurons obtained from three independent experiments ( = 9 cells/treatment within one experiment). Labeling F-actin with TRITC-conjugated phalloidin identified dendritic spines. Where noted, double labeling was performed using phalloidin together with antibodies against ST, synaptophysin, or GluR1. A total dendrite length of 40 μM analyzed per cell and dendritic spine number was determined based on the phalloidin. Where applicable, the determination of dendritic spine number also took into consideration only those F-actin protuberances exhibiting apposed presynaptic terminals (as detected with the presynaptic markers). In all cases, the total number of dendritic spines per cell was averaged for each experiment (population). These population means and the respective SDs were then pooled together and plotted as bar graphics in percentages, thus reflecting the mean variation between individual neurons across populations. Analysis within each separate population revealed similar variations, indicating that the pooled population data reflect the variations across individual populations. Multivariate analysis was performed with one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test (confidence interval = 95%). Signal intensities of proteins (in autoradiogram) detected by Western blot analysis were measured using NIH 1.63 software. Unless noted differently, each experiment was repeated three times independently. Such measurements were individually normalized against the background of each individual autoradiogram and renormalized against the signal obtained with the unrelated protein α-tubulin (CP06; Oncogene Research Products) unless immunoprecipitated or activation assayed where total protein was loaded for each lane analyzed. The resulting individual, normalized signal densities were averaged for each experiment (population). The data were analyzed by paired test (two-tailed distribution and two-sample unequal variance). In case of multivariate tests, one-way ANOVA followed by Tukey's multiple comparison test (confidence interval = 95%) was performed. The means and SDs of each experiment were pooled together and represented as graphs in percentages. Fig. S1 shows immunofluorescent images of mature hippocampal neurons labeled with TRITC-conjugated phalloidin. The F-actin pattern and its signal intensity are drastically modified in 55 mM KCl-treated neurons in respect to controls. Fig. S2 shows immunoprecipitations from synaptosomes, illustrating the specificity of the interaction between RhoA and GluR1/NMDAR2a. Fig. S3 illustrates that in synaptosomes, AMPA and NMDA mimic the effect of 55 mM KCl on the RhoA-dependent recruitment of ROCK and PIIa. Online supplemental material is available at .

Myelin is a multilamellar, tightly compacted membrane that surrounds axons in the central nervous system (CNS) and peripheral nervous system (PNS). Myelin helps concentrate voltage-gated Na channels at nodes of Ranvier (), the short unmyelinated regions between myelin segments. The nerve impulse jumps from node to node by a process called saltatory conduction, which facilitates rapid nerve communication in an energy-efficient manner. Although mammalian CNS and PNS myelin serve similar functions, they can be distinguished by two major features. First, oligodendrocytes form multiple myelin internodes in the CNS, whereas Schwann cells form single myelin internodes in the PNS. Second, myelin proteolipid protein (PLP), a four-transmembrane-domain protein, represents >50% of the protein in mammalian CNS myelin (), whereas P protein, a type I integral membrane glycoprotein and member of the immunoglobulin gene super family, represents >70% of the total myelin protein in mammalian PNS myelin (). P was initially the primary structural protein of CNS and PNS myelin, which first appeared ∼440 million years ago in cartilaginous fish (; ; ; ; ). The DM20 isoform of the gene also appeared in cartilaginous fish myelin, where it was apparently coopted by duplication of an ancestral gene (DMα family) that originated in () and is expressed today in neurons and epithelial cells (). The PLP protein appeared after the divergence of the bony fish ∼400 million years ago () and differs from DM20 by the addition of 35 amino acids into exon III (; ). Both PLP and P had high mutational rates until ∼300 million years ago. During this period, it is likely that the function of PLP was evolving, with a requirement for the continued coexpression of P. However, with the appearance of reptiles/aves, the function of PLP became fully established, allowing the silent dropout of P from CNS myelin (; ). Once PLP and P expression was separated exclusively into CNS and PNS myelin, their mutation rates dropped dramatically, and both are highly conserved (almost 100%) across all mammalian species analyzed (; ; ). This suggests an essential role for PLP and/or a detrimental role for P protein in CNS myelin of higher species. Myelin also provides trophic support, which is essential for axonal survival. Axonal degeneration is the major cause of neurological disability associated with inherited and acquired diseases of myelin (; ). Although the molecular mechanisms responsible for this trophic support are not well understood, mice with a null mutation in the gene have a late-onset axonopathy (). It remains to be determined whether this axonopathy results from loss of PLP-related trophic support or from alterations in the periodicity and/or stability of PLP-deficient myelin. In addition, mutation and duplication of the human PLP gene is a major cause of the inherited disease of myelin, Pelizaeus-Merzbacher disease. To investigate the possible benefits for the P–PLP evolutionary conversion, we “reversed” that evolutionary step using transgenic mice and introduced P expression in exchange for PLP into the mammalian CNS. In the absence of PLP, P protein conferred a highly regular and compact PNS-like structure to CNS myelin. However, because of degeneration of myelinated axons, the lifespan of these mice (compared with wild-type [WT] or PLP-null mice) was reduced by >50%. Based on these data, reversing a discrete step in vertebrate brain evolution demonstrates that the emergence of a new myelin protein was associated with a vital neuroprotective function of myelin-forming CNS glia. Technically, the preferred strategy for expressing P and removing PLP is a direct knockout and knockin of genes by homologous recombination in ES cells. However, when the full-length cDNA encoding P was placed in-frame into exon 2 of the gene, mutant mice expressed very little P mRNA, presumably because of the altered spacing of cis-regulatory elements. Therefore we chose to cross mice carrying a P transgene, strongly expressed in oligodendrocytes, on a -null background. We generated transgenic mice in which the mammalian P cDNA was driven by 9.1 Kb of the murine myelin basic protein (MBP) promoter (). These mice were bred to mice null for PLP (). The heterozygous F1 pups were interbred to generate pups that were genotyped for PLP and the P transgene. The resulting pups were genotyped and interbred further, and at the F3 generation, mice were identified to be WT, PLP-null, P homozygous transgenic (PLP/P-CNS), and P homozygous transgenic plus PLP-null (P-CNS; ). Two lines of mice homozygous for the P transgene were generated, and neither showed a neurological phenotype. The changes described below in mice homozygous for the P transgene and null for PLP are therefore not caused by insertional mutagenesis. These four mouse lines were maintained and included in the present study. To determine whether the genotypes translate to the mRNA level, total brain RNA was isolated from all four lines at postnatal day (P) 60 and probed with P and PLP cDNAs by Northern blot (). These data establish that the P transgene is abundantly transcribed in the CNS and that mice null for PLP mRNA express significant levels of P mRNA. Myelin was also prepared from P60 brains, and its protein composition was analyzed on SDS gels (). As expected, WT CNS myelin contained PLP, myelin from PLP/P-CNS mice contained P and PLP, P-CNS myelin contained P, and PLP-null mice contained neither P nor PLP. The levels of P and PLP were similar in the PLP/P-CNS myelin, and the level of P in the P-CNS myelin appeared similar to PLP in WT myelin. Our goal of replacing PLP with similar levels of P was therefore achieved. We also compared levels of three other myelin proteins, myelin-associated glycoprotein (MAG), 2′, 3′-cyclic nucleotide 3′-phoshodiesterase (CNP), and MBP in CNS myelin purified from the four lines of mice by Western blot (). There were no significant differences in levels of these myelin proteins except their increase in PLP-null myelin, which reflects their relative contribution to total myelin proteins after the loss of PLP, i.e., 50% of the total myelin protein. We next determined the cellular distribution of P and PLP in the different mice by immunocytochemistry (). P protein was synthesized by oligodendrocytes in P-CNS and PLP/P-CNS mice and targeted to myelin internodes. At the light microscopic level, the distribution of P () was indistinguishable from PLP in WT () or PLP/P- CNS () brains. To determine whether P was targeted to compact myelin, we performed electron microscopic immunocytochemistry using immunogold procedures (). P was not detected in compact myelin from WT () or PLP-null () mice but was abundant in PLP/P-CNS () and P-CNS () compact myelin. When expressed, P protein did not accumulate in mouse oligodendrocyte perinuclear cytoplasm, nor was it targeted to paranodal loops or oligodendrocyte plasma membranes. Similar to their amphibian ancestors, mammalian oligodendrocytes maintain their ability to appropriately and exclusively target P to compact myelin. To investigate the possible impact of P protein on the periodicity or membrane spacing of CNS myelin, we analyzed optic nerves from the mice by x-ray diffraction and transmission EM. X-ray diffraction measures myelin periodicity and membrane packing of unfixed, freshly dissected nerves. As documented previously (), typical CNS and PNS myelin diffraction patterns are readily distinguished from one another by the spacing between reflections (which signals the periodicity) and by the number of reflections. Thus, for WT animals, CNS myelin shows two strong Bragg orders (the second and fourth) from an ∼155-Å periodicity, whereas PNS myelin shows four distinctive Bragg orders (the second through fifth) from an ∼174-Å periodicity (). Optic nerves from P-CNS mice gave a diffraction pattern that was identical to WT sciatic nerve; i.e., the positions of the reflections and the relative intensities were the same (). This was true even for the higher resolution region of the patterns, where higher Bragg orders were detected (, inset). The membrane profiles () that were calculated from the diffraction data showed similar dimensions: as measured from the positions of the centers of the electron-dense peaks, the membrane bilayers were ∼47 Å wide, the cytoplasmic appositions were ∼32 Å wide, and the extracellular spaces were ∼48 Å wide. The only significant difference between the myelin diffraction from WT sciatic and P-CNS optic nerves was the overall stronger intensity of the WT sciatic nerve. This difference reflects thicker myelin internodes (more lamellae) in the PNS than in the CNS. For optic nerves from PLP/P-CNS mice, the diffraction patterns were virtually identical to WT CNS myelin (), i.e., with the second and fourth Bragg orders being the strongest and located at positions corresponding to a 155-Å periodicity. The membrane profiles calculated from these patterns showed similar dimensions, with ∼47-Å-wide bilayers, ∼32-Å-wide cytoplasmic spaces, and ∼32-Å-wide extracellular spaces (). Thus, despite its known homophilic interactions in trans (; ), P appeared to have little affect on myelin spacing when expressed at equal amounts with PLP. The x-ray scatter recorded from PLP-null optic nerves () indicated expanded arrays of disordered membranes, with a 194-Å periodicity compared with the 155-Å period of WT optic nerve myelin. Consistent with the diffraction data, electron micrographs showed that the periodicity of compact myelin in P-CNS optic nerves () was identical to compact myelin in WT sciatic nerves () and greater than that of compact myelin in WT optic nerves () or PLP/P-CNS optic nerves (). To determine the effect of replacing PLP with P on motor function, we compared the performance of P-CNS mice on a standard rota-rod treadmill with that of WT, PLP/P-CNS, or PLP-null mice at 3, 6, and 12 mo of age (). At 3 mo of age, all strains of mice performed similarly. At 6 mo of age, the performance of WT, PLP/P-CNS, and PLP-null mice was indistinguishable, whereas that of the P-CNS mice was reduced by 70%. Compared with WT mice at 12 mo of age, the performance of P-CNS and PLP-null mice was reduced by 90 and 60%, respectively, whereas the performance of PLP/P-CNS mice was unchanged. We terminated these experiments at 12 mo of age because of the high mortality rate of the P-CNS mice. The mortality rate of P-CNS mice was compared with that of WT, PLP/P-CNS, and PLP-null mice (). There was no difference in the mortality rate of WT, PLP/P-CNS, and PLP-null mice. The mortality rate of P-CNS mice, however, was significantly increased. Approximately 50% of the P-CNS mice died by 12 mo of age. 80% were dead by 16 mo, and none lived past 18 mo, when 70% of the other strains were still surviving. Thus, replacing PLP with P in mouse CNS myelin causes the premature death of these mice. The reduced lifespan of P-CNS mice and the earlier demonstration of axonal pathology in PLP-null mice () prompted histological examination of the P-CNS mice for underlying neurodegenerative changes. We performed a detailed analysis of the amyloid precursor protein (APP) in the brains from the four lines of mice. APP detection is a reliable indicator of axonal pathology in primary myelin disease affecting PLP-null mice (; ) and humans with multiple sclerosis (). APP is only detected in axons with compromised axonal transport (). A dramatic increase in APP immunoreactivity occurred in the P-CNS brains compared with the brains from the other three lines (). This APP immunoreactivity appeared predominately as small ovoids. This is consistent with previous identification of APP accumulation in axonal swellings (; ). We quantified APP-positive ovoid densities in the cerebral cortex from 1-, 3-, 6-, and 12-mo-old mice. WT () and PLP/P-CNS () mice contained few APP-positive swellings. In contrast, APP swellings were abundant in the cortices from the P-CNS () and PLP-null () mice. The density of these swellings was greater in P-CNS cortices than in PLP-null cortices and thus correlated with the more severe neurological phenotype and reduced lifespan. A small but statistically significant increase in APP swellings was detected in P30 P-CNS cortices when compared with the other three lines. Compared with PLP-null cortices, P-CNS cortices contained approximately twice the number of axonal swellings at 3 and 6 mo of age and over three times the number at 1 yr. We confirmed the identification of APP-positive ovoids as myelinated axonal swellings using postembedding electron microscopic immunocytochemistry (). Transmission EM detected a dramatic increase in the number of organelle-filled axonal swellings in the cerebral cortex from 1-yr-old P-CNS cortices (). As described in PLP-null mice (; ), APP-positive swellings occurred predominantly at distal paranodes. To determine whether axonal pathology occurred throughout the neural axis in P-CNS mice, we quantified axonal swellings and axonal degeneration in the dorsal cervical spinal cord (corticospinal tracts) at 6 mo of age (). Myelinated axonal pathology was not detected in WT and PLP/P-CNS dorsal columns (not depicted) but was present in PLP-null () and P-CNS () dorsal columns. The density of swollen axons and myelinated axons undergoing Wallerian degeneration () was three times greater in P-CNS than in PLP-null mice. These and data not shown detected axon ovoids in the optic nerves, diencephalons, and brain stems of P-CNS mice. The data described in the previous paragraph indicate that replacement of PLP by P either accelerates the formation paranodal axonal ovoids or increases the number of axonal ovoids formed. Because axonal ovoids occur at paranodal regions, we investigated whether P-CNS mice have greater nodal densities than WT and PLP-null mice. Sections from month-old P-CNS and WT optic nerves were double labeled with antibodies specific for Na channels and caspr, molecules enriched in nodal and paranodal axolemma, respectively (; ; ). As described previously (), Na channels were clustered between caspr-positive paranodal regions in WT and P-CNS sections (). Quantification of Na channel cluster density detected a 50% increase in nodes of Ranvier in the P-CNS optic nerves (). This interpretation was confirmed by electron microscopic examination of optic nerve sections. Nodes of Ranvier were rarely detected in electron micrographs of WT optic nerves, consistent with long internodal lengths. In contrast, it was common to view short internodes in P-CNS optic nerve (). In addition, axon organelles had begun to accumulate at distal paranodes and form intra-axonal ovoids (). Nodal density was increased twofold in P-CNS optic nerves when compared with WT nerves. Because a previous study () reported similar internodal distances in PLP-null and WT mice, these observations establish that replacement of PLP with P induces a gain-of-function mutation that inhibits longitudinal growth of internodal CNS myelin. This generates more paranodes, which cause more axonal pathology, severe neurological deficits, and early death. Although short internodes can result from demyelination/remyelination, we saw no evidence of macrophage-mediated myelin stripping or asymmetric internodal lengths suggestive of segmental demyelination/remyelination. Some fibers were undergoing Wallerian degeneration, in which myelin breakdown is secondary to axonal degeneration. The purpose of this study was to investigate whether the shift from P to PLP during CNS myelin evolution was related to a new function of myelinating CNS glia. After genetic reversal of this shift in mice, we conclude that PLP expression in mammalian oligodendrocytes and/or CNS compact myelin has a neuroprotective benefit for axons. This conclusion is based on the severe neurological deficits, significant axonal degeneration, and dramatically reduced lifespans of mice whose oligodendrocytes express P instead of PLP. Mice expressing both PLP and P in CNS myelin had normal lifespans and no neurological disability or axonal degeneration. Therefore, the mere presence of P in CNS myelin does not appear to be responsible for the phenotypes in P-CNS mice. These data suggest that mammalian CNS myelination sets up an axonal dependency on glial trophic signals that requires PLP and that cannot be replaced by reintroduction of the ancestral CNS myelin protein P. P and PLP family members coevolved with myelinating cells in the CNS of fish where P mediates membrane adhesion of compact myelin (; ). The earliest PLP family members are in fact more closely related to the alternatively spliced DM20 isoform of mammalian PLP. PLP/DM20 evolved and coexisted with P in amphibian compact CNS myelin (). The positively charged 35–amino acid sequence that distinguishes PLP from DM20 () has been proposed to play an important role in stabilizing myelin membrane compaction and to permit the phenotypically silent dropout of P from terrestrial vertebrate CNS myelin (). We propose an alternate and/or additional hypothesis that highlights a CNS axon trophic role for PLP established during oligodendrocyte evolution. A role for PLP in maintaining the integrity and long-term survival of mammalian CNS axons was proposed previously, based on axon ovoids and axonal degeneration in the CNS of PLP-null mice (). In those studies, however, it was impossible to determine whether the axonal degeneration was a primary effect of PLP loss or a secondary response caused by alterations in the periodicity and/or stability of PLP-deficient myelin. This issue is resolved in the current study, where the neuroprotective effect of PLP was uncoupled from its role in CNS myelin compaction by stabilizing compact CNS myelin with P protein. Our Western blot analysis of myelin fractions isolated from transgenic and control mice indicated that total amounts of P and/or PLP were similar in P-CNS, PLP/P-CNS, and WT mice. When expressed with or without PLP, P was appropriately targeted to compact myelin. When overexpressed in oligodendrocytes, PLP becomes toxic and eventually kills the oligodendrocyte (; ). When overexpressed in Schwann cells, P is mistargeted to Schwann cell surface membranes and inhibits the spiral wrapping of myelin, which causes amyelination and severe neurological disability (; ). The axon ovoids described here in P-CNS mice are not present in either P- or PLP-overexpressing mice. In addition, oligodendrocyte death and amyelination did not occur in P-CNS or PLP/P-CNS mice. The phenotypes and pathologies of P-CNS mice result from changes in myelin protein composition and not from increased myelin protein dosage. The periodicity of compact CNS and PNS myelin differs, and each reflects molecular features of PLP and P, respectively (). The extracellular domain of P is larger than that of PLP, and thus the space between the extracellular leaflets and overall periodicity of PNS myelin are each ∼20 Å greater than that found in CNS myelin. Our immunocytochemical and Western blot studies demonstrate that when expressed by oligodendrocytes, P protein is targeted to CNS myelin. When P and PLP coexist in compact mouse CNS myelin, they have the periodicity of CNS myelin (). This was unexpected, as one might predict that obligate P homophilic adhesions (; ) would dominate putative weaker electrostatic trans-binding of PLP to charged lipids (). Because P homophilic adhesion may occur by trans-binding between cis-linked P tetramers (), it is possible that PLP interferes with P tetramer formation in cis. P monomers or dimers would have little effect on the periodicity of myelin because they cannot bind in trans. In contrast, in the absence of PLP, P tetramers would form and bind in trans, thus dominating the spacing of compact CNS myelin, giving it the periodicity of PNS myelin, as seen ultrastructurally and when examined by x-ray diffraction (). Lipid bilayer thickness and cytoplasmic spacing of CNS and PNS myelin are identical and were unchanged in P-CNS and PLP/P optic nerves. P, therefore, can replace but not compete with PLP as the major structural protein of CNS mammalian myelin. These observations support the concept that PLP evolved to serve functions unrelated to compact myelin formation. How does PLP maintain axonal integrity? Although it remains to be determined whether PLP plays a direct or indirect role, observations from the PLP-null and P-CNS mice provide clues to biological mechanisms. PLP-null mice have reduced anterograde and retrograde axonal transport that manifest as organelle accumulations at nodes of Ranvier (; ). Mice chimeric for the X-linked PLP-null allele (50% of myelin internodes lack PLP) have numerically 50% of the axonal ovoids present in PLP-null mice (). This implies that PLP facilitates paranodal axonal cytoskeletal organization at the level of individual myelin internodes. Unraveling the precise mechanism by which PLP provides tropic support to the axon presents a significant challenge as the axonopathy evolves over several months and is likely to involve altered molecular complexes that organize the axon cytoskeleton in paranodal regions of myelinated fibers. Based on the chimeric mouse data cited above, a generalized axonal defect or altered neuronal gene expression seem unlikely causes of initial axonal pathology in PLP-null mice. This issue is further complicated as a null mutation in another myelin-specific protein, CNP, which produces axon ovoids similar in appearance to those in PLP-null mice. Like the P-CNS mice, CNP-null mice have earlier and more extensive axonal pathology, decreased motor performance at earlier ages, and reduced life spans (). Some CNP-null mice have enlarged ventricles (), a feature not present in P-CNS or PLP-null mice. Whether hydrocephalus contributes to or results from axonal loss in CNP-null mice remains to be determined. Ovoids are not the only myelin-induced axonal pathology, as axonal atrophy occurs in MAG-deficient mice (). These changes also have late onset (after 6 mo), manifest in paranodal axoplasm, and include reduced phosphorylation and spacing of neurofilaments (). Because P-CNS mice have more axonal pathology than PLP-null mice, the PLP–P shift in CNS myelin causes a gain-of-function mutation that accelerates and/or increases focal reductions in axonal transport caused by PLP loss of function. Immunocytochemical and electron microscopic examination of myelin internode length in optic nerves established that P-CNS mice have shorter myelin internodes, resulting in a greater density of paranodal specializations, which increased and accelerated alterations in axonal transport, formation of axonal ovoids, axonal degeneration, and neurological disability. The addition of the positively charged 35–amino acid sequence that distinguishes PLP from DM20 () may play a crucial role in axonal trophic support, as DM20 was unable to rescue the late-onset axonopathy in PLP-null mice (). Furthermore, it is possible that this 35–amino acid sequence evolved solely for this function, as this PLP-specific sequence shows no significant homology with any other mammalian protein. The data presented here have important implications for human diseases of myelin, as a spectrum of neurological disabilities is associated with null mutations, duplications, and various point mutations in the PLP gene (; ). Recent studies have reported axonal degeneration in individuals with PLP deletions or point mutations (). Genotype–phenotype correlations focusing on axonal degeneration and point mutations may identify specific regions of PLP that mediate axonal survival in the human CNS. Such information may lead to the development of novel neuroprotective therapies in inherited and acquired diseases of CNS myelin, such as Pelizaeus-Merzbacher disease and multiple sclerosis. We generated transgenic mice that expressed the mouse P cDNA () ligated to a 9.1-kb region of the mouse MBP promoter (). These mice were maintained in the animal colony as homozygous animals, and they displayed no obvious behavior abnormalities. null mice () were crossed with homozygous mice carrying the MBP-P transgene. These mice were interbred for three generations to obtain mice that were PLP-null and homozygous for the MBP-P transgene (P-CNS), as determined by genomic DNA analysis and outbreeding. DNA was prepared from tail clips by standard proteinase K/phenol chloroform extraction (). Samples were analyzed by Southern blot or PCR to track the P transgene and the gene, respectively. For Southern blot analysis, 10 μg of tail DNA was digested with EcoRI overnight and then separated on 1% agarose gels. Gels were blotted and then probed with a random primed 1.8-kb P probe (). WT DNA contained four bands, and additional novel bands were found with the P transgene, most prominently a diagnostic 1.8-kb fragment. The PLP-null mutation was analyzed by PCR, which generated a 620-bp fragment from WT DNA and a 542-bp fragment from PLP-null DNA. The PCR program was 94°C × 3 min; 40 cycles of 94°C × 30 s; 55°C × 30 s; 72°C × 30 s; and 72°C × 7 min. The PCR primers used were 5′-ACGAGCAGTGAGAGTTGGGT-3′ and 5′-AGTCTGTTTTGCGGCTGACT-3′. Total RNA was prepared from P60 brains using the RNAgents total RNA isolation system (Promega). 5-μg samples were separated on 1% agarose gels and probed with P or PLP cDNA as described previously (; ). Myelin was prepared from homogenates of P60 brains from WT, PLP-null, PLP/P-CNS, and P-CNS mice according to established procedures (). In brief, brains were homogenized in 0.32 M sucrose and myelin was prepared by differential and gradient centrifugation as a fraction that floats on 0.88 M sucrose. This fraction was osmotically shocked and centrifuged. Samples were analyzed on 12% polyacrylamide gels and stained (10 μg of protein) with Coomassie blue or blotted (2 μg of protein) for Western analysis with PLP/DM20 antibody (clone AA3; a gift from S. Pfeiffer, University of Connecticut Health Science Center, Farmington, CT), P antibody (), MAG antibody (a gift from R. Quarles, National Institutes of Health, Bethesda, MD), CNP, or MBP antibodies (Sternberger Monoclonals). To access motor function, we used the rota-rod test as described previously (). Five mice from each strain were tested at 3, 6, and 12 mo of age. All mice were placed on the roller at the initial speed of 2 rpm. The speed was continuously increased over a period of 5 min to a final rate of 20 rpm. The time each mouse remained on the roller was recorded, and differences between strains were determined by performing a test. A Kaplan-Meier analysis of mortality was conducted for a minimum of 30 mice from each strain. WT, PLP-null, PLP/P-CNS, and P-CNS mice were perfused with 4% paraformaldehyde and 0.08 M Sorenson's phosphate buffer. Brains were removed, immersion fixed overnight, cryoprotected, frozen, and sectioned (30 μm thick) on a sliding microtome. Free-floating sections were treated with PBS containing 3% normal goat serum and 1% Triton X-100 for 1 h at room temperature. Sections were incubated overnight at 4°C with PLP/DM20 antibody (1:100) or P antibody (1:800) in PBS containing 1% normal goat serum/0.01% Triton X-100. Sections were developed with the avidin biotin complex kit (Vector Laboratories) and diaminobenzidine as described previously (). Sections were photographed in a photomicroscope (Axiophot; Carl Zeiss MicroImaging, Inc.) with standard 2.5–20× dry lenses using a Magnafire (Optronics) digital camera and software. For analysis of APP accumulation, three mice from each strain were processed at 1, 3, 6, and 12 mo of age. Sections were immunostained with APP antibodies (Zymed Laboratories) as described above. The density of APP-positive axonal swellings was quantified in three sections from each mouse at all ages. Differences in APP densities were determined by test. To identify APP-positive structures as axonal swellings, several APP-stained 30-μm sections from P-CNS mice were processed for EM as described previously (). Sections were viewed and photographed in an electron microscope (CM-100; Phillips). Light and electron microscope images were prepared for publication using Photoshop 7.0 software (Adobe). The distribution of P protein was also determined at the ultrastructural level. P30 mice were perfused with 2.5% glutaraldehyde, 4% paraformaldehyde, and 0.08 M Sorensen's phosphate buffer. The ventral cervical spinal cord was removed and infiltrated with 30% polyvinlpyrolidine and 2.3 M sucrose. The tissue was cut in a Ultracut S ultracryomicrotome (Reichert) maintained at −110°C. Sections were placed on carbon- and formvar-coated grids and immunostained with P antibodies and immunogold procedures as described previously (). Sections were examined in an electron microscope. Three WT, PLP-null, PLP/P-CNS, and P-CNS mice were perfused with 2.5% glutaraldehyde, 4% paraformaldehyde, and 0.08 M Sorensen's phosphate buffer at 1, 3, 6, and 12 mo of age. Optic nerves, coronal slices of cerebral cortex, and cervical spinal cords were removed, placed in fixative overnight, osmicated, dehydrated, and processed to Epon 812 as described previously. 1-μm-thick sections were cut on glass knives in an Ultracut Eultramicrotome, mounted on glass slides, stained with toluidine blue, and examined on a light microscope (Axiophot; Carl Zeiss MicroImaging, Inc.). Transmission EM was performed on select Epon blocks. Sections were cut on a diamond knife, placed on formvar-coated grids, stained with uranyl acetate and lead citrate, and examined in an electron microscope. Optic nerve sections were obtained from all four lines and compared with sciatic nerves from WT mice. The periodicity of compact myelin was determined. The nature and density of axonal ovoids were determined in thin sections from the cerebral cortex blocks of 12-mo-old mice. In addition, the densities of axonal swellings and/or degenerating myelinated fibers in the cortical spinal tract of 6-mo-old mice were compared by test. Sciatic and optic nerves were dissected from WT and transgenic mice that had been killed by decapitation. During dissection, the tissue was continually rinsed with physiological saline (154 mM NaCl and 5 mM Tris buffer, pH 7.4). Nerves were tied off at both ends with fine silk suture and inserted into medium-containing quartz capillary tubes, which were then sealed at both ends with wax. Diffraction experiments were performed as described previously () using nickel-filtered, single-mirror-focused CuKα radiation from a fine-line source on a 3.0-kW Rigaku x-ray generator (Rigaku/MSC, Inc.) operated at 40 kV by 16 or 22 mA. The x-ray diffraction patterns for each nerve were recorded first for 10 min (to assess myelin integrity) and then for 2 h using a linear, position-sensitive detector (Molecular Metrology, Inc.). The myelin periodicity was determined from the positions of the peaks (or Bragg orders) in the diffraction patterns, and the membrane profiles (from which the intermembrane spacings at the extracellular and cytoplasmic appositions and membrane bilayer thickness were measured) were calculated by Fourier synthesis from the intensities of the Bragg orders after background subtraction (). The profile shows a periodic fluctuation in electron density through the multilamellar stacking of the myelin membranes. High electron density corresponds to the positions of the lipid polar head groups, and the low density trough corresponds to the centers of the lipid hydrocarbon. The intermediate level of electron density between the membranes indicates protein and water in the spaces at the extracellular and cytoplasmic appositions. Thus, peak- to-peak distances measured off the profiles relate to structural parameters of the membrane structure and its packing. Density of nodes was quantified in two ways in WT and P-CNS optic nerves. Free-floating 30-μm sections of an optic nerve from each of three 1-mo-old WT and P-CNS mice were double immunostained as previously described () using antibodies against Na channels (mouse pan Na channel; Sigma-Aldrich) and caspr (rabbit antibody; a gift from J.A. Trimmer, University of California, Davis,. CA), a paranodal marker. Six nonadjacent areas midway between retinal and chiasmal ends of the optic nerve were imaged using a confocal microscope (TCS-NT; Leica) and software. The fields were collected as a z series of six slices covering 6-μm tissue thickness using 100× (1.3 NA) lens, and nodes were counted within an area 50 × 50 μm in the projected image. Z series projection was necessary for interpretation, as in single slices it was sometimes difficult to unequivocally interpret single spots of staining. Nodal density was also determined by EM. Optic nerves from three WT and three PLP/P mice aged 1 mo were embedded and mounted on Gilder 300 mesh ultrahigh transmission hexagonal grids (hexagonal edges were 43 μm long, and the grid area was 4,800 μm). The number of nodes in each of six grid squares was counted per optic nerve.

Tom Rapoport (Harvard University, Boston, MA) presented his group's efforts to understand how organelle shapes are formed and maintained. The findings identify two protein families that create tubules out of ER membrane. A sphere is the most stable membrane shape; deformations that increase membrane curvature cost energy and must be actively stabilized. Rapoport and colleagues, including Gia Voeltz, used an in vitro system for ER network formation to identify factors that create curved membrane shapes. With just membranes, salt, and GTP, their system produces a network of ER tubules. This simple, self-contained system was not ideal for identifying the shape-creating components. “When we first saw [that] everything was already in the membrane,” said Rapoport in his talk, “we thought, ‘what a bummer.’” But the group found a way around this difficulty by using small molecule inhibitors to block in vitro ER formation, and then identifying the inhibitors' targets. One such target was an integral membrane protein called Reticulon4a (Rtn4a), previously named for its localization to ER membranes. All eukaryotes express at least one homologue of Rtn4a, and the proteins are the first known markers specifically localized to the tubular ER and absent from sheets. Cells overexpressing Rtn proteins formed more tubules, but loss of the two yeast members did not prevent tubule formation under normal conditions. Only when mutant cells were subjected to osmotic stress were their tubules lost. Rtn proteins form homo- and hetero-oligomers, so the group figured that another Rtn-interacting protein might be required for tubule formation. Indeed, they found that Rtn pulled down another ubiquitous integral membrane protein called DP-1. Loss of both the yeast DP-1 and the more abundant of its two Rtns now blocked tubule formation. The group has proposed that Rtn and DP-1 might be wedge-shaped, with their wider sides in the outer membrane leaflet. The presence of these proteins would thus favor a highly curved membrane. They now plan to test whether purified Rtn and DP-1 can turn liposomes into tubules. Reference: Transport within a cell occurs with the help of three classes of motor proteins: myosin, kinesin, and dynein, which carry their cargo along cytoskeletal tracks. Conventional kinesin and myosin V move processively—that is, they remain bound to their tracks for many steps. But the minus end–directed motor dynein, because of its large size and many ATP binding sites, has been difficult to study. Samara Reck-Peterson (University of California, San Francisco, CA) described recent experiments that demonstrate that, despite major structural differences, dynein's stepping mechanism has at least some similarities to that of kinesin and myosin. Using Reck-Peterson and colleagues, led by Ronald Vale, engineered a recombinant version of dynein that could be tagged at various positions to image its molecular motion at the single molecule level in vitro. This technique allowed the researchers to visualize directly dynein's ability to move processively. As single monomers of the construct were not processive, Reck-Peterson showed that dynein achieves its processivity as a dimer. Previous studies showed that both conventional kinesin and myosin V walk in a “hand-over-hand” manner, with step sizes that are based on the distance between binding sites on the microtubule (∼8 nm) or the actin filament (∼37 nm). Dynein, which is an AAA ATPase, differs both evolutionarily and structurally from kinesin and myosin. By measuring the change in position of a single-labeled head and of a central portion of the protein, Reck-Peterson and colleagues found that dynein, too, coordinates the action of its two motor domains and takes approximately 8-nm steps. As rings on some AAA ATPases arrange in stacks, Reck-Peterson suggested that dynein might achieve these small steps by stacking and shuffling the position of the two rings of the dimerized protein. Reference: Formins nucleate new actin filaments and accelerate polymerization rates. Elizabeth Harris (Dartmouth University, Hanover, NH) presented her studies from Henry Higgs lab that reveal a new function for mammalian formins—the ability to bundle existing filaments. Formin's actin activities rely on its dimeric, donut-shaped FH2 domain. This domain, Harris showed, is sufficient for bundling for the mDia2 and FRL1 formins. mDia1, in contrast, had no bundling ability. According to current models, the FH2 dimer sits on actin barbed ends as a washer would on the tip of a screw. This interaction did not seem like the sort that would lead to bundling, so Harris wondered whether formins might also bind further down actin filaments. Indeed, she found that dimers of the two bundling formins bind in a one-to-one ratio with actin subunits in the filament. FH2 dimers are thought to be very stable, based on structural and biochemical studies on the yeast formin Bni1p. Harris wondered how so many dimers could bind at the tip and slide down the filament quickly enough to account for their rapid bundling kinetics. She has now found that the FH2 domains of mDia2 and FRL1, unlike Bni1p (and mDia1), are able to dissociate and reassociate. Harris imagines that the unstable FH2 domains are able to clamp an actin filament anywhere along its length as the monomers reassociate. Her in vitro experiments support this idea for FRL1, but suggest that mDia2 acts differently. As her theory predicts that barbed-end binding and side binding should be competitive (because both should require interactions with the same residues), Harris increased the concentration of barbed ends. The bundling activity of FRL1, as predicted, decreased accordingly. But the same was not true of mDia2. And mutations in the inner ring of mDia2's donut—where barbed-end binding occurs—did not interfere with its bundling activity. Thus, mDia2 might use residues on the outside of the donut to bind to and bundle actin filaments. Two other formins (yeast Bnri1p and an formin) have also been shown to bundle in vitro. Still, the field has yet to prove that formins bundle in cells. But if the test tube studies hold true, it may be that formins rearrange actin networks even as they assemble the filaments (as in filopodia, for example). More conventional bundlers, such as fascin, might then be brought in to stabilize the networks. Reference: In the past few years, several mutations causing polycystic kidney disease (PKD) were shown to affect cilia proteins in both human cases and mouse models of the disease. But the functional link between epithelial cilia and the overgrowth of kidney tissue that marks this disorder was not clear. Thomas Weimbs (University of California, Santa Barbara, CA) reported that cellular overgrowth in PKD is driven by a transcription factor after it escapes from an unusual sequestration site—epithelial cilia. Weimbs and colleagues were studying the cellular effects of dominant-negative versions of a PKD-associated protein called polycystin-1 (PC-1). Although PC-1 is a cilial integral membrane protein, one of the dominant-negative versions was found in the nucleus. A nuclear PC-1 fragment also occurs naturally—expression of full-length PC-1 in cell culture resulted in its proteolytic cleavage and the nuclear accumulation of the cytoplasmic tail. In separate immunoprecipitation experiments, the group had discovered that PC-1 interacted with a transcriptional coactivator called P100. The interaction was also seen by colocalization at basal bodies and in cilia. One of P100's well-known partners is the proliferation-inducing transcription factor STAT6. The group now finds that, under normal kidney apical flow conditions, STAT6 localizes to epithelial cilia along with PC-1. But when they disrupted flow, as might occur upon kidney injury, STAT6 and PC-1 moved to the nucleus. Weimbs' model is that cleavage of PC-1 is flow-sensitive, such that the loss of flow frees the cytoplasmic domain. This portion then moves to the nucleus, bringing along its bound P100 and STAT6 and inducing the gene expression needed for wound healing. Indeed, overexpression of the cytoplasmic PC-1 fragment induced STAT6-dependent gene expression in cell culture and caused renal cysts to form in zebrafish embryos. As the team found that PKD patients had high levels of nuclear PC-1, STAT6, and P100, the disease might be due to constitutive activation of this injury response pathway. Small molecule inhibitors, perhaps of the as-yet unidentified protease that cleaves PC-1, might therefore be promising PKD therapeutics. Reference: The start of mitosis involves major architectural changes in the cell, culminating in the construction of a mitotic spindle. Yixian Zheng (Carnegie Institute of Washington and Howard Hughes Medical Institution, Baltimore, MD) proposed that an interphase nuclear lamina protein called Lamin B gives the spindle crucial structural support as it takes shape in mitosis. To understand spindle assembly, Zheng's group turned to nuclear components, reasoning that they may have become fixed during the evolution of the eukaryotes because they helped the cell to divide. Indeed, her team and others have identified a handful of nuclear proteins that are important during mitosis. In earlier work, Zheng's team and others found that Ran, a GTPase required for nuclear trafficking during interphase, forms a gradient along the chromosome that kick-starts mitosis by stimulating spindle microtubule assembly. In addition, Zheng's lab showed that Ran also activates a mitotic kinase, called Aurora A, which regulates microtubules to form the spindle with the help of other proteins. Using a novel assay for spindle assembly involving the coupling of Aurora A antibodies to magnetic beads, the group's new experiments in egg extracts show that Ran has another regulatory function in spindle assembly: it regulates Lamin B during mitosis. Lamin B localized to the spindle, and depleting it inhibited spindle formation. But Lamin B did not bind to microtubules, suggesting that it affects spindle assembly indirectly. In the nucleus, lamins form a filament system that is important for various nuclear functions. The researchers found that Lamin B may also have a structural role during mitosis, as it assembled around the spindles as a polymer. “A mitotic scaffold has been hypothesized for decades, but molecular components have remained elusive,” Zheng said in her talk. Lamin B may be just one of such components. It probably acts on spindle assembly factors, tethering them in place so that mitosis can proceed. Reference: What do fly photoreceptor clusters and wing cells have in common with vertebrate inner ear hair cells and convergent extension? In all cases, tissue function depends on orderly cell packing and planar cell polarity (PCP). All four also require a highly conserved gene cassette encoding PCP proteins. Anne-Kathrin Classen (Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany) suggested how PCP proteins control such diverse processes by directing vesicle transport. The fly's epithelial wing cells reorganize into a perimeter-minimizing hexagonal packing arrangement at the pupal stage. Observing the process with confocal microscopy, Classen and colleagues, led by Suzanne Eaton, found that repacking occurred via E-cadherin–mediated remodeling of cell contacts at intercellular junctions. Remodeling requires both E-cadherin internalization and its deposition via exocyst vesicles. PCP proteins, which are located near the cell membrane, attracted exocyst vesicles carrying E-cadherin and thus directed how the cell remodeled its contacts. Mutant cells lacking PCP proteins formed fewer new boundaries with neighboring cells and had impaired hexagonal repacking. When cells orient, they must make certain connections and release others. “The unifying feature [of PCPs] may be to regulate the polarized trafficking of exocyst components,” Eaton says. The cargo that exocysts carry could differ in each case. For example, whereas the cargo for convergent extension and wing cell packing should include cadherins, signaling proteins such as Delta may be the relevant cargo for orienting photoreceptor clusters. Reference: It's old news that, during cytokinesis, actin and myosin accumulate at the cell's equator, forming a contractile ring that pinches the cell together like a purse string to divide it in two. What is less well-known is that these proteins localize and constrict asymmetrically. Ingression starts on one side of the division plane and progresses almost to its center before the other side starts to furrow in as well. Amy Maddox (University of California, San Diego, CA) reported that this asymmetry is generated by two conserved components of the contractile ring: anillin and the septins. The group, led by Karen Oegema, had previously observed this asymmetry and surmised that it might be important for cell division. First, they imagined, one-sided ingression may rapidly create a barrier between the two masses of segregated chromosomes, ensuring that both do not end up in the same daughter cell. Second, in dividing epithelial cells, the upwards ingression of a cleavage furrow from the basal membrane might allow apical cell–cell junctions to remain intact. In the present study, Maddox and colleagues imaged dividing cells and found that the septins, anillin, and actomyosin localize asymmetrically around the circumference of the contractile ring. Depleting cells of septins or anillin via RNAi left cytokinesis intact—but symmetrical. Without these proteins, actin and myosin were evenly distributed around the contractile ring, which contracted isometrically to cleave the cell. Further experiments showed that, whereas actin and myosin are required for cytokinesis, anillin and the septins seem to make the process more robust. When the researchers weakened the contractile ring by partially compromising myosin function, the cell was still able to divide. However, when anillin was depleted from this sensitized background, cytokinesis failed and the cells became multinucleated. Septins are GTP-binding, filament-forming proteins that were recently demonstrated to form a structural meshwork in the cell cortex. Anillin is a structural protein that appears to stabilize the cytoskeleton during cell division. Maddox suggested that, as anillin can bind to septins, actin filaments, and myosin, it may cross-link them, perhaps helping to arrange actin and myosin in relation to each other both to promote their uneven distribution around the contractile ring and to increase the efficiency of constriction. Reference: Membranes, microtubules, and actin might be connected via an Arp2/3 activator. Matt Welch (University of California, Berkeley, CA) described this newly identified actin nucleation activation factor, which he calls WHAMM, for WASP homology actin microtubules and membranes. Known Arp2/3 activators include the WASP and WAVE/Scar families. In flies, mammals, and yeast, Arp2/3 and many of its known activators are essential proteins. Since they are so important to survival, Welch wondered if other regulators exist. For his search, Welch looked to the Arp2/3-binding region (called the WCA domain) that is common to WASP and WAVE proteins. He then used a bioinformatics approach to identify other evolutionarily conserved proteins containing WCA sequences. Among others, his search turned up WHAMM. As Welch had hoped, WHAMM promoted actin filament nucleation in vitro by activating Arp2/3. But he wondered what makes WHAMM necessary, as WASP and WAVE proteins are already doing this same job. The subcellular localization of the new protein gave him some ideas. Although WAVE is located at the plasma membrane, where it regulates cell migration, WHAMM was found in a perinuclear compartment that might correspond to Golgi cisternae or late endosomes. It was also found on as-yet unidentified tubular membrane structures. Domain-mapping experiments showed that, in addition to a domain for perinuclear membrane binding, WHAMM also contains a domain that interacts with microtubules. When expressed alone, this coiled-coil domain bundled and rearranged microtubules. Welch has yet to show that the full-length protein can also reorganize microtubules in vitro, but it does align with microtubules and with actin filaments in cells. The binding of WHAMM to microtubules and to membranes might regulate its Arp2/3 activation function. If so, WHAMM might theoretically coordinate transport to and from the Golgi, for example, by synchronizing membrane movements with cytoskeletal rearrangements. Reference: In many species, chromatin signals can induce spindle assembly. But whether features of spindle shape depend on the chromatin itself or on other components of the cytoplasm has remained under debate. Jedidiah Gaetz (Rockefeller University, New York, NY) described new results showing that several properties of spindle shape are indeed regulated independently of chromatin. Gaetz and colleagues, headed by Tarun Kapoor, placed DNA-coated magnetic beads into a magnetic field to construct chromatin strands in cell extracts missing key organizing components such as centrosomes and kinetochores, and imaged spindle assembly in this simplified system. Spindle length and organization did not change, regardless of the length of the chromatin, suggesting that the chromatin signal merely induces assembly, leaving the details to other players. Some of these players may be microtubule-associated proteins, such as dynein, which have been linked with spindle assembly. Previous work from the group showed that dynein, a microtubule minus end–directed motor, controls spindle length by regulating tubulin subunit depolymerization at the poles. In the present study, Gaetz reported, inhibiting dynein led to spindle fibers extending as strands along the entire length of the chromatin, their minus ends failing to join up to form a pole at all. Dynein is known to grab microtubules and pull them along other microtubules to bring their minus ends together. The researchers thus speculate that dynein is depleting surrounding regions of microtubules or microtubule associated proteins to ensure that spindle organization occurs in just one place. Reference: With too few genes to set up its own export system, viruses deftly appropriate their host cell's endosomal machinery to get themselves repackaged in vesicles and shipped on to the next cell. Most viruses hook into the process via one of a series of protein complexes that form a membrane-bound lattice essential for multivesicular body formation, but the mechanism of subsequent viral budding is poorly understood. Earlier research by Wes Sundquist (University of Utah, Salt Lake City, UT) showed that HIV packaged itself for export by recruiting an early-acting lattice component called ESCRT-I. In his talk, Sundquist described new research on a novel HIV budding pathway that hooks into later lattice steps that depend upon the ESCRT-III complex and an AAA ATPase protein called Vps4. Structure and function studies of Vps4 suggested that, when the protein is activated by ATP, it forms hexameric rings that are recruited to the site of vesicle formation through interactions with ESCRT-III proteins. Sundquist's team then compared the structure of Vps4 to other proteins in the AAA ATPase family, including pore pumps that denature DNA and RNA, and bacterial chaperones that denature misfolded and aggregated proteins. Based on this homology, Sundquist proposed that Vps4 similarly acts as a protein pump, binding to ESCRT-III proteins and pulling them up into the narrow central pore, where they are unfolded. When Vps4 was mutated by disabling a tag on its inner pore, HIV budding was strongly impaired, suggesting that ESCRT-III proteins must travel up into the Vps4 pore for viral escape. The group is now studying how exactly ESCRT-III is coupled to vesicle formation. One possibility is that the release of ESCRT-III subunits may be necessary for membrane fission. “We are starting to understand what parts of Vps4 are important,” says Sundquist. “But we can't say in mechanistic terms why ESCRT-III has to be unfolded for budding to take place, because we don't understand how you make a vesicle.” Reference: Premature aging and abnormal DNA methylation patterns both occur in cloned animals such as Dolly the sheep. In a talk by Dale Shumaker (Northwestern University, Chicago, IL), aging and histone methylation were linked by the disease progeria. Progeria—the rare and mysterious disease of rapid childhood aging—is caused by dominant-negative mutant versions of the nuclear intermediate filament protein lamin A. Many nuclear morphological defects arise in cell lines that express this mutant lamin A: pore complexes cluster, the lamina thickens, lobulation occurs, and heterochromatin is lost. But exactly which consequence of the lamin mutation leads to disease is far from untangled. Shumaker and colleagues, working in Robert Goldman's laboratory, now find that histone methylation abnormalities appear long before these nuclear changes take place. As heterochromatin is lost over time in progeria cells, the group examined whether specific chromatin modifications, such as histone methylation, might also be altered. Changes were measured by monitoring methylation on the inactive X chromosome. The group found that fibroblasts isolated from a progeria patient had fewer H3K27me3 methylation marks on the inactive X chromosome. Transcript levels of the EZH2 transmethylase responsible for this epigenetic mark were strongly reduced in progeria cell lines. Lamins that comprise the lamina lining the inner nuclear membrane, as well as those dispersed throughout the nucleoplasm, act as a platform that supports and regulates DNA replication and Pol II–mediated transcription. Shumaker hypothesized that they might also support the multisubunit methylase complex. Mutant lamin A in progeria cells is known to remain inappropriately farnesylated, and thus has an unusually high affinity for the nuclear membrane (most likely causing the thickened lamina of progeria nuclei). If the lamina is indeed a large molecular scaffold, these physical changes might impair EZH2 function, possibly leading to negative feedback of EZH2 transcription. How methylation and heterochromatin influence aging, however, remains to be seen. Reference: David Ron (New York University, New York, NY) is dealing with stress. In his talk, Ron discussed two ways that cells cope with the stress of unfolded or misfolded proteins in the ER. In the first part of his talk, Ron showed that mutant fibroblasts lacking the CHOP transcription factor are better able to survive ER stress that would kill normal cells. As CHOP activates apoptotic pathways, the group figured it killed normal stressed cells via its induction of death effectors such as Bax and Bak. But instead they found that CHOP counteracted the translational dampening that ER stress induces. Normally, translation is down-regulated by stress to give the ER a bit of time to deal with the excess unfolded proteins. A negative feedback loop then leads to translational recovery. This loop, Ron showed, includes CHOP-mediated induction of a phosphatase that reactivates eIF2α. The “lazy” CHOP knock-out cells, as Ron called them, keep translation down longer, and thus survive acute stressors such as tunicamycin because they have fewer proteins to fold. He speculated, however, that CHOP mutations might not be helpful under a more physiological stress such as wounding, which requires the translational up-regulation of collagen to help in healing. Another protein that helps cells control the ER protein load, Ron showed, is P58. His group has found that P58 mutants turn on the unfolded protein response even in the absence of other stresses. P58 is part of a family of cochaperones that work with Hsp70 proteins, and it is associated with the cytoplasmic side of the translocon. Ron proposed that P58 monitors the normal passage of ER proteins through the translocon. Those that stall as they pass through (due to unfavorable lumenal ER conditions, for instance) would be exposed to P58, which might lead to their extraction from the translocon and degradation. To this effect, Ron showed that the degradation of at least one stalled protein, ApoB, requires P58. Reference: A family of zinc finger proteins plays matchmaker for worm chromosomes, as presented by Abby Dernburg (University of California, Berkeley, CA). Her work reveals site-specific DNA binding proteins that pair up homologues during meiosis. Homologues must first come together before they can be equally distributed to daughter cells during meiosis. Pairing centers are the chromosomal sites that, in worms, are necessary for this coupling. Dernburg's group has now identified the trans-acting factors that recognize these sequences and make the correct matches. The findings were led off by the group's studies of a mutation that results in the birth of mostly male worms. This mutation created segregation defects due to ineffective meiotic pairing and synapsis. But unlike most pairing mutants, the problems did not extend to the other chromosomes. Only X chromosome segregation was affected, which got Dernburg excited. Her group then cloned and determined that it encodes a Zn-finger protein, perfect for interacting with DNA. As expected, HIM-8 concentrated at the X chromosome's pairing center. It also had an unexpected nuclear envelope localization. Dernburg imagines that the nuclear envelope might be a scaffold for pairing and synapsis. Her more recent work has revealed that relatives of HIM-8 pair the other worm chromosomes. Some even control two different sets of homologues, which implies that initial pairing by HIM-8 relatives is not the final word. In her talk, Dernburg hypothesized that pairing centers and the Zn-finger proteins “create stable intermediate states where chromosomes assess their homology.” But these pairs are readily reversible in case the wrong sets are brought together. Reference: During cytokinesis, membrane must be quickly deposited at the scission site to close the gap between the daughter cells, often resulting in excess membrane accumulation at the site. Jayne Squirrell (University of Wisconsin, Madison, WI) presented new data suggesting that this membrane deposition involves proper ER organization by the protein CAR-1. The sequence of suggests that it encodes an RNA-associated protein, but CAR-1 has also been implicated in cytokinesis. Its homology to a suppressor of clathrin deficiency in yeast, along with its role in cytokinesis, suggested that this protein might be involved in membrane trafficking during cell division. When Squirrell and colleagues, led by John White, depleted CAR-1 in , they found that embryos failed to complete the last stage of cytokinesis and that membrane did not accumulate at the scission site. These CAR-1–depleted embryos had other problems, including defects in ER dynamics. Recent studies identified an arrangement of ER around the spindle and centrosomes during cytokinesis in worm and fly embryos. CAR-1 depletion led to the disruption of this structure. Further studies showed that CAR-1 localized to the spindle region with a similar distribution to that of the ER. The group is further investigating CAR-1's association with the ER and the function of the ER at the spindle. Squirrell suggested that CAR-1 may be one of the organizing forces behind this ER arrangement, which might contribute to the maintenance of membrane at the scission site. Additionally, the two may work in tandem to stabilize the spindle structure. The ER is a regulator of calcium levels, and calcium causes spindle instability. “Perhaps the ER regulates calcium levels in a very local region for maintaining proper spindle structure at the right time,” Squirrell said. Reference:

xref #text By screening for signaling targets of repulsive axon guidance factors, semaphorins (described in Materials and methods), we identified a novel GTPase and named it CRAG, after collapsin response mediator protein (CRMP)–associated molecule (CRAM [CRMP-5])–associated GTPase. The full-length CRAG cDNA presents an open reading frame of 369 amino acid residues containing a glutamine-rich domain at the NH terminus, a Ras homology domain in the middle, and an NLS at the COOH terminus (). indicates a comparison of structure between CRAG and other related GTPase proteins. The amino acid sequence of CRAG shows 95% identity with centaurin-γ3 and 43% identity with the nuclear GTPase phosphatidylinositol 3-kinase enhancer, short isoform (alignment is not depicted; ). Analysis of human genomic databases suggests that CRAG may be an alternative splicing variant of centaurin-γ3. Northern blot analysis indicated that the CRAG gene was dominantly expressed in brain and slightly in heart (). A band higher than CRAG that is present in various tissues may be centaurin-γ3. Immunohistochemical analysis revealed a diffuse cytoplasmic distribution of CRAG in rat hippocampal neurons at rest (, top). Upon UV irradiation, an NI formed of CRAG was detected at 10 min (). These NIs exhibited a doughnut shape under the large-scale microscopic analysis (Fig. S1 A, available at ). This phenomenon was reproduced in UV-stimulated HeLa cells expressing HA-tagged CRAG (HA-CRAG) wild type (WT) or GTPase-deficient mutants (S114N; ). In contrast, NLS-disrupted mutants of CRAG (KR342-343EE within the NLS motif) formed inclusions but were cytosolic even after UV stimulation. These results demonstrated that NLS, but not GTPase activity, was required for nuclear translocation and NI formation by CRAG. We found that GFP fused to the NH terminus of CRAG WT or GTPase mutants spontaneously translocated to the nucleus and formed NI without any stimulation. The absence of nuclear localization of NLS-disrupted mutants of GFP-CRAG also supported an NLS-dependent NI formation by CRAG (). This suggested that GFP fusion at the NH terminus of CRAG caused a drastic conformational change in CRAG and converted it from an inactive to an active form. An in vitro GTPase assay revealed CRAG activation by UV irradiation and spontaneous activation of GFP-CRAG (unpublished data). Collectively, these data demonstrate that CRAG is a unique GTPase that forms NI under UV stress in an NLS-dependent and GTPase-independent manner, and a conformational change in CRAG may induce its intrinsic activity to form NI in an NLS-dependent manner. Several nuclear domains have been reported, including PML body, a major component of nuclear bodies (; ). Merged images demonstrated the colocalization of endogenous CRAG and PML bodies with an enlarged ring-like structure in a UV- stimulated dorsal root ganglion (DRG) neuron ( and Fig. S1 B). Similar results were obtained in UV-stimulated HeLa cells expressing HA-CRAG or unstimulated cells expressing GFP-CRAG. In contrast, HA-CRAG GTPase mutants or GFP-CRAG GTPase mutants did not colocalize with PML. Statistical analysis indicated that GTPase activity was essential for colocalization of GFP-CRAG with PML. A coimmunoprecipitation assay demonstrated that endogenous CRAG associated with PML in UV-stimulated neurons (). No association of CRAG with PML by antigen peptide against anti-CRAG antibody demonstrated the specificity of anti-CRAG antibody. Furthermore, in HeLa cell expression system, GFP-CRAG associated with HA-PML in a GTPase-dependent manner (). We noticed that ubiquitin signals were accumulated in these GFP-CRAG–associated inclusions but not in GTPase mutants (). Therefore, we examined whether CRAG stimulated ubiquitin ligase activity in PML immunoprecipitates. An in vitro ubiquitin ligase assay revealed that the activity in PML immunoprecipitates was undetectable in the absence of GFP-CRAG, but GFP-CRAG coexpression induced ubiquitin ligase activity in immunoprecipitates of PML WT but not PML RING (really interesting new gene)–finger mutants (C51S/C54S; ). For in vivo ubiquitination, ubiquitinated proteins were purified from cells coexpressing HA-PML and Flag ubiquitin with or without GFP-CRAG and analyzed by immunoblot with anti-HA and anti-Flag antibodies. As shown in , ubiquitinated proteins significantly increased by GFP-CRAG expression. Collectively, these results suggest that CRAG is an inducer for the association of unknown ubiquitin ligases with PML or a direct activator of PML-associated ubiquitin ligase. Because PML contains a RING-finger domain that confers on it an E3 ubiquitin ligase activity, it is possible that PML is an E3 ubiquitin ligase. The fact that polyglutamine diseases are characterized by the presence of ubiquitinated, PML-associated NIs suggested the possible involvement of CRAG in polyglutamine diseases. To ascertain a possible involvement of CRAG in polyglutamine diseases, we examined subcellular localization of CRAG in the brains of Machado-Joseph disease (MJD) patients and detected specific CRAG inclusion (). Similar CRAG inclusions were observed in the brains of other MJD patients (unpublished data). To further confirm this phenomenon, we generated GFP-Q12 as control and GFP-Q69 as polyQ and examined whether CRAG could interact with polyQ. Subcellular distribution analysis revealed the diffuse cytoplasmic localization of Q12 and perinuclear aggregation of misfolded Q69 in HeLa cells (, left). Compared with the cytoplasmic distribution of both Q12 and CRAG, Q69 and CRAG translocated to the nucleus and formed NIs (, right). An immunoprecipitation assay also demonstrated that CRAG interacted with GFP-Q69 but not -Q12 (, c and e). Similarly, in hippocampal neurons expressing GFP-Q69 or -Q12, endogenous CRAG associated with GFP-Q69 but not -Q12 (). NI formation by CRAG was induced by various stress stimuli generating reactive oxygen species (ROS) such as UV. Indeed, we found that HO induced CRAG nuclear translocation and that this was blocked by ROS scavenger edaravone (3-methyl-1-phenyl-2-pyrazolin-5-one; Radicut; Fig. S2, available at ). An in vitro GTPase assay indicated that HO activated CRAG GTPase (unpublished data). To understand the mechanism by which CRAG recognizes and interacts with polyQ, we focused on the role of ROS generation by polyQ in the activation of CRAG. An in vitro GTPase assay demonstrated that coexpression with Q69 but not Q12 activated CRAG GTPase, and this activation was blocked by the treatment of ROS scavenger Radicut (). Actually, Radicut inhibited Q69-mediated ROS generation (unpublished data). Furthermore, in Radicut-treated cells, CRAG failed to colocalize with Q69 and induce Q69 nuclear translocation (). A coimmunoprecipitation assay also indicated that Radicut blocked the association of CRAG with Q69 in a Radicut dose–dependent manner (). Thus, ROS may be required for CRAG activation and interaction with polyQ. To confirm this, doxycycline (DOX)-inducible Tet-On HeLa cell lines expressing HA-CRAG WT and NLS mutants were established. Two clones each selected from WT and NLS mutants were treated with DOX, and induction of HA-CRAG expression was checked by immunoblot analysis using anti-HA antibody (). That equal amount of total protein was loaded is shown by tubulin immunoblots (, bottom). Using these cell systems, we monitored the time-dependent nuclear translocation of Q69 by CRAG. As shown in , no nuclear localization of Q69 was detected before DOX treatment, but 3 h after DOX treatment, a major part of CRAG was colocalized with Q69 aggregates at perinuclear sites and a small part of the CRAG–Q69 complex was detected in the NIs. After 6–12 h, almost all CRAG and Q69 translocated to the nucleus and formed NIs. Merged images of CRAG and Q69 (, insets, yellow) demonstrated the colocalization of CRAG and Q69 in the nucleus. As a negative control, cells treated with solvent for 12 h revealed no nuclear translocation of Q69 (unpublished data). Similar results were obtained in other Tet-On HeLa cell lines (unpublished data). On the other hand, NLS mutants of CRAG could colocalize with Q69 but not induce nuclear translocation of Q69 at all (, right). Merged images (, insets, yellow) showed the perinuclear aggregates of the CRAG–Q69 complex. Subcellular fractionation analysis also indicated CRAG-dependent nuclear translocation of Q69 and no nuclear localization of Q69 in cells coexpressing NLS mutants (). Statistical analysis also indicated that CRAG WT, but not NLS mutants, induced Q69 nuclear translocation (). These results demonstrated that CRAG promoted Q69 nuclear translocation in an NLS-dependent manner. The disappearance of polyQ at PML body has been previously reported as a pathological finding of polyglutamine diseases (). We examined CRAG-dependent degradation of Q69 using a DOX-inducible Tet-On HeLa cell line expressing HA-CRAG. Upon DOX treatment, CRAG expression was up-regulated, whereas Q69 protein level was concomitantly down-regulated. That equal amount of total protein was loaded is demonstrated by tubulin immunoblots. A distinct background band sometimes appeared as a cross-reacting protein in the anti-HA immunoblots (, left). We confirmed that DOX alone did not affect Q69 protein level in Tet-On HeLa cells without CRAG (, right). As shown in , the proteasome inhibitor MG132 blocked Q69 down-regulation by CRAG WT, suggesting that CRAG may promote Q69 degradation through the ubiquitin–proteasome pathway. Furthermore, FACS analysis of cells stained with annexin-V/propidium iodide demonstrated that CRAG WT expression suppressed Q69-induced cell death (). In contrast, DOX alone could not rescue Q69-mediated cell death in Tet-On HeLa cells without CRAG (). To confirm ubiquitination of Q69 by CRAG, an in vivo ubiquitination assay was performed. As shown in , the ubiquitinated Q69 significantly increased by CRAG expression. Degradation of Q69 was not enhanced by NLS or GTPase mutants, and DNA ladder formation assays consistently revealed that the two mutants could not rescue Q69-mediated cell death (unpublished data). Because NLS and GTPase activity in CRAG are critical for its interaction with PML, CRAG-mediated activation of PML-associated ubiquitin ligase may be responsible for this ubiquitination. We next examined the effect of CRAG knockdown by RNA interference method on nuclear translocation of polyQ in neurons. Several small interfering RNA (siRNA) oligonucleotides specifically targeted to the CRAG sequence were generated, and their inhibitory effects were estimated using the COS-7 cell expression system. All three RNA interference oligonucleotides, but not scramble oligonucleotides, suppressed the expression of HA-CRAG (, left). Among them, one siRNA (No. 1) showing the strongest inhibitory effect on CRAG expression was used for the following experiments. This siRNA or scramble with Q69 plus pEGFP vector as a marker were introduced into cultured rat hippocampal neurons, and the inhibitory effect on endogenous CRAG expression was evaluated by CRAG immunoblotting and immunostaining. As shown in (a [right] and b) this siRNA, but not scramble, was found to suppress endogenous CRAG expression, indicating that this siRNA against the CRAG gene is useful to assess CRAG function. To determine whether endogenous CRAG is involved in nuclear translocation, inclusion body formation, and decreased cell toxicity of polyQ in neuronal cells, the effects of CRAG knockdown on subcellular distribution and cell toxicity of Q69 were examined. In scramble siRNA–cotransfected DRG neurons, endogenous CRAG and Q69 translocated to the nucleus and formed NIs (, top). In contrast, perinuclear aggregations of Q69 were dominantly detected in siRNA-mediated CRAG-deficient cells (, middle). Moreover, almost all CRAG-deficient cells revealed a typical apoptotic phenotype including chromatin condensation as shown in (bottom). Statistical analysis indicated that siRNA-mediated CRAG depletion blocked the nuclear translocation of Q69 in DRG neurons (, top). In addition, analysis of a cell death assay, judging from chromatin condensation, showed that >80% of CRAG-deficient cells underwent cell death, whereas only 30% of scramble siRNA–transfected cells died by Q69 expression 48 h after transfection (, bottom). CRAG knockdown by siRNA did not cause death in cells that were not expressing Q69 (unpublished data). These results demonstrate that endogenous CRAG mediates nuclear translocation and NI formation by polyQ and confers resistance to cell death under the conditions of polyQ accumulation. A schematic model of CRAG action on polyQ is shown in . Patients with polyglutamine diseases and transgenic mouse model carrying polyQ showed the late-onset and gradually progressive neurological pathology (; ). Indeed, CRAG expression is very high in the developing brain and decreased thereafter in the adult brain (Fig. S3, available at ). This developmentally regulated expression of CRAG may be closely related to the appearance of polyglutamine disease; a decreased level of CRAG expression fails to scavenge unfolded proteins at PML bodies and permits an accumulation of polyQ aggregates in the nucleus, thereby conferring a toxic gain of function that is selectively deleterious to neurons. If so, CRAG is a rate-limiting factor in the degradation of pathological forms of polyQs and targeted expression of CRAG is a potential gene therapy for polyglutamine disease. Anti-Flag M2 monoclonal and anti–α-tubulin antibodies were obtained from Sigma-Aldrich. Anti-HA high affinity, anti-HA affinity matrix, and anti–c-myc mouse mAbs were obtained from Roche. Anti-GFP rabbit polyclonal antibody, annexin V, propidium iodide (PI), and secondary antibodies conjugated with Alexa Fluor 488, 594, and 647 were obtained from Invitrogen. Anti-6×His monoclonal, anti-GFP mouse monoclonal, and Hoechst 33258 were obtained from Nacalai Tesque. DOX and mouse MTN blot were obtained from CLONTECH Laboratories, Inc. HA-probe was purchased from Santa Cruz Biotechnology, Inc. Nuclear mitotic apparatus protein antibody 1 was purchased from Biocarta. Anti-ubiquitin antibody was obtained from Santa Cruz Biotechnology, Inc. Anti-HA mouse mAb was purchased from Covance. Anti-PML mouse mAb was purchased from Santa Cruz Biotechnology, Inc. CRAM (CRMP-5) has been implicated in semaphorin signaling (). We have searched for CRAM-interacting proteins from developing rat brain. The purification procedure was performed by using anti-CRAM antibody affinity column chromatography and specific elution with antigen peptides for anti-CRAM antibody. Among purified proteins, we focused on a 42-kD protein, and partial amino acid sequence analysis revealed several peptide sequences, including KSALVHRYLTGTYVQEESPEGGRF. Based on this information, we cloned a novel gene encoding a GTPase from the mouse brain cDNA library (available from GenBank/EMBL/DDBJ under accession no. ). CRAG WT, GTPase mutants, NLS mutants, and murine PML1 isoform (available from GenBank/EMBL/DDBJ under accession no. ) cDNA tagged with the HA epitope at the NH terminus were subcloned into pCMV5 expression vectors. Murine PML1 shows 67% homology to human PML1 at the amino acid level. CRAG GTPase mutation was generated by missense mutation S114N. CRAG NLS mutation was generated by missense mutation KR342-343EE. The plasmid encoding NH-terminal–truncated ataxin-3 with Q69 together with the COOH-terminal myc epitope and the NH-terminal HA epitope was described previously (). In this plasmid, 286 amino acid residues of ataxin-3 were deleted from the NH-terminal side. For our experiment, the HA epitope was removed. These procedures were performed as described previously (; ). Fluorescence images were analyzed on a confocal microscope (LSM 510 META; Carl Zeiss MicroImaging, Inc.) equipped with three lasers (UV.Ar. 364, Ar. 488, and HeNe 543) using Plan-Apochromat 63× oil-immersion (NA 1.40) objective. LSM 510 META 3.0 software (Carl Zeiss MicroImaging, Inc.) was used for image acquisition from confocal microscopy. Photoshop 6.0 software (Adobe) was used for minor adjustments to contrast and overlaying. Postmortem brain tissues were obtained from two MJD patients and analyzed as described previously (). The GTPase assay was performed as described previously (). Immunoprecipitates were washed three times with lysis buffer and once with ubiquitination buffer (50 mM Tris-HCL, pH 7.4, 10 mM MgCl, 5 mM ATP, and 2 mM dithiothreitol) and incubated in 50 μl of the same buffer supplemented with 100 ng E1 (), 500 ng E2ubcH5(a,b,c) (), and 2.5 μg His-tagged ubiquitin (Calbiochem) for 30 min at 25°C. Samples were analyzed with anti-His antibody. Fig. S1 a shows the hollow, doughnut-like nuclear bodies of CRAG in UV-irradiated hippocampal neurons, and Fig. S1 b shows colocalization of CRAG doughnut-like bodies with PML1 in a UV-irradiated DRG neuron. Fig. S2 shows NIs of CRAG in response to HO stimulation in HeLa cells. Fig. S3 shows a comparison of CRAG expression in adult and developing mouse brains. Online supplemental material is available at .

xref italic #text Both Dgp71WD and NEDD1 are suspected to contain WD repeats in their amino-terminal half. To further investigate their structural similarity, we performed sequence alignment and structure prediction using bidimensional hydrophobic cluster analysis (). This method allowed the identification of two WD repeats in addition to the five repeats described in Dgp71WD by . We used as a template the sequence of human transducin β chain 1, of which the three-dimensional structure has been solved (pdb identifier 1got). shows that the three proteins are structurally similar within a region of seven repetitions of four β strands—A, B, C, and D. These strands are predicted to form seven blades of a β propeller structure, with strand C carrying the WD signature (Fig. S1, available at ). Within their WD domain, NEDD1 and Dgp71WD are more similar to each other than to any other WD protein. The two proteins show an overall identity of 20.8% (), have a similar length, and possess a coiled-coil domain of ∼100 amino acids at their carboxy-terminal ends (unpublished data). These similarities suggest that they perform similar functions. To test this, we raised antibodies against the carboxy-terminal domain of the protein and performed localization studies in HeLa cells. Small amounts of NEDD1 are present at the centrosome during interphase (). However, NEDD1 recruitment to the centrosome increases significantly at the onset of mitosis (). Immunoblotting revealed that NEDD1 in arrested mitotic cells has a reduced mobility (). Phosphatase treatment suggests that this shift in mobility is attributable to the mitotic phosphorylation of NEDD1. In metaphase, most of NEDD1 is concentrated at the centrosome, but small quantities are also seen along the spindle fibers (). During anaphase, the centrosomal localization of NEDD1 is gradually reduced (), and in telophase, small amounts are also seen along the midbody microtubules (). This dynamic distribution resembles γ-tubulin localization during the cell cycle (; ; ; ). Indeed, we observed a perfect colocalization of NEDD1 with γ-tubulin (). We performed quantitative immunoblotting of NEDD1 in HeLa extracts using the carboxy-terminal fusion protein of NEDD1 for calibration and determined that the abundance of NEDD1 is ∼0.002% of total cellular protein (unpublished data). In comparison, we found γ-tubulin to be significantly more abundant, with an estimated 13–20 molecules of γ-tubulin per molecule of NEDD1. To study the function of NEDD1, we reduced the levels of the protein using small interfering RNA (siRNA). We show by immunoblotting that after 24 h of siRNA treatment, NEDD1 levels are reduced to 20% (). Considering a transfection efficiency of ∼85%, as determined by cytoplasmic incorporation of rhodamine-labeled RNA oligomers (unpublished data), we reached a nearly quantitative depletion in individual cells. These low levels were maintained also after prolonged siRNA treatment of up to 72 h. We show as a control that the levels of γ-tubulin remain unaffected after NEDD1 depletion (). After 72 h of treatment, cell cultures show a significantly increased mitotic index of 47% ( = 200), compared with 5% ( = 550) of mitotic cells in control cultures. NEDD1-depleted cells accumulate mitotic aberrations and are arrested in a prometaphaselike state of mitosis (). In the majority of these cells, an aberrant mitotic apparatus is formed, with unseparated or poorly separated poles and with microtubules arranged in a monoastral pattern (). The chromosomes are distributed randomly in the cytoplasm, and MAD2 staining reveals the absence of microtubule attachment to kinetochores (unpublished data). The overall density of microtubules in these cells seemed to be reduced. Few mitotic cells were observed where NEDD1 was reduced to a lesser extent. These cells still form bipolar spindles but lack astral microtubules and show an enlarged pole-to-pole distance (). Identical results were obtained with two different siRNA oligomers (unpublished data), confirming the specificity of this phenotype. We then investigated whether NEDD1 depletion affects the recruitment of γ-tubulin to the centrosome. shows that partial reduction of NEDD1 in mitosis results in partial loss of γ-tubulin staining. More efficient removal of NEDD1 results in almost complete loss of γ-tubulin at the centrosome, correlated with the formation of monoastral microtubule organization in mitosis (). Consistently, the localization of the γTuRC proteins GCP2 and -4 to the centrosome is suppressed in these monoastral structures (). Our data leave open the question of whether NEDD1 depletion prevents γTuRC assembly or whether γTuRCs are still assembled but fail to be recruited to the centrosome in the absence of NEDD1. To address this, we examined the association between NEDD1 and the γTuRC by coimmunoprecipitation. Purified antibody against NEDD1 coprecipitated components of the γTuRC, including γ-tubulin, GCP2, and GCP4 (). In the reverse experiment, antibody against γ-tubulin also coprecipitated NEDD1, in addition to the γ-complex proteins GCP2 and -4 (). Further, NEDD1 sediments at ∼32S together with γTuRC proteins in sucrose-gradient centrifugation experiments (, control). In addition, we coimmunoprecipitated NEDD1 and γ-tubulin from the peak fraction of the sucrose gradient containing the γTuRC (). This indicates that NEDD1 associates with the γTuRC. We performed additional control experiments using gel filtration of HeLa cell lysates, demonstrating that all γTuRC proteins elute in a peak together with NEDD1 and that NEDD1 coimmunoprecipitates with γ-tubulin from this peak (Fig. S2, available at ). Sucrose-gradient centrifugation showed a significant amount of γTuRC components sedimenting at sizes smaller than 19S (). These might represent breakdown products of γTuRCs preferentially formed under the conditions of sucrose-gradient experiments, as compared with gel filtration. We then examined the composition of the γTuRC in cells depleted of NEDD1. RNA silencing in three different experiments reduced the levels of NEDD1 coimmunoprecipitating with γ-tubulin; in contrast, GCP4 was only slightly affected and GCP2 coprecipitated at nearly regular levels (; coprecipitating protein amounts compared with controls: NEDD1, 28% ± 6; GCP2, 96% ± 4; GCP4, 88% ± 21, after normalization over amounts of immunoprecipitated γ-tubulin). In addition, immunoblots of cell extracts separated on sucrose gradients reveal that the sedimentation of the majority of γTuRC proteins is unaffected after depletion of NEDD1, with small levels of GCP4 and γ-tubulin sedimenting more slowly (, RNAi [RNA interference]). We conclude that the majority of γTuRCs can still form in the absence of NEDD1, although we cannot exclude the possibility that residual levels of NEDD1 after depletion support the assembly of reduced amounts of γTuRCs. To determine whether the association between NEDD1 and γ-tubulin complexes is cell cycle dependent, we also investigated NEDD1 silencing effects in interphase cells. As seen before in mitosis, γTuRC localization to the centrosome was also dependent on NEDD1 in interphase cells. Reduced levels of NEDD1 resulted in reduced amounts of γ-tubulin at the centrosome (). Localization of other centrosomal components, such as pericentrin, was not affected (). The reduction ofγ-tubulin correlated with a delay in microtubule regrowth after cold treatment (). Moreover, in U2OS cells that usually contain a well-focused microtubule network, microtubules appeared disorganized and regrowth did not originate from a single centrosomal focus, as seen in control cells (, min). Next, we wanted to determine which region of the NEDD1 protein mediates an interaction with the γTuRC. We separately expressed the amino- and carboxy-terminal halves of the protein (amino acids 1–319 and 321–660, respectively), fused to GFP, in U2OS cells. The amino-terminal half corresponds to the WD-repeat domain. 24 h after transfection, overexpressed full-length NEDD1 localizes to the centrosome, in addition to diffuse localization in the cytoplasm (). Interestingly, the soluble pool seemed to increase with the expression level of the protein, indicating that the fraction associated with the centrosome is limited, probably via an endogenous anchoring factor. The amino-terminal domain also localizes to the centrosome. In contrast, the carboxy-terminal domain does not exhibit any centrosomal localization. In mitosis, the amino-terminal domain localizes to the centrosome and to the spindle poles, whereas the carboxy-terminal domain remains cytoplasmic (). We further noticed that overexpressing the carboxyterminal domain caused a dominant-negative phenotype in mitosis, inhibiting spindle pole separation, similar to NEDD1-silencing experiments (). Colocalization of γ-tubulin shows that overexpression of the carboxy-terminal domain, but not the amino-terminal domain, displaces γ-tubulin from the centrosome (). Full-length NEDD1, when highly expressed, also partly displaces γ-tubulin. This suggests that excess NEDD1 associates with the centrosome and blocks the attachment of NEDD1–γTuRC complexes or, alternatively, that excess NEDD1 is able to sequester γ-tubulin in the cytoplasm via its carboxy-terminal domain. We further performed biochemical experiments to determine whether γTuRC components could be coprecipitated with the amino- or carboxy-terminal half of NEDD1. Coimmunoprecipitations of the GFP-tagged, overexpressed NEDD1 fragments were performed with an antibody against GFP. These experiments revealed that complexes containing γ-tubulin, GCP2, and GCP4 are able to bind to the carboxy-terminal fragment of NEDD1, although less efficiently than to full-length NEDD1 (). Our mapping experiments thus indicate that the amino-terminal domain of NEDD1 is responsible for centrosome targeting, whereas the carboxy-terminal domain mediates the interaction with the γTuRC (). The loss of γTuRC recruitment to the centrosome after NEDD1 silencing led us to examine in more detail the resulting spindle defects, particularly the structure of the spindle poles. We noticed that, in mitotic cells lacking detectable amounts of NEDD1, microtubules were often organized from broad irregularly shaped areas rather than from focused organizing centers. This was also clear in cells overexpressing the carboxy-terminal domain of NEDD1 (). Staining of nuclear mitotic apparatus protein (NuMA) to indicate the localization of microtubule minus ends confirmed the presence of dispersed microtubule organizing centers (). Consistently, the pericentriolar material was found scattered, as shown by pericentrin staining, which distributed either more diffusely or dispersed in multiple foci (). In contrast, components of the γTuRC were almost completely absent from the microtubule organizing centers and instead found diffuse throughout the cytoplasm (). This suggested that although γTuRC proteins failed to localize to the centrosome both in interphase and mitosis, pericentrin was able to target to microtubule organizing centers, although less focused in mitosis (; , A and B; and ). Dispersal of microtubule organizing proteins correlated with defects of centrosome-dependent microtubule regrowth after cold treatment. After short times (30 s), microtubule nucleation occurred from multiple sites in the cytoplasm, in contrast to centrosomal nucleation in controls (). We then investigated whether NEDD1 depletion also affected the formation of the centrosomal core structures, i.e., the centrioles. We used centrin staining to visualize the centrioles in mitotic cells (). In control cells, a pair of centrioles can be clearly distinguished at both spindle poles (<10% of the poles present a single centrin dot, and <1% of the spindles show only one dot at both poles). In NEDD1-depleted cells, we selected spindles showing incomplete pole separation. In those spindles, we counted only one centrin dot at each pole, at a frequency of >40% (). This suggests that NEDD1 is necessary for daughter centriole assembly during duplication or for centriole maturation. Electron microscopy analysis of serial sections of multiple NEDD1-depleted cells confirmed that only one centriole was present at each pole (). To determine whether centriole duplication was dependent on NEDD1, we performed an assay previously published by : U2OS cells were arrested in S phase in the presence of aphidicolin, allowing the centrioles to undergo multiple rounds of duplication. In the vast majority of control cells (84% after control RNA and 92% after mock treatment; ), this led to the formation of more than four structures per cell containing centriolar markers, as detected by immunostaining of centrin and polyglutamylated tubulin (). Because NEDD1 was not present on all of these structures, these could either represent imperfectly assembled centrioles or immature centrioles that might accumulate visible amounts of NEDD1 later in the cycle. In contrast, cells depleted of NEDD1 mainly contained four centriolar structures or less (only 35% contained more than four; ). Because our experiments on NEDD1 depletion or overexpression of the carboxy-terminal half of NEDD1 both resulted in displacement of γ-tubulin from the centrosome, we suspected that the NEDD1-dependent defects on centriole assembly and spindle formation might have been indirectly caused by the absence of γTuRCs at the centrosome. To test this idea, we performed γ-tubulin depletion by siRNA treatment and investigated the effects on spindle formation and on centriole duplication (). Our RNA oligomers were directed against a sequence present in both human forms of γ-tubulin and resulted in a nearly quantitative reduction of γ-tubulin levels (). As seen previously in NEDD1-depleted cells, γ-tubulin depletion also led to defects in spindle pole separation and to a high number of cells with unreplicated centrioles in mitosis (). Only 33% of γ-tubulin–depleted cells displayed centriole pairs at each pole, whereas the remaining cells contained at least one pole with an unreplicated centriole (). When investigating the morphology of the spindle poles, we observed scattering of pericentrin and NuMA staining as seen in NEDD1-depleted cells (; and ). NEDD1 is still found localized to the area of microtubule organization in mitosis, although more dispersed and at variable amounts (). In interphase, regular levels of NEDD1 are recruited to the centrosome, even in the absence of γ-tubulin (). We have shown that NEDD1 is an essential protein for the recruitment of γ-tubulin complexes to the centrosome. NEDD1 depletion prevents the localization of γ-tubulin and GCPs to the centrosome both in interphase and in mitosis. In contrast, the absence of γ-tubulin does not prevent centrosomal targeting of NEDD1 itself. Our biochemical studies as well as previous work on the homologue Dgp71WD show that NEDD1 associates directly with the γTuRC (). It has previously been suggested that the homologue is a scaffold protein of the γTuRC.Although we see the majority of the γTuRC still forming after NEDD1 depletion, our experiments are consistent with the idea that NEDD1 could be required for efficient γTuRC assembly because cosedimentation and coimmunoprecipitation of GCP4 with other γTuRC components seem slightly reduced at low levels of NEDD1. However, NEDD1 is different both structurally and functionally from proteins of the GCP/grip family, of which several were shown to have a strong effect on γTuRC assembly (; ). Because the majority of γTuRCs still sedimented at ∼32S after reducing NEDD1 levels to ∼20%, we conclude that NEDD1 associates with γTuRCs at a low stoichiometry; otherwise, its absence would have resulted in a significant shift. Consistently, comparison of γ-tubulin and NEDD1 amounts indicates a large excess of γ-tubulin over NEDD1 in the cytoplasm. If γ-tubulin assembled into γTuRCs in excess over NEDD1, multiple γ-tubulin molecules and associated GCPs could build a ring-shaped template for microtubule nucleation, whereby a single molecule of NEDD1 bound to the periphery of the γTuRC could increase the binding competence of the complex to the centrosome. Our biochemical data suggest that the carboxy-terminal half of NEDD1 binds to the γTuRC, whereas the amino-terminal half corresponding to the WD-repeat domain mediates its attachment to the centrosome. Additional interactions between γTuRC proteins and the centrosome are thought to be mediated by the binding of GCP2 and -3 to pericentrin and AKAP450/CG-NAP (; ; ) and by ninein or ninein-like protein (; ; ). A different group of proteins known to mediate γ-tubulin interaction includes centrosomin in and the fission yeast proteins Mto1 and Pcp1, all sharing a conserved 60–amino acid motif that putatively binds to the γ-tubulin complex (; ; ). The significant increase of NEDD1 on mitotic centrosomes led us to wonder whether NEDD1 is responsible for the increased recruitment of γ-tubulin complexes at the onset of mitosis (). Consistently, the most striking defects after depletion of NEDD1 are monopolar spindles or spindles with poorly separated poles, which are reminiscent of the spindle defects found in cells depleted of γ-tubulin (; ; ; ; ; this study). Moreover, similar phenotypes are found in cells depleted of CeGrip-1 and -2 and pericentrin and in mutants of Dgrip91 (, ; ; ). In all these examples, cells lack proper targeting of γ-tubulin to the centrosome, which is thought to affect microtubule nucleation and microtubule minus end dynamics. Altered microtubule-end dynamics have been reported to cause the formation of monopolar spindles (). Consistently, defects at microtubule ends are a possible explanation for the for-mation of monopolar spindles in NEDD1- or γ-tubulin–depleted cells. However, loss of bipolar spindle organization is a complex phenotype, and additional factors might contribute to this defect: loss of astral microtubules in NEDD1-depleted mitotic cells might eliminate astral pulling forces and therefore favor spindle collapse. In addition, NEDD1-depleted cells show defects in centriole duplication, which in turn might favor monopolarity (), although acentriolar pathways have been reported to support bipolar spindle formation, even in somatic vertebrate cells (). Besides NEDD1, a large number of kinases and structural proteins have been implicated in centriole duplication. In particular, proteins such as SPD-2; SAS-4, -5, and -6; and centrin are not only present in the pericentriolar material but also form part of the centriolar cylinder and could therefore be structural components directly responsible for the formation of new centrioles (; ; ; ; ; ; ; ). Other proteins, such as SPD-5 or ɛ-tubulin, are located at the pericentriolar material or at subdistal appendages of the centriole, and it was suggested that their role in the duplication process could be to recruit additional protein for daughter centriole assembly (Chang et al., 2003; ). Likely, one of the key components that need to be recruited for centriole duplication is γ-tubulin itself: several kinases and structural proteins that are involved in centrosome duplication are also involved in attracting γ-tubulin, such as aurora A, SPD-2, and SPD-5 (; ; ). Moreover, γ-tubulin has been shown to be present in the centriolar cylinder, besides localizing to the pericentriolar material (; ; Klotz et al., 2003). It is quite possible that γ-tubulin plays a role in nucleating the microtubule triplets of the centriolar cylinder during centriole duplication, similar to the way it nucleates regular microtubules. Thus, the failure of centriole duplication after NEDD1 depletion can be explained by an indirect effect that is attributable to the failure of γ-tubulin recruitment. Our data on centriole duplication defects after depletion of γ-tubulin support this idea. In agreement with this concept are further reports on defective centriole formation in Dgrip91 mutants that fail to recruit γ-tubulin to the centrosome, in embryos depleted of γ-tubulin by RNAi, as well as observations in the ciliates and , where the absence of γ-tubulin leads to defects in basal body formation (; ; ; ). Further, it appears that duplication of centrioles and basal bodies is regulated by posttranslational modifications of γ-tubulin: in the flagellate , basal body duplication is believed to require phosphorylation of γ-tubulin, and in mammalian cells ubiquitination of γ-tubulin is thought to prevent excess rounds of centriole duplication (; ). We have shown that NEDD1 is another key component in controlling γ-tubulin levels at the mammalian centrosome. Considering the putative relationship between aberrant centrosome duplication and cancer, the regulation of γ-tubulin recruitment to the centrosome will be of continued prime interest to cell biologists. EST clones DKFZp313N211Q2 and IRATp970F0222D6 containing the full-length cDNA of NEDD1 were obtained from Deutsches Ressourcenzentrum für Genomforschung GmbH. The full-length NEDD1 construct was amplified by PCR and cloned into the pCR BluntII-TOPO vector (Invitrogen), and the internal NdeI site was removed without change of amino acid sequence. Sequencing of the first EST revealed two mutations that induced the amino acid changes P97T and S379N. To correct the sequence, a SwaI–AclI fragment encoding amino acids 97–379 was swapped between the two ESTs. NEDD1 was then subcloned into pEGFPC2 (CLONTECH Laboratories, Inc.) using the SalI and KpnI restriction sites flanking the cDNA in pCR BluntII-TOPO. GFP-NEDD1Nt (corresponding to amino acids 1–319) and Ct (corresponding to amino acids 321–660) were generated by introducing the 5′ SalI–AleI and the 3′ AleI–KpnI fragments, respectively, into pEGFPC2. Transfection of the GFP constructs into cells was performed with Fugene reagent (Roche) according to the manufacturer. HeLa, U2OS, and Cos-7 cell lines were grown at 37°C in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. HeLa-GFP tubulin-expressing cells were generated by stable transfection with pEGFP-Tub (CLONTECH Laboratories, Inc.). Mitotic indexes were ascertained by determining the percentage of live cells containing a fluorescent spindle, to prevent underestimation because of loss of mitotic cells during staining procedures. Double-stranded siRNA oligomers were transfected using oligofectamine (Invitrogen) according to the manufacturer. In brief, cells were seeded in 6-well dishes at a density of 170,000 cells/well and transfected after 24 h of growth with 2.5 μl of 100 μM siRNA. Two siRNAs targeting NEDD1 mRNA were used (Ambion). Results presented here correspond to the targeting of nucleotides 229–247. Targeting of nucleotides 286–303 induced similar depletion levels and cellular phenotypes. Note that a dTdT overhang was added to the 3′ of the RNA oligomers. For depletion of γ-tubulin, we used two siRNAs, both targeting the two human isoforms, at nucleotide positions 259–279 and 1257–1281. Maximal depletion was reached by transfecting the second siRNA twice with a 96-h interval and stopping the cells 48 h after the second transfection. However, both induced similar cellular phenotypes. Control (nonsilencing) siRNA was provided by Xeragon (QIAGEN). Mock depletion was performed with oligofectamine but no siRNA. In general, except for the centriole duplication assay where the control siRNA showed a weak effect, we did not observe any phenotypic difference between transfection in the presence of control siRNA or without siRNA. The NEDD1-His fusion protein corresponding to amino acids 279–660 was expressed in and affinity purified on Ni-NTA agarose (Invitrogen) under denaturing conditions. The eluted protein was injected into a rabbit for antibody production. The resulting antibody was affinity purified over the same antigen. A rabbit antibody recognizing GCP2 was raised against a synthetic peptide, LRGPPAPAPRVA, corresponding to amino acids 887–898 at the carboxy terminus of the protein. Rabbit anti-GCP2 and anti-GCP4 antibodies () were affinity purified over the respective full-sized proteins produced in . Other primary antibodies used in this study were anti–α-tubulin (mouse B-5-1-2 [Sigma-Aldrich] or rabbit [Abcam]), anti–γ-tubulin (mouse GTU-88 [Sigma-Aldrich] or rabbit serum R75 []), anti-pericentrin (mouse [] or rabbit [Covance]), anti-NuMA (mouse Ab-2 [Calbiochem] or rabbit serum [a gift from D. Compton, Dartmouth Medical School, Hanover, NH]), anti-centrin (mouse 20H5; a gift from J. Salisbury, Mayo Clinic, Rochester, MN), and anti–polyglutamylated tubulin (mouse GT335; a gift from B. Eddé, Centre de Recherches de Biochimie Macromoléculaire, Montpellier, France). Different mouse anti-GFP antibodies have been used for immunoprecipitation (mAb 3E6; Invitrogen) and immunodetection (Roche) of the GFP-NEDD1 derivatives. Cells grown on coverslips were fixed in methanol at −20°C and processed for immunofluorescence following conventional protocols. For detection, secondary antibodies conjugated to Alexa 488 or 568 (Invitrogen) were used. Fluorescence microscopy was performed on a microscope (Axiovert; Carl Zeiss MicroImaging, Inc.) equipped with a Z motor, using 63× or 100× 1.4 NA objectives. Z series images were acquired with a camera (AxioCam MRm; Carl Zeiss MicroImaging, Inc.) and AxioVision software (Carl Zeiss MicroImaging, Inc.). Images were subsequently deconvolved using AxioVision, and z planes were projected onto a single view. Image processing and quantification of fluorescence were done using Photoshop (Adobe). Note that , A, 7 D, and 8 I were not deconvolved and single representative planes are shown. In (B–E), 2 (C–G), and 5 B, gray-scale images show single representative planes. For electron microscopy analysis, NEDD1-depleted and control cells were fixed in 0.5% glutaraldehyde in 60 mM Pipes, 25 mM Hepes, 1 mM EGTA, and 2 mM MgCl, pH 6.9; scraped and pelleted; and subsequently postfixed in 2% osmium tetroxide. After dehydration, cells were embedded in araldite, serially sectioned, and contrasted with uranyl acetate and lead citrate. Pictures were taken with an electron microscope (CM120 Biotwin; Philips). Cells were scraped in the culture medium and rinsed twice in PBS. They were resuspended in IP buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EGTA, 1 mM MgCl, and 0.25 mM GTP) with 0.2–0.5% Triton X-100 and a protease inhibitor mix (Sigma-Aldrich). Lysates were prepared after 5 min of incubation on ice by repeated pipetting and centrifugation in a microfuge for 10 min at 12,000 , at 4°C. For coimmunoprecipitations, appropriate antibodies (2 μg anti-GFP, 25 μg anti-NEDD1, 22 μg anti-NuMA Ab-2, or 10 μl anti–γ-tubulin R75 serum) were immobilized on 50 μl Dynabeads/protein A (Dynal) and incubated with 700 μg of cell extract supernatant for 1 h at 4°C. The immunoprecipitates were washed three times in IP buffer and eluted in 30 μl of gel sample buffer containing SDS. For detection of protein in total extracts, 40 μg of extract was loaded per gel lane. For sedimentation analysis, 750 μg of total cell extracts was fractionated on gradients of 5–40% sucrose in IP buffer, in a SW55Ti rotor (Beckman Coulter) at 55,000 rpm for 3 h. Proteins were detected by immunoblotting using an ECL kit (GE Healthcare). NEDD1 protein levels were determined with a Li-Cor imager (Odyssey). Cells grown on coverslips were transferred into precooled medium on ice for 1 h and then into prewarmed medium at 37°C. Regrowth was stopped at different time points by methanol fixation. Cells grown on coverslips were treated with 1.6 μg/ml aphidicolin (Coger) 12 h after transfection of siRNAs. Cells were fixed 72 h after transfection (i.e., after 60 h of incubation in the presence of the drug) and processed for immunofluorescence. Fig. S1 shows the predicted three-dimensional structure of the NEDD1 WD-repeat domain. Fig. S2 shows that NEDD1 copurifies with the γTuRC. Online supplemental material is available at .

In metazoans, the centrosome organizes the microtubule cytoskeleton. The molecular mechanisms responsible for the initiation and the regulation of microtubule assembly remain unclear, although γ-tubulin appears critical to these processes. In addition to a centrosomal fraction, γ-tubulin is present in cytosolic high-order protein structures (; ; ; Fujita et al., 2002). In , two main complexes have been characterized. The simplest ones, called γ-tubulin small complexes (γ-TuSCs), are salt-stable tetramers of ∼10S that are composed of two γ-tubulin molecules and two associated proteins, Dgrip84 and Dgrip91 (). They represent the basic components of the larger complexes, the γ-tubulin ring complexes (γ-TuRCs). γ-TuRCs, whose sedimentation coefficients range from 25 to 35S, contain at least four other proteins (Dgrip75, Dgrip128, Dgrip163, and Dgp71WD) in addition to the γ-TuSC subunits, in a yet unknown stoichiometry. Dgrip75, Dgrip128, and Dgrip163 exhibit sequence homologies called grip motifs, with the two γ-tubulin–associated proteins of the γ-TuSC (Fava et al., 1999; ; ). In contrast, Dgp71WD does not posses any grip motifs, but contains seven WD (tryptophan–aspartic acid) repeats (). In vitro, this protein directly interacts with the grip motif containing γ-TuRC subunits, suggesting that it may play a scaffolding role in γ-TuRC organization (). Current models suggest that γ-TuRCs that have been previously assembled in the cytoplasm are recruited to the centrosomes, where they play a role in microtubule nucleation and stabilization (; ; Fava et al., 1999; ). The functions of γ-tubulin and its associated proteins in the γ-TuSC have been extensively studied. Deletion of either corresponding gene is lethal, resulting in an accumulation of cells in mitosis. This is usually correlated with the appearance of strong mitotic defects such as monopolar structures or bipolar spindles exhibiting unfocused poles, impairment of pole maturation, and increase in aneuploidy (; ; ; ; ; ; ; ; ; ). In contrast, our knowledge about the function of γ-TuRC–specific components is limited. In , is essential for fertility, but not for viability. loss-of-function mutants specifically affect localization of the maternal determinant during oogenesis, suggesting a role in the organization or in the dynamics of a subset of microtubules (). In , mutants in (a orthologue) and in (a orthologue) exhibit defects associated with altered microtubule function, but without any effect on cell viability (Fujita et al., 2002; ). Current models cannot explain these data. Instead, they raise questions not only about the redundancy or the specificity of the γ-TuRC–specific proteins but also about the respective functions of γ-TuSCs and γ-TuRCs. In this work, we have tested whether γ-tubulin is only recruited to the centrosome in the form of γ-TuRCs or if γ-TuSCs can be recruited and subsequently matured into functional γ-tubulin complexes by attracting additional components. To this aim, we have developed two strategies: RNA interference (RNAi) in cultured cells involving individual or concomitant depletion of γ-TuRC components, and genetic analyses by taking advantage of the availability of mutant strains (Dgrip75, Dgrip163, and Dgp71WD). We demonstrate that the γ-TuRC–specific subunits display functional specificities and that the γ-TuSCs could be targeted to the centrosome where basic microtubule assembly functions are maintained. First, we characterized the consequences of the depletion of a γ-TuRC–specific protein on the assembly of cytosolic γ-tubulin complexes. Cultured S2 cells were treated by RNAi to deplete Dgrip75, a grip protein specifically present in the γ-TuRC (Fava et al., 1999). The treatment led to a strong decrease of the protein level (; >95% of the control level, as judged by Western blot analysis). This effect was specific, as determined by examining the amount of the three γ-TuSC proteins (γ-tubulin, Dgrip84, and Dgrip91) and actin (). Immunofluorescence analysis of control cells showed that although Dgrip75 was undetectable at the interphase centrosome, it localized to the poles at the onset of mitosis, where it was maintained throughout cell division (, c). In marked contrast, the protein was absent from the mitotic centrosomes in Dgrip75-depleted cells, consistent with Western blot quantification (, d; and ). When extracts from treated cells were submitted to sucrose gradient sedimentation, γ-TuRCs were severely reduced, as indicated by immunoblotting of soluble fractions with antibodies against γ-tubulin, Dgrip84, Dgrip91, Dgrip128, and Dgrip163 (). The main remaining complexes appeared to be γ-TuSCs, as judged by their protein content and their sedimentation coefficient. In addition, the total level, as well as the soluble and cytoskeletal fractions of the three γ-TuSC proteins, are unchanged in control and Dgrip75-depleted cells (), suggesting a redistribution of these proteins in the different complexes rather than a change in quantity. In contrast, after Dgrip75-RNAi treatment, we noticed a decrease in the total level of the two other grip-motif proteins of the γ-TuRC, Dgrip128 and Dgrip163. The remaining Dgrip128 protein was distributed on the gradient in the form of heterogeneous and uncharacterized complexes with apparent masses equal or slightly higher than the mass of the γ-TuSC. Dgrip163 protein migrated mainly in light fractions (<10S). One hypothesis could be that this protein was present as a monomeric or dimeric form. Thus, Dgrip75 appears to be required for efficient assembly or stability of cytoplasmic γ-TuRCs. We next investigated whether Dgrip75 depletion, and thus the subsequent decrease of γ-TuRCs, affected mitotic progression. The mitotic index was significantly increased by ∼2.6-fold (2.9%; = 4,766; probability of 95% [P95] = 2.5–3.3 in treated cells, compared with 1.1%; = 9,235; P95 = 0.9–1.3 in control cells). This mitotic accumulation coincided with the maintenance of an active mitotic checkpoint, as judged by a strong BubR1 signal at kinetochores (not depicted), which was indicative of a transient block in prometaphase. α-Tubulin immunostaining analysis confirmed an accumulation of cells in premetaphase stages (increased 1.8-fold in frequency), whereas postmetaphase figures exhibited a 2.6-fold decrease relative to controls (). Aberrant mitotic figures were observed, mainly monopolar spindles and bipolar spindles that were either elongated or with lagging chromosomes ( and ). Approximately one third of prometaphases and metaphases were still organized as bipolar structures, albeit with a longer interpolar distance (average increase of 50%; 7.9 ± 0.9 in treated cells vs. 5.3 ± 0.7 in controls) and a poor microtubule density. However, treated bipolar spindles exhibited astral microtubules (, and l). Their poles were normally focused, consistent with polar localization of Asp (, a), a marker of the spindle microtubules minus ends (). They seemed to properly separate their centrosome, as 97% ( = 112; P95 = 92–99) exhibited centrosomin (Cnn) labeling at both poles (, b; ). Surprisingly, in Dgrip75-depleted cells, γ-tubulin was still recruited to the poles at all stages of mitosis (; and ). The maximal fluorescence values raised by γ-tubulin antibodies at the two poles of symmetrical spindles did not differ between Dgrip75-treated and control cells (). The two γ-tubulin partners in the γ-TuSC, Dgrip84, and Dgrip91 were also targeted to the poles in an efficient manner (; and ). At the same time, pole localization of γ-TuRC–specific proteins (Dgrip128, Dgrip163, and Dgp71WD) was significantly inhibited, but not completely abolished ( and ). The reductions in Dgrip128 or Dgrip163 stainings (, ) may result, at least partly, in the overall decrease in the levels of these proteins after Dgrip75 depletion. Dgp71WD staining was present, but reduced in intensity in most of the cases (, insets). To illustrate these changes, we performed double-labeling immunofluorescence on cells depleted of Dgrip75, showing colocalization of γ-tubulin with γ-TuSC–associated proteins (Dgrip84 or Dgrip91) or with Dgp71WD, whereas the γ-TuRC proteins Dgrip128 or Dgrip163 were no longer detectable (Fig. S1, available at ). The localization of γ-tubulin along spindle microtubules, observed in 97% ( = 138; P95 = 93–99) of control cells, was detected in only 10% ( = 146; P95 = 5–15) of the depleted cells (). Similarly, the two γ-TuSC proteins, Dgrip84 and Dgrip91, were no longer detected along spindle microtubules (). Moreover, staining of γ-TuSC proteins at the midbody was impaired in treated cells (not depicted). Hence, the assembly of cytoplasmic γ-TuRCs does not appear as a prerequisite for centrosomal recruitment of the γ-TuSCs, but seems essential for its localization along spindle microtubules and at the midbody. To examine the functional significance of in vivo, we took advantage of the availability of the mutant allele 175–14 (), which was either a null or a strong allele (). mutant was viable, although adults of both sexes exhibited a slight increase in lethality some days after hatching. Moreover, they showed abnormalities in the abdominal cuticle segmentation and the thoracic macrochaete pattern (unpublished data), which were common in mutations affecting mitosis. Finally, mutant females were sterile because of a failure to undergo normal oogenesis. Although Dgrip75 did not appear essential for viability, the phenotypes observed after RNAi treatment of cultured cells prompted us to study the mitotic processes in mutant L3 larval brains. Western blot analysis showed that Dgrip75 levels were reduced below the threshold of detection in brain extracts (). The mitotic index was elevated approximately three-to fourfold (2.5%; = 12,410; P95 = 2.2–2.8) compared with wild-type cells (0.7%; = 16,134; P95 = 0.6–0.8). Mutant cells accumulated in prometaphase stages, whereas postmetaphase figures were reduced (unpublished data). More than half of the mutant cells exhibited overcondensed chromosomes (60%; = 186; P95 = 53–67) compared with wild type (2%; = 211; P95 = 0–4; ). These results were consistent with prolonged prometaphase and metaphase stages. However, only very few mutant cells showed an aneuploid phenotype (6%; = 112; P95 = 3–9) compared with wild type (1%; = 116; P95 = 0–3), and this low frequency was consistent with adult viability. Transient prometaphase delays observed in the mutant could be a consequence of defects in the mitotic apparatus. Nevertheless, immunofluorescence analysis revealed neither a clear difference in spindle morphology nor a significant decrease of cells stained positively for γ-tubulin or Dgrip84 (). Given the limitations of immunofluorescence in neuroblasts, we cannot exclude defects in microtubule organization, such as changes in microtubule density or interpolar distance. These data suggest that in brains, most of the cells complete cell division even though mitosis might be slowed down, consistent with our observations in RNAi-treated cultured cells. γ-TuSC components are still recruited to the centrosomes and most of the mitotic apparatus are able to ensure correct chromosome segregation. To evaluate whether the phenotypes induced by Dgrip75 depletion were specific of this grip-motif protein, we performed an RNAi treatment against Dgrip128 or Dgrip163, the two other grip-motif γ-TuRC–specific components. Depletion was efficient in either case: after Dgrip128 RNAi, no poles ( = 142) were labeled, and after Dgrip163 RNAi, <4% of the poles ( = 150) were labeled with respective Dgrip128 and Dgrip163 antibodies (). Western blot analysis confirmed this view (not depicted). Sucrose gradient analysis of extracts depleted for Dgrip128 or Dgrip163 showed a marked decrease in γ-TuRC content (). These experiments show that these proteins, like Dgrip75, appear involved in the assembly or stability of γ-TuRCs. Moreover, silencing of either protein led to accumulation of cells in mitosis (approximately three- to fourfold), which resulted from a higher frequency of prometaphases concomitant with a decrease in postmetaphase stages (). Mitotic phenotypes were mainly characterized by elongated bipolar spindles and less frequently by monopolar structures. γ-Tubulin () as well as Dgrip84 and Dgrip91 (not depicted) were efficiently recruited to the centrosomes, even in mitotic cells that exhibited severe phenotypes (i.e., monopolar spindles). However, γ-tubulin labeling along spindle microtubules was no longer observed (). It was striking that these phenotypes were similar to the defects obtained after depletion of Dgrip75. Furthermore, we next performed an in vivo study using a insertion mutant () in which the transposable element was inserted into exon 4 (out of 5) between codons 822 and 823. This mutant behaved genetically as a null or strong allele (unpublished data). mutants exhibited reduced viability. The emerging homozygous or hemizygous adults displayed abdominal abnormalities, and mutant females were sterile. Altogether, these studies reveal that inhibition of each individual grip-motif protein specific of the γ-TuRC leads to similar phenotypes in cultured cells and in vivo. In contrast to the other proteins specific of the γ-TuRCs, Dgp71WD does not belong to the same structural family (). RNAi treatment induced a strong inhibition, as judged by Western blot analysis (). In agreement with this quantification, Dgp71WD centrosomal recruitment was severely impaired in treated cells (1%; = 385; P95 = 0–2) compared with control cells (95%; = 296; P95 = 93–97; ). In treated cells, we noticed the systematic loss of Dgp71WD staining on other mitotic structures, such as spindle microtubules (, inset) and midbodies (not depicted). As Dgp71WD contains WD repeats and interacts in vitro with the four grip-motif subunits Dgrip84, Dgrip91, Dgrip128, and Dgrip163 (), it might play a role as a scaffold for cytosolic γ-TuRC assembly. When control extracts were subjected to sucrose gradient sedimentation, this protein appeared not only as part of γ-TuRCs but also of smaller uncharacterized complexes (). In Dgp71WD-depleted conditions (), no significant modification in the quantity of ∼30S complexes is observed, as judged by γ-tubulin or Dgrip75 labeling. This result suggests that Dgp71WD, in contrast to Dgrip75, appears to be dispensable for the assembly or the stability of large γ-tubulin complexes. However, we cannot exclude the possibility that the assembly of these complexes is mediated by residual amounts of Dgp71WD. Studies of the mitotic phenotypes observed after Dgp71WD down-regulation revealed similarities to the defects recorded after the removal of Dgrip75, Dgrip128, or Dgrip163; an increase of the mitotic index (2.7%; = 7,970; P95 = 2.3–3.1) compared with control cells (0.9%; = 19,650; P95 = 0.8–1); and an accumulation of cells in the prometaphase and metaphase stages (Fig. S2 A, available at ). Dgp71WD-depleted cells showed defects in spindle morphology, such as bipolar spindles with unfocused poles or monopolar structures (, B and D; and Fig. S2 B) with unseparated poles, as determined by Cnn staining (not depicted). γ-TuSC components (γ-tubulin, Dgrip84, and Dgrip91), as well as γ-TuRC–specific subunits (Dgrip75 and Dgrip163), were still targeted to the mitotic centrosomes, but often the stainings were weak ( and Fig. S2 C). These proteins were no longer detected at the spindles. These results suggest that Dgp71WD is dispensable for γ-TuRC assembly and recruitment, but is required for the efficient function of this complex. A genetic approach was performed to test this hypothesis in vivo. We used a mutant strain that carried a element insertion (GE30807) in the 5′ untranslated mRNA region at position 49 from the translation initiator ATG. As shown by Western blot, the level of Dgp71WD was strongly reduced in a mutant larval brain extract compared with a wild-type extract (). An immunofluorescence analysis performed in larval brains sustained that Dgp71WD was no longer detectable in mutants ( = 120), whereas 96% of wild-type poles ( = 110; P95 = 92–100) exhibited a clear staining (). These results indicate that is at least a strong allele. This was confirmed by analysis of hemizygous flies, using a chromosomal deficiency that covers . Most of the homozygous or hemizygous mutants reached the adult stage. However, they showed a shorter life, morphological abnormalities, and female sterility. Mitotic phenotypes had been analyzed in L3 larval brains confirming RNAi-mediated phenotypes. We observed a fourfold increase in the mitotic index, an accumulation in prometaphase stages associated with a significant hypercondensation of chromosomes and a high incidence of disorganized or monopolar spindles (Fig. S3, available at ). However, γ-tubulin is detected at all the centrosomes of neuroblasts ( = 130; ). Moreover, mutant neuroblasts exhibited an increase in aneuploidy (12%; = 139; P95 = 7–17) versus wild-type brains (1%; = 125; P95 = 0–5), suggesting that spindles without Dgp71WD were not fully functional. Whatever their role in the assembly and recruitment of cytosolic γ-tubulin complexes, the inhibition of each individual γ-TuRC subunit allowed efficient γ-TuSC anchoring to the centrosomes. It was possible that some grip-motif proteins were in part redundant and could complement each other to some extent. Alternatively, the residual γ-TuRC–specific proteins could account for the formation of a γ-TuRC–like structure that was unstable in the cytoplasm, but stabilized upon assembly with the pericentriolar matrix. Hence, we performed cosilencing of the four proteins Dgrip75, Dgrip128, Dgrip163, and Dgp71WD. The levels of the four proteins were dramatically reduced as judged by Western blot () and immunofluorescence analyses ( and ). The mitotic index was increased in treated cells (3.9%; = 2,552; P95 = 3.4–4.4) compared with control cells (1.1%; = 8,567; P95 = 0.8–1.4). Most of the mitotic structures consisted of monopolar prometaphases/metaphases (70%; = 80; P95 = 60–80) compared with control cells (1%; = 104; P95 = 0–3; ). A striking feature of this treatment was the significant accumulation of polyploid cells and the appearance of very large interphase cells (9%; = 281; P95 = 5–12), which appeared to account for only 0.6% of control interphase cells ( = 320; P95 = 0–1.4) (, left arrowhead). Most of the poles exhibited γ-tubulin, Dgrip84, and Dgrip91 stainings ( and ), but at reduced intensities. Spindle labeling was no longer detectable (). These parameters showed that codepletion of these proteins induced more severe phenotypes than after any individual down-regulation. This result was supported by the synthetic lethality observed when the - and -viable alleles were combined in the same genotype (unpublished data). Altogether, these data confirm that the γ-TuSC assembled in the cytoplasm can, at least to some extent, be directly targeted to the poles. #text RNAi was performed in S2 cells () according to . Cells were treated twice with RNAi at days one and five and harvested on day seven for immunoblottting and immunofluorescence staining. To perform cosilencing, cells were incubated in the same way, with the four double-stranded RNAs (dsRNAs) against , , and , and Dgp71WD together (20 mM each). The dsRNAs used correspond to positions 717–1,577, relative to start of translation, for (cDNA clone LD 19773); 157–918, relative to the start position for (clone GH 21414); 141–846, relative to the start position for (clone RE 43571); and 145–849, relative to the start position for (clone RE59956). These dsRNAs were generated from the cDNA plasmid clones as described in . or were used as control strains, and mutant strains (), , and (GenExel, Inc.) were used for the study. Each mutant chromosome had been balanced over , , or (). The GFP-balanced strains were used to select the homozygous mutant larvae. Viability of 100 selected larvae was studied during 28 d. The deficiencies and uncovering or , respectively, were used to produce hemizygous flies. Note that is referred as in the Flybase database (). The mouse monoclonal antibodies T-5168, GTU88 (Sigma-Aldrich), and 1501 (CHEMICON International, Inc.) were used to stain α-tubulin, γ-tubulin, and actin, respectively. The following rabbit antibodies were raised: R62 specifically directed against 23C γ-tubulin (); R403 against the COOH-terminal peptide of Dgrip91 (-CRLDFNEYYKKRDTNLSK) for Western blot and R7075 against the recombinant COOH-terminal Dgrip91 (415–917 aa) for immunofluorescence (); and R367 against the recombinant full-length Dgrip84. R300 was raised against recombinant full-length Dgrip75 and affinity purified on recombinant Dgrip75, which is produced in , R267 was raised against the recombinant COOH-terminal Dgp71WD (486–647 aa), and R360 against the recombinant COOH-terminal Dgrip163 (1150–1352 aa) was used for Western blot. For γ-tubulin detection, we used GTU88 for Western blot analysis and R62 for immunofluorescence experiments, except for double labeling, where monoclonal GTU88 antibody was used instead of polyclonal R62 antibody. We also used Rb666 against BubR1 (; gift from C. Sunkel, Universidad do Porto, Porto, Portugal), Rb3133 against Asp (; gift from D. Glover, University of Cambridge, Cambridge, UK), R19 against Cnn (; gift from T.C. Kaufman, Howard Hughes Institute, Indiana University, Bloomington, IN), and antibodies against Dgrip128 and Dgrip163 (; gift from Y. Zheng, Howard Hughes Institute, Carnegie Institution of Washington, Baltimore, MD). 5–40% sucrose gradient was prepared as described previously (). Cell extracts (500 μg of protein) were overlaid onto the gradient and centrifuged for 4 h and 30 min at 4°C in a swinging rotor (SW55.1; Beckman Coulter) at 45,000 rpm (average 150,000 ). 10 0.5-ml fractions were collected and 250 μl of each fraction were precipitated with cold methanol (−20°C). The nine fractions corresponding to the soluble part of the extracts were analyzed by immunoblotting. The calibration of the gradient was determined by running in parallel 0–3-h embryonic extracts that contain γ-TuSCs and γ-TuRCs. Protein extracts from cultured cells () and from total larval brains () were prepared and subjected to Western blot analyses (7.5% polyacrylamide gel and SDS [Sigma-Aldrich]). Apparent masses were determined by comparison with the SDS-PAGE molecular weight standards (broad range), which were obtained from Bio-Rad Laboratories. For immunostaining, S2 cultured cells and semisquashed L3 larval brains were fixed and permeabilized as previously described (). DAPI staining of squashed third instar larvae was performed as described previously (). For detection, secondary antibodies conjugated to Alexa Fluor 488 or 568 (Invitrogen) were used. Fluorescence microscopy was performed on a microscope (Axiovert; Carl Zeiss MicroImaging, Inc) equipped with a Z-motor, using 100×, 1.4 NA, objectives. Z-series images were acquired with a camera and software (Axiocam MRm and Axiovision; Carl Zeiss MicroImaging, Inc.). S2 cell or brain images were subsequently deconvolved using Axiovision, and Z-planes were projected onto a single view. The percentages of the different mitotic phenotypes were determined with confidence intervals calculated for a P95. Usually, image processing and quantification of fluorescence were done using Photoshop (Adobe). For quantification of polar γ-tubulin by immunofluorescence, we measured the maximal signal obtained after γ-tubulin staining. The signal of the camera was proportional to the fluorescence intensity, and the background and the maximal fluorescence values were adjusted to 0 and 25, respectively (). Fig. S1 shows colabeling of γ-tubulin with associated proteins in Dgrip75-depleted cells. Fig. S2 shows phenotypes observed after Dgp71WD depletion in S2 cells. Fig. S3 shows mitotic phenotypes in the mutant. Fig. S4 shows the number of microtubule protofilaments in mitotic cells after Dgrip75 RNAi treatment. Online supplemental material is available at .

Maintenance of genetic stability requires faithful segregation of duplicated chromosomes during mitosis. Most human cancers consist of cells that have an abnormal chromosome content, which is also known as aneuploidy. However, how the aneuploidy develops remains poorly understood (; ; ). The spindle assembly checkpoint, commonly referred to as the mitotic checkpoint, is a surveillance mechanism that ensures accurate segregation of mitotic chromosomes by delaying metaphase-to-anaphase progression until each kinetochore has properly attached to the mitotic spindle (). Kinetochores that are not yet attached to mitotic microtubules and chromosome pairs that lack tension across sister chromatids generated by the spindle poles activate the spindle assembly checkpoint (; ). Various mitotic checkpoint proteins, including Bub3, Bub1, BubR1, Mad1, and Mad2, bind to kinetochores that lack attachment or tension and generate a “stop anaphase” signal that diffuses into the mitotic cytosol (). This signal is believed to consist of Bub3, BubR1, and Mad2 protein complexes (; ; ). These complexes bind to Cdc20 to prevent premature ubiquitination of cyclin B and securin by the anaphase-promoting complex (APC; ; ). Checkpoint inactivation occurs after proper alignment of the mitotic chromosomes at the metaphase plate. This triggers release of Bub3, BubR1, and Mad2 protein complexes from Cdc20 and subsequent destruction of cyclin B and securin by the APC. Separase, which in mammalian cells is inhibited through association with securin and also by cyclin B/Cdk1-mediated phosphorylation, triggers sister chromatid disjunction by cleavage of the cohesin subunit Scc1 (). Various studies have addressed the physiological relevance of the spindle assembly checkpoint by deleting Mad or Bub genes in the mouse (for review see ). Mice that are homozygous null for Mad2, Bub3, or BubR1 die during the early stages of embryogenesis (; ; ; ; ), with a small subset of Mad2-null embryos remaining viable until embryonic day 10.5 (). On the other hand, mice that are heterozygous null for these genes are born alive and seem to exhibit no overt phenotypes. Thus, the challenge for studying the function of individual mitotic checkpoint genes in the mouse is to disrupt their function significantly but not so severely that the embryo dies. Recently, we have reported the generation of a series of mice in which the expression of BubR1 is reduced in a graded fashion from normal levels to zero by the use of wild-type, knockout, and hypomorphic alleles (). Of this series, mice with BubR1 levels as low as 10% of normal (BubR1 mice) are born alive and develop into adult mice. Remarkably, despite having severe aneuploidy, these animals do not have an escalated spontaneous tumor burden (). Instead, these mice develop a variety of progeroid (resembling old age) features. In addition, mouse embryonic fibroblasts (MEFs) from BubR1 mice show profound premature cellular senescence. These observations, combined with the demonstration that BubR1 levels decline in ovary, testis, and spleen tissue as wild-type mice age, have suggested that this mitotic checkpoint protein is a key regulator of the normal aging process. However, whether aging is a general feature of spindle assembly checkpoint dysfunction remains unknown. The mitotic checkpoint protein Bub3 is a seven-blade β propeller that has substantial sequence similarity with the mRNA export factor protein Rae1; the two proteins have 34% identity and 52% similarity in humans (). Like Bub3 (), Rae1 is essential for early mouse embryogenesis, but mice that are haploinsufficient for Bub3 or Rae1 are viable and have a normal appearance (). These heterozygous mice show remarkably similar mitotic defects, including spindle assembly checkpoint impairment and chromosome missegregation (). Mice that are haploinsufficient for both Bub3 and Rae1 are also viable but develop much more severe mitotic phenotypes than single haploinsufficient mice. In this study, we have investigated the long-term phenotypes of mice in which the mitotic checkpoint proteins Bub3 and Rae1 are disrupted individually or in combination. We find that phenotypes associated with aging appear early in double haploinsufficient mice but not in single haploinsufficient mice. We also find that single and double haploinsufficient mice are not predisposed to spontaneous tumor development despite the accumulation of substantial numbers of aneuploid cells. To examine the long-term biological consequences of impaired spindle assembly checkpoint function, we generated cohorts of mice in which the mitotic checkpoint genes Bub3 and Rae1 were haploinsufficient individually or in combination (). We monitored 70 wild-type, 56 Bub3, 96 Rae1, and 100 Bub3/Rae1 mice daily for spontaneous tumor development or ill health for a period of >2 yr. During the first year of life, Bub3/Rae1, Bub3, and Rae1 mice were indistinguishable from wild-type mice and exhibited no overt abnormalities (; and not depicted). However, in the second year, a substantial proportion of Bub3/Rae1 mice began to develop an aged appearance (). Wild-type mice did not show this appearance (), nor did Bub3 or Rae1 mice. The median overall survival of Bub3 and Rae1 mice was slightly reduced in comparison with wild-type animals (24 vs. 25 mo, respectively), but these differences were not significant (). In contrast, the median survival of Bub3/Rae1 mice was significantly reduced to 22 mo, which translates into a 12% decrease in lifespan. We note that the reduced lifespan of Bub3/Rae1 mice may be the result of the sum of the lifespan reductions in Bub3 and Rae1 mice. We performed biopsies on moribund mice from our cohorts to investigate whether chromosome number instability predisposes mice to spontaneous tumorigenesis. The tumor incidences of Bub3, Rae1, and Bub3/Rae1 mice were not significantly different from wild-type mice (). In addition, the tumor spectra exhibited in these groups were nearly identical. Comparison of the tumor-free survival curves of the various genotypes revealed a slight decrease in tumor latency in Bub3/Rae1 mice compared with wild-type mice; however, this decrease is not significant (). The observed slight reduction in tumor latency most likely relates to the lifespan reduction of ∼3 mo in the Bub3/Rae1 group. This difference is mirrored by the reduction in tumor latency by ∼4 mo. Thus, mice in which the mitotic checkpoint genes Bub3 and Rae1 are disrupted individually or in combination are not predisposed to spontaneous tumorigenesis. Chromosome counts on splenocytes from 5-mo-old mice have shown that Bub3 and Rae1 single heterozygotes have significant numbers of aneuploid cells and that combined heterozygosity dramatically increases these numbers (; ). We sought to determine whether aneuploidy progressed with age in Bub3, Rae1, and Bub3/Rae1 mice. Indeed, aneuploidy in single haploinsufficient Bub3 and Rae1 mice increased dramatically over time. Specifically, the aneuploidy in Bub3 splenocytes increased from 9% at 5 mo to 29% at 27 mo, whereas in Rae1 splenocytes, it increased from 9% at 5 mo to 33% at 24 mo (). In splenocytes of Bub3/Rae1 mice, aneuploidy was already 37% at 5 mo of age but further increased by 10% over the next 19 mo. Wild-type mice had no aneuploidy at 27 mo, but 3% aneuploidy was detected at 35 mo. Aside from aneuploidy, another feature of defective or weakened spindle assembly checkpoint activity in mice is premature sister chromatid separation (PMSCS; ; ). Mitotic figures from Bub3 and Rae1 single heterozygotes showed no PMSCS at 5 mo, but 10–11% of metaphases examined at 24–27 mo displayed this feature. As Bub3/Rae1 mice aged, the percentage of cells exhibiting PMSCS also increased (), going from 14% at 5 mo to 24% at 24 mo. The increases in PMSCS in Bub3, Rae1, and Bub3/Rae1 mice suggest that the activity of the spindle assembly checkpoint further declines as these animals age. Splenocytes from wild-type mice displayed no PMSCS at 27 mo, but 4% of splenocytes had PMSCS at 35 mo, suggesting that checkpoint activity declines as normal mice reach an extremely old age. We observed no chromosome breaks or fusions in metaphase spreads from 27-mo-old wild-type and 24-mo-old Bub3, Rae1, and Bub3/Rae1 mice (). Thus, aneuploidy and PMSCS progress as mice in which Bub3 and Rae1 are disrupted individually or in combination age. Furthermore, mild aneuploidy and PMSCS are features of very old wild-type mice. The finding that several Bub3/Rae1 mice displayed an aged appearance at a relatively young age prompted us to screen for aging-associated phenotypes in a systematic fashion. All mice of our cohort were screened every 2 wk for the development of cataracts and lordokyphosis (abnormal convexity in the curvature of the spine when viewed from the side). This monitoring strategy revealed that the incidence of cataract formation was significantly higher in Bub3/Rae1 mice than in Bub3, Rae1, and wild-type mice (). Furthermore, the latency of cataract formation was shorter in Bub3/Rae1 mice than in mice of the other genotypes. Histological examination confirmed that cataractous eyes from Bub3/Rae1 mice had features reminiscent of human age–related cataracts (). Bub3/Rae1 mice developed lordokyphosis at an earlier age and with greater severity than Bub3, Rae1, and wild-type mice (). Mutant mice with early lordokyphosis, but not age-matched control animals, showed clear signs of skeletal muscle atrophy and degeneration (). We next screened for additional aging-associated phenotypes by comparative analysis of mice from our Bub3/Rae1 and wild-type cohorts. A well-established feature of aging is loss of body weight (). Although the weight of Bub3/Rae1 mice at 5 mo was similar to that of age-matched wild-type mice, Bub3/Rae1 mice had significantly lower body mass than wild-type mice at 24 mo (). Dual energy X-ray absorptiometry was used to determine whether this loss of weight was caused by a reduction in total body fat. At 5 mo of age, Bub3/Rae1 and wild-type mice had similar percentages of total body adipose tissue; however, at 27 mo, Bub3/Rae1 showed a dramatic reduction in body fat compared with wild-type mice (). Reduced dermal thickness and subdermal adipose are characteristics of aged mice (; ). Histological examination of the skin revealed significant reductions in subcutaneous adipose thickness in 27-mo-old Bub3/Rae1 mice but not in age-matched control animals (). The mean dermal thickness of 27-mo-old Bub3/Rae1 mice was 21% lower than that of 5-mo-old Bub3/Rae1 mice (), whereas in 27-mo-old wild-type mice, it was only 9% lower than in 5-mo-old wild-type mice (). We then investigated whether the age-related phenotypes present in Bub3/Rae1 mice could also be observed in very old wild-type mice. We found that wild-type mice of 35 mo had lordokyphosis (), muscle atrophy (), and osteoporosis (not depicted). These mice also had significantly lower amounts of total body fat and subcutaneous adipose tissue than 5-mo-old wild-type mice (). As expected, the dermis of 35-mo-old wild-type mice was significantly thinner than that of 5-mo-old wild-type mice (). Together, the aforementioned data show that several aging-associated phenotypes appear early in mice that are double haploinsufficient for the mitotic checkpoint genes Bub3 and Rae1 but not in mice that are single haploinsufficient for these genes. The data also show that most of these phenotypes do occur in wild-type mice of extremely advanced age. We next investigated whether combined haploinsufficiency of Bub3 and Rae1 triggers cell death or cellular senescence, which are two processes that have been linked to aging (; ). We intercrossed Bub3 and Rae1 mice to produce Bub3/Rae1, Bub3, Rae1, and wild-type MEFs from 13.5-d-old fetuses. At passage 5 (P5), we assayed Bub3/Rae1 and wild-type MEFs for cell death by TUNEL staining. We observed no difference in amounts of TUNEL-positive cells in Bub3/Rae1 and wild-type MEFs cultures (), indicating that combined haploinsufficiency of Bub3 and Rae1 did not increase the rate of cell death. Combined annexin V–FITC and propidium iodide (PI) staining of P5 Bub3/Rae1 and wild-type MEFs showed that neither apoptotic nor nonapoptotic cell death was significantly increased in Bub3/Rae1 MEFs () despite high rates of chromosome missegregation in mitosis. This is remarkable because high rates of cell death have been observed in studies in which the mitotic checkpoint proteins Mad2 and BubR1 were depleted by short inhibitory RNA (; ; ). Thus, it seems that cell death only occurs when mitotic checkpoint proteins drop below a critical threshold level and that aberrant chromosome segregation itself is insufficient to trigger cell death. Cellular senescence is characterized in vitro by a decline in growth rate (). At P3, Bub3/Rae1 growth rates were similar to those of Bub3, Rae1, and wild-type MEFs (). On the other hand, at P7, Bub3/Rae1 MEFs showed significantly reduced growth rates compared with MEFs of the other three genotypes (). At P3, MEF cultures of all four genotypes had comparable numbers of cells that were positive for senescence-associated (SA) β-galactosidase activity (). However, at P5 and P7, Bub3/Rae1 MEFs showed a profound increase in the number of cells staining positively for SA β-galactosidase in comparison with Bub3, Rae1, and wild-type MEFs. Consistent with these data, the senescence response genes p53, p21, p19, and p16 were induced earlier in Bub3/Rae1 MEFs than in Bub3, Rae1, and wild-type MEFs (). Together, these findings indicate that compound heterozygosity of Bub3 and Rae1 causes early onset of cellular senescence. The induction of p53, p21, and p19 indicates that the lesion triggering cellular senescence in Bub3/Rae1 double haploinsufficient MEFs activates the p53 DNA damage response pathway. To investigate whether activation of the p53 pathway might be caused by defective DNA repair, we analyzed the fidelity of distinct DNA damage repair pathways by measuring cell survival and colony formation ability after exposing Bub3/Rae1 MEF cultures to various kinds of DNA-damaging agents. As shown in , the DNA repair capacities of Bub3/Rae1 and wild-type MEFs were very similar, suggesting that increased accumulation of DNA damage is an unlikely cause of increased senescence in Bub3/Rae1 cultures. The early aging–associated phenotypes that develop in Bub3/Rae1 mice resemble those detected in mice with low levels of BubR1 (). However, BubR1 mice had a much earlier onset of aging phenotypes than Bub3/Rae1 mice. Consistent with this, BubR1 hypomorphic MEF cultures contained significantly higher numbers of SA β-galactosidase–positive cells than Bub3/Rae1 MEF cultures (). In addition, BubR1 MEFs showed a more robust activation of the p53 and p16 pathways than Bub3/Rae1 MEFs (). These data draw a clear correlation between the level of induction of senescence response genes and the rate of premature aging. To further test this correlation, we prepared MEFs from Mad2 haploinsufficient mice and measured the degree of senescence at P3, P5, and P7. At each passage, the percentage of SA β-galactosidase–positive Mad2 MEFs was similar to that of wild-type MEFs (Fig. S1 A, available at ). Furthermore, at each passage, p53, p21, p19, and p16 levels of Mad2 MEFs were comparable with those of wild-type MEFs (Fig. S1 B). Consistent with these findings, we observed no early aging–related phenotypes in the small cohort of Mad2 mice that we followed for a period of 19–27 mo (Fig. S1 C). Chromosome counts showed that 18% of Mad2 splenocytes were aneuploid when mice were 5 mo of age (Fig. S1 D). Thus, Mad2 haploinsufficiency does not promote cellular senescence in MEFs and does not seem to trigger aging-related phenotypes in mice, although aneuploidy is increased. Our chromosome counts suggested that there is a rather weak link between checkpoint dysfunction and the early onset of senescence and aging. To further investigate this issue, we obtained two additional measures for checkpoint activity by methods involving live cell imaging. The first assay is a nocodazole challenge assay. MEFs carrying various mitotic checkpoint gene defects were first transduced with a retrovirus carrying a YFP-tagged H2B gene to allow visualization of chromosomes by fluorescent microscopy. We then challenged the MEFs with nocodazole and measured the duration of the ensuing checkpoint-dependent mitotic arrest. This duration of arrest in mitosis was defined as the interval between nuclear envelope breakdown (NEBD; onset of mitosis) and chromatin decondensation (exit from mitosis without cytokinesis). We used the time at which 50% of the cells had exited mitosis for comparison. Nocodazole-challenged wild-type MEFs typically remained arrested in prometaphase for 7.2 h (). However, single haploinsufficient MEFs of Bub3 and Rae1 showed a profound decrease in their ability to maintain this arrest, with 50% of the cells exiting at 3 h. Bub3/Rae1 MEFs had an even further reduction in their ability to sustain arrest, with 50% of the cells exiting mitosis at 2.2 h. BubR1 and BubR1 MEFs, which express ∼60 and 30% of normal BubR1 protein, respectively (), also had significant reductions in their ability to maintain a mitotic arrest, with 50% of the cells exiting at 5.3 and 3.8 h (). BubR1 MEFs (∼10% BubR1) had the most profound inability to maintain mitotic arrest, with 50% of the cells exiting at 1.1 h. It surprised us that the duration of arrest of BubR1 MEFs was substantially shorter than that of Bub3/Rae1 MEFs, because BubR1 MEFs have less severe aneuploidy than Bub3/Rae1 MEFs (; also see Discussion). As an alternative measure for spindle checkpoint activity, we measured the accuracy of chromosome segregation during mitosis. In essence, we followed YFP-H2B–positive MEFs through an unchallenged mitosis by live cell imaging and determined the fraction of mitotic cells with chromosome segregation defects. We observed little or no misaligned chromosomes in wild-type, Bub3, Rae1, Bub3/Rae1, and BubR1 metaphases (). However, a relatively large proportion of Bub3/Rae1 and BubR1 anaphases had lagging chromosomes (, Videos 1–3, and Fig. S4, available at ). The incidence of lagging chromosomes was also increased in Bub3 and Rae1 anaphases, but to a lesser extent. It should be noted that lagging chromosomes are not necessarily a sign of failed mitotic checkpoint signaling because they could result from merotelically attached kinetochores that do not activate the mitotic checkpoint (). Bub3/Rae1 MEFs had not only a somewhat higher percentage of anaphases with lagging chromosomes than BubR1 MEFs but also a higher percentage of abnormal anaphases with more than one lagging chromosome ( and Video 3). This may explain why the spectrum of chromosome losses and gains is wider for Bub3/Rae1 MEFs than for BubR1 MEFs (). Collectively, the aforementioned data confirm that spindle assembly checkpoint dysfunction and chromosome missegregation do not correlate well with cellular senescence and aging. In this study, we used mice in which the mitotic checkpoint proteins Bub3 and Rae1 were disrupted individually or in combination to address whether premature aging is a common consequence of mitotic checkpoint gene defects. We found that Bub3 and Rae1 single haploinsufficient mice exhibit no signs of early aging despite their weakened checkpoint and progressive aneuploidy. Conversely, several aging-related phenotypes appear early in mice that are haploinsufficient for both Bub3 and Rae1. However, BubR1 hypomorphic mice age much faster but have less aneuploidy than Bub3/Rae1 haploinsufficient mice. This suggests that spindle assembly checkpoint dysfunction, which leads to the accumulation of aneuploidy, is unlikely to cause premature aging in BubR1 and Bub3/Rae1 mutant mice. We find that only mitotic checkpoint gene defects that can provoke cellular senescence in MEFs cause the early onset of aging-associated phenotypes in mice, with levels of induction of p19, p53, p21, and p16 highly correlating with the rate of cellular senescence and aging. In the study of hypomorphic BubR1 mice (), it became apparent that as these mice began to show signs of premature aging, they also started to accumulate aneuploid splenocytes. This led to the hypothesis that aneuploidy might be the primary lesion triggering cellular senescence and the process of aging. The use of Bub3 and Rae1 haploinsufficient mice allowed for rigorous testing of this hypothesis. At 5 mo, BubR1 hypomorphic mice have an aged appearance with detectable aneuploidy in 15% of their splenocytes (). Age-matched single heterozygous Bub3 and Rae1 mice each have 9% aneuploid splenocytes but show no signs of early aging. One could argue that this may be the result of having aneuploidy that remains below a critical level. Splenocytes from 5-mo-old compound Bub3/Rae1 mice have 36% aneuploidy and a wide spectrum of chromosome losses and gains (), but these animals show no overt features of aging at this age. If aneuploidy alone was a driving force of aging, these mice should have shown phenotypes of aging earlier than BubR1 hypomorphic mice. It is possible that the degree of aneuploidy in splenocytes is not representative of the degree present in tissues and cell types that develop early aging phenotypes. Although we cannot exclude this possibility with our current methods for measuring aneuploidy in vivo, the view that aneuploidy does not correlate with aging is further underscored by the finding that Bub3/Rae1 MEFs are more aneuploid than BubR1 hypomorphic MEFs but exhibit lower levels of cellular senescence. In contrast, in mice with mitotic checkpoint gene defects, the rate of aging strongly correlates with the rate of cellular senescence. The induction of p53, p21, and p19 in both BubR1 hypomorphic and combined Bub3/Rae1 haploinsufficient cells but not Bub3 and Rae1 single haploinsufficient cells suggests that the lesion triggering cellular senescence and aging activates the p53 pathway. Like BubR1 MEFs (), Bub3/Rae1 MEFs have no detectable defects in DNA repair, suggesting that the lesion that triggers p53 activation is unlikely to be the accumulation of DNA damage or mutations. Various stress signals other than DNA damage are known to activate the p53 pathway (; ), and future experiments will be necessary to determine whether any of these signals might be active in BubR1 hypomorphic and Bub3/Rae1 mutant mice. Rae1, Bub3, and BubR1 have all been implicated in functions outside of the spindle assembly checkpoint. For example, Rae1 was originally discovered as a nucleocytoplasmic transport factor that regulates the export of mRNA from the nucleus (; ; ). How Rae1 regulates nuclear transport and whether Bub3 functions also in the transport machinery has not yet been established. We found that combined Bub3/Rae1 haploinsufficiency does not cause overt defects in global mRNA export or nuclear pore complex biogenesis (Fig. S2, available at ). However, our experiments by no means represent a comprehensive analysis of the transport machinery, and it is certainly possible that Bub3/Rae1 haploinsufficient cells have subtle trafficking defects. A recent study has suggested that Bub3 can interact with histone deacetylases and that, in this context, Bub3 might act to repress gene transcription in interphase (). BubR1 can also repress gene transcription in a heterologous DNA-binding context, although not in conjunction with histone deacetylases (). BubR1 also seems to mediate apoptosis of cells that exit mitosis without chromosome segregation (), and, in yeast, the BubR1 homologue Mad3 and Bub3 have both been linked to the accumulation of gross chromosomal rearrangements (). Thus, it is possible that functions outside of the spindle assembly checkpoint are compromised in BubR1 hypomorphic and Bub3/Rae1 haploinsufficient mice and that early onset of aging-related phenotypes is linked to such defects. Although BubR1 hypomorphic mice age much faster than Bub3/Rae1 mutant mice, both models have a common spectrum of age-related phenotypes that is distinctive from that of mouse strains in which DNA damage repair or telomere maintenance defects cause premature aging (; ). For example, mice with mitotic checkpoint gene defects develop cataracts at high incidence, whereas the other models do not. Conversely, mice with defects in DNA damage repair or telomere maintenance show hematopoietic stem cell depletion and glucose intolerance, whereas mice with a weakened mitotic checkpoint do not (unpublished data). This distinction supports the idea that the lesion that triggers aging in BubR1 hypomorphic mice and Bub3/Rae1 mutant mice is distinct from that of other aging models. This suggestion is further supported by the demonstration that DNA damage repair and telomere maintenance functions are seemingly intact in Bub3/Rae1 haploinsufficient mice and BubR1 hypomorphic mice (). In this study, we have used a modified nocodazole challenge assay to measure the duration of sustained spindle checkpoint activity of MEFs with various mitotic checkpoint gene defects. In essence, we treated MEFs with nocodazole and determined how long individual cells remained in mitosis by live cell imaging and then used this time as a measure for spindle checkpoint activity. We validated our nocodazole challenge assay by using a series of MEFs with a graded reduction of BubR1 protein level. We found that the duration of arrest in mitosis declined as the amount of BubR1 decreased. We found further that this time of arrest correlated well with the rate of aneuploidy and chromosome missegregation. However, the finding that Bub3/Rae1 double-insufficient MEFs are more aneuploid than BubR1 MEFs (the percentage of aneuploid cells is higher and the spectrum of chromosome losses and gains is wider) appears inconsistent, as we find that Bub3/Rae1 haploinsufficient MEFs sustain their mitotic arrest in nocodazole for twice as long as BubR1 MEFs. A possible explanation for this discrepancy is that the nocodazole challenge assay may not provide a reliable measure for spindle assembly checkpoint activity; the reasoning is that the checkpoint functions to protect cells against chromosome missegregation rather than against microtubule poisoning. Alternatively, increased aneuploidy in Bub3/Rae1 haploinsufficient cells might be caused by the dysfunction of proteins involved in chromosome segregation but not in the spindle assembly checkpoint. Examples of such proteins are AdAPC (adenomatous polyposis coli; ), securin (), and separase (; ). Abnormal numbers of chromosomes, termed aneuploidy, is one of the most common properties of cancer cells (). Because aneuploidy occurs with such high frequency, it has been suggested that aneuploidy is an essential step in the process of tumorigenesis (). In this study, we have demonstrated that Bub3, Rae1, and Bub3/Rae1 mice do not show any differences in both the rate of tumor formation and the type of tumor formed in comparison with wild-type animals. An earlier cancer susceptibility study on Bub3 mice yielded similar data (). That study further showed that even on a p53 or Rb1 heterozygous background, Bub3 haploinsufficiency had no significant effect on tumorigenesis. A suggested explanation for these findings was that the degree of impairment of the checkpoint machinery was too low to cause substantial aneuploidy (). However, although the aneuploidy in our double heterozygous Bub3/Rae1 mice is quite severe at 5 mo, we find no increased susceptibility to lifetime spontaneous tumor development. This result is consistent with our earlier observation that BubR1 hypomorphic mice are not tumor prone despite having severe aneuploidy (). So far, only Mad2 haploinsufficiency has been shown to increase the risk for spontaneous tumors in mice (). However, several studies have shown that impaired spindle checkpoint function increases the risk for carcinogen-induced tumors. For instance, we have shown previously that Rae1 and Bub3/Rae1 mice are susceptible to lung adenoma formation after early postnatal treatment with dimethylbenzanthrene (DMBA; ). Others have shown that BubR1 mice have a susceptibility to azoxymethane-induced intestinal and lung tumor development (), and, consistent with this finding, we observed that DMBA-treated BubR1 mice are prone to lung tumors (Fig. S3, available at ). All of these studies illustrate an important point that challenging these mice with carcinogenic agents introduces cooperating mutations that synergize with mitotic checkpoint failure to cause tumorigenesis and that these mutations may not occur under normal situations. It seems that the identification of gene mutations that cooperate with spindle assembly checkpoint gene alterations in tumorigenesis will be important to provide a better understanding of the role of aneuploidy in cancer. The link between spindle assembly checkpoint impairment and cancer predisposition is further supported by a recent report showing that a high proportion of individuals with mosaic variegated aneuploidy (MVA) syndrome carry bi-allelic mutations in the BubR1 gene (). MVA is a rare recessive disorder that is characterized by severe aneuploidy and a high risk for neoplasms such as rhabdomyosarcoma, Wilms tumor, and leukemia (; ). Other characteristics of MVA syndrome include microcephaly, growth retardation, and eye abnormalities such as cataracts. Whether MVA patients with bi-allelic mutations in BubR1 develop progeroid features other than growth retardation and cataracts remains to be established: so far, only four patients with BubR1 mutations have been identified, three of which are <3 yr of age. In a recent study, confirmed that MEFs from Bub3 haploinsufficient mice have increased aneuploidy, but the increase was not as profound as the one we previously observed (). The smaller increase most likely results from the fact that chromosome counts were performed at P3 instead of P5. Furthermore, BubR1 mice generated by typically exhibited splenomegaly and extramedullary megakaryopoiesis, but we did not detect these phenotypes in our BubR1 and BubR1 mice (). It is unclear why that is, but it might be the result of the different methods used to create the models (enhancer trapping vs. homologous recombination), differences in the mouse housing facilities, and/or possible differences in genetic background. Bub3 and Rae1 heterozygous knockout mice and BubR1 hypomorphic mice were generated as described previously (; ). These mice were derived from 129Sv/E embryonic stem cells and maintained on a mixed 129Sv/E × C57BL/6 genetic background. Mad2 mice were gifts from R. Benezra (Weill Medical College of Cornell University, New York, NY; ). All mice were housed in a pathogen-free barrier facility for the duration of the study. Experimental procedures involving laboratory mice were reviewed and approved by the Institutional Animal Care and Use Committee of the Mayo Clinic. Prism software (GraphPad Software, Inc.) was used for the generation of survival curves and for statistical analysis. DMBA treatments were performed as described, except the duration of treatment was 4 mo instead of 5 (). Metaphase spreads were prepared from splenocytes as described previously (). The aneuploidy data for Bub3, Rae1, and Bub3/Rae1 MEFs presented in were first published in ; the data for BubR1 MEFs and mice were first published in . Moribund mice were killed, and major organs were screened for overt tumors using a dissection microscope. Collected tumors were processed for histopathology by standard methods. A chi square or Fisher's exact test was used to compare tumor incidence proportions across the genotypes for mice that developed tumors. Board-certified pathologists assisted in tumor analysis. Biweekly, mice were screened for overt cataracts by examining dilated eyes with a slit light. At biopsy, cataractous eyes were collected and processed by standard methods for histological evaluation. Body fat content was measured using a small animal densitometer (Lunar PIXImus Corp.) as described previously (). A Mann-Whitney test was used for statistics. Dorsal skin was dissected, fixed in 10% formalin, processed, and embedded in paraffin. 5-μm cross sections were prepared and stained with hematoxylin and eosin using standard procedures. Sections were stained, and the thickness of the dermal and adipose layers was calculated using a calibrated computer program (Spot Advanced; BioSpot). 40 random measurements were taken for each mouse of each genotype at the indicated age ( = 4 males). Histopathology on gastrocnemius muscle was performed as previously described (; ). A Mann-Whitney test was used for statistics. MEFs were generated from wild-type Bub3, Rae1, or Bub3/Rae1 13.5-d-old embryos as previously described (). Growth curves of MEFs were generated as previously described (). BubR1, BubR1, and BubR1 MEFs were previously generated (). TUNEL/Hoechst stainings were performed on P5 MEFs according to the manufacturer's protocol (Roche), as were annexin V/PI stainings (BD Biosciences). FACS analysis was performed as previously described (). DNA damage survival experiments were performed as described previously (). MEFs were stained with an SA β-galactosidase kit according to the manufacturer's protocol (Cell Signal Technology). Nuclei were stained with Hoechst for visualization. The percentage of senescent cells is the total number of senescent cells divided by the total number of cells. We performed Western blot analysis as previously described (). Antibodies for SA proteins were used at a 1:200 dilution (purchased from Santa Cruz Biotechnology, Inc. unless otherwise noted): p16 (M-156), p19 (NB200-106; Novus Biologicals), p21 (M-19), and p53 (FL-393-G). β-actin was used as a loading control (1:50,000 dilution; Sigma-Aldrich). To allow visualization of chromosomes by fluorescent microscopy on live cells, we used a retrovirus expressing YFP-tagged H2B that was constructed as previously described (). P2 MEFs cells were seeded in T25 flasks at 50% confluence and cultured in DME/10% FBS. The next day, these MEFs were grown for 24 h in medium harvested from EcoPACK cells that produce pMCSV-puro H2B-YFP virus. 24 h after transduction with pMCSV-H2B-YFP retrovirus, MEFs were seeded in 35-mm glass-bottomed culture dishes (MatTek Corporation). The next day, Hepes was added to the culture medium at a final concentration of 20 mM. 4 h later, the dish was placed in a heat-controlled stage of a microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.). The temperature was maintained at 37°C. CO levels were maintained at 10% using a controller (CTI 3700; Carl Zeiss MicroImaging, Inc.). For nocodazole challenge experiments, nocodazole was added to a final concentration of 100 ng/ml. During the next 30 min, 20–30 cells undergoing NEBD were marked by the use of a mark/find module (AV4 MOD; Carl Zeiss MicroImaging, Inc.). Subsequently, time-lapse sequences were captured by using a plan Apo 63× NA 1.4 differential interference contrast (D = 0.18) oil objective (Carl Zeiss MicroImaging, Inc.) and a camera (AxioCam Hrm; Carl Zeiss MicroImaging, Inc.). The exposure times in nocodazole challenge experiments were always 100 ms at 2 × 2 binning. Interframe intervals were 15 min except for wild-type, BubR1, and BubR1 MEFs (intervals were 30 min). Imaging WS/20A software (Carl Zeiss MicroImaging, Inc.) was used to determine the time of arrest in mitosis of individual cells. The time of arrest in mitosis was defined as the interval between NEBD (onset of mitosis) and chromatin decondensation (exit from mitosis without cytokinesis). For analysis of mitotic defects, cells entering mitosis were marked, and images were acquired at 2- or 3-min interframe intervals until the completion of mitosis. Indirect immunofluorescence for mAb414 was performed as described previously (), as were stainings for poly(A) RNA (). Fig. S1 shows that Mad2 haploinsufficiency does not accelerate cellular senescence or aging. Fig. S2 shows that Bub3/Rae1 MEFs have no overt defects in nuclear pore complex architecture and mRNA export. Fig. S3 shows that BubR1 mice are highly susceptible to DMBA-induced tumor formation. Fig. S4 shows examples of BubR1 anaphases with lagging chromosomes. Video 1 shows a Bub3/Rae1 cell undergoing normal mitosis. Videos 2 and 3 show mitotic Bub3/Rae1 cells with one and two lagging chromosomes, respectively. Online supplemental material is available at .

Several recent studies have demonstrated that the cytoskeletal protein actin has important functions within the nucleoplasm (; ; ). The first identified function that unambiguously required a nuclear pool of nonmuscle actin was chromatin remodeling. β-Actin and a series of recently discovered actin-related proteins have been characterized both biochemically and genetically as components of chromatin remodeling complexes (). The role of actin in the nucleus is not limited to chromatin remodeling. β-Actin was recently found to bind specifically to hrp65-2; blocking this interaction in vivo resulted in a dramatic reduction in RNA polymerase II transcription (). β-Actin also resides as a component of the Balbiani ring gene messenger RNP (mRNP) throughout its nuclear maturation and export to the cytoplasm (). These complexes appear to require β-actin in its monomeric form. More recently, studies have shown that actin is associated with and required for transcription by RNA polymerase I (; ), II (; ), and III (). The cellular concentration of actin is typically at least 100–1,000 times higher than the concentration required for actin to spontaneously polymerize in vitro (). The concentration of nuclear actin is therefore sufficient to spontaneously polymerize and will require active processes to prevent polymerization. Indeed, several critical regulators of actin polymerization can be found in the nucleoplasm. For example, phosphoinositide signaling occurs within the nucleus and is initiated by a nucleus-specific isoform of phospholipase C (). The machinery for phosphoinositide signaling is enriched in splicing factor compartments (; ; ). Much of the nuclear monomeric actin pool has recently been reported to be bound in a complex with actin depolymerizing factor/cofilin (). Other proteins that sequester monomers of actin (), cap filaments (; ), or are directly involved in regulating the polymeric state of actin (; ; ) are present in the nuclei. The regulation of nuclear actin is therefore likely to include the controlled polymerization and depolymerization of actin and involve both mechanisms that stimulate polymerization, such as phosphoinositides, and molecules that inhibit actin polymerization or regulate the size of actin polymers. Specific affinity reagents for globular species of actin and filamentous species of actin have been used to detect nuclear actin and nuclear actin filaments (; ; ; ; ; ; ; ; ; ; ). Unfortunately, the conclusions of these studies are complicated by the fact that the different reagents used in these studies lead to different results and different conclusions on localization and polymerization state of the nuclear actin pool. Consequently, current models of nuclear actin function are speculative, with most discussions on the subject proposing that nuclear actin exists and functions as a monomer () or as short oligomers (; ; ). In this study, we take the first step toward quantifying the properties of actin within the living nucleus. Using fluorescently tagged actins and actin binding proteins, we measured the steady-state distribution and dynamics of nuclear actin in living HeLa cells. In fluorescence recovery after photobleaching (FRAP) experiments, nuclear actin, like cytoplasmic actin, presented as both rapidly and slowly moving kinetic populations. Based on the loss of this slow moving population when cells are treated with latrunculin and the absence of this population when a mutant actin incapable of polymerization is used in place of wild-type β-actin, we conclude that this slow recovering population is comprised of polymeric (F) actin. Thus, rather than polymerization, the only significant difference between the pools of cytoplasmic and nuclear actin is the bundling of cytoplasmic actin that generates the stress fibers that are prominent in cells stained with fluorescent phalloidin. Under normal physiological conditions, actin bundles are exclusively found in the cytoplasm. The kinetics of actin polymerization and cytoskeletal dynamics can be quantified by microinjection of purified and fluorescently tagged β-actin (; ,; ; ) or by the expression of a GFP-fusion protein (; ; ; ). Both fluorescent probes have been used successfully to estimate the dynamics of actin and F-actin in cell types ranging from yeast to mammals (; ; ; ; ). Stable expression of GFP–β-actin (or spectral variants of GFP fused to actin) has been achieved in a wide variety of cell lines with different motility properties, cytoskeletal organizations, and dynamics of cytoskeleton remodeling (; ; ). Thus, fluorescent actins are well-established probes for investigating actin dynamics in living cells. Because actin is normally found in the nucleoplasm, it stands to reason that fluorescent actins can also be used to measure the dynamic properties of nuclear actin. We first determined whether GFP-actin could be detected in sufficient quantity to allow kinetic measurements in living cells. shows a field of HeLa cells that are stably transfected with GFP–β-actin (, green). We were unable to simultaneously detect GFP–β-actin with endogenous β-actin by immunoblotting despite using several independent antibody preparations recognizing different epitopes on β-actin. Because the protein was readily detected using an anti-GFP antibody, we believe that this is likely due to the low relative abundance of the EGFP-tagged actin relative to the endogenous actin pool. It was necessary to use a separate approach to estimate the nuclear actin concentration based on the concentration of diffusing EGFP-actin molecules. Using fluorescence correlation spectroscopy (FCS), we measured the diffusing pool of EGFP–β-actin to be present at 0.14 ± 0.11 μM. Measurements of actin concentrations in human cell lines () indicate that the endogenous concentration of actin is at least 1,000 times higher than that of EGFP-actin. The drug resistance–based selection of cells for stable expression reduced the heterogeneity in the expression level of the transfected proteins; therefore, stable transfectants were used in most experiments. The nuclei (, red) are visibly depleted in GFP-actin, but actin is nonetheless easily detected within the nucleus. Consistent with previous studies, GFP-actin was incorporated into a dynamically organized cytoskeleton. This was confirmed by counterstaining cells with fluorescently labeled phalloidin (Fig. S1, available at ). shows a subregion of this field over time. By examining the cell surface, the dynamics of the bundles of filaments at the cell surface are easily illustrated. This time series demonstrates that the GFP-actin is incorporated into the expected F-actin–containing structures and that these GFP-actin–containing structures can both assemble and disassemble filaments that incorporate the GFP-tagged actin hybrid protein. We employed a commonly used and simple extraction procedure to estimate the efficiency of GFP-actin incorporation into actin filaments and compare it to the efficiency of incorporation of the endogenous actin within the same cell population. The results of this experiment are shown in . The top three images show the distribution of GFP-actin among soluble (S) and filamentous (P) pools isolated from HeLa cells, HeLa cells transfected with EGFP, or HeLa cells transfected with EGFP–β-actin. The bottom two images show the distribution of EGFP–β-actin and EGFP in the soluble and cytoskeletal fractions. When the distribution between soluble and cytoskeletal fractions was quantified from these immunoblots, the distribution for the actin and EGFP actin ranged from 52% of the total pool found in the F-actin fraction for EGFP actin to 56% of the total pool found in the F-actin fraction for endogenous actin in HeLa cells. As expected, EGFP was found entirely within the soluble fraction. Stable transfection of EGFP or EGFP–β-actin did not significantly alter the distribution of the endogenous β-actin. Collectively, these results demonstrate that EGFP–β-actin is an effective probe for β-actin dynamics in vivo. This is consistent with the conclusions of previous studies of EGFP–β-actin as a probe for actin dynamics in living cells (; ; ). Proteins that bind or sequester actin have been reported to be constituents of the nucleoplasm (). Hence, it might be anticipated that the nucleoplasm is rich in proteins that regulate actin polymerization. To address this possibility, we generated nuclear extracts from cells treated for 60 min with 20 μM latrunculin B and tested the ability of added purified actin to polymerize in vitro. Under these conditions, nuclei are released more easily and, as judged by brightfield microscopy, appear free of cytoplasmic contaminants (unpublished data). To control for effects solely due to increased protein concentration, purified actin was incubated with an identical amount of BSA. After incubation with either nuclear extract or BSA, polymerized actin was separated from soluble actin by centrifugation. shows the results of an assay with duplicate reactions for each sample. In the presence of BSA, very little actin polymerizes under the assay conditions. In contrast, when nuclear extract is mixed with purified actin, most of the actin polymerizes and consequently is found in the pellet fraction. The involvement of the nuclear actin pool in chromatin remodeling or transcription/processing of RNAs may result in the compartmentalization of the nuclear actin population. Consistent with this hypothesis, there are several reports of the specific enrichment of actin within transcription or splicing-associated compartments (; ; ; ). Because of the relative ease in identifying subnuclear domains within the mouse embryonic fibroblast nucleus, we have used this cell line to illustrate the subnuclear distribution of β-actin (). All cell lines examined, including HeLa, showed an identical subnuclear organization. illustrates that GFP-actin is homogeneously distributed throughout the nucleoplasm. Line scans were used to quantify actin distribution within the nucleus (). The nuclear regions of lowest GFP-actin concentration, the nucleoli, maintain concentrations of actin that are ∼10% of the cytoplasmic concentration (Fig. S4, available at ). As a control, we demonstrated that an exclusively cytoplasmic fluorescently tagged dextran does not show any evidence of nuclear fluorescence (Fig. S5 B). Because of the small size of the nucleolus, however, it is not possible to rule out the possibility that the source of the nucleolar signal originates in the nucleoplasm. We confirmed that the failure to enrich in specific subnuclear domains was not solely a property of the GFP-tagged version of β-actin. shows the distribution of rhodamine actin in living HeLa cells. Like GFP-actin, rhodamine actin is present in significant quantities in the nucleus and is homogeneously distributed outside of the nucleoli (; examples of nucleoli are illustrated with asterisks). The generation of subnuclear patterns of molecular distribution reflects both molecular exclusion events () and binding interactions (). Binding interactions, if they occur with molecules that are much more massive than the free protein, will reduce the mobility of a protein through the local environment. If these binding sites are of sufficient affinity and spatial density, the concentration of the protein within the local environments that contain binding sites will be higher at steady state than the concentration that is unbound and diffusing through the nucleoplasm. When the spatial information obtained from live cell imaging is combined with kinetic information on the underlying molecular flux, it becomes possible to investigate the relationships between organization and function in a quantitative manner. Therefore, we examined the mobility properties of actin within the nucleoplasm of living cells to determine whether kinetic populations of actin could be resolved that may differ functionally. The molecular flux of actin has been studied in the living cytoplasm. These experiments, which used microinjection of fluorescently labeled β-actin, easily resolved monomeric and polymeric pools of actin (; ; ). This is expected when the cytoplasm is measured because the actin becomes incorporated into filaments that are sufficiently large that the rate of filament turnover, rather than diffusion, limits the rate of fluorescence recovery in these studies. We confirmed that the cytoplasmic pool resolved into two distinct populations during recovery from photobleaching of GFP–β-actin in living HeLa cells. When the recovery profiles from many individual cells are pooled and the mean kinetic properties of the cytoplasmic actin pool are compared with the nuclear pool, the recovery profiles are very similar (). The x axis (time) is plotted on a log scale to better illustrate both the fast and slow phases of the recovery profile. Like the cytoplasm, the nucleoplasm clearly resolves into at least two distinct kinetic pools. The principal difference between the recovery kinetics of β-actin in the cytoplasm and the recovery kinetics in the nucleus is a small difference in the distribution between the fast and slow recovering species. This is most evident near the 10-s mark, at which time the rapid recovery phase has reached equilibrium. At this point, 84% of the nucleoplasmic signal has equilibrated, whereas only 76% of the cytoplasmic pool has equilibrated. This difference was statistically significant (P < 0.02). In the cytoplasm, the slow phase of the recovery profile is generated by the relatively slow turnover of fluorescence within the F-actin pool relative to the diffusion rate of the much larger diffusing pool (,).To determine whether the slow recovering phase of nucleoplasmic actin is rate limited by the turnover of actin filaments, we pretreated cells with latrunculin A, which depletes the cellular F-actin pool, and then analyzed the FRAP recovery kinetics. shows a comparison between the recovery of GFP-actin in control cells and in cells treated for 1 h with latrunculin A. Latrunculin A completely eliminates the slow recovery phase that is otherwise found in the nucleoplasmic actin FRAP profile (; ). A second drug, swinholide, which can fragment existing polymeric actin, also resulted in the loss of the slow recovering fraction of nuclear GFP-actin (unpublished data). In contrast, a third drug, jasplakinolide, stabilizes actin polymers, and treatment with this drug resulted in a time-dependent decrease in the recovering fraction of nuclear actin (). Latrunculin, by destabilizing stress fibers, causes the treated cells to detach from the coverslip and adopt a more spherical morphology. To rule out the possibility that changes in cell shape either directly or indirectly alter the mobility of nucleoplasmic actin independent of any direct effects of nuclear actin, we also examined the FRAP recovery profile of nuclear actin in cells that have been pretreated with EDTA. EDTA treatment caused the cells to adopt a rounded shape similar to that of the latrunculin-treated cells. However, unlike the latrunculin-treated cells, the recovery profile of nuclear actin in cells detached by EDTA treatment is indistinguishable from that of the adherent control cells (). We devised several approaches to confirm the presence of a nuclear pool of polymerized actin. Fluorescently labeled phalloidin was initially used but was found to be unsuitable because the binding was reversible when it was studied by FRAP after fixation with paraformaldehyde. shows the different probes used and the proportion of fluorescence signal that has recovered by 11.5 s into the recovery curve. This time point represents a period sufficient to allow recovery of freely diffusing molecules but insufficient time for significant amounts of the slow recovering molecules to equilibrate. As such, where fluorescent actins are used, this point represents a good approximation of the size of the pool of unpolymerized actin. The probes were chosen to either incorporate into (FITC- and EGFP-actin) or bind to (FITC–anti-actin antibody and moesin-GFP; ) F-actin. The subcellular distribution of these probes is illustrated in Fig. S5 A. In each case, a slow recovering population of nuclear actin was detected and, where tested, was lost or reduced upon treatment with latrunculin. The diffusion of the fluorescent anti–rabbit IgG antibody serves as a useful control for latrunculin-dependent changes that may indirectly alter diffusion in the nucleoplasm. There was no significant difference in the diffusion of microinjected anti–rabbit IgG antibody in the presence or absence of latrunculin treatment (). To directly test the requirement for polymerization in defining the low-mobility pool of nuclear actin, we made use of a previously characterized R62D mutant of actin (). This mutant actin does not incorporate into actin filaments. We prepared an EGFP fusion of the R62D mutant of actin and transfected cells with either the R62D variant or wild-type fusion proteins. This EGFP fusion protein was not evident in any of the more prominent components of the actin cytoskeleton (stress fibers and cell cortex). Because the expression of this protein altered cellular morphology, we chose to examine the kinetics of the mutant protein in transiently transfected cells. shows time-lapse images of transiently transfected GFP–β-actin (top) and GFP–R62D–β-actin (bottom) collected during the first few seconds of the recovery phase. The differences between the two proteins are most evident at these earliest stages of the recovery period. The first image of each time lapse shows the cell before photobleaching a 2-μm-wide line across the cell nucleus. The differences in the mobility of the two proteins are obvious in the first image collected after photobleaching. In the wild-type GFP-actin images, the photobleached region is obvious, whereas in the R62D mutant GFP-actin, inhomogeneity is evident but there is no clearly defined photobleached region. The absence of a clear photobleached region reflects the rapid mobility of the protein being examined. The graph in shows the mean recovery profiles of HeLa cells transiently transfected with GFP-actin or R62D mutant GFP-actin. For comparison, the recovery profile of GFP-actin obtained from cells treated with latrunculin is also plotted. We analyzed the data using mathematical models to simulate two possible mechanisms for explaining the recovery profile of nuclear β-actin. The development of the mathematical simulations of nuclear GFP-actin and the fitting of these models to the experimental data has been published separately (,). In simple terms, we assumed that the recovery curve reflected a mixture of rapidly diffusing complexes containing one or a few actin molecules and much slower moving polymeric actin species. In this model, both species recover through diffusion of new fluorescently labeled macromolecules or polymers into the region that has undergone photobleaching. The second model assumes that although polymeric actin may diffuse, its rate of diffusion is slower than the rate of remodeling of the polymer with new fluorescent monomers through treadmilling of actin polymers. In this case, the first phase of the recovery curve reflects diffusion of actin monomers and complexes, whereas the second phase of the recovery curve represents the turnover time of monomers within larger polymeric actin structures. Both mathematical simulations produced a very good approximation of the experimental data (), demonstrating that either a turnover process or a diffusion-dependent process can explain the observed recovery rates. The effective diffusion coefficients for the fast phase and the slow phase were estimated at 0.47 μm s and 0.009 μm s, respectively. The measured rate of diffusion obtained by FCS, ∼30 μm s is more consistent with the expected diffusion rate of a protein of this size. In vitro, the measured rate of actin monomer diffusion is ∼70 μm s (). Thus, even this fast phase migrates at least 50 times slower than expected for a molecule of monomeric size diffusing through the nucleoplasm. If we simulate the recovery of the slow phase as a reflection of the underlying kinetics of actin polymerization and depolymerization, the data can be fit by a population consisting of 84% of the nuclear β-actin molecules residing in the G-actin pool, whereas the remaining 16% are incorporated into actin filaments for a mean duration of 76 s (). Polymeric actin found within the cytoplasm is known to be dynamic and show different turnover rates in different regions of the cell. For example, actin cables, which represent bundles of actin microfilaments, require several minutes to turn over their constituent actin monomers and represent relatively stable polymeric actin structures (; ). In contrast, the polymeric actin assembled at the leading edge of the cell turnover with half-lives of <30 s (; ). The rate of actin turnover in the nucleus is relatively high but similar to what is observed for polymeric actin outside of actin cables (; ; ; ). To further assess the properties of the diffusing forms of GFP-actin within the nucleoplasm, we used FCS. FCS is a technique that measures stochastic fluctuations in the fluorescence emission that corresponds to molecules moving into and out of the focal point of the objective lens. Because the focal plane has a defined volume, this information can be directly used to quantify the diffusion rate of the fluorescent species through the focal volume. In principal, current mathematical modeling of FCS data enables no more than three diffusing components to be estimated. When the nucleoplasm and the cytoplasm were simultaneously measured by FCS, the diffusing species resolve into roughly four separate effective diffusion rates (Fig. S2, available at ). Approximately 70% of both the nuclear and the cytoplasmic actin pools diffuse at a rate of ∼30 μm s, consistent with the diffusion of monomeric actin or small complexes containing actin. The remainder of the detected GFP-actin–containing molecules diffuse at rates that are between 10 and 500 times slower than the monomeric pool. There is now evidence that actin is a required component of the RNA polymerase holoenzymes (; ; ). Based on this association, it is possible that the regulation of actin polymerization and nuclear transcription are interdependent. To address this possibility, we first determined whether inhibiting transcription altered the kinetic properties of the nuclear actin pools. Inhibition of RNA polymerase I transcription using actinomycin D or RNA polymerase II using 5,6-dichloro-1-β--ribofuranosylbenzimidazole (DRB) yielded similar results. shows the FRAP recovery curves of nuclear actin after 4 h of pretreatment with the RNA polymerase II transcriptional inhibitor DRB relative to the FRAP recovery curves of nuclear actin under normal growth conditions or when cells are pretreated with latrunculin B. These results demonstrate that there is a similar equilibrium between polymeric and monomeric actin, regardless of whether transcription is occurring. We next examined whether depolymerization of cellular polymeric actin altered transcription rates. In this experiment, a significant effect was seen. Mouse 10T1/2 cells were the most sensitive of the cell lines tested. Upon treatment with latrunculin, the incorporation of [H]uridine into nuclear RNA was inhibited by >90%. This was also evident in experiments incorporating 5-fluorouridine into nascent RNA, but analysis by indirect immunofluorescence is complicated by rounding and detachment of cells (unpublished data). A B cell line (Raji) was also examined for 5-fluorouridine incorporation in nascent RNA. These cells grow in suspension and do not contain the extensive cytoskeleton present in adherent cell lines. When Raji cells were analyzed by flow cytometry for the incorporation of 5-fluorouridine into nascent RNA, although less sensitive than the mouse embryonic fibroblasts, we also observed significant inhibition after treatment with latrunculin (). The inhibition of transcription by the actin binding drug latrunculin could be due to the drug interfering with the association of actin monomers with the RNA polymerase complexes. To determine whether this was the mechanism responsible for the inhibition of transcription that we observed, we performed coimmunoprecipitation experiments. We observed that an antibody directed against β-actin coimmunoprecipitates RNA polymerase II independent of whether the cell extracts were obtained from latrunculin-treated or control cells (Fig. S3, available at ). In this study, we sought to define the kinetic properties of the nuclear actin pool in living cells. Our results demonstrate that nuclear actin exists in several different kinetic populations, including a significant pool of polymeric actin. The effective diffusion coefficient of the rapidly recovering pool of nuclear actin was ∼0.5 μm s. This mobility is ∼50–100 times slower than expected of actin monomers or multimolecular complexes incorporating actin and could contain some oligomeric complexes of actin. When translated into mass equivalents, this represents up to a 1,000,000-fold increase in mass that is required to explain the reduced diffusion of the rapidly recovering population of actin molecules. It is more reasonable to explain differences of this magnitude based on binding interactions (; ; ). In this respect, it is notable that the rate of effective diffusion is similar to what is observed for many chromatin binding proteins, including those that modify chromatin structure (). Thus, the diffusion of monomers or oligomers that are incorporated into protein complexes as well as the reversible binding interactions that these complexes undergo in the nucleoplasm are likely to all contribute to the fast phase of actin recovery observed during FRAP. The more slowly recovering population of nuclear actin was directly correlated with the ability of actin to polymerize and was also detected among endogenous actin pools when photobleaching experiments were performed on fluorescent proteins that specifically bind actin in the F-actin state. We found that the equilibrium between the monomer and polymer state differs only slightly between the nucleus and the cytoplasm. The nuclear pool of polymeric actin turns over slightly faster than what is observed in the cytoplasm. In addition, the nucleoplasm contained a slightly lower proportion of polymeric actin than the cytoplasm. Based on the geometry of the interchromatin space, have suggested that it is unlikely that long filaments of actin exist in the nucleoplasm. Others have speculated that oligomeric actin may associate with the nuclear lamina and/or RNA polymerase II transcription sites (; ). The slow recovering phase containing the polymeric fraction of nuclear actin could be explained by more than one potential mechanism. First, oligomeric actin or relatively small actin polymers could be complexed with larger structures within the nucleoplasm. In this case, polymers could be relatively small and yet essentially immobile within the nucleoplasm. Alternatively, the mass of the polymer, either as a filament or as a network of filaments, may be too large to diffuse through the nucleoplasm whether or not it binds to other nuclear structures. Given that the concentration of cellular actin is hundreds to thousands of times greater than the critical concentration for spontaneous polymerization of actin, even if there were a 10-fold reduction in the concentration of actin within the nucleoplasm, it would still far exceed what could be maintained in a monomeric form without regulation. Although the FRAP recovery curve of nuclear actin revealed that ∼20% of the nuclear actin pool recovered with the kinetics of a polymeric actin pool, we wanted to verify that this fraction contained polymeric actin. The conclusion that this pool is predominantly or entirely polymeric actin is supported by the following observations: (a) the detection of the slow migrating population using fluorescently tagged actin, regardless of the fluorescent tag used or the method of introduction into the cell; (b) the detection of this same pool of slow migrating actin when fluorescent probes that bind specifically to endogenous actin or F-actin are used; (c) the loss of this population, independent of whether actin is fluorescently tagged or a fluorescent tag is present on an actin binding protein, when cells are incubated with latrunculin; (d) the failure to detect this slow phase of nuclear actin when an actin mutant that does not incorporate into filaments is measured using FRAP; and (e) the stabilization of this fraction in cells treated with jasplakinolide. One of the principal mechanisms for initiating actin polymerization in the cytoplasm is through phosphoinositide release and the generation of phosphatidylinositol-4,5-bisphosphate; thus, the cell membrane is where much of the cytoplasmic actin polymerization is initiated (). Although the nuclear membrane may similarly function as a site where actin polymerization is initiated (; ; ), the nuclear phosphoinositide signaling machinery and nuclear phosphoinositides colocalize with splicing factor compartments (; ) and thereby provide a potential site to nucleate actin polymerization distant from the nuclear membrane. Because these sites are centrally located within the euchromatic domains of the nucleus, these sites are well positioned to regulate the polymerization of actin throughout the nucleoplasm. With the exception of depletion in the nucleolus, we did not find evidence of significant inhomogeneities in the presence of either diffusing or polymeric and immobilized nuclear actin analogous to what are observed for proteins that enrich in splicing factor compartments or other nuclear bodies. Although this result is consistent with an essentially homogeneous distribution of actin and polymeric actin outside of the nucleolus, the large size of the monomer pool may require large differences in nucleoplasmic concentrations of polymeric actin to be evident using this approach. When actin was depolymerized with latrunculin, we observed a significant, although not complete, inhibition of transcription. β-Actin has recently been found to directly associate with both RNA polymerase I and II holoenzymes and is required for transcription in vitro (; ; ; ). Although our result is consistent with a function of F-actin in the transcription process, until better tools are developed for specifically targeting nuclear polymeric actin without altering the cytoplasmic fraction, the experiment cannot be considered conclusive. In summary, our results demonstrate that actin polymerization occurs within the living nucleoplasm and highlight the necessity to consider monomeric, oligomeric, or polymeric forms of actin as studies continue to reveal new functions and nuclear complexes that contain actin. To make significant progress in discriminating between these possibilities, it is essential that methods be developed that selectively target the state of polymerization within the nucleus without directly altering polymerization within the cytoplasm. Human β-actin containing an NH-terminal EGFP was purchased from CLONTECH Laboratories, Inc. Indian muntjac fibroblast, mouse 10T1/2, HeLa, and SK-N-SH human neuroblastoma cells were each transfected with the GFP-actin construct, and stable expression of the plasmid was selected for using G418. In some instances, drug-resistant cultures were further selected by fluorescence-activated cell sorting for GFP expression. R62D actin was provided by R. Treisman (Cancer Research UK, London, UK), and moesin-GFP was provided by D. Kiehart (Duke University, Durham, NC). Our protocol for FRAP has been described in detail previously (). In brief, living cells were photobleached in 2-μm-diameter circles or 1.5-μm strips (for mathematical modeling) across the width of the nucleus using a laser-scanning confocal microscope (LSM 510; Carl Zeiss MicroImaging, Inc.) and a 488-nm laser line at 100% intensity for 20 iterations. Photobleaching was completed in <250 ms. The recovery of the fluorescence signal over time was then monitored and used to measure the mobility of the actin. Because the stress fibers are the brightest GFP-actin–containing structures within the cells, it was necessary to close the pinhole aperture on the LSM 510 to the extent that light was collected from sections no thicker than 2 μm. Quantification of data was not performed on mouse 10T1/2 or Indian muntjac fibroblasts because the flat profile of these cells made it difficult to exclude the possibility that some of the fluorescence was contributed by stress filaments located above or below the nucleus. It was also necessary, in some instances, to perform image summing or frame averaging to improve the signal-to-noise ratio and maximize the dynamic range during data collection. Finally, the number of scans during collection was optimized by increasing the time interval between scans from <1 to 5 s after the first 5–10 s of recovery. Recovery curves were generated from between 20 and 30 individual cells recorded from at least three separate experiments. P-values were determined using the test. For quantification, each image was normalized for total fluorescence intensity relative to the first image collected after photobleaching to correct for any photobleaching that occurred during the collection of the postbleach time series. This was accomplished by measuring the total cellular fluorescence at each time point. Mathematical modeling of mobility using the compartment model (,) used a 2-μm-wide line photobleached across the width of the cell nucleus and extended the duration of the experiment to ∼2 min. Images were collected using MetaMorph (Universal Imaging Corp.) to control an Axiovert 200 M (Carl Zeiss MicroImaging, Inc.) equipped and acquired with a 12-bit charge-coupled device camera (Sensicam; Cooke Corp.) or a 14-bit charge-coupled device camera (Cascade; Photometrics). In some cases, confocal sections were acquired using a laser-scanning confocal microscope and a pinhole aperture setting of 1 Airy unit. GFP was excited using a Xenon lamp or a 488-nm laser line. Rhodamine was excited using a 514-nm laser line. The spatial sampling ranged from 0.07 to 0.15 μm per pixel in the xy plane and 0.2 to 0.4 μm in the z plane. For images acquired on fields of cells, a 1.3 NA Plan Fluor 40× objective (Carl Zeiss MicroImaging, Inc.) was used. For higher magnification images, a 1.4 NA Plan Apo 63× objective (Carl Zeiss MicroImaging, Inc.) was used. Time-lapse experiments involving living cells were typically acquired at 23°C in standard DME with added fetal calf serum. Images from the LSM 510 were imported into MetaMorph software if it was necessary to apply a 3 × 3 median filter to reduce pixel noise in the figure images. Otherwise, images were directly exported as 16-bit TIFF files and rescaled over an 8-bit data range. In most cases, the background fluorescence of the medium and the base signal from the detector are minimized to better represent the dynamic range of the data content in the image. In some instances, three-dimensional image sets were imported into Imaris 4.12 (BitPlane) and three-dimensional image sets were generated. In this instance, the image was scaled to map the data over the range of the display and the screen capture function in Imaris 4.12 was used to capture the image used in the figure. Figures were prepared in Photoshop CS (Adobe) for Windows. In general, images were scaled to span the 8-bit data depth, reducing background in the process, and then pasted into a composite canvas that was either 8-bit grayscale or 24-bit RGB color. If necessary, images were interpolated to 300 dpi using Photoshop. Cells growing in culture, either in culture flasks or plated onto glass coverslips, were treated with latrunculin A or B (Calbiochem) at the concentrations indicated. No difference was observed between cells treated with latrunculin A or B, with the exception that treatment with latrunculin B required a 10-fold increase in concentration. This was expected based on the known properties of the two latrunculins. Jasplakinolide was used at a final concentration of 1 μM. Alexa 546–conjugated phalloidin was used as recommended by the manufacturer (Invitrogen). Fluorouridine was used to determine the synthesis of RNA as described previously (; ). For these studies, incubations of between 10 and 20 min were used for RNA labeling, depending on the rate of incorporation into the cells (which varies between cell types). For flow cytometry analysis, cells were fixed with 100% methanol. Unless otherwise indicated, all other fixations were performed with 4% paraformaldehyde. HeLa cells were plated on glass coverslips embedded in 35-mm dishes. Microcapillaries were drawn from 1.0-mm outer diameter aluminosilicate filaments on a flaming/brown micropipette puller (model P-87; Sutter Instrument Co.). They were loaded from the rear with fluorescent actin at 0.5 μg/ml in 2 mM Tris HCl, pH 8, 0.1 mM ATP, 0.1 mM DTT, and 0.1 mM CaCl. Injection was attempted in the nucleus, but occasionally the actin appeared to go in the cytoplasm. However, when looking at the injected cells under fluorescent light, the two were indistinguishable. The cells were then incubated in fresh media for 1–2 h before FRAP experiments. FCS data was collected with an LSM 510 NLO using a 488-nm excitation source and collected with a 40× C-Apochromat 1.2 NA water immersion objective (Carl Zeiss MicroImaging, Inc.) and a ConfoCor 2 detector (Carl Zeiss MicroImaging, Inc.). Autocorrelation analysis was performed using the software provided with the ConfoCor 2 module. Curve fitting was performed by comparing the fit (r value) for single, two, and three parameter fits. The data was summarized and tabulated as described previously (). The distribution of actin in soluble and insoluble fractions was determined as described previously (). Nuclear extracts were prepared as described previously (). An Ultrafree-MC centrifugal filter (10-kD cutoff; Millipore) was used to concentrate the nuclear extract approximately fivefold and perform a buffer exchange back to buffer A. The concentrated extract was precleared by centrifuging at 125,000 for 1 h at 22°C. Lyophilized nonmuscle actin was resuspended in 5 mM Tris, pH 8, 0.2 mM CaCl, 0.2 mM MgCl, and 0.2 mM ATP at 2 mg/ml and incubated on ice for 30 min. Nonmuscle actin was partially polymerized on ice for 60 min by diluting the actin to 1 mg/ml and adding 1 mM ATP, 2 mM MgCl, and 50 mM KCl (all final concentrations). This solution was then used directly in the spin assay. Nuclear extract from ∼8 × 10 − 1.0 × 10 cells was gently mixed with 40 μg F-actin and allowed to incubate at room temperature for 30 min. As controls, buffer A alone or 8 μg BSA solubilized in buffer A was incubated with 40 μg F-actin under the same conditions. The samples were then centrifuged at 125,000 for 90 min at 22°C. The supernatant was removed, and an appropriate amount of 3× concentrated SDS sample buffer was added. The pellets were solubilized in 3× concentrated SDS sample buffer and heated at 60°C for 15 min. Approximately one third of the supernatant and pellet from each sample condition was then separated by 10% SDS-PAGE and visualized by Coomassie blue staining. HeLa S3 cells were either treated with 20 μM latrunculin B for 60 min before extraction or left untreated and used for control extracts. Whole cell extracts were prepared using modified RIPA buffer (50 mM Tris, pH 7.4, 1% NP-40, 150 mM NaCl, and 1.0 mM EDTA) for 60 min on ice. Insoluble material was removed by centrifugation at 21,000 for 10 min. The supernatant was used as the extract for the immunoprecipitation experiments. Immunoprecipitations were performed as described by Hoffman et al. with minor alterations. In brief, 10 μg anti–β-actin (AC-15) antibody were added to precleared, untreated, or latrunculin B–treated whole cell extracts (300 μg diluted 10-fold in 10 mM Tris-HCl, pH 7.9, 20 mM Hepes, pH 7.9, 8% glycerol, 45 mM KCl, 8 mM MgCl, 5 mM (NH)SO, 2% polyethylene glycol, 4.5 mM β-mercaptoethanol, 0.05 mM EDTA, and 0.025% sodium lauryl sarcrosine) and incubated for 2 h at 4°C. Protein G–Sepharose (150 μl of a 50% solution; GE Healthcare) was added, and the mixture was incubated for an additional 2 h at 4°C. The beads were washed five times with five volumes of IP dilution buffer. Bound proteins were recovered by boiling the washed beads in SDS sample buffer. The eluted proteins were separated by 10% SDS-PAGE and analyzed by protein immunoblotting using antibodies to RNA polymerase II or β-actin. For mathematical modeling of FRAP recovery curves, see ,). Fig. S1 shows costaining of cells expressing GFP-tagged β-actin with phalloidin. Fig. S2 shows an FCS analysis of GFP-tagged β-actin. Fig. S3 shows that RNA polymerase II coimmunoprecipitates with β-actin in the presence and absence of latrunculin. Fig. S4 shows a series of laser-scanning confocal optical sections through cells expressing EGFP–β-actin. Fig. S5 shows fluorescence microscopy of indirect actin probes and fluorescent dextrans. Online supplemental material is available at .

Mitochondria are double membrane–bound organelles nearly ubiquitous in eukaryotic cells. One important function of mitochondria is the production of ATP through oxidative phosphorylation, and defects in this process have been found to lead to several severe human diseases (; ). Over 98% of mitochondrial proteins are encoded in the nucleus (; ) and transported posttranslationally into the organelle. Such proteins use the TOM (translocase of outer mitochondrial membrane) and TIM (translocase of inner mitochondrial membrane) translocation machineries in the outer and inner mitochondrial membranes for their transport into mitochondria (; ; ; ). The mitochondrial genome typically encodes a small number of proteins, most of which are polytopic membrane proteins of the respiratory chain complexes. In the yeast , these are cytochrome of complex III; the subunits Cox1, -2, and -3 of complex IV; and the subunits Atp6, -8, and -9 of complex V. Although much is known about how proteins are targeted to and imported into mitochondria, little is known about how mitochondrially encoded proteins are transported and assembled into the inner membrane. Only a limited number of proteins that are involved in this process have been identified, including Oxa1, Cox18/Oxa2, Pnt1, Mss2, and Mba1 (, , 2001; ; ; ; ; ; ; ). Oxa1 is a member of the Oxa1/YidC/Alb3 family of proteins (; ) and is required for growth of on nonfermentable carbon sources. Inactivation of Oxa1 leads to defects in the insertion of mitochondrially encoded proteins into the inner membrane. Oxa1 interacts both with newly synthesized mitochondrial proteins () and mitochondrial ribosomes (; ). It has been suggested that Oxa1 links mitochondrial translation to membrane insertion of the newly translated proteins (; ). However, it has been shown that preproteins differ in their dependency on Oxa1. Specifically, Cox2 and -1 strictly require Oxa1 for membrane insertion, but cytochrome and Atp6 show little Oxa1 dependency for transport across the inner membrane (). The membrane insertion pathway for these Oxa1-independent proteins is currently unclear. Mba1 is an additional component of the mitochondrial export machinery that shares substrate specificity with Oxa1 but can either cooperate with or function independently of Oxa1 (). Finally, Cox18 (Oxa2 in ) is distantly related to Oxa1 and has been shown to be required for the transport of the COOH-terminal region of the Cox2 protein across the inner membrane, together with Pnt1 and Mss2 (; ; ; ; ). Thus, given the substrate specificity of the known constituents of the export machinery, it has been proposed that alternative transport pathways must exist to handle Oxa1-independent proteins, particularly cytochrome and Atp6 (). () was identified as a gene associated with Wolf-Hirschhorn syndrome (WHS; ). This disorder affects 1 in 50,000 live births and results in pre- and postnatal growth retardation, severe mental retardation, and developmental delay with microcephaly. In addition, this disease is associated with an impairment of muscular tone and seizures. WHS is caused by partial deletion of chromosome 4 at locus 4p16.3. This region encompasses multiple genes, among them ; thus, direct genotype–phenotype correlations are difficult to determine (). deletions occur in almost all patients with WHS (), and those patients with mild forms of the disease, defined as the absence of microcephaly, seizures, and severe mental retardation, lack a deletion of this gene (; ). Letm1 is predicted to contain two EF-hand Ca binding domains, a transmembrane domain, a leucine zipper, and coiled-coil domains (). Two orthologues of Letm1 exist in , Mdm38 () and Ypr125w (; ), which we have named for yeast LETM1 homologue of 47 kD. Letm1, Mdm38, and Ylh47 are mitochondrial proteins, and Mdm38 has been reported to localize to the inner membrane (; ; ). However, it is unclear how these proteins are transported into mitochondria and where in the mitochondria Ylh47 is localized. mutant mitochondria have been reported to display various pleiotropic defects such as altered mitochondrial morphology () and defects in K homeostasis (). Yet, the specific function of Mdm38 and Ylh47 and how their function correlates to these phenotypes is unclear. Because some of the phenotypes seen in WHS patients and are reminiscent of mitochondrial disorders (; ), we wanted to investigate the role of Mdm38 and Ylh47 in mitochondrial function. We report that Mdm38, Ylh47, and human Letm1 are transported across the inner mitochondrial membrane and processed to a mature form in a membrane potential (Δψ)–dependent manner. mitochondria exhibit a severe reduction in the amounts of a subset of mitochondrially encoded proteins. Export of cytochrome and Atp6 from the matrix across the inner membrane is especially affected in mitochondria. In agreement with this defect, mitochondria have reduced amounts of respiratory complexes III and IV and accumulate unassembled Atp6 of complex V. Moreover, both Mdm38 and Ylh47 interact with mitochondrial ribosomes. Our results indicate that Mdm38 acts as a component of the mitochondrial export machinery in an Oxa1-independent pathway and is particularly required for biogenesis of cytochrome and Atp6. The yeast proteins Mdm38 (Yol027c) and Ylh47 (Ypr125w) display significant sequence similarity to the human Letm1 protein (), which has been implicated in WHS. A high degree of sequence similarity is observed in the NH-terminal portion of the proteins, which includes a predicted single transmembrane helix that is rich in proline residues and contains a conserved glutamate residue, uncommon for transmembrane segments (, underlining). The COOH-terminal portions of the proteins contain predicted coiled-coil motifs and, in the case of Letm1, two predicted EF-hand Ca binding motifs (). However, this region, especially the critical amino acid residues required for Ca binding (), is not conserved among the three proteins. Mitochondrial targeting signals (presequences) and potential cleavage sites for the matrix processing peptidase were identified in all three proteins (MitoProt II; ). Presequences direct proteins across outer and inner mitochondrial membranes into the matrix or inner mitochondrial membrane via the TIM23 complex (; ; ; ). To assess whether Mdm38, Ylh47, and Letm1 possess targeting information for transport across the inner membrane, we synthesized the precursor proteins in rabbit reticulocyte lysate in the presence of [S]methionine/cysteine and imported them into isolated yeast mitochondria. Mdm38, Ylh47, and the human Letm1 were efficiently imported and processed to a protease-protected mature form in a Δψ-dependent manner (). The imported mature forms of the proteins migrated similarly to the authentic proteins in yeast and human (). Protein transport across the inner membrane requires Δψ, suggesting that Mdm38, Ylh47, and Letm1 were localized to the mitochondrial matrix or inner membrane. To address this, we treated with proteinase K both intact mitochondria and osmotically swollen mitochondria with a disrupted outer membrane. Both Ylh47 and Mdm38 remained stable under these conditions, and no stable cleavage products of these proteins were observed ( and not depicted). When mitochondrial membranes were disrupted by sonication or Triton X-100 treatment, both proteins became accessible to protease (). Thus, Mdm38 and Ylh47 are inner membrane proteins exposed to the mitochondrial matrix, which is in agreement with previous observations for the epitope-tagged Mdm38 (). showed that tagged Mdm38 was an inner membrane protein. To determine whether Ylh47 behaved in a similar way or represented a soluble protein of the matrix, mitochondria were subfractionated. After sonication, soluble proteins such as Mge1 were released into the supernatant, in contrast to Mdm38 and Ylh47, which remained associated with the membranes in the pellet fraction (, lanes 2 and 3). Carbonate treatment at both pH 11.5 and 10.8 led to a partial release of Mdm38 and Ylh47 from the membranes, whereas the peripheral membrane protein Tim44 was completely extractable (). Similar to Mdm38 and Ylh47, the single membrane–spanning Rieske Fe/S-protein (Rip1) of complex III of the respiratory chain remained partially resistant to alkaline extraction at pH 10.8, although it was fully extracted at pH 11.5. This result is explained by the fact that the transmembrane helix of the protein has been shown not to be completely embedded in the lipid environment of the inner membrane but rather located in close proximity to other proteins (), thus making it more sensitive to alkaline extraction than conventional membrane-spanning proteins. Based on these results, we conclude that Ylh47 and Mdm38 are proteins that may span the inner membrane in a complex with other proteins, as in the case of Rip1, or may not fully penetrate the inner membrane. In the course of analysis of the wild-type, , , and / strains, we determined that Mdm38 was required for efficient growth on nonfermentable carbon sources (, right). Although cells display slow growth on medium containing glycerol, cells showed no significant growth defect under the same growth conditions (, right). In contrast to previous work by , we did not observe a temperature-dependent growth defect for cells. Surprisingly, the / cells displayed a growth phenotype that was similar to that of at all temperatures. The high degree of homology between Mdm38 and Ylh47 makes it likely that these proteins perform similar functions in mitochondria. Yet, in light of our results, it seems that Mdm38 is more important for mitochondrial biogenesis than Ylh47, as Ylh47 is unable to compensate for the loss of Mdm38. The growth defect of cells on nonfermentable carbon sources suggested that Mdm38 was required for respiration. To support this, we assessed the Δψ in wild-type, , and mitochondria by fluorescence quenching. Although mitochondria showed only a marginal reduction of the Δψ when compared with wild-type mitochondria (, top and bottom), mitochondria from displayed a significant Δψ reduction (, middle). One explanation for the observed Δψ defect was that the respiratory chain of mitochondria was functionally compromised. To determine whether this was the case, we compared the steady-state protein levels of various mitochondrial proteins from wild-type, , and mitochondria. Most of the proteins tested, such as components of the outer and inner membrane protein translocases (Tom40, Tim23, and Tim22), complex V (Fβ, Atp3, and Atp6), porin, matrix chaperones (Ssc1, Mge1, and Hsp60), and the mitochondrial export machinery (Oxa1, Pnt1, and Mba1), were present in similar amounts in all three strains. However, mitochondria showed a significant reduction in the steady-state levels of the mitochondrially encoded proteins Cox1, Cox2, and cytochrome (Cob), the nuclear encoded Rieske Fe/S-protein (Rip1) of complex III, and a slight reduction in Cox3 levels (, lanes 3 and 4). In contrast, the steady-state protein levels in mitochondria were similar to those of wild-type in all cases. (Cyt1). Next, we solubilized mitochondria in digitonin buffer and separated respiratory chain protein complexes on blue native (BN) PAGE. Complexes III and IV form supercomplexes consisting of a dimer of complex III together with one or two complex IV monomers (III/IV or III/IV). We found that the levels of complexes III and IV were significantly decreased in mitochondria (), whereas other inner membrane complexes, such as the TIM22 complex, were not affected (). In wild-type, , and mitochondria, the monomeric and dimeric forms of the FF–ATPase (complex V) were similar, as judged by decoration with antibodies against Fβ, Atp21, and Atp6 (). However, although Atp6 was mainly present in the monomeric and dimeric forms of the complexes in wild-type and mitochondria, a low–molecular weight form of the protein selectively accumulated in mitochondria (, lane 8). Because the steady-state levels of Atp6 were not reduced in the mutant (), this finding suggests that a stable pool of Atp6 accumulates in an unassembled form in the mitochondria. We conclude that Mdm38 is required for the biogenesis of respiratory chain complexes. In yeast, complexes III, IV, and V of the respiratory chain contain seven mitochondrially encoded subunits: cytochrome (complex III); Cox1, -2, and -3 (complex IV); and Atp6, -8, and -9 (complex V). Because Cox1, Cox2, Cox3, and cytochrome were reduced in mitochondria, we wanted to determine whether translation of these proteins was affected in the mutant strain. To address this, we incubated mitochondria in the presence of [S]methionine/cysteine under conditions that promote mitochondrial translation. After labeling, a chase was performed by adding excess unlabeled methionine to allow translation to go on to completion. and did not display a significant difference with regard to the amounts of mitochondrial translation products compared with wild type (). An alternative explanation for the decreased steady-state protein levels in mitochondria was that posttranslational steps in the biogenesis of mitochondrially encoded proteins involved Mdm38. We therefore tested to determine whether the mitochondrial translation products were correctly transported across the inner mitochondrial membrane. After in organello translation, wild-type and mutant mitochondria were incubated in hypotonic buffer to disrupt the outer mitochondrial membrane, generating mitoplasts. If transport across the inner membrane occurs correctly, newly synthesized proteins are accessible to protease in mitoplasts (; ; ). In mitoplasts, the membrane insertion of newly synthesized cytochrome and Atp6 were affected, as these proteins remained significantly more resistant to protease treatment than in wild-type mitochondria (). In contrast to this, the mitoplasts only displayed a small increase in the protease resistance of cytochrome and Atp6 (, lanes 5 and 6). Other proteins, such as Cox1, -2, and -3 showed little or no increase in protease resistance in or mitochondria. To exclude the possibility that the observed protease resistance in mitoplasts resulted from the inefficient swelling of mitochondria, we analyzed the protease accessibility of marker proteins. The intermembrane space domain of the inner membrane protein Pam18 was efficiently degraded in all strains upon protease treatment. In contrast, the matrix protein Tim44 remained protected from protease by the inner membrane in all cases (). One concern was that the reduced Δψ in mitochondria could have indirectly caused the observed export defects. To address this, we first analyzed the import of proteins into wild-type and mitochondria. Import of matrix proteins such as Fβ or the precursor b-DHFR (NH-terminal portion of cytochrome fused to dihydrofolate reductase), which possesses an inner membrane sorting signal, was similar between wild-type and mutant mitochondria (Fig. S1, available at ). Thus, the reduced Δψ in mitochondria was sufficient for import of proteins into mitochondria, yet it did not rule out the possibility that Δψ was insufficient to promote protein export. Therefore, we isolated mitochondria from cells. These cells exhibit a severe reduction of Δψ because of loss of complex IV (Fig. S2, A and B). When mitoplasts were treated with proteinase K after in organello translation, the accessibility of newly synthesized proteins was not decreased, as compared with wild type (Fig. S2 C). This is in contrast to mitochondria, for which an increased protease resistance for cytochrome and Atp6 was observed (). reported that addition of the K/H exchanger nigericin restored the Δψ of mitochondria. Therefore, we added nigericin to mitoplast and analyzed the protease accessibility of cytochrome and Atp6. However, the phenotype of increased resistance of these proteins to proteinase K was not suppressed (unpublished data). Thus, we conclude that the export defects observed in mitochondria were not due to a reduction of the Δψ. The NH terminus of Cox2 is transported across the inner membrane into the intermembrane space and then processed by Imp1 of the inner membrane peptidase complex. When Oxa1 is defective, Cox2 accumulates as an unprocessed precursor (; ; ). Cox2 processing by the Imp1 protease was unaffected in and mitochondria, indicating that export of the NH terminus of Cox2 occurred despite the lack of Mdm38 or Ylh47 () and that Oxa1 function was not compromised. Thus, we conclude that efficient transport of cytochrome and Atp6 across the inner membrane requires Mdm38, whereas the export of Cox1 and -2 was only mildly affected in mitochondria (, bottom). Moreover, we observed a similar but much weaker effect for Ylh47. We next addressed whether newly synthesized mitochondrially encoded proteins interact with Mdm38. Mitochondrial translation products were labeled as described in the previous section in wild-type mitochondria and mitochondria containing Mdm38 with a COOH-terminal Protein A tag. Mitochondria were then solubilized in digitonin-containing buffer, and Mdm38, together with associated proteins, was isolated (; ). As a control for specificity of the purification procedure, Western blots were performed and decorated for Mdm38, the abundant inner membrane protein AAC, and Ssc1 (matrix Hsp70). As expected, neither AAC nor Ssc1 were purified with Mdm38 (). In contrast to the mock-purification where no radiolabeled proteins were purified (), we found that Cox1, Cox2, Cox3, cytochrome , Atp6, and Atp9 were all associated with Mdm38 (). To determine whether the interaction between Mdm38 and newly synthesized proteins reflected a possible chaperone-like activity for Mdm38, Qcr8 and Atp16, subunits of complex III and V, respectively, were imported into mitochondria containing Mdm38 or Tim18 of the TIM22 complex (). After removal of unimported precursor proteins by protease treatment, mitochondria were lysed and Protein A–tagged proteins were isolated. None of the imported proteins was found to be associated with Mdm38 or the TIM22 complex (). There are several reasons why Mdm38 could copurify with newly synthesized proteins. One of these possibilities is that Mdm38 could be directly associated with respiratory chain complexes. To test this, mitochondria expressing Mdm38 or Ylh47 were solubilized in digitonin buffer and subjected to IgG chromatography. Bound proteins were eluted, separated by SDS-PAGE, and subjected to Western blot analyses. Both Mdm38 and Ylh47 efficiently bound to and eluted from the IgG-Sepharose (). To develop a control, we tested for the presence of Ssc1 and AAC, neither of which were found in the eluate fractions. Interestingly, a fraction of Ylh47 coeluted with Mdm38 and vice versa, suggesting that these two proteins interact either directly or indirectly. Although newly synthesized Cox2 and -3 were found to be associated with Mdm38 (), components of complex IV were not copurified with Mdm38 or Ylh47 at steady state (). Similarly, Cyt1 and Rip1 of complex III were not detected in the eluate fraction (), indicating that neither complex III nor complex IV was significantly bound to Mdm38 or Ylh47 at steady state. In light of this observation, we speculate that the interaction of Mdm38 with newly synthesized mitochondrial proteins reflects a role for this protein in the early stages of membrane insertion before assembly of these proteins into the respiratory chain complexes. Oxa1, a central component of the export machinery, had been found to interact with mitochondrial translation products and to associate with ribosomes (; ; ). Thus, we tested to determine whether mitochondrial ribosomes were bound to Mdm38 and Ylh47. Indeed, both Mdm38 and Ylh47 copurified with Mrp49, a component of the large subunit of the mitochondrial ribosome (). Surprisingly, the interaction of Mdm38 and Ylh47 with ribosomal proteins was stable enough to be maintained over the course of the purification. We estimated from the Western blot analyses of Mdm38 purifications that only a fraction of ribosomal proteins copurified with Mdm38 under our experimental conditions. Previous analyses have shown that, similar to what we found for Mdm38, Oxa1 associates with a subfraction of the mitochondrial ribosomes (). To obtain further support for the ribosome association, we performed sucrose density–gradient centrifugation of mitochondrial lysates. While soluble mitochondrial proteins such as Mge1 migrated at the top of the gradient, Mdm38 was found at higher sucrose concentrations. It comigrated with the ribosomal peak, which included proteins such as Mrpl22, a result that is consistent with the idea that Mdm38 and ribosomes form a complex (Fig. S3 A, available at ). To support the idea that the interaction of Mrp49 with Mdm38 was specific, we compared the purification of Mdm38 to that of Tim23 and Oxa1. Although Tim50 and Ssc1 selectively copurified with Tim23 (; ), Mrp49 was found predominantly in association with Mdm38 (). Thus, it appeared that ribosomal proteins such as Mrp49 were specifically associated in a protein complex with Mdm38. We did not detect significant amounts of Mrp49 associated with Oxa1, although the protein had been previously reported to associate with ribosomes (; ). Conceivable explanations for this discrepancy are that the ribosome–Oxa1 interaction is too labile to survive the purification procedure and that tagging of Oxa1 at the COOH terminus had a destabilizing effect on the ribosome interaction. The copurification of Mrp49 with Mdm38 and Ylh47 suggested that mitochondrial ribosomes were associated with Mdm38 and Ylh47. To further verify this, we performed Protein A purifications of Mdm38 and Ylh47 and separated the eluted proteins by SDS-PAGE. The gels were either stained with Coomassie or subjected to Western blotting with antibodies against Protein A. The Coomassie-stained gels showed a large number of proteins that selectively copurified with both Mdm38 and Ylh47 but were not present in the mock-isolation (). Western blotting for Protein A demonstrated that the bands observed by Coomassie staining were not breakdown products of the tagged proteins (, lanes 5 and 6). To determine the identity of the coisolated proteins, the gel lanes were cut into pieces, subjected to in-gel digestion, and subsequently analyzed by mass spectrometry. The major proteins identified were constituents of both the large and small subunits of the mitochondrial ribosome ( and Table S1, available at ). To determine whether Mdm38 interacts with translating ribosomes, Mdm38 mitochondria were incubated before initiation of translation with or without puromycin, an antibiotic that inhibits protein synthesis and induces the release of nascent chains from ribosomes. After treatment, mitochondria were solubilized and Mdm38 was isolated by IgG chromatography. The puromycin treatment efficiently inhibited translation under these conditions (, top). In contrast, copurification of the ribosome was not significantly affected (, bottom). Thus, association of ribosomes with Mdm38 occurs in the absence of nascent chains. In light of this finding, we asked whether the association of newly synthesized precursors with Mdm38 reflected its association with translating ribosomes. Although mitochondrial translation products copurified with Mdm38 in control isolations, treatment with puromycin after translation and subsequent isolation under low-salt conditions significantly reduced the amount of radiolabeled proteins that copurified with Mdm38 (). In contrast, puromycin treatment did not affect the association of ribosomes with Mdm38 (). When isolations were performed in the presence of high salt, even in the absence of puromycin, both ribosomes and translation products were no longer associated with Mdm38 (, lane 6). However, after chemical cross-linking, Mdm38 could be coisolated with mitochondrial ribosomes via Mrp20, even in the presence of high salt (Fig. S3 B). Thus, based on the puromycin treatment, we conclude that the copurification of translation products with Mdm38 reflects a transient interaction that is mediated through association with translating ribosomes. To gain insight into the molecular role of Letm1 in mitochondria, we analyzed the function of its orthologues, Mdm38 and Ylh47, in mitochondria. First, we addressed the transport of the human Letm1, Mdm38, and Ylh47 into mitochondria and found that all three proteins could be efficiently transported into yeast mitochondria in a Δψ-dependent manner. Transport of the human protein, as well as Mdm38 and Ylh47, is accompanied by proteolytic cleavage of the predicted presequence, indicative of protein translocation across the inner membrane. Further analyses of the submitochondrial localization of the yeast proteins showed that both Ylh47 and Mdm38 were inner membrane proteins exposed to the mitochondrial matrix. This finding is in agreement with a recent analysis in which a tagged Mdm38 protein was shown to be an inner mitochondrial membrane protein (). Thus, Ylh47 and Mdm38 are both tightly associated with the inner membrane and exhibit similar topologies. Letm1 has been previously localized to mitochondria, and subsequent analyses of truncation constructs have suggested that targeting information resides in the NH-terminal portion of the protein (). Based on these findings and our analyses, it appears likely that the localization of Letm1 in human cells is similar to that of Mdm38 and Ylh47 in yeast. identified Mdm38 in a screen for yeast cells with defects in mitochondrial distribution and morphology. Moreover, work by suggested that Mdm38 is involved in mitochondrial K homeostasis. Therefore, we wanted to analyze the function of Mdm38 and Ylh47 in more detail. Our analyses have demonstrated that Mdm38 is associated with newly synthesized mitochondrial proteins and that this interaction occurs via the ribosome. Furthermore, Mdm38 is required for efficient membrane insertion of cytochrome and Atp6, whereas other proteins appear to be less dependent on Mdm38 function. In agreement with a role for Mdm38 in export, the steady-state levels of cytochrome were reduced in mitochondria, leading to reduced amounts of complexes III and IV. Defects in human and yeast Atp6 have been found to only mildly affect the stability of the FF–ATPase (; ). However, besides the fraction of Atp6 in the FF–ATPase, a stable pool of Atp6 was also found to accumulate in an unassembled state in mitochondria. In contrast, mitochondria lacking Ylh47 exhibited a significantly weaker phenotype with regard to protein insertion and did not display a growth defect on nonfermentable carbon sources. In addition to the observed defect in the biogenesis of cytochrome and Atp6, cells also displayed reduced steady-state amounts of other respiratory chain proteins. The reduction of both nuclear and mitochondrially encoded proteins can be explained by an increased instability of unassembled proteins (; ). In this regard, it is interesting to note that the stability of Atp6 is decreased in mitochondria because of increased turnover via the membrane-embedded protease Yme1 (). The accumulation of a stable Atp6 pool in mitochondria suggests that the protein is inaccessible to the protease in the membrane. Mdm38 and Ylh47 were found in physical association with mitochondrial ribosomes in the absence of translation products. Yet, this interaction was not critical for the synthesis of mitochondrial proteins. Because a subset of newly synthesized mitochondrial proteins showed defects in membrane insertion and Mdm38 was shown to interact with newly synthesized proteins indirectly via the ribosome, it is likely that ribosome binding to Mdm38 is important for Oxa1-independent protein export. We found that Mdm38 lacking amino acids 361–573 was still able to bind ribosomes (unpublished data), implicating the most conserved part of the protein in this function. Given that not all Mdm38 molecules seem to be bound to ribosomes at steady state, it is conceivable that a free pool of Mdm38 exists or that Mdm38 forms additional protein complexes with as-yet-undefined proteins. reported morphological defects for yeast cells with compromised respiratory chain assembly and/or function. This agrees well with the observed respiratory chain defects in mitochondria. Moreover, the respiratory chain defects might also account for an indirect effect on ion transport (). Mdm38 displays similarities to the mitochondrial Oxa1 protein in several regards. Oxa1 is a central component of the mitochondrial protein insertion machinery and is highly conserved between yeast and man (; ; ; ). Oxa1 interacts with newly synthesized mitochondrial proteins () and associates with mitochondrial ribosomes (; ), similar to Mdm38. In addition, mutant mitochondria are defective in inner membrane insertion of mitochondrially encoded proteins. Export of Cox2 strictly depends on Oxa1 function, whereas membrane insertion of cytochrome , Cox3, and Atp6 are only mildly affected in mutants (; , ). This suggests the existence of alternative pathways for inner membrane protein insertion (; ). In contrast to mutant mitochondria, mitochondria display selective protein insertion defects for cytochrome and Atp6. Thus, it appears that Mdm38 is a candidate protein for an Oxa1-alternative protein insertion pathway in mitochondria. Further support for this hypothesis comes from the fact that Cox2 transport and processing were not affected in mitochondria, whereas mutant mitochondria accumulate the precursor of Cox2 (; , ). The biogenesis of respiratory chain complexes requires the synthesis of the proteins encoded by the mitochondrial genome and their subsequent insertion into the inner membrane. Several human diseases have been linked to mutations in mitochondrial genes encoding structural components of complexes I, III, IV, and V of the respiratory chain. In addition, a limited number of nuclear encoded proteins that are mainly involved in the assembly of the respiratory chain complexes have been identified and linked to human diseases (; ). defects likely contribute to the neuromuscular features characteristic of most WHS patients, as patients with smaller chromosomal deletions that excluded the gene showed weaker phenotypes and lacked the WHS characteristic seizures (; ). Based on these findings, it appears likely that these neuromuscular features result from defects in mitochondrial function because of the loss of functional Letm1. However, the molecular function of the Letm1 protein remains unclear. The Letm1 orthologue Mdm38 plays a role in respiratory chain function at the cellular level, as demonstrated by the growth defects and reduced Δψ observed in mitochondria. Letm1 partially rescues the growth defect of cells, suggesting that both proteins fulfill similar cellular functions (). Thus, it is probable that Letm1 is involved in the biogenesis of the respiratory chain in humans similar to Mdm38 in yeast. Therefore, the phenotype of WHS patients, especially the neuromuscular defects and the seizures, likely reflects defects in oxidative phosphorylation and thus resembles classical mitochondrial encephalomyopathies such as MELAS (mitochondrial encephalomyopathy, lactic acidosis, and strokelike episodes), MERRF (myoclonic epilepsy with ragged-red fibers), or MILS (maternally inherited Leigh's syndrome; ). However, future work on Letm1 in human cells and especially in WHS patient cells is clearly needed to support this hypothesis. strains used in this study were derivatives of YPH499, with the exception of Mrp20 (Open Biosystems), which is a derivative of BY4741. For the isolation of mitochondria, cells were grown in liquid YPG medium (1% [wt/vol] yeast extract, 2% [wt/vol] bacto-peptone, and 3% [wt/vol] glycerol). For mitochondrial isolation, cells and the corresponding wild type were grown on media containing 2% sucrose as a carbon source. Tagging of Mdm38, Ylh47, and Oxa1 was performed by chromosomal integration in YPH499. Deletion of the (AFY23) and (AFY24) open reading frames in YPH499 was performed by homologous recombination using a PCR-generated cassette containing the marker gene flanked by regions homologous to the 5′ and 3′ end of the coding region. To generate an / strain (AFY27), the open reading frame was replaced by in the strain. Mitochondrial precursor proteins were synthesized in rabbit reticulocyte lysate in the presence of [S]methionine/cysteine (GE Healthcare). Import of radiolabeled precursor proteins into mitochondria was performed as previously described (). For localization of Ylh47 and Mdm38 in mitochondria, mitochondria that had been subjected to osmotic shock in a buffer containing 1 mM EDTA and 10 mM MOPS, pH 7.2 (mitoplasts), or mitochondria sonicated (3 × 30 s with 40% duty cycle in a sonifier [model 250; Branson]) in buffer containing 10 mM Tris/HCl, pH 7.4, and 500 mM NaCl, were treated with proteinase K at the indicated concentrations. Samples were subjected to TCA precipitation and analyzed by SDS-PAGE and Western blotting. To determine whether Mdm38 and Ylh47 were membrane proteins, mitochondria were subjected to carbonate extraction essentially as described previously (). In brief, mitochondria were suspended in 0.1 M sodium carbonate (pH 11.5 or 10.8) for 30 min at 0°C and centrifuged at 45,000 rpm in a TLA45 rotor (Beckman Coulter) for 30 min at 2°C. Samples were precipitated with TCA and subjected to SDS-PAGE and Western blotting for various marker proteins. For in organello translation of mitochondrially encoded proteins, mitochondria were isolated from cells grown in media containing glycerol as the carbon source. Translation was performed in the presence of [S]methionine/cysteine essentially as described previously (). After translation and reisolation, mitochondrial proteins were either precipitated with TCA and directly analyzed by tris-tricine SDS-PAGE or swollen in EM buffer (1 mM EDTA and 10 mM MOPS, pH 7.2) and divided, and one half was treated with 10 μg/ml proteinase K for 15 min on ice before TCA precipitation. After staining with Coomassie brilliant blue, gels were cut into slices. In-gel digestion of proteins was performed with trypsin in 25 mM of ammonium bicarbonate buffer (pH 7.8) at 37°C overnight. Tryptic peptide fragments were extracted twice with 50:50 acetonitrile/5% formic acid and subjected to online reversed-phase capillary HPLC separation with HPLC systems (Dionex/LC Packings; ). Tandem mass spectrometry (MS/MS) spectra were obtained using a QStar XL (Applied Biosystems) system equipped with a nano electrospray ionization source (MDS Sciex). Alternatively, MS/MS experiments were performed with a Finnigan LCQ XP instrument (Thermo Electron Corporation) equipped with a nano electrospray ionization source (PicoView 100; New Objective) and distal coated SilicaTips (FS360–20–10-D; New Objective). Data-dependent software (QStar XL, Analyst QS [Applied Biosystems], LCQ XP, and Xcalibur [Thermo Electron Corporation]) was used for online MS/MS analyses. Mitochondrial proteins with a known function that were identified with >25% sequence coverage were selected, whereas proteins frequently found as contaminants in Protein A purifications were omitted. For protein identification, uninterpreted MS/MS spectra were correlated with the National Center for Biotechnology Information protein sequence database ( www. ncbi. nlm. nih. gov) restricted to applying the SEQUEST algorithm. Mrp20 mitochondria were suspended in 20 mM Hepes/KOH, pH 7.4, 50 mM NaCl, 1% digitonin, and 10% glycerol and were incubated with the thio-cleavable cross-linker dithiobis(succinimidyl propionate) or left untreated for 30 min at 4°C. The cross-linker was quenched, and samples were subjected to a clarifying spin. Subsequently, the samples were loaded on IgG-Sepharose. Bound proteins were washed with buffer (20 mM Hepes/KOH, pH 7.4, 750 mM NaCl, 0.3% Triton X-100, and 10% glycerol) and eluted from IgG-Sepharose with SDS sample buffer containing 1% β-mercaptoethanol. Protein A–tagged proteins were isolated from mitochondria essentially as described previously (). In brief, mitochondria (8–10 mg for preparative scale, 1 mg for analytical scale, and 0.5 mg for in organello labeling and subsequent isolation) were solubilized in solubilization buffer (30 mM Tris/HCl, pH 7.4, 80 mM KCl, 5% [wt/vol] glycerol, 5 mM MgCl, and 1% digitonin) at 4°C, subjected to a clarifying spin, and applied to IgG-Sepharose. After binding to the column and extensive washing with solubilization buffer containing 0.3% digitonin, bound proteins were eluted with SDS sample buffer without β-mercaptoethanol to avoid release of the IgG chains from the Sepharose. After elution from the column, samples received reducing reagents and were subsequently analyzed by SDS-PAGE, Western blotting, or digital autoradiography. Antibodies were generated in rabbits against selected peptides of Letm1, Ylh47, Mdm38, Oxa1, Pnt1, and Mba1. Antibodies against Cox1, -2, and -3 were purchased from Invitrogen. For the separation of proteins after in organello translations, tris-tricine SDS-PAGE was used. Western blots were performed with PVDF or nitrocellulose membranes according to standard procedures and detected by ECL. Δψ measurements were determined using the potential-sensitive fluorescent dye DiSC (). BN-PAGE analyses were performed essentially as described previously (). Fig. S1 shows that the import of nuclear encoded proteins into mitochondria is not affected. Fig. S2 demonstrates that mitochondria lack intact complex IV and have a reduced Δψ but are not affected in the export of cytochrome or Atp6. Fig. S3 shows that Mdm38 comigrates with mitochondrial ribosomes on a sucrose gradient and that it can be cross-linked to mitochondrial ribosomes. The supplemental text gives a list of nonribosomal proteins identified by mass spectrometry in the Protein A purification analyses. Online supplemental material is available at .

The folding of nascent proteins is an extremely error-prone process, and cells must deal with malfolded proteins, which tend to form aggregates, by using molecular chaperones and protein degradation machinery. The membrane of the ER in mammalian cells contains three sensors (PKR-like ER-resistant kinase [PERK], activating transcription factor 6 [ATF6], and inositol requiring enzyme 1 [IRE1]) that can monitor the accumulation of unfolded proteins in the ER (ER stress) and activate elaborate defense mechanisms known collectively as the ER stress response to alleviate the burden of unfolded proteins (; ; ; ). The first sensor molecule, PERK, is a transmembrane kinase that is activated in response to ER stress () and phosphorylates the α subunit of eukaryotic translational initiation factor 2, leading to translational attenuation to avoid further accumulation of unfolded proteins in the ER (). The second sensor, ATF6, a transmembrane transcription factor, is transported to the Golgi apparatus upon ER stress and is sequentially cleaved by site-1 and -2 proteases (; Haze et al., 1999, ; ). The liberated cytoplasmic fragment of ATF6, containing a basic leucine zipper motif (pATF6α(N)), translocates into the nucleus, binds to the cis-acting ER stress response element (ERSE), and activates transcription of ER chaperones such as BiP, GRP94, and calreticulin (, , ). The third sensor, IRE1, is a transmembrane RNase (; ; ; ) involved in the splicing of XBP1 pre-mRNA (; ). XBP1 is a basic leucine zipper–type transcription factor containing a DNA-binding domain and a transcriptional activation domain, each encoded by a separate open reading frame on the pre-mRNA. Upon ER stress, XBP1 pre-mRNA is cleaved by the activated IRE1 and ligated by an unidentified RNA ligase to form mature (spliced) XBP1 mRNA, which encodes pXBP1(S) (; ). pXBP1(S) binds to ERSE to induce transcription of ER chaperones, and to another cis-acting element, unfolded protein response element, to induce transcription of other genes (probably genes involved in ER-associated protein degradation [ERAD]; ). The IRE1 signaling pathway is well conserved from yeast to mammals. In the budding yeast , Ire1p converts HAC1 pre-mRNA to mature mRNA, which allows translation of the active transcription factor Hac1p to induce transcription of ER chaperones and ERAD components (; , ; ). The splicing of HAC1 and XBP1 pre-mRNAs by IRE1 is quite unconventional (; ; ). The conventional splicing involves an elaborate complex of proteins and RNAs, called the spliceosome, and occurs exclusively in the nucleus, whereas the splicing reaction of HAC1 and XBP1 pre-mRNA simply requires IRE1 and RNA ligase, which is completely independent of the spliceosome, and takes place in the cytoplasm (). Because the removal of an intron from the HAC1 and XBP1 pre-mRNAs causes a switching of the reading frame in the COOH-terminal portion of the respective proteins, such splicing could be called “frame switch splicing” () or “cytoplasmic splicing” (). One of the unresolved issues regarding XBP1 is whether XBP1 pre-mRNA encodes a functional protein. In yeast, HAC1 pre-mRNA has a long (252 nt) intron that inhibits translation (; ; ). In contrast, unspliced (U) XBP1 pre-mRNA contains a much shorter (26 nt) intron and is actively translated to produce a protein (pXBP1(U)), although pXBP1(U) is rapidly degraded by the proteasome and not detected by immunoblotting (; ). It remained possible, however, that pXBP1(U) expression was enhanced in certain situations and played an important physiological role. reported that pXBP1(U) mutants whose lysine residues were replaced with arginine residues were resistant to degradation by proteasomes and that overexpression of the mutants repressed transcriptional induction by pXBP1(S), suggesting that pXBP1(U) could modulate function of pXBP1(S) if its expression was induced. We examined this problem and revealed an elaborate regulatory mechanism of mammalian ER stress response taking advantage of the dynamic interplay between pXBP1(U) and pXBP1(S) that appears critical for swiftly adapting to physiological changes in the ER. It had been postulated that XBP1 pre-mRNA is not translated into a functional protein because pXBP1(U) encoded in the pre-mRNA is rapidly degraded by the proteasome and because endogenous pXBP1(U) cannot be detected by immunoblotting (; ). By improving the protein extraction protocol, it became possible to detect pXBP1(U) protein accumulated in the cells even without prior proteasome inhibitor treatment. In cells not treated with ER stress inducers, a very small amount of 29-kD pXBP1(U) was detected (, lanes 1 and 6). Northern blot analysis revealed a certain amount of XBP1 mRNA in these cells (, lanes 1 and 6). When RT-PCR analysis was performed to differentiate spliced XBP1 mRNA from unspliced pre-mRNA, using a pair of primers designed to encompass the intron containing a PstI site (), only the bands corresponding to unspliced XBP1 mRNA cut with PstI (63 and 297 nt) were detected in uninduced cells (, lanes 1 and 6). Upon addition of tunicamycin (an inhibitor of N-linked glycosylation) or thapsigargin (an inhibitor of calcium ATPase in the ER), the band of spliced mature mRNA (334 nt) was markedly induced (, lanes 2–4 and 7–9). This induced splicing was shortly followed by a marked expression of pXBP1(S) and XBP1 mRNA, whereas the amount of pXBP1(U) increased gradually (, lanes 2–4 and 7–9, respectively). At the later phase of ER stress, the amount of pXBP1(S) gradually decreased, whereas the amount of pXBP1(U) continued to increase (, lanes 5 and 10). At this stage, a large amount of XBP1 mRNA was still accumulated (, lanes 5 and 10), and the amount of unspliced XBP1 mRNA was increased (, lanes 5 and 10; see the bands of 63 nt). These results implied that XBP1 transcription is still vigorous at this stage, but an appreciable portion of newly transcribed XBP1 pre-mRNA remains unspliced, possibly because of the gradual inactivation of IRE1, leading to the enhanced synthesis of pXBP1(U). This suggested the interesting possibility that pXBP1(U) expression is induced at the recovery phase of ER stress. To exclude the possibility that induction of pXBP1(U) at the later phase is caused by stabilization of pXBP1(U), we examined the stability of pXBP1(U) during ER stress. HeLa cells expressing pXBP1(U) were treated with cycloheximide to block de novo protein synthesis, and degradation of pXBP1(U) was monitored by immunoblotting with anti–XBP1-A antiserum. Degradation of pXBP1(U) was enhanced in the later phase of ER stress (). We also examined pXBP1(U) expression in human embryonic kidney 293 (HEK293) cells and obtained results similar to those observed in HeLa cells (). Interestingly, the kinetics and duration of ER degradation–enhancing mannosidase-like protein (EDEM) mRNA induction, one of XBP1's targets (), correlated well with the expression level of pXBP1(S) in HeLa and HEK293 cells (), and EDEM mRNA expression actually waned in the later phase (, lanes 6–8 and 14–16), though the peak of EDEM mRNA level was delayed as compared with that of pXBP1(S) level. It should be noted that the level of pXBP1(S) determines the rate of transcription of EDEM mRNA, whereas the data observed in the Northern blots reflects accumulation of EDEM mRNA. To confirm the notion that pXBP1(U) expression is induced at the recovery phase of ER stress, the expression of pXBP1(U) was examined in cells recovering from ER stress induced with DTT, which is a potent inducer yet easily washed out to accelerate recovery. As expected, DTT markedly induced transcription and splicing of XBP1 mRNA and expression of pXBP1(S) protein (, lanes 1–5). In contrast, when DTT was washed out after 30 min of treatment, transcription and splicing of XBP1 mRNA gradually decreased to the basal level, whereas the accumulation of pXBP1(U) increased (, lanes 6–10). This clearly showed that pXBP1(U) expression is induced during recovery from ER stress, when XBP1 mRNA splicing is halted, leading to accumulation of pre-mRNA. This temporal regulation of pXBP1(U) expression suggested that pXBP1(U) may have an important regulatory role in mammalian ER stress response, especially during the recovery phase. We next analyzed subcellular localization of pXBP1(U) using immunofluorescent microscopy. When HeLa cells were transfected with plasmid expressing HA-tagged pXBP1(U) or pXBP1(S), pXBP1(U) was found in the cytoplasm as well as in the nucleus (), whereas pXBP1(S) was specifically localized to the nucleus, as expected (). We confirmed this result by subcellular fractionation and immunoblotting of HeLa cells (). To determine the amino acid sequence responsible for this unexpected localization of pXBP1(U), we constructed and analyzed a set of deletion mutants of pXBP1(U) (). The COOH-terminal deletion mutant pXBP1(U)-[1–185] was exclusively localized to the nucleus (), suggesting that the anterior ([1–185]) and posterior ([186–261]) regions of pXBP1(U) contain the NLS and cytoplasmic localization signal, respectively. Analysis of further COOH-terminal deletions revealed that pXBP1(U)-[1–133] and pXBP1(U)-[1–92] containing the basic domain were localized to the nucleus (), whereas pXBP1(U)-[1–74], which was lacking the basic domain, resided in both the cytoplasm and nucleus (), indicating that the basic domain ([75–92]) functions as a NLS. On the other hand, pXBP1(U)-[1–208] was predominantly localized to the cytoplasm (), as opposed to pXBP1(U)-[1–185], which was found only in the nucleus. This indicates that the [186–208] region contains a strong nuclear exclusion signal (NES) that overcomes the function of the NLS. Indeed, this small fragment of 22 amino acid residues was exclusively localized to the cytoplasm (). To further examine the function of the NES, this segment was fused to the transcription factor pATF6α(N), which contains a basic domain and is solely expressed in the nucleus (; Haze et al., 1999). When expressed in HeLa cells, the fusion protein [186–208]-pATF6α(N) was localized to the cytoplasm (), whereas [209–261]-pATF6α(N) resided in the nucleus (). The 22-residue NES region represents a leucine-rich sequence that is conserved between human and zebrafish pXBP1(U), and is similar to a conventional NES (Lx-[LIVFM]-x-L-x-[LI]; ; ). To confirm that the nuclear export of pXBP1(U) is mediated by the conventional nuclear export machinery, HeLa cells transfected with plasmid expressing pXBP1(U)-[1–208] were treated with leptomycin B (LMB), an inhibitor for the nuclear export receptor CRM1 (), and the localization was analyzed. pXBP1(U)-[1–208] was expressed in the cytoplasm in the absence of LMB (), whereas it was clearly concentrated in the nucleus in the presence of LMB (). This suggested that the nuclear export of pXBP1(U) is mediated by the conventional nuclear export machinery and that pXBP1(U) shuttles between the nucleus and cytoplasm. Finally, a comparison of pXBP1(U)-[1–208] (localized to the cytoplasm; ) with pXBP1(U)-[1–261] (found in both the cytoplasm and nucleus; ) suggested that the [209–261] region partially attenuates the NES activity of the [185–208] region, possibly by sterically hindering the NES from interacting with NES receptors. Because pXBP1(U) is very unstable and is rapidly degraded by the proteasome (; ), we analyzed the region involved in rapid degradation. Plasmids carrying deletion derivatives of HA-tagged pXBP1(U) were transfected into HeLa cells, and degradation of each mutant protein was evaluated by comparing the protein level in the presence or absence of the proteasome inhibitor MG132. The amount of intact pXBP1(U) was much increased when cells were treated with MG132, which is consistent with rapid degradation by the proteasome (compare , lanes 1 and 7), the level of COOH-terminal deletion mutants lacking the [209–261] region was little affected (, lanes 2–6 and 8–12). This indicated that the [209–261] region is required for rapid degradation by the proteasome. The level of NH-terminal deletion mutants retaining the [209–261] region also increased by MG132 (, lanes 13–15 and 19–21), whereas that of deletion mutants lacking this region was little affected (, lanes 16–18 and 22–24), confirming our previous conclusion. We also examined the turnover of pXBP1(U) and its deletion mutants. Cells expressing pXBP1(U) or its derivatives were treated with cycloheximide, and whole cell lysates were subjected to immunoblotting. pXBP1(U) was rapidly degraded (), whereas pXBP1(U) mutants lacking the [209–261] region were more stable ( and ). This confirmed the aforementioned conclusion that the [209–261] region is indispensable for rapid degradation of pXBP1(U). The aforementioned results prompted us to examine whether pXBP1(U) can modulate the function of pXBP1(S), possibly as a negative regulator, to shut off transcriptional induction by pXBP1(S). Thus, we investigated whether pXBP1(U) binds to pXBP1(S) when coexpressed in vivo. HeLa cells were transiently transfected with a plasmid expressing Histidine-tagged pXBP1(S), together with a plasmid expressing XBP1(U), and whole cell lysate was incubated with Ni-NTA resin to bind His-pXBP1(S). Evidently, pXBP1(U) was specifically copurified with His-pXBP1(S) in this pull-down assay (, lanes 5–8), suggesting that pXBP1(U) associates with pXBP1(S) under these conditions. We then tested whether pXBP1(U) binds to pXBP1(S) synthesized in vitro. His-pXBP1(S) and pXBP1(U) (or a deletion mutant pXBP1(U)-[1–208]) were simultaneously translated in the same reaction using rabbit reticulocyte lysates, and the resulting products were treated with Ni-NTA resin. pXBP1(U), as well as pXBP1(U)-[1–208], was specifically copurified with His-pXBP1(S) (, lanes 7–12). These results suggested that pXBP1(U) synthesized in vitro can directly bind to pXBP1(S). We next examined the physiological implication of the interaction between pXBP1(S) and pXBP1(U). Based on the results presented in , most pXBP1(S) was expected to interact with pXBP1(U) in the presence of excess pXBP1(U). To compare the stability of the pXBP1(S)–pXBP1(U) complex with pXBP1(S), HeLa cells were transfected with a plasmid expressing pXBP1(S), together with a ninefold excess of a plasmid expressing pXBP1(U) or vector alone, and the accumulation of pXBP1(S) was examined by immunoblotting using anti–XBP1-C that specifically recognizes pXBP1(S) (). Under these conditions, pXBP1(U) was expressed in great excess over pXBP1(S) (, lane 2), and the level of pXBP1(S) was markedly reduced by overexpression of pXBP1(U) as compared with cells cotransfected with vector alone (, lanes 1 and 2). This effect of pXBP1(U) overexpression was greatly compromised by the proteasome inhibitor MG132 (, lanes 4 and 5), suggesting that pXBP1(S)–pXBP1(U) is more susceptible to degradation by the proteasome than pXBP1(S). Interestingly, overexpression of pXBP1(U)-[1–185] lacking the “degradation domain” identified in had little effect on the stability of pXBP1(S) (, lanes 3 and 6), though pXBP1(U)-[1–185] was expressed at a level similar to pXBP1(U) (, lane 3). This is also consistent with the notion that the degradation domain of pXBP1(U) is indispensable for rapid degradation of the pXBP1(S)–pXBP1(U) complex. We confirmed these notions by examining the turnover of pXBP1(S) in the presence or absence of abundant pXBP1 (U). Cells transfected as in were treated with cycloheximide, and whole cell lysates were subjected to immunoblotting. Coexpression of pXBP1(U) significantly accelerated pXBP1(S) degradation (, left and middle), whereas that of pXBP1(U)-[1–185] hardly effected pXBP1(S) stability (, right). We also analyzed subcellular localization of the pXBP1(S)–pXBP1(U) complex using HeLa cells cotransfected with plasmid expressing pXBP1(S) and a ninefold excess of plasmid expressing pXBP1(U), which resulted in speckled staining in the cytoplasm and reduced staining in the nucleus (). In contrast, pXBP1(S) was located solely in the nucleus (). Interestingly, coexpression of pXBP1(U)-[1–185] lacking the NES did not affect pXBP1(S) localization (). This suggested that pXBP1(S)–pXBP1(U) was exported from the nucleus to cytoplasm and that the NES contained in pXBP1(U) is essential for this export, although the significance of the speckled staining observed remained unknown. We confirmed this notion by subcellular fractionation experiments. Cells transfected as in were separated into the nuclear and postnuclear (cytoplasmic) fractions and subjected to immunoblotting with anti–XBP1-A antiserum. pXBP1(S) was localized in the nuclear fraction when it was expressed (, lanes 5 and 6). When pXBP1(U) was coexpressed, most of pXBP1(S) was actually found in the postnuclear fraction (, lanes 7 and 8), whereas coexpression of pXBP1(U)-[1–185] did not affect pXBP1(S) localization (, lanes 1 and 2). Finally, we examined the effect of pXBP1(U) coexpression on expression of pXBP1(S) target genes. Total RNA was extracted from cells transfected as in , and subjected to Northern blotting. Accumulation of BiP mRNA was enhanced in cells expressing pXBP1(S) (, lanes 1 and 2). In contrast, this induction was abolished by coexpression of pXBP1(U) (, lane 3), whereas it was not affected by coexpression of pXBP1(U)-[1–185] (, lane 4). These results strongly suggested that pXBP1(U) functions as a negative feedback regulator of pXBP1(S). We revealed that pXBP1(U) translated from the unspliced pre-mRNA represents a functional protein with an important role in mammalian ER stress response. One interesting characteristic of pXBP1(U) is that it contains three distinct subcellular localization determinants, which are the NLS, NES, and NES attenuator (). The functional balance among these signals appears to determine the subcellular location of XBP1 proteins under diverse physiological and environmental conditions. pXBP1(U), which contains all of these signals, can shuttle back and forth between the nucleus and the cytoplasm (). pXBP1(S) contains only the NLS and is located exclusively in the nucleus (). Interestingly, the pXBP1(S)–pXBP1(U) complex was localized to the cytoplasm (). Conformational changes evoked by the formation of this complex might cancel the NES attenuator activity. Although the molecular mechanism behind the attenuation remains unknown, it is possible that the attenuator sterically prevents NES receptors such as CRM1 from approaching the NES. The XBP1(U)-[186–261] region contains two highly conserved domains, a conventional NES motif ([186–208]), and a SWKPLMN motif at the very end (), and the latter might be important for the attenuation or for degradation by the proteasome. pXBP1(U) also contains a “degradation domain” in the [209–261] region that makes it more susceptible to degradation by the proteasome. Because pXBP1-[1–208], which lacks the degradation domain and is exclusively localized to the cytoplasm, is clearly more stable than pXBP1(U) (), the cytoplasmic localization by itself is not sufficient for rapid degradation. Although pXBP1(S) does not contain the degradation domain, it becomes less stable when it forms a complex with pXBP1(U), possibly because the degradation domain of pXBP1(U) is presented to the degradation machinery. Taking our findings into account, as well as the fact that pXBP1(U) expression is significantly induced during recovery from ER stress, we propose a working model of pXBP1(U) function in mammalian ER stress response (). In the absence of ER stress, pXBP1(U) protein is translated from XBP1 pre-mRNA and shuttles between the cytoplasm and the nucleus to monitor the level of pXBP1(S), and possibly that of inopportune expression of pATF6(N) as well because pXBP1(U) can form a heterooligomer with pATF6(N) (). Upon ER stress, activated IRE1 splices XBP1 pre-mRNA into mature mRNA, from which pXBP1(S) is translated. pXBP1(S) is then translocated into the nucleus, and activates transcription of targets such as ER chaperones and ERAD components. When enough ER chaperones and ERAD components are produced and ER stress has subsided, IRE1 becomes inactive, but a certain level of pXBP1(S) remains in the nucleus, leading to further transcription of XBP1 pre-mRNA and production of pXBP1(U). pXBP1(U) shuttles between the cytoplasm and the nucleus and forms a complex with pXBP1(S) left over in the nucleus. The pXBP1(S)–pXBP1(U) complex that may expose NES is efficiently exported from the nucleus to the cytoplasm and degraded by the proteasome by virtue of the degradation motif in pXBP1(U), resulting in a complete shutoff of transcription of the target genes. Thus, it seems very likely that pXBP1(U) functions as a negative feed-back regulator specific to pXBP1(S) (and possibly to pATF6(N) as well). Such a shutoff mechanism would clearly be beneficial for cell survival because constitutive activation of the ER stress response would be harmful and is known to retard cell growth in yeast (; ). To our knowledge, this study identified the first case in which a functional protein was translated from pre-mRNA. The findings suggest that cytoplasmic mRNA splicing is a very sophisticated mechanism of gene regulation as compared with conventional nuclear mRNA splicing, which is catalyzed by the spliceosome (). In conventional splicing, which occurs in the nucleus, it would be difficult to resplice mRNA if it has already been exported to the cytoplasm. Thus, it would be necessary to transcribe and splice the primary transcript de novo to accommodate possible change in amino acid sequence of product protein translated from the mRNA. In contrast, in the case of cytoplasmic splicing, pre-mRNA that is transported to the cytoplasm and used for translation can be respliced when necessary to change the structure of product protein translated from the mRNA, in response to extracellular or intracellular signals. Thus, cytoplasmic splicing would be a very rapid, versatile, and energy-efficient mechanism with minimum garbage, as compared with conventional mRNA splicing. The ER stress response may require such a sophisticated system to effectively exploit the interplay between pXBP1(U) and pXBP1(S) to deal with malicious unfolded proteins accumulated in the ER. Moreover, it is conceivable that other systems or cellular processes, such as development or antiviral response, adopt similar mechanisms in which cytoplasmic splicing plays an essential role. Identification of such systems would greatly increase our understanding of cellular response mechanisms to cope with various physiological and pathological situations. HeLa cells were grown in DME/glucose at 4.5 g/liter supplemented with 10% fetal calf serum, 2 mM glutamine and antibiotics, 100 U/ml penicillin/100 μg/ml streptomycin. Cells were maintained at 37°C in a humidified 5% CO/95% air atmosphere. Transient transfection was performed by the standard calcium phosphate method (; ). In brief, HeLa cells cultured in 60-mm dishes were incubated with precipitates of calcium phosphate, including plasmids, for 6 h at 37°C. After washing with PBS to remove calcium phosphate precipitates, cells were incubated in fresh medium for 24 h and harvested for analysis. For subcellular fractionation, transfected cells were suspended in 100 μl of ice-cold PBS containing protease inhibitors and 5% NP-40 and separated into the nuclear and postnuclear fractions by centrifugation at 14,000 rpm for 1 min. For turnover analysis, transfected cells were treated with 40 μM cycloheximide for the indicated period and subjected to immunoblotting. LMB was provided by M. Yoshida (Institute of Physical and Chemical Research, Saitama, Japan). To construct pCMV-HA-pXBP1(U) and pCMV-HA-pXBP1(S), 1,787- and 1,761-bp fragments, respectively, of XBP1 cDNA encoding pXBP1(U) and pXBP1(S) were cloned into an XhoI site of pCMV-HA vector (CLONTECH Laboratories, Inc.). Expression plasmids for a series of pXBP1(U) deletion mutants fused with HA tag were made by ligating the PCR product of the corresponding region with pCMV-HA. pCMV-HA-pATF6α(N) was constructed by inserting a cDNA encoding the [1–373] region of human ATF6α into BglII–XhoI sites of pCMV-HA vector. For construction of pCMV-HA-pXBP1-[165–261]-pATF6α(N), cDNA encoding the [165–261] region of human pXBP1(U) was cloned into a BglII site of pCMV-HA-pATF6α(N). pcDNA-His-pXBP1(S) was made by inserting pXBP1(S) cDNA into an XhoI site of pcDNA3.1-His vector (Promega). Cells grown in a 60-mm culture dish were harvested with a rubber policeman and pelleted by centrifugation. The pellet was suspended in 20 μl of ice-cold PBS containing protease inhibitors (100 μM 4-(2-Aminoethyl)benzenesulphonyl fluoride, 80 μM aprotinin, 1.5 μM E-64, 2 μM leupeptin, 5 μM bestatin, and 1 μM pepstatin A), mixed with 20 μl of 4× SDS sample buffer (200 mM Tris-Cl, pH 6.8, 400 mM DTT, 8% SDS, and 40% glycerol), and immediately boiled at 100°C. 10-μl portions of samples were subjected to SDS-PAGE using 4–20% gradient gels, transferred onto a Hybond-P filter (GE Healthcare), and incubated with various antisera according to the standard protocol (). Anti–XBP1-A detects both pXBP1(U) and pXBP1(S), whereas anti–XBP1-C detects only pXBP1(S) (). An ECL Western blotting detection kit (GE Healthcare) and lumino-image analyzer (model LAS-3000; Fuji) were used to detect antigens. Total RNA extracted with guanidine-phenol was separated by electrophoresis with 2% agarose gel containing 2.2 M formaldehyde, blotted onto a Hybond-N+ filter (GE Healthcare), hybridized with alkaline phosphatase-conjugated cDNA probes, and detected with a LAS-3000 using the Gene Images AlkPhos direct labeling and detection system (GE Healthcare). RT-PCR of XBP1 mRNA was performed essentially as described previously (). In brief, 10 μg of total RNA was reverse-transcribed with M-MLV reverse transcriptase (Invitrogen) and amplified with Ex Taq polymerase (TaKaRa) using a pair of primers that correspond to nt 493–512 (CGCGGATCCGAATGAAGTGAGGCCAGTGG) and 834–853 (GGGGCTTGGTATATATGTGG) of XBP1 mRNA, respectively (). Amplified fragments covering a 26-nt intron (nt 531–556) and flanking exon fragments were digested with PstI (a unique PstI site existed at nt 556) and separated on 4–20% polyacrylamide gels. cDNA was visualized by staining with SYBR Gold (Invitrogen) and detected by a Fluor-image analyzer (model FLA-3000; Fuji). HeLa cells grown on coverslips were transiently transfected with appropriate expression plasmids by the calcium phosphate method. Cells were fixed with 2% paraformaldehyde for 10 min, permeabilized with 0.2% Triton X-100 for 10 min, and stained with the appropriate antisera. Coverslips were mounted with 90% glycerol/10% PBS containing 100 ng/ml DAPI. Images were acquired using a microscope (model TE2000; Nikon) and a digital charge-coupled device camera (ORCA-ER; Hamamatsu Photonics). For in vivo pull-down assays, HeLa cells were cotransfected with expression plasmids for His-pXBP1(S) and pXBP1(U), harvested by rubber policeman, lysed in binding buffer (20 mM Hepes, pH 7.9, 100 mM KCl, 10% glycerol, 1 mM MgCl, 1 mM mercaptoethanol, 0.1% Tween-20, and 20 mM imidazole), and centrifuged. Supernatants were mixed with Ni-NTA agarose in binding buffer containing protease inhibitors for 1 h at 4°C, washed with binding buffer three times, and then suspended in 1× SDS sample buffer. Samples were boiled at 100°C and subjected to 4–20% SDS-PAGE. pXBP1(U) coprecipitated with His-pXBP1(S) was detected by immunoblotting with anti–XBP1-A antiserum. For in vitro pull-down assays, both His-pXBP1(S) protein and pXBP1(U) protein were cotranslated using TNT Quick Coupled Transcription/Translation systems (Promega), incubated with Ni-NTA agarose for 1 h at 4°C, and processed like the HeLa cells.

mRNA localization is a commonly used intracellular trafficking mechanism that provides the means to restrict the translation of specific proteins to discrete cytoplasmic regions (for review see ). Localized mRNAs are directed to their destinations by cis-acting localization elements (LEs) that generally reside in the transcript's 3′ untranslated region (UTR). These elements must be recognized by specific RNA-binding proteins (RBPs) that link the mRNA to the localization machinery. However, this has only been clearly established in yeast, where four stem-loops in Ash1 mRNA are recognized by She2p, which links the mRNA to the myosin Myo4p through She3p (; ; ; ). Apart from yeast, a definitive relationship has not been established between a localization signal, its cognate RBP, and the protein's function. Although genetic screens indicate that RBPs such as Staufen (Stau; ), HRP48 (), and Squid () are required for localizing particular mRNAs, the elements that these proteins recognize are not well defined. On the other hand, RBPs such as hnRNPI (), 40LoVe (), hnRNPA2 (), and VgRBP71 and Prrp (; ) bind specific LEs, but in these cases, it remains to be conclusively proven that the protein is actually responsible for localizing the RNA. One of the best candidates for an RBP that plays a direct role in mRNA transport is chicken and rat zip code–binding protein 1 (ZBP-1), as well as its homologue Vg1 RNA and endoplasmic reticulum–associated protein (VERA)/Vg1 RNA-binding protein (RBP), because it is highly conserved and binds to the localization signals of several different localized mRNAs (; ; ). ZBP-1 was first identified because it binds specifically to a 54-nt LE in chicken β-actin mRNA, called the zip code, and colocalizes with actin mRNA in the leading lamellae of motile fibroblasts (; ). Several lines of evidence support the hypothesis that this interaction is important for β-actin mRNA localization. The overexpression of a truncated version of ZBP-1 reduces the proportion of cells in which the RNA is localized, and the introduction of ZBP-1 into cells that do not express it can induce β-actin mRNA localization (; ). ZBP-1 colocalizes with β-actin mRNA in the growth cones and dendrites of cultured neurons, and both the localization of the mRNA and its colocalization with ZBP-1 are reduced by antisense oligonucleotides directed against either the zip code or ZBP-1 RNA (; ; ). The ZBP-1 homologue VERA/Vg1RBP was identified through its binding to the Vg1LE (, ; ). VERA/Vg1RBP recognizes a motif, UUCAC (called E2), which is repeated in the Vg1LE, where it is required for the RNA's localization to the vegetal pole of the oocyte. The same E2 motif occurs five times in the VegTLE, and these sites are likewise required for the accumulation of VegT mRNA at the vegetal pole (; ). Consistent with a role for VERA binding, the injection of anti-VERA antibodies inhibits the localization of both Vg1 and VegT mRNAs by 50% (). Although there is convincing evidence that mRNA localization requires the motifs recognized by VERA and ZBP-1, it is much harder in these experimental systems to demonstrate conclusively that the proteins themselves are required. Anti-sense treatments, antibody injections, and dominant-negative constructs against ZBP-1/VERA appear to inhibit RNA localization, but the effects are partial and variable (; ; ). Therefore, we have addressed whether the ZBP-1/VERA orthologue, insulin-like growth factor II mRNA–binding protein (IMP), is required for RNA localization in , where it is possible to evaluate mRNA localization in mutants that lack the protein completely. One of the best systems to examine mRNA localization in is in the oocyte, where the localizations of (), (), (), and mRNAs define the anterior–posterior and dorsal–ventral axes of the embryo (; ; ; ; ). The most relevant to our study is mRNA, which localizes to the oocyte posterior pole. Once there, Osk protein nucleates assembly of the polar granules, which contain the posterior determinant mRNA, as well as the germline determinants (; ; ). RNA accumulates in the oocyte from early oogenesis onwards, localizes transiently to the anterior at stage 8, and then translocates to the posterior pole over a period of several hours during stages 8–9 (; ). Posterior localization involves two substeps, initial transport and long-term anchoring (). mRNA anchoring requires Osk protein, whose synthesis is triggered upon the RNA's arrival at the posterior pole (; ; ; ). Premature translation of mRNA produces a bicaudal phenotype in which an ectopic abdomen develops in place of the head and thorax, illustrating the importance of restricting translation to the posterior pole (; ). This is achieved by repressing the translation of unlocalized mRNA and relieving this repression once the mRNA reaches the posterior pole (; ). Many gene products are required for repressing the translation of unlocalized RNA (). In contrast to repression, very little is understood about the localization-dependent translational activation of , other than the potential involvement of the Aubergine, Orb, and Stau proteins and the requirement of sequences at the 5′ end of the mRNA (; ; ). We have addressed whether the ZBP/VERA homologue IMP is required for maternal mRNA localization in the oocyte. Upon finding that IMP localizes at the posterior with mRNA, we focused our analysis on the role of the protein and its binding sites in the regulation of mRNA localization and translation. IMP contains the four signature KH-type RNA-binding domains and the glutamine-rich COOH terminus () that are present in the vertebrate orthologues (; ). Affinity-purified antibodies against IMP reveal the protein in nurse cells and the oocyte early in oogenesis. However, the high concentration of IMP in the follicle cells blocks the penetration of the antibody into the oocyte after stage 4. Therefore, we evaluated IMP localization in a homozygous, viable, and fertile GFP–IMP protein trap line (). GFP–IMP is enriched around the nurse cell nuclei (, inset) and accumulates in the oocyte as soon as it is specified in the germarium, where it shows a uniform distribution until stage 7 (). IMP accumulates transiently at the anterior of the oocyte during stages 8–9 and then localizes in a crescent at the posterior pole at stage 9, where it remains for the duration of oogenesis (). This pattern of localization is very similar to that observed for mRNA and Stau protein, which colocalize with IMP throughout oogenesis (; ; ; ). To ascertain whether IMP localization depends on , we examined whether it is perturbed in various mutants that affect the posterior accumulation of mRNA and protein. IMP does not localize to the posterior of the oocyte in , , and mutants, which block the transport of mRNA to the posterior pole ( and not depicted; ; ; ). Furthermore, IMP colocalizes with RNA to an ectopic dot in the center of the oocyte in a mutant that disrupts microtubule polarity (; ). Together, these results demonstrate that the localization of IMP to the oocyte posterior pole requires the localization of mRNA. IMP could localize to the posterior through a direct interaction with mRNA or protein or could be recruited to the posterior by a downstream component of the pole plasm. To distinguish between these possibilities, we examined IMP localization in a strong hypomorph (), which prevents the posterior recruitment of Vasa by Osk and disrupts all subsequent steps in pole plasm assembly (; ; ). IMP localizes normally to the posterior of these oocytes (), suggesting that its posterior accumulation depends on directly. Finally, we addressed whether IMP localization depends on Osk protein rather than mRNA by examining a nonsense mutation () that disrupts the anchoring, but not the initial localization, of mRNA (; ; ; ). IMP still localizes to the posterior of these oocytes at stage 9, but the posterior crescent is weaker than in wild type (WT) and disappears at stage 10 (). Thus, IMP behaves like mRNA in every mutant combination examined, suggesting that it localizes to the posterior in association with the mRNA. KH domains recognize short, single-stranded RNA motifs (, ; ) similar to the motifs that are required for localization of RNAs in chicken embryo fibroblasts (; ) and oocytes (; ; ). To identify the motifs recognized by IMP, we performed in vitro selection experiments on a large pool of ∼7 × 10 RNAs containing random 25-nt sequences. Because we were unable to obtain the first and second KH domains of IMP as soluble proteins, we selected RNAs that bind to the third and fourth KH domains. This seemed justified, as the vertebrate homologue ZBP-1 binds the β-actin zip code primarily through its third and fourth KH domains (). The structural basis of RNA recognition by KH domains was established through biochemical and x-ray diffraction studies of the KH domains from another protein, NOVA (; ). Those studies used 11 rounds of in vitro selection against their isolated KH domain to identify its preferred recognition element, which is a particular sequence of four bases. On this basis, we chose to evaluate the 11th and 12th round “winner sequences” selected by the IMP KH domains in respect to the frequency of all tetramers. The most common tetramer retained by either KH3 or KH4 was UUUA, which occurred in 43% of the winning KH3 sequences and 46% of the KH4 winners. The base preferred by the IMP KH domains after the principal tetranucleotide was most frequently C (35%) or U (32%). Thus, SELEX indicates that the optimal binding sequence for both IMP KH3 and KH4 is UUUAY, which was present in 36% of clones bound by KH3 and 37% of clones bound by KH4 (). To quantify the binding of KH3 and full-length IMP to UUUAY-containing RNA, we performed filter-binding assays using three tandem repeats of the 25-nt winner RNA, 4-12-13 (). When all five nucleotides of the motif are changed to GGGCG, the affinity of the RNA for the KH3 domain diminishes by an order of magnitude; and even a single nucleotide change, UUUAY to UUgAY, decreases the affinity significantly (). Full-length IMP binds to the UUUAY-containing RNA with an even higher affinity, and the mutations in the motif decrease binding to a similar extent to that observed for the single KH domains (). Electrophoretic mobility shift assays confirm the results of filter-binding assays; IMP shifts the mobility of RNAs with UUUAY, but not the mutant motifs (). These results indicate that IMP's KH domains 3 and 4 specifically recognize the UUUAY motif, which we refer to as an IMP-binding element (IBE). The IBE motif occurs 13 times in the 3′UTR of mRNA (), which is significantly more frequent than would be expected by chance. This contrasts with the 3′UTRs of other localized mRNAs, such as , which contains only two copies of the motif. Indeed, mRNA associates specifically with IMP in vivo because it coimmunopreciptitates with IMP from ovary extracts, whereas mRNA does not (Fig. S1, available at ). To characterize the interaction between IMP and the 3′UTR, we performed UV cross-linking assays with ovary extracts and a P-labeled RNA probe of the 3′UTR (). Using a procedure that optimized the binding of VERA/Vg1RBP to the LEs of Vg1 and VegT in oocyte extracts (), we found that a single 65-kD polypeptide cross-links to the 3′UTR, but not to the 3′UTR. This polypeptide co-migrates with the band detected by anti-IMP antibodies on immunoblots (). Similar cross-linking experiments, using extracts from embryos expressing the GFP–IMP fusion protein, labeled a second polypeptide, whose slower mobility corresponds to that expected of the GFP–IMP fusion (). This confirms that the protein cross-linked in the experiments is IMP. To address whether the binding of IMP to the 3′UTR depends on the IBEs, we mutated all 13 copies of the motif to UUgAY or gggcg. Both mutant RNAs are significantly impaired in their ability to complete UV cross-linking of the WT 3′UTR to IMP in ovary extracts (). The predicted secondary structures () of the WT and UUgAY 3′UTRs are virtually identical, suggesting that the single-base substitutions in 's IBEs inhibit IMP binding not through a nonspecific effect on the RNA's folding, but instead through abrogation of sequence-selective binding of the IBEs by IMP's KH domains 3 and 4. The very specific effects of IBE base substitutions on RNA localization and translation provide additional, much stronger, evidence that the mutations do not affect RNA folding significantly. To address whether the IBEs in the 3′UTR are required for the posterior localization of IMP, we created transgenic lines in which all 13 copies of the IBE are mutated from UUUAC/U to UUUgC/U ( ) in an otherwise WT genomic fragment. , into an RNA-null background (/; unpublished data). and the control unmutated transgene ( ) rescue the stage 6 oocyte-arrest phenotype of RNA-null flies completely (unpublished data). mRNA is comparable to that of endogenous WT mRNA (). Thus, the IBEs are not necessary for mRNA's initial transport to the posterior pole. or transgenes. mRNA, although Stau still localizes normally (). The IBEs in the 3′UTR are therefore essential for the posterior localization of IMP, confirming that these UUUAY motifs are bona fide IMP-binding sites in vivo. mRNA is similar to that of the control mRNA until the end of stage 9, the mutant mRNA disappears from the posterior at stage 10 b (). Furthermore, Stau protein displays an identical phenotype; it forms a WT posterior crescent at stage 9 and then disappears from the posterior at stage 10 b (). The IBEs in the 3′UTR are therefore necessary for the anchoring of mRNA at the posterior cortex. mRNA at the posterior at stage 10 could reflect a direct role for the IBEs in the anchoring of the mRNA. However, the maintenance of mRNA at the posterior requires Osk protein, which is only translated once the mRNA has been localized (). Thus, an alternative possibility is that the IBEs are required for the activation of mRNA translation at the posterior, and that the anchoring defect is secondary to a lack of Osk protein. ovaries with an anti-Osk antibody. mRNA (). Thus, the IBEs are essential for the derepression of mRNA translation at the posterior pole. The embryos from /; mothers display a fully penetrant maternal-effect phenotype in which the abdomen fails to form, consistent with the failure to translate Osk protein (). transgene (). suggests that the IBEs are important for mRNA translation and anchoring, an alternative possibility is that one of the IBE mutations prevents translation for some other reason; e.g., by chance one IBE might overlap the actual translational control element or all 13 IBE mutations might alter the folding of RNA. We therefore created four sets of transgenic lines in which nonoverlapping subsets (A–D) of three or four consecutive IBEs are mutant (). . They display a fully penetrant maternal-effect defect; Stau and the mutant RNAs localize to the oocyte posterior pole at stage 9, but appear dislodged from the posterior or disappear altogether during stage 10; and Osk protein is absent (, and G; and not depicted). In contrast, the fourth construct ( B) rescues the mRNA–null phenotype completely, and the localizations of mRNA and protein and Stau are normal (, and F; and not depicted). These findings support the hypothesis that multiple IBEs, and not some other control element that overlaps one IBE, are responsible for RNA translational activation and anchoring. To test whether IMP is required for mRNA translation and anchoring, we generated null mutations in the protein through imprecise P excision. Screening by PCR revealed that three of these lines, , , and , correspond to imprecise excisions that specifically removed parts of the IMP-coding region (). Both and remove a large portion of the IMP-coding region and are presumably null alleles, whereas removes both the alternate initiation codons, but may produce some protein from downstream in frame ATGs (). Furthermore, there is no detectable IMP staining in mutant germline clones, marked by the absence of GFP (). Although mutants are zygotic lethal, the complete removal of IMP from the germline has no obvious effect on oogenesis. Most importantly, mRNA localizes normally to the posterior of the oocyte at stage 9 in germline clones of all three alleles and remains anchored there throughout oogenesis (). Furthermore, the mRNA is translated at the posterior pole and produces a normal crescent of Osk protein (). Thus, despite being a bona fide component of the RNA localization complex and binding to the motifs required for translation and anchoring, IMP plays no essential role in the assembly or function of the pole plasm. However, maternal IMP is essential for embryogenesis, as 100% of the embryos from germline clones die in late embryogenesis and this phenotype is not rescued by a WT paternal copy of the gene. Our objective was to address whether IMP is required for mRNA localization, as previous studies of its vertebrate homologues, ZBP-1 and VERA/Vg1RBP, had not resolved this question definitively (; ; ; ; ). We have demonstrated that IMP binds directly to mRNA at well defined sites that are required for translation and anchoring. The best evidence that these sites are bona fide IBEs is that IMP is not recruited to the posterior by mRNA in which all 13 IBEs have been mutated with a single base change. Indeed, this is one of the only cases we are aware of where it has been possible to demonstrate that an RBP interacts in vivo with well defined elements identified biochemically in vitro. In vitro, mutant RNA still competes for binding of IMP, albeit less effectively than the WT RNA, suggesting that the 3′UTR may contain other lower affinity sites. However, these sites are not involved in the recruitment of IMP to the posterior in vivo, nor are they sufficient for translational activation. Although the IBEs are thus bona fide in vivo IMP-binding sites, their role in RNA translation and anchoring is independent of IMP, which is not required for these activities. Two outcomes of this investigation seem particularly surprising. First, IBEs are required not for the initial localization of mRNA, but instead for its translational activation once it is localized and its subsequent anchoring at the posterior pole. Second, mRNA localization-dependent translation and anchoring require the IBEs in its 3′UTR, but not IMP itself. Because Osk protein defines where the pole plasm forms, and hence where the pole cells and abdomen develop, it is essential that mRNA is only translated at the oocyte posterior. Indeed, translational control is arguably more important than localization in restricting Osk to the posterior, as normally only 18% of mRNA is actually localized (), and mRNA localization mutants such as () produce a normal abdomen. The translational repression of unlocalized mRNA occurs in different ways, depending on the stage of oogenesis. Mutants in RNA interference pathway components cause premature translation of mRNA during early oogenesis (). Repression at later stages does not depend on these components, but instead requires the binding of Bruno and Hrp48 to three elements in the 3′UTR called Bruno response elements (; ; ). This repression may occur at the level of translation initiation through the binding of Bruno to Cup protein and of Cup to the Cap-binding protein eIF4E, implying that the 5′ and 3′ ends of the mRNA are linked (; ). Much less is known about how mRNA translation is derepressed at the posterior, apart from the findings that a 297-nt element at the 5′ end is required for the localization-dependent activation of a reporter RNA fused to the 3′UTR () and that the 3′UTR, although sufficient to repress the translation of heterologous coding sequences, is insufficient to activate their translation at the posterior (). Our data now provide direct evidence that the 3′UTR, through its IBEs, is required for translational derepression. Therefore, like activation, repression involves both the 5′ and 3′ ends. Moreover, three transgenes ( A, C, and D) with only 3 out of 13 sites mutated at a single base prevent translational derepression. These are much more subtle mutations than the deletions that have previously been used to define derepression elements () and will be useful for identifying the corresponding derepressor proteins. Although the CPEB homologue, Orb, and the RISC component, Aubergine, have been proposed to play a role in translational activation (; ), mutants in these proteins also affect the initial localization of mRNA to the posterior, and this may account for the observed reduction in Osk protein levels (; ). is -null mutants that have been rescued by a transgene-expressing Stau protein that lacks the fifth double-stranded RNA–binding domain (). However, Stau is unlikely to be the putative factor that interacts with the IBEs in the 3′UTR to activate translation, both because it recognizes double-stranded RNA rather than short-sequence motifs () and because the IBE mutations prevent mRNA translation without affecting Stau localization to the posterior pole at stage 9. This brings us to the most significant outcome of our investigation: RNA translational activation and anchoring is disrupted by mutants in the IBEs, but not by the loss of IMP itself. The possibility that the IBE mutations prevent mRNA derepression and IMP localization indirectly by altering the structure of the RNA seems extremely unlikely, as single-base substitutions within three nonoverlapping sets of three IBEs in widely separated regions of the >1-kb 3′UTR produce an identical and very specific defect in translation, without affecting any of the earlier functions of the 3′UTR, such as the maintenance of oocyte fate, the transport of the mRNA from the nurse cells into the oocyte, the translational repression of unlocalized mRNA, or its localization to the posterior pole. Thus, none of these mutations disrupt the binding of any of the factors that mediate these earlier steps, including Staufen, which is thought to recognize the secondary structure of the RNA through the interaction of its double-stranded RNA-binding domains with multiple stem loops. This strongly argues against the possibility that the single base changes to the IBEs inhibit RNA translation through a nonspecific effect on RNA folding. This leads us to conclude that the IBEs play a direct role in the derepression of mRNA translation. Because IMP itself is not necessary for derepression, this implies that the IBEs are also recognized by another factor, which we will call factor X. IMP and factor X could function redundantly to derepress translation, i.e., the two proteins might share 's IBEs and compensate for each other's loss. However, factor X cannot be a ZBP-1/VERA family member because, unlike mammals, no such relatives are evident in the genome. Alternatively, IMP and factor X might function independently, i.e., derepression might occur exclusively through factor X binding. Rather than implementing 's translational derepression, IMP's actual function might be to compete with factor X for IBE binding. In support of this, we have found that overexpression of IMP reduces Osk protein levels at the posterior (Fig. S2, available at ). Although the purpose of IMP competition is presently unclear, one possibility is that IMP serves to bind, and thereby mask, IBEs that occur by chance in RNAs for which factor X binding would be unnecessary or even detrimental. According to this view, competition with IMP would restrict factor X binding to those mRNAs, such as , that contain many copies of IBEs clustered within a restricted region. In the absence of IMP, factor X could bind to mRNAs with fewer IBEs and inappropriately regulate their translation. This may explain why embryos from -null oocytes always die, but from defects that appear unrelated to Osk function. Our analysis of the interaction of IMP with mRNA closely parallels that of ZBP-1 and VERA/Vg1RBP with β-actin and Vg1 mRNA, respectively. (a) In each case, the protein has been shown to colocalize with the localized mRNA and can be UV cross-linked to it in extracts; (b) the precise binding sites of each protein have been determined and reveal that it recognizes a repeated motif in the target mRNA; (c) the function of these sites has then been analyzed by introducing specific point mutations that abrogate the binding of the protein, and these have been found to have a dramatic effect on translation or localization. In this study, we have gone one step further, and have compared the phenotype of the IBE mutants with that of mutations in IMP itself. The observation that the former gives a fully penetrant defect in mRNA translation, whereas the latter has no phenotype in the germline, conclusively demonstrates that IMP is not responsible for the function of the IBEs in the 3′UTR. This is important in light of the observation that many RBPs have been implicated in the posttranscriptional regulation of particular mRNAs by studying the effects of mutations in their binding sites. Our results highlight the potential limitations of this approach by demonstrating that one cannot necessarily infer the function of a protein from the phenotype of mutations in the cis-acting sequences that it recognizes. The clear similarities between the localizations and functions of and mRNAs in oocytes, and of mRNA in oocytes, suggest that binding motifs for ZBP-1 proteins have a fundamental role in embryogenesis. , , and localize as mRNAs to one pole of the oocyte, which is the site where the germ or pole plasm forms, and all three proteins play key roles in the formation of the primary body axis (; ; ; ). Our findings extend this parallel by showing that the localized expression of all three proteins also depends on a repeated RNA motif, defined by its interaction with IMP or its homologues. Because our results rule out a function for IMP in the regulation of mRNA, this calls into question the role of VERA/Vg1RBP1 in the localization of Vg1 and Veg T mRNAs, and it may therefore be worth considering the possibility that there is also a factor X in . The cDNAs encoding IMP KH3 (residues Leu 301–Ala 396) and KH4 (residues Val 387–Gln 482) were subcloned into ProEX HTb (Life Technologies). These KH constructs included the canonical KH domain, as well as 20 additional residues at the COOH termini that, in a previous study of a different protein, were found essential for high affinity binding of the RNA recognition element (). The constructs were expressed in and recovered by extraction of the bacteria with a solution of 8 M Urea, 100 mM NaHPO, and 10 mM Tris-Cl, pH 8.0. The fusion proteins were bound to Ni-NTA agarose (QIAGEN), eluted at pH 4.5, and dialyzed against 50 mM NaHPO, pH 8.0, 300 mM NaCl, 5% glycerol, and 2 mM DTT. To create a random 25mer RNA pool for SELEX, we used 1.2 nmol of the oligonucleotide 5′-GCGAATTCAGATAGTAAGTGCAATCT{25N}AATTGAATAAGCTGGTATCTCCC-3′ (Invitrogen), where N indicates the incorporation of nucleotides at random. EcoRI sites and sequences for RT-PCR amplication are included. This provided an oligonucleotide pool with an estimated complexity of 7.2 × 10 sequences. To generate a double-stranded DNA library suitable for in vitro transcription, we PCR amplified the pool using 5′-GCGAAGCTTTAATACGACTCACTATAGGGAGATACCAGCTTATTCAATT-3′ and 5′-GCGAATTCAGATAGTAAGTGCAATCT-3′ as the forward primer containing the T7 promoter and HinDIII sites and the reverse primer containing an EcoRI site. We synthesized RNA from the double-stranded DNA using T7 RNA polymerase in the presence of α-P-UTP and then purified the RNA pool on an acrylamide gel run under denaturing conditions. P-RNAs consisting of three tandem repeats of the winner sequence 4-12-13 () were synthesized in vitro using the oligonucleotide 5′-GTTGAAAAAAATAAAGTTGAAAAAAATAAAGTTGAAAAAAATAAACTATAGTGAGTCGTATTA-3′ annealed to 5′-TAATACGACTCACTATAG-3′, which contains the T7 promoter. To create the corresponding motif mutants (UUgAY and gggcg) in the same context, the nucleotides encoding the IBEs (underlined) were altered (5′-ATcAA-3′ or 5′-CGCCC-3′) accordingly. RNA synthesis was performed with α-P-UTP and an AmpliScribe T7 transcription kit (Epicentre Biotechnologies). Filter-binding assays were performed in 100-μl reactions consisting of 1XB buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM MgCl, 0.5 mM DTT, and 20 μg/ml tRNA), 100 fmol of P-RNA, and varying concentrations of KH3 or IMP. Reactions were incubated at RT for 15 min. Protein-bound and -unbound RNA fractions were separated by filtration of the reactions through 0.22- (KH domain) or 0.45-μm (IMP) nitrocellulose filters. For electrophoretic mobility shift assays, 60 fmol of P-RNA was incubated with Histidine-tagged IMP (30, 100, 300, and 900 nM) at RT for 30 min before electrophoresis under nondenaturing conditions using 8% acrylamide (37.5:1) gels. The gels were run at 100 V for 4 h at 4°C. Gels were dried and imaged on a PhosphorImager SI (Molecular Dynamics). UV cross-linking assays were performed as described previously (). We engineered GFP fusions of IMP using cDNA that we obtained either from ESTs (provided by K. Korey and D. Van Vactor, Harvard Medical School, Boston, MA) or a ovarian cDNA library (provided by N. Brown, The Gurdon Institute, Cambridge, UK). We cloned the ORF into pUMAT-GFP downstream of the maternal α-4-tubulin promoter, which drives expression in the germline (), or into pUAS-p, which allows for the tissue-specific expression of the transgene using the Gal4/UAS system (; ). The constructs were made from a 10-kb Xho1–Apa1 fragment of genomic DNA (a gift from U. Irion, The Gurdon Institute, Cambridge, UK). The 3′UTR TTTAY motifs were mutated using the Transformer site-directed mutagenesis kit (CLONTECH Laboratories, Inc.). excision mutants were recombined onto y,w,v,P{FRT(w{hs})}101. Germline clones were then generated by crossing the FRT recombinant lines to or . Other lines used in this study include GFP-tagged IMP protein trap line G080 (), , () and , (), (), , (), (), and (). We generated rabbit antisera against full-length recombinant IMP or peptides. Antibodies were affinity purified against peptide immobilized to Sulfolink resin (Pierce Chemical Co.) or recombinant IMP immobilized to CnBr–Sepharose (Roche). Dilutions for immunoblots were as follows: 1:300 for anti-IMP antibodies, 1:3,000 for anti-Osk antibody (a gift from A. Ephrussi, European Molecular Biology Laboratory, Heidelberg, Germany), and 1:5,000 for anti-tubulin antibody (Sigma-Aldrich). Dilutions for immunostaining were 1:100 for anti-IMP, 1:500 for anti-Osk, and 1:500 for anti-Stau (). Secondary antibodies were obtained from Jackson ImmunoResearch Laboratories. We performed RNA in situ hybridization as previously described, using dig-UTP–labeled RNA (Roche) and Cy3–anti-Dig (Jackson ImmunoResearch Laboratories; ). Cuticle preparations were mounted in 1:1 Hoyers/lactic acid, and images were collected using a SPOT camera and software (Diagnostic Instruments) on an Axioplan microscope (Carl Zeiss MicroImaging, Inc.) using a 10× objective at RT. Fluorescent samples were mounted in Vectorshield (Vector Laboratories). Images were collected on a confocal system (models 1024 or Radiance 2100; BioRad Laboratories) with Lasersharp 2000 software (BioRad Laboratories), attached to a microscope (Eclipse E800; Nikon) using a 40×, 1.3 NA, objective at RT. Images were subsequently processed with Photoshop (Adobe). We used standard methods to generate and isolate element excision lines that lack IMP gene segments. We obtained the EP(X) 760 P-insertion line (w, P{w = EP}IMP) generated by the Berkeley Gene Disruption Project from the Bloomington Stock Center. To characterize the excisions molecularly, we extracted DNA from homozygous mutant larvae and performed PCR with primers designed to identify lines that lack regions of the gene. Fig. S1 depicts an experiment showing that RNA is specifically immunoprecipitated with IMP. Fig. S2 shows that overexpression of IMP in the germline decreases the amount of Oskar protein at the posterior, as well as causing actin defects late in oogenesis. Online supplemental material is available at .

Neurodegenerative disorders are characterized by the slow exacerbation of symptoms and by gradual progression of brain pathologies. Patients suffer for 5–20 yr from the onset of the disease to the bed-ridden state. Even fast-progressing amyotrophic lateral sclerosis takes 2–5 yr to render the patient bed ridden. Regarding the pathology, the total number of neurons and neural networks among them decrease. However, some of the neurons survive for an extensive period of time despite their expression of abnormal structures that are derived principally from the pathogenic disease-causing products. Typically, nigral neurons that express Lewy bodies in Parkinson's disease, hippocampal neurons that carry paired helical filaments in Alzheimer's disease, and motor neurons bearing Bunina bodies in amyotrophic lateral sclerosis can partially survive until the death of the patient. The mutant protein aggregates that characterize many of these diseases are known to trigger multiple cellular responses, including ER stress and mitochondrial abnormality. These stress responses are clearly sufficient to induce apoptosis in nonneuronal cell lines, whereas the brain pathology of patients indicates that neurons survive for a long period before their demise. A lengthy period of cell death is also observed in the polyglutamine (polyQ) diseases, a major group of neurodegeneration that includes nine disorders (for reviews see ; ; ; ; ). Again, a fraction of the neurons that possess nuclear and/or cytoplasmic inclusions of mutant polyQ peptides survives even in affected regions of the brain until the time of necropsy. So far, there is no model that fully explains the lengthy period of cell death in neurodegeneration. In addition to ER and mitochondrial stresses, transcriptional dysfunction is suggested as a critical pathological component of polyQ diseases (for reviews see ; ; ; ; ). Translocation of mutant proteins to the nucleus seems essential for neuronal dysfunction or cell death in polyQ diseases (; ; ). Numerous transcription-related factors, including LANP, PQBP-1, N-CoR, ARA24, p53, mSin3A, ETO/MTG8, P160/GRIP1, A2BP1, TAF130, CA150, CRX, Sp1, CtBP, PML, TAF30, NF-κB, and SC35, are known to interact or colocalize with mutant polyQ disease proteins (for reviews see ; ). Interaction with polyQ proteins may impair physiological functions of these transcription factors (; ), and, finally, even the general transcription level could be repressed (). Some of these polyQ pathology-mediating factors bind directly to the core of transcription machinery, RNA polymerase II (Pol II; ). Therefore, one of the paramount issues in the field of polyQ diseases is the relationship between transcriptional dysfunction and neuronal death. However, as yet, the role of transcriptional disruption in neuronal death is unclear, as is the mode of neuronal death when transcription is severely impaired. In this study, we found that inhibition of Pol II–dependent transcription leads neurons to undergo a slowly progressive atypical cell death (transcriptional repression-induced atypical death [TRIAD]) distinct from apoptosis, necrosis, or autophagy in morphological and biochemical analyses. Transcriptome analysis of TRIAD suggested that yes-associated protein (YAP), a known transcriptional cofactor, might be relevant to the death process. YAP, which was originally found as a binding protein to Src homology domain 3 of the yes proto-oncogene product (for review see ), acts as a transcriptional cofactor of p73, mediates the expression of cell death–promoting genes, and induces apoptosis (; ; ). We found that in TRIAD, full-length YAP (FL-YAP) is down-regulated, and novel neuron-specific YAPΔC isoforms lacking the cell death–promoting activity sustain to protect neurons in a dominant-negative manner. The shift of balance in YAP isoforms seemed to slow down the cell death signaling pathway of p73 activated by α-amanitin (AMA), at least partially. We further questioned the relevance of YAP and p73 to Huntington's disease (HD) by using cellular, , and mouse models as well as human brain samples. Our data suggest that these molecules might be involved in neuronal death triggered by mutant Htt, the causative gene product of HD. To address the role of transcriptional disruption in neuronal death, we first made multiple short inhibitory RNAs (siRNAs) against RNA Pol II to suppress Pol II–dependent transcription. However, suppression of Pol II was inadequate and reminiscent of recent efforts to suppress basic transcription machinery by similar approaches (). Therefore, we used a specific inhibitor of Pol II (AMA) whose three-dimensional molecular structure is exactly complementary to the groove of Pol II, through which mRNA is elongated (; ). Different concentrations of AMA were added to the culture medium of HeLa cells, primary rat embryonic (embryonic day [E] 15) cortical neurons, rat E15 striatal neurons, and rat pup cerebellar neurons (postnatal day [P] 7). BrdU up-take assay () showed significant repression of transcription at 6 h of AMA treatment in primary neurons () and HeLa cells (not depicted). The survival of AMA-treated cells estimated by trypan blue assay () revealed that AMA induces a slowly progressive cell death in a dose-dependent fashion. This was most pronounced in primary neurons, with half-lives of nearly 5 d. AMA-induced neuronal death was much slower than low potassium–induced apoptosis of cerebellar neurons, whose half-life was ∼12 h (not depicted). The slow progression of AMA-induced neuronal death was confirmed independently by MTT (3-[4, 5-dimethylthiazol-2-yl]-2, 5-diphenyltetrazolium bromide) assay (Fig. S1, available at ). A population of HeLa cells (10–30%) began to show cytoplasmic vacuoles proximal to the nucleus (, HeLa-TRIAD) from 6–12 h after the addition of AMA. Similar vacuoles were also observed in cortical neurons treated with AMA for 2 d (, CTX neuron-TRIAD), although with a diminished frequency (1–5%). It is important to note that the vacuoles did not possess double-membrane structures reminiscent of autophagosomes. No classic apoptotic features such as chromatin condensation, nuclear fragmentation, or apoptotic bodies () were found in these neurons by electron microscopic analysis (). In addition, no necrotic features such as mitochondrial dilatation (, CTX neuron necrosis) or cytoplasmic ballooning and rupture (, CTX neuron necrosis) were observed in primary neurons under TRIAD. Immunohistochemical analyses using organelle-specific antibodies excluded the idea that the cytoplasmic vacuole was derived from the Golgi apparatus, endosome, lysosome, and mitochondria (Fig. S2, available at ). Autophagosomes induced by rapamycin and labeled with EGFP-LC3, a marker protein of the autophagosome (, top and middle), failed to colocalize with AMA-induced vacuoles (, bottom). In addition, the size of AMA-induced vacuoles was larger than that of autophagosomes (, bottom). EGFP-LC3 actually expresses the LC3 peptide (, arrow), verifying the morphological result. Note that the immunoblot shows a nonspecific band that is consistently detected by this antibody (, asterisk; unpublished data). Furthermore, the addition of rapamycin to the medium increased LC3-positive vacuoles but did not affect the formation of LC3-negative vacuoles induced by AMA (Fig. S3 A).Collectively, these data suggested that AMA-induced cell death is distinct from autophagy. Finally, we found colocalization of the vacuoles with ECFP-ER fusion constructs (expressing calreticulin ER-targeting sequences and KDEL ER retrieval tags at the 5′ and 3′ ends, respectively, of ECFP; ). It suggested that vacuoles might be derived from expanded ER. In agreement with the absence of morphological features of apoptosis, genomic DNA analyses of cell lines and primary neurons did not show ladder formation after AMA treatment (). Caspase-3, -7, and -12 were not remarkably activated in primary neurons by AMA (). AMA induced neither the release of cytochrome c into the cytosol from these neurons () nor the interaction of annexin-V with the membrane of these neurons at an early stage (Fig. S3 B). Caspase inhibitors z-DEVD-fmk and z-VAD-fmk did not repress AMA-induced cell death in neurons or in HeLa cells (not depicted). As expected, cycloheximide did not affect the cell death (not depicted). Calpain inhibitors, including ALLN and ALL, showed no remarkable effect on cell death. Pretreatment of cells with different concentrations of ATP in the medium did not affect AMA-induced cell death (not depicted). Although AMA is a highly specific inhibitor of Pol II, as confirmed by molecular structural analyses (; ), to further verify that AMA-induced cell death is mediated by transcriptional repression, we examined the effect of another type of transcription inhibitor, actinomycin D, on primary neurons (Fig. S4, available at ). Actinomycin D binds directly to DNA and inhibits transcription () by stalling the rapidly moving fraction of Pol II (). We found that actinomycin D also induced a slowly progressive neuronal death (Fig. S4 A), in which some neurons show cytoplasmic vacuoles similar to those by AMA (Fig. S4 B). Neither DNA fragmentation nor caspase activation was induced by actinomycin D (Fig. S4, C and D). Collectively, our results suggest that AMA induces a slowly progressive TRIAD of neurons that is distinct from apoptosis, necrosis, and autophagy. To understand the molecular basis of TRIAD, we conducted microarray analysis and compared gene expression profiles between TRIAD and low potassium–induced apoptosis in primary neurons. To detect initial changes, neurons were harvested at 1 h for RNA preparation. Duplicate experiments allowed us to extract eight genes whose expression levels changed in both apoptosis and TRIAD and a further 11 genes whose expression was changed specifically in TRIAD (). The latter group included YAP (), a transcriptional coactivator of p73 mediating apoptosis (). Detailed information of the selected genes is provided in Fig. S5 (available at ). Northern blotting confirmed that AMA treatment down-regulates YAP expression at the level of transcription (). Surprisingly, however, we identified novel isoforms of YAP containing 13-, 25-, and 61-nt inserts () in addition to the full-length form by PCR cloning with RNA extracted from nontreated normal cortical and cerebellar neurons. The insert sequences matched genomic sequence with consensus junction sequences (). All three insertions lead to a reading frame shift, causing truncation of the COOH-terminal transcriptional activation domain (, D and E; ). Therefore, we designated them YAPΔCs. Tissue expression profiling by RT-PCR revealed that the 13- and 61-nt insert isoforms (denoted here as Ins13 and Ins61, respectively) relatively specific to neurons (). Brain tissue (, third lane; not CTX or CBL neurons), including many glial and nonneuronal cells, showed only faint signals of the 13-nt variant comparable with those seen in other tissues (). Ins61 was highly specific to cortical neurons (). The 25-nt insert could not be detected by RT-PCR. Supporting the expression of YAPΔCs in neurons, a truncated YAP isoform–specific antibody stained cortical and striatal neurons in immunohistochemistry with human and mouse brains (see and ). In addition, temporal regulation of YAP isoforms during TRIAD was observed by Western blot analysis. Interestingly, although FL-YAP decreased before day 3 in cortical neurons, YAPΔCs were expressed at a relatively constant level (). It is also important to note that the levels of YAPΔCs were significantly lower in HeLa cells (). These data prompted us to test the function of YAP isoforms in TRIAD. p73 and YAP mediate cisplatin (CDDP)-induced apoptosis of a cancer cell line, MCF-7 cells (). In this case, DNA damage induced by CDDP leads to activation of p73, and the transcription cofactor YAP promotes p73-mediated transactivation of cell death genes, including Bax and possibly PUMA (). Truncation of the transcriptional activation domain () in YAP may impede transduction of the cell death stimulus, and YAPΔCs may act as dominant negatives against FL-YAP. As expected, luciferase assay showed that expression of YAPΔC isoforms represses p73-mediated activation of the p21/WAF1 gene promoter in MCF-7 cells by CDDP (, left; ). Overexpression of FL-YAP did not promote transcriptional activation any more () probably because the function of endogenous FL-YAP was saturated. YAPΔCs also showed repressive effects on CDDP-induced apoptosis of MCF-7 cells () mediated by FL-YAP (). In these assays, the expression of each truncate was confirmed in parallel (; right). Next, we tested whether YAPΔCs could repress TRIAD of primary cortical neurons (). Before the addition of AMA, neurons were infected with adenovirus vectors for YAPΔCs or the empty adenovector (AxCA) as a negative control (, left). Expression of YAPΔCs was confirmed by Western blot analysis simultaneously (, right). To further test whether YAPΔCs are involved in TRIAD, we transfected a siRNA targeting a sequence shared by three YAPΔC isoforms but not FL-YAP (). The siRNA accelerated TRIAD to ∼90% (), supporting the idea that YAPΔCs suppress the cell death process in TRIAD at least partially. The suppression of TRIAD by YAPΔCs suggested, in turn, that p73, the target transcription factor of FL-YAP, would be activated in TRIAD. Therefore, we analyzed the amount and phosphorylation of p73 in AMA-treated cortical neurons at day 2. As expected, AMA accelerated the phosphorylation of p73, whereas the total amount of p73 was not changed (). Together with the former results, YAPΔCs might inhibit the action of p73, leading neurons to apoptosis by antagonizing FL-YAP, especially at the early phase of TRIAD when FL-YAP is still expressed (). To investigate the relevance of YAP isoforms to the HD pathology, we infected primary cortical neurons with adenovirus vectors of YAPΔCs and found that expression of the truncated isoforms repressed Htt111-induced cell death of cortical neurons at 4 d after the infection of adenovirus vectors (; ). Consistently, YAPΔC-specific siRNA promoted Htt-induced cell death of cortical neurons (). We also found that mutant Htt induced p73 phosphorylation in cortical neurons at 2 d after infection (). Suppression of p73 by siRNA repressed cell death of mutant Htt-expressing neurons at day 4 (), suggesting the relevance of p73 to Htt-induced neuronal death. To examine the possible involvement of p73 in vivo, we analyzed p73 activation with brain samples of human HD patients. Western blotting with human brain samples suggested higher levels of p73 phosphorylation in HD brains than in control brains (). Correspondingly, immunohistochemical analysis revealed an increase of phosphorylated p73 in striatal neurons of mutant Htt transgenic mice (R6/2) at 4 wk (, middle). It is noteworthy that antiphosphorylated p73 antibody stained both the nucleus and cytoplasm of striatal neurons in R6/2 mice (, middle), although anti-p73 antibody detecting both nonphosphorylated and phosphorylated p73 proteins (full-length and NH-terminus deletion forms) dominantly stained the cytoplasm (, left). On the other hand, YAPΔCs were expressed in striatal neurons of both normal and R6/2 transgenic mice, whereas the signal was relatively stronger in transgenic mice (, right). Furthermore, phosphorylation of p73 was detected in striatal neurons of human HD patients (), suggesting that p73 is activated in human HD pathology. In this experiment (), because we used the antibody detecting full-length p73 but not ΔNp73, the full-length form of p73 was considered to be phosphorylated (, top). YAPΔCs were shown to exist in striatal neurons of human HD patients by a specific antibody (, bottom right) and to be colocalized with activated p73 in striatal neurons (, bottom). It is important to note that phosphorylated p73 and YAPΔCs were at very low levels in control human brains (, top). Collectively, these results suggest the possibility that p73 and YAPΔCs might be involved in the HD pathology. Finally, we examined the in vivo effect of YAPΔCs on Htt-induced neurodegeneration in models (). We generated more than three transgenic fly lines of human YAPΔCs. In the transgenic flies, the expression of YAPΔC protein was triggered by GMR-GAL4 that directs expression in the developing and adult eyes. To analyze the effects on photoreceptor neuron degeneration and/or the characteristic eye phenotype induced by the expression of human Htt120Q, we compared eye phenotypes between the F1 sibling fl ies at 10 d directly under the microscopy or by toluidine blue staining of 2-μm sections of epon-embedded eye tissues. Ommatidia structure and photoreceptor neurons were severely disrupted in GMR-Htt120Q/GMR-GAL4 double-transgenic flies (BL8533; ), whereas the expression of YAPΔC with a 61-nt insert (YAPΔC61) markedly preserved structure in triple-transgenic flies (GMR-Htt120Q/GMR-GAL4/UAS-YAPΔC61; ). Expression levels of YAPΔC61 and Htt120Q were checked in the same fly in parallel (). Quantitative analysis of rhabdomere numbers per ommatidium in four independent transgenic fly lines supported the repression of neurodegeneration by YAPΔC61 (). We observed similar improvement of neurodegeneration in other YAPΔC transgenic flies (not depicted). Collectively, these in vivo data further suggest the possibility that YAPΔC isoforms might play a protective role against the toxicity of mutant Htt in HD pathology. In this study, we report atypical neuronal death induced by transcriptional repression (TRIAD). Transcriptional repression by Pol II–specific inhibitors leads to a very slow atypical neuronal death whose progression is clearly different from the well-known cell death prototypes. A morphological feature of TRIAD might be the vacuolization of ER, although it should be stressed that the majority of neurons (>90%) do not show remarkable morphological changes. These findings might be relevant to the roles of transcriptional disturbance in HD disease (for review see ; ; ; ; ; ; ). In addition, the lengthy progression of TRIAD might cast light on the basic question of why neurons stay alive under neurodegeneration for a long period. To the best of our knowledge, there are a few atypical cell deaths that might be partially analogous to TRIAD. One is a lengthy cell death of during sorocarp formation, in which dying cells show cytoplasmic vacuolization (). The second, termed paraptosis, is induced by the overexpression of the intracellular domain of insulinlike growth factor I (IGF-I) receptor in 293T cells (). Paraptosis is characterized by vacuolization of the ER but no nuclear fragmentation, cellular blebbing, or apoptotic body formation (). These two atypical cell deaths might share molecular pathways (). Although TRIAD shows a related morphological change, TRIAD is clearly different from paraptosis, as the latter is inhibited by both actinomycin D and cycloheximide (). Another point that distinguishes TRIAD from paraptosis is the cell death stimulus. Paraptosis was reported only in ectopic expression of truncated IGF-I receptor in nonneuronal cells (). Furthermore, the role that we find for YAP in TRIAD has not been demonstrated in paraptosis or cell death. It is noteworthy that has recently reported a new type of cell death—necroptosis. They showed that in the absence of intracellular apoptotic signaling, extrinsic TNF stimulation triggers nonapoptotic cell death, showing necrotic morphology and autophagy. Although rapamycin did not increase typical LC3-negative vacuoles of TRIAD (Fig. S3) negating the autophagic component in TRIAD, we need to analyze carefully the relationship between TRIAD and necroptosis, including the viewpoint of cell death speed. It is also necessary to consider TRIAD with previous classifications of cell death. classified three types of cell death. Type 1 was manifested as nuclear condensation and pyknosis, reduced cytoplasmic volume, and late cell fragmentation/phagocytosis. Type 2 was an autophagic vacuolization in the cytoplasm, and type 3 was described as cytoplasmic cell death in which general organelle breakdown was apparent. Type 1 is apoptosis, and types 2 and 3 were necrotic (). In 1990, Peter Clarke redefined an earlier model of cell death developed by Schweichel and Merker (). Clarke's modification was to expand the forms of cytoplasmic cell death into types 3A and 3B. 3A is a nonlysosomal breakdown, and 3B is cytoplasmic (). Cells undergoing the 3A type of cell death show an initial swelling of cytoplasmic organelles and the generation of vacuoles that eventually fuse with the extracellular space. A breakup of cell structure without autophagic or heterophagic activity occurs. In type 3B death, which is also known as the cytoplasmic form of cell death, swollen organelles (dilated perinuclear space, ER, and Golgi apparatus) are apparent as well as vacuoles. The cell membrane retracts, and the nucleus becomes karyolytic/edematous. Heterophagic elimination can occur. Type 3B has also been termed paraptosis/oncosis. Among these, TRIAD is close to type 3B. However, in addition to the aforementioned reason, TRIAD seems to be different from type 3B because vacuolization of ER is far more remarkable than morphological changes of other organelles in TRIAD. In HD models, several studies have reported atypical cell death with cytoplasmic vacuolization. reported that mutant Htt accumulates in punctate structures mimicking endosomal–lysosomal organelles of affected HD neurons. They further showed by extensive analyses, including immunoelectron microscopy, that mutant Htt appears in autophagosomes (). Other studies also pointed out the possible involvement of autophagy in the HD disease pathology (; ; ). Meanwhile, isolated VCP (valosin-containing protein)/p97, a member of the AAA+ family of ATPase proteins, as a HD-interacting protein. The expression of the mutant form of VCP leads to cytoplasmic vacuolization, which might be homologous to vacuoles in TRIAD because they were fused to ER (). Collectively, although our results so far seem to negate the identity of the TRIAD vacuoles to autophagosomes, we cannot exclude the possibility that they might share certain characteristics with the vacuoles reported in HD models. As for the molecular pathway of TRIAD, YAPΔCs and p73 might modify the process. Up- or down-regulation of YAPΔCs suppresses or enhances TRIAD in cortical neurons, respectively (). Together with evidence that AMA treatment increases active p73 in neurons () and that YAPΔCs remain during TRIAD of cortical neurons (), these data suggest that p73-mediated cell death signaling might be attenuated by YAPΔCs in TRIAD. Consistently, the percentage of morphologically changed neurons (vacuole-possessing neurons) was very low. It might be a reason why TRIAD does not progress rapidly like apoptosis. p73 was activated in human and mouse HD pathology in vivo ( and ). YAPΔC isoforms were coexpressed in affected neurons of human HD patients (). Repression of p73 and expression of YAPΔCs attenuated Htt-induced neuronal cell death of primary neurons (), whereas YAPΔC repression enhanced the neuronal cell death (). Furthermore, YAPΔC isoforms suppressed neurodegeneration of photoreceptor cells of in vivo (). These findings suggest that YAPΔCs and p73 might be relevant to the HD pathology. p53 has been implicated in the HD pathology because p53 coaggregates with mutant Htt (). recently reported that mutant Htt interacts with, translocates, and activates p53. They also showed that mating mutant Htt transgenic mice with p53-null mice ameliorates neurological symptoms by mutant Htt (). These results suggest that p53 activation promotes the HD pathology. Because p73 and p53 belong to the same family of transcription factors recognizing a similar consensus sequence on genomic DNA (for review see ), the common cascade shared by the two factors should be investigated in the HD pathology. For instance, upstream signals activating these two factors and target gene activation by these transcription factors in the HD pathology should be analyzed in the future. On the other hand, because p53 is suggested to have a direct effect on mitochondria (), it might be necessary to test whether p73 also plays a similar role. It is important to note that hyperactive p73 could trigger vacuolar changes of ER in nonneuronal cells (). If this is true, the vacuole formation in TRIAD might be triggered by activated p73. In this case, although ER stress could be induced by mutant polyQ protein (; ), ER stress might also be evoked by a signal from the nucleus in parallel. Investigation on the possible connection between the nucleus and ER might contribute to understanding the polyQ pathology. The hypothetical pathway should be examined and elucidated in the future. In summary, our results present a novel model of cell death that might cast more light on the HD pathology. Cerebral cortex tissues isolated from E17 Wistar rat embryos and cerebellar tissues isolated from P7 Wistar rat pups were minced (with razors) and treated with 0.25% trypsin (Invitrogen) in PBS, pH 7.5, at 37°C for 20 min with gentle shaking every 5 min. After stopping the reaction with DME containing 50% FBS, DNase I (Boehringer) was added to the solution at a final concentration of 100 μg/ml, and tissues dissociated gently by pipetting with blue tips. Cells filtered by nylon mesh (pore size of 70 mm; Falcon; BD Biosciences) were collected by centrifugation, resuspended in DME supplemented with 20 mM glucose, 16 mM sodium bicarbonate, 4 mM glutamine, 25 μg/ml gentamicin, and 10% FBS, and plated on 24-well dishes (Corning) coated by poly--lysine (Sigma-Aldrich) at 3 × 10 cells/well. 12 h after plating, cytosine arabinoside was added to the culture medium at 4 M of final concentration to prevent the growth of glial cells. Cerebellar neurons were cultured at high potassium (25 mM) ordinarily but were cultured at 5.4 mM potassium to induce apoptosis. Cortical neurons were cultured at low potassium condition (5.4 mM). Necrosis of cortical neurons was induced by the freeze and thaw treatment. To induce TRIAD, AMA (Sigma-Aldrich) was added to the medium at a final concentration of 10 or 25 μg/ml, except for dose–response survival experiments in which the final concentration was 10–250 μg/ml. Actinomycin D (Sigma-Aldrich) was added to the medium at 0.1, 0.5, or 2.5 μg/ml. Cell death assays were performed either by trypan blue dye exclusion assay or MTT assay as described in each figure legend. For trypan blue assay, cells were incubated for 5 min in 0.4% trypan blue (Invitrogen). Blue-stained (nonviable) and nonstained (viable) cells were counted (at least 2,000 cells for each condition) in 10–20 visual fields randomly selected at 100× from each of three dishes, as described previously (). MTT assay was performed with MTT cell proliferation/viability assay (R&D Systems) according to the commercial protocol. At each time point, the value of drug-treated cells was corrected to the value of nontreated cells as 100%. Regarding electron microscopic observation, cells were washed with PBS three times, fixed in 2.5% glutaraldehyde/0.1 M phosphate buffer, and treated with 1% OsO/0.1 M phosphate buffer for 2 h. Fixed cells were dehydrated through a graded ethanol series and embedded in epoxy resin. Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a transmission electron microscope (H-9000; Hitachi) at 24°C (5,000–50,000×). Numerical aperture of the objective lens was 4, and the imaging medium was air. Data acquisition was performed by electron microscope film. 24 h after transfection of pEGFP-LC3, HeLa cells were treated with 10 μg/ml AMA and observed by fluorescence microscopy (). To further analyze the relationship between LC3-positive phagosomes and AMA-induced vacuoles, autophagy was induced by 200 ng/μl rapamycin for 2 h (Sigma-Aldrich; for review see ). LC3-positive and -negative vacuoles were counted in the presence or absence of AMA. HeLa cells were transfected with pEGFP-LC3 by SuperFect (QIAGEN), collected 36 h after transfection, and subjected to Western blot analysis. Anti-EGFP polyclonal antibody (BD Living Colors) and anti-LC3 antibody were used at dilutions of 1:1,000 and 1:2,000, respectively. pEGFP-LC3 and anti-LC3 antibody were gifts from T. Yoshimori (National Institute of Genetics, Mishima, Japan) and N. Mizushima (Tokyo Metropolitan Institute for Medical Science, Tokyo, Japan). After treatment of AMA (Sigma-Aldrich) for 6 h, HeLa cells were washed with PBS and fixed using 4% PFA for 15 min at RT. Cells were incubated for 1 h at RT with the following primary antibodies: anti-CCO1 mouse monoclonal antibody (1:100; Invitrogen); anti-EEA1 mouse monoclonal antibody (1:100; Transduction Laboratories); anti-Golgi58k mouse monoclonal antibody (1:100; Sigma-Aldrich); and anti-CD63 (1:100; Cymbus Biotechnology Ltd.). Secondary antibodies conjugated with AlexaFluor488 (Invitrogen) were used at a dilution of 1:1,000 and hybridized for 30 min at RT. HeLa cells were transfected by pEGFP-LC3 (a gift of N. Mizushima and T. Yoshimori; ) or pECFP-ER (BD Biosciences) using Superfect (QIAGEN) according to the manufacturer's instructions. Primary neurons were treated with 25 μg/ml AMA as indicated and were dissolved in 62.5 mM Tris-HCl, pH 6.8, 2% (wt/vol) SDS, 2.5% (vol/vol) 2-mercaptoethanol, 5% (vol/vol) glycerin, and 0.0025% (wt/vol) bromophenol blue on culture dishes. Positive controls for caspase-3 and -7 were prepared from HeLa cells treated with 1 μM staurosporin (Sigma-Aldrich) for 5 h. For a caspase-12 control, HeLa cells were treated with 20 μM (Calbiochem) for 24 h. Primary and secondary antibodies were diluted as follows: anticaspase-3 polyclonal rabbit antibody (Cell Signaling) at 1:1,000; anticaspase-6 polyclonal antibody (Cell Signaling) at 1:500; anticaspase-7 polyclonal antibody (Cell Signaling) at 1:500; anticaspase-12 polyclonal antibody (14F7; Sigma-Aldrich) at 1:1,000; HRP-conjugated anti–rabbit IgG (GE Healthcare) at 1:3,000; and HRP-conjugated anti–rat IgG (Sigma-Aldrich) at 1:20,000. 10 primary cortical neurons were treated with 10 μg/ml AMA as indicated (). As a positive control, the same amount of primary cortical neurons were treated with 1 μM staurosporin (Sigma-Aldrich) for 8 h. The cells were washed twice with ice-cold PBS on the dish, collected, and suspended in 500 μl of ice-cold buffer (20 mM Hepes, pH 7.4, 10 mM KCl, 1.5 mM MgCl, 1 mM EDTA, 1 mM DTT, 1 mM PMSF, and 250 mM sucrose) and disrupted by moderate strokes in a homogenizer. The homogenate was centrifuged twice at 1,300 for 5 min to remove nuclei, unbroken cells, and large membrane fragments. From the supernatant, mitochondria were isolated by further centrifugation at 17,000 and 4°C for 15 min. Pellets were dissolved in the sample buffer described above, separated by 15% SDS-PAGE, blotted to polyvinylidene difluoride membranes (Fine Trap; Nihon Eido), incubated with cytochrome c monoclonal antibody (1:1,000; Santa Cruz Biotechnology, Inc.), and subjected to HRP-coupled detection. The supernatant of the final centrifugation was used as a cytosolic fraction. Cells in culture dishes are harvested in TRIzol reagent (Invitrogen) after rinsing with PBS twice, and total RNA was prepared according to the manufacturer's protocol. Labeling and amplification of RNA was performed using the Agilent Fluorescent Linear Amplification Kit (G2554A; Agilent Technologies) according to the manufacturer's protocol. First, double-stranded cDNAs with a T7 promoter were synthesized from 2 μg of total RNA by Moloney murine leukemia virus reverse transcriptase using an oligonucleotide dT-primer, which contains the T7 promoter sequence, and random hexamers (40°C for 4 h). Then, using these double-stranded cDNAs as templates, cRNA was synthesized by T7 RNA polymerase using Cy3- or Cy5-labeled CTP (40°C for 1 h). cRNAs from AMA-treated cortical neurons, AMA-treated cerebellar neurons, or low potassium–exposed cerebellar neurons were labeled with Cy3 or Cy5. Synthesized cRNA was precipitated with lithium chloride, ethanol rinsed, and dissolved in nuclease-free water. To check the quality of cRNA, OD, OD, A (for Cy3), and A (for Cy5) measurements were taken. Then, OD/OD, amplification rates and dye incorporation rates (pmol/μg RNA) of cRNA were calculated. Using these criteria, we found that our samples were of high quality (OD/OD, <2.0; amplification rate, <400; Cy3 incorporation, <15 [pmol/μg RNA]; and Cy5 incorporation, <12 [pmol/μg RNA]). Hybridization procedures were performed using the In situ Hybridization Kit Plus (5184–3568; Agilent Technologies) according to the manufacturer's protocol. First, Cy3- and Cy5-labeled cRNAs (1 μg each) were mixed and incubated with fragmentation buffer (Agilent Technologies) at 60°C for 30 min. Mouse Development Oligo Microarray (G4120A; Agilent Technologies), which contains 20,371 60-mer oligonucleotides from mouse cDNA, was hybridized with fragmented cRNA targets at 60°C for 17 h using CHBIO (Hitachi). Hybridized microarrays were rinsed twice and dried by spraying N gas (99.999%) using a filter-equipped air gun (mycrolis KK; Nihon). RT-PCR cloning of YAP was conducted with cDNA reverse transcribed from 1 mg of total RNA prepared from rat cortical neurons by using the RNA LA PCR Kit (Takara) and primers F (5′-GGAATTCTATGGAGCCCGCGCAA-3′) and R (5′-ACGCGTCGACCTATAACCACGTGAG-3′). PCR amplification was performed for 35 cycles (94°C for 30 s, 52°C for 30 s, and 72°C for 90 s). The resultant cDNAs were subcloned between EcoRI and SalI sites of pBluescriptII SK+. Nucleotide sequences were determined by using M13 or synthesized internal primers and the ABI PRISM BigDye Terminator Cycle Sequencing Kit version 3.1 (Applied Biosystems) and ABI PRISM 310 DNA Sequencer (Applied Biosystems). pBluescript plasmids containing 13-, 25- and 61-nt insert forms of YAP were named pBSins13, pBSins25, and pBSins61, respectively. The cDNA of each YAP insert was subcloned into pCI neo (Promega) and denoted pCIins13, pCIins25, and pCIins61, respectively. 5 × 10 cells MCF-7 cells were transiently transfected with 5 μg of pGL3-Bax-Luc () with pCI–FL-YAP, -YAPΔC (pCIins13, pCIins25, and pCIins61), or control pCI-neo using LipofectAMINE 2000 (Invitrogen) according to the protocol described previously (). Cells were resuspended in 62.5 mM Tris-HCl, pH 6.8, 2% (wt/vol) SDS, 2.5% (vol/vol) 2-mercaptoethanol, 5% (vol/vol) glycerin, and 0.0025% (wt/vol) bromophenol blue on culture dishes. Cell lysates prepared from wells containing either 3.3 × 10 HeLa cells or 1.0 × 10 primary neurons were subject to SDS-PAGE gels, transferred onto polyvinylidene difluoride membranes (Fine Trap; Nihon Eido), incubated with each primary antibody for 1 h and the corresponding HRP-conjugated secondary antibody for 30 min, and visualized using the ECL Western Blotting Detection System (GE Healthcare). The dilution conditions for primary and secondary antibodies were as follows: anti-YAP polyclonal rabbit antibody (H-125; Santa Cruz Biotechnology, Inc.) at 1:1,000; anti-GAPDH mouse monoclonal antibody (Chemicon) at 1:100,000; HRP-conjugated anti–mouse IgG (GE Healthcare) at 1:5,000; and HRP-conjugated anti–rabbit IgG (GE Healthcare) at 1:3,000. Anti-YAPΔC rabbit polyclonal antibody was raised against the common COOH-terminal peptide (SVFSRDDSGIEDNDNQ) by immunizing rabbits and was used for Western blotting at a 1:1,000 dilution. The replication-defective adenovirus vectors were constructed by using the Adenovirus Expression Vector Kit (Takara) according to the manufacturer's instructions. In brief, cDNAs of YAP isoforms were digested with EcoRI and SalI from pBSYAP-FL (containing rat FL-YAP), pBSins13, pBSins25, and pBSins61. Ends were blunted using the blunting high kit (Toyobo), and each insert was subcloned into the SwaI site of the pAxCAwt cosmid (Takara). The resultant cosmids were transfected into 293T cells by the calcium-phosphate method with digested DNA of adenovirus and the medium containing dead cells recovered as the virus solution. After two or three rounds of amplification (5 × 10 and ∼5 × 10 plaque-forming units/ml), clonality was checked by restriction with endonucleases and PCR. We designated the adenovirus vectors AxCAYAP-FL, AxCAins13, AxCAins25, and AxCAins61. The vectors were used for infection of HeLa cells and primary neurons at a multiplicity of infection (MOI) of 100. Preliminary examination of the efficiency of protein expression and toxicity of adenovirus was performed by infecting primary neurons with a vector for EGFP and a mock vector at multiple MOI, respectively. More than 90% of the neurons expressed EGFP at an MOI of 100. The difference in cell death percentage between noninfected and mock-infected neurons estimated by trypan blue staining was <3% when the MOI did not exceed 500. 10 μg of total RNA from primary culture neurons was subjected to electrophoresis using a MOPS/formaldehyde gel. Separated RNAs were capillary blotted to Hybond-N (GE Healthcare) and fixed by UV cross-linking (120,000 μJ/cm). Full-length cDNA of ins61 was digested from pBSins61, purified from gel, and radiolabeled using α-[P]dCTP (GE Healthcare) and a random primer DNA labeling kit (Takara). P-labeled probes were hybridized to nylon membrane at 60°C overnight with shaking. Hybridized membrane was rinsed with 1× SSC, 0.1% SDS at 50°C for 20 min twice, and with 0.1× SSC and 0.1% SDS at 60°C for 20 min twice. The membrane was then exposed to X-ray film for an appropriate time at −80°C. Cells were transfected with siRNA oligonucleotides by RNAiFect (QIAGEN) according to the manufacturer's instructions. 2.5 × 10 cells in six-well dishes were infected at 0.5 μg siRNA/well 24 h after plating. 24 h after infection, AMA was added to a final concentration of 10 μg/m. The cell death assay was performed after another 24 h. Sequences of siRNAs of YAP and p73 were the same as those published previously (). Sequences of the YAPΔC isoform-specific siRNAs were 5′-r(ACCGTCAGAGCGGGAATTAGCTC)d(TT)-3′ and 5′-r(GAGCTAATTCCCGCTCTGACGGT)d(TT)-3′, corresponding to the common exon among three YAPΔC isoforms. HeLa cells and cortical neurons were treated with 10 μg/ml AMA for 6 h and 2 d, respectively. For the Htt experiments, HeLa cells and primary cortical neurons were harvested 2 d after infection. HeLa cells or primary cortical neurons were dissolved in TNE buffer (10 mM Tris-HCl, pH 7.8, 1% NP-40, 0.15 M NaCl, 1 mM EDTA, 10 μg/ml aprotinin, 1 mM PMSF, 1 mM NaVO, and 1 mM NaF), and the supernatant was collected after centrifugation. Nonspecific binding proteins were removed by preincubation with protein G–Sepharose beads (GE Healthcare), and anti-p73 goat polyclonal antibody (S-20; Santa Cruz Biotechnology, Inc.) was added to the supernatant at 1:200. The mixture was incubated overnight at 4°C and precipitated by protein G–Sepharose beads for 1 h at 4°C. After washing five times with TNE, the precipitate was boiled in 2× loading buffer and subjected to Western blot analysis. The filter was blotted with the anti-p73 goat polyclonal antibody (S-20; 1:1,000; Santa Cruz Biotechnology, Inc.) or antiphosphorylated p73 rabbit polyclonal antibodies (1:1,000; Cell Signaling) followed by HRP-coupled detection. For analysis of p73 phosphorylation in human brain, each sample of striatum was homogenized in 20× vol TNE and subjected to the detection of p73 phosphorylation by Western blotting and immunohistochemistry. Brain tissues were prepared from 4-wk-old R6/2 transgenic mice and the littermates. After deparaffinization and rehydration, the sections were incubated sequentially with 3% hydrogen peroxide for 30 min to inhibit endogenous peroxidase, 1.5% normal goat serum in PBS for 1 h at RT, and either a rabbit polyclonal antibody specific for full-length p73 that was raised against the NH-terminal 80 amino acids of p73 (H-79; 1:100; Santa Cruz Biotechnology, Inc.), an antiphospho-p73 rabbit polyclonal antibody (1:50; Cell Signaling), or an anti-YAPΔC rabbit polyclonal antibody against the common COOH-terminal peptide (SVFSRDDSGIEDNDNQ) of YAPΔC isoforms (1:100). These incubations were overnight at 4°C. The slides were incubated with anti–rabbit EnVision conjugates of secondary antibody (DakoCytomation) for 1 h at RT and visualized with DAB (Sigma-Aldrich). The same protocol was applied for immunohistochemistry of human brain sections. For double staining, each section after the first staining was agitated in stripping buffer (0.05 M glycine-HCl, pH 3.6) for 3 h at RT, hybridized with anti–glial fibrillary acidic protein polyclonal antibody (1:1,000; Chemicon) overnight at 4°C and with anti–rabbit EnVision conjugates of secondary antibody (DakoCytomation) for 1 h at RT, and visualized with DAB (Sigma-Aldrich) containing NiCl 6HO. Postmortem brain tissues were prepared from HD patients diagnosed by CAG repeat expansion. The paraffin-embedded section was deparaffinized, rehydrated, and blocked with 5% skim milk in PBS for 30 min at RT. Single staining was performed as described in the previous section. For double staining, the section was incubated with anti-p73 rabbit polyclonal antibody specific for full-length p73 (H-79; 1:500; Santa Cruz Biotechnology, Inc.) or with or an anti-YAPΔC rabbit polyclonal antibody overnight at 4°C, washed with TNT (0.1% Tween 20–TBS) buffer twice, incubated with HRP-conjugated secondary antibody (1:3,000; GE Healthcare) for 1 h at RT, washed with TNT buffer twice, and visualized by incubation with FITC-tyramide (1:200; PerkinElmer) for 10 min. The tyramide complex was stripped off by incubation with 0.05 M glycine-HCl, pH 3.6, for 3 h at RT. After complete stripping, antiphospho-p73 rabbit polyclonal antibody (1:500; Cell Signaling) was hybridized and visualized with Cy3-conjugated secondary antibody (1:1,000; Chemicon). Fly culture and mating were carried at 25 and 60% humidity. P(GMR-GAL4) (BL8121) and P(GMR-HD120Q) (BL8533) () were obtained from the Bloomington Stock Center. The UAS-YAPins13, 25, and 61 transgenic flies were generated by cloning the corresponding human cDNA into pUAST transformation vector and injecting the construct DNA into cantonized w(cs10) () by standard methods (). Genotypes of the YAP transgenic flies were determined by mating them with double balancer flies, and they were kept with a balancer gene before use. To analyze the effects of YAPins61 on photoreceptor neuron degeneration and/or the characteristic eye phenotype induced by the expression of human Htt 120Q, we compared eye phenotypes between the F1 sibling flies (GMR-HD120Q/UAS-YAPΔ61;GMR-GAL4/+ vs. GMR-HD120Q/+; GMR-GAL4/+) directly under the microscopy VH5000 (Keyence) or by toluidine blue staining of 2-μm sections of epon-embedded eye tissues. For sections of fly photoreceptor neurons, adult heads (0–10 d) were prefixed overnight in 2% formaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer overnight at 4%, postfixed in 1% osmium at RT for 3 h followed by dehydration in ethanol and embedding in epon for vertical and transversal semi-thin sections (2 μm), and stained with toluidine blue. At least five individuals were examined in each fly line and at each time point. Fig. S1 shows an MTT assay of TRIAD. Fig. S2 shows immunocytochemical analysis of TRIAD-associated vacuoles. Fig. S3 shows that TRIAD is neither autophagy nor apoptosis. Fig. S4 shows that actinomycin D also induced TRIAD. Fig. S5 shows transcriptome analysis of TRIAD. Online supplemental material is available at .

Huntington's disease (HD) is a neurodegenerative disorder caused by expansion of the CAG repeat in the gene encoding Huntingtin (Htt), which confers to the protein an expanded NH-terminal polyglutamine (polyQ) stretch of >35 residues (for review see ). The function of Htt is largely unclear. It has been shown to interact with microtubules () and to display anti-apoptotic activity (, ). Insights into its function came from studies of Htt-interacting proteins (HIPs) and Htt-associated proteins (HAPs). For example, interactions with HIP1, HIP1R, PACSIN1, SH3GL3, and HIP14 have implicated Htt in clathrin-mediated endocytosis. Studies of HAP1 have suggested a role for Htt in axonal transport in neurons by linking vesicles to the dynein–dynactin motor complex (; ). The polyQ expansion confers the adjacent proline-rich sequence in Htt alterations in binding affinity to HIPs/HAPs. Thus, release from or sequestration of these molecules by mutant Htt has been implicated in the pathogenetic mechanisms. For example, the tighter binding of HAP1 to mutant Htt is thought to impair the correct dynactin–dynein motor complex assembly and cause a trafficking defect, leading to neuronal degeneration (; ; ). Consistently, mutant Htt was recently shown to release dynein from microtubules and reduce the motility of EGFP–brain-derived neurotrophic factor–containing vesicles in vivo (). However, the upstream events that target Htt and its partners to their various sites of function and the mechanisms whereby they regulate intracellular trafficking remain elusive. In this study, we report an unexpected link between Htt and the small GTPase Rab5 via the adaptor protein HAP40. Rab5 is a key regulator of endocytosis that orchestrates the recruitment of multiple effector proteins on the early endosome membrane to regulate organelle tethering, fusion, and microtubule-dependent motility (). Our data extend the analysis of the Rab5 effector machinery by functionally implicating the interaction between Rab5 and Htt in the regulation of the differential association of early endosomes with the actin and tubulin cytoskeleton. Affinity chromatography revealed several downstream effectors of the small GTPase Rab5 (). Surprisingly, among the proteins specifically eluted from the GST-Rab5–GTPγS but not from the Rab5–guanosine diphosphate (GDP) nor GST-Rab4 affinity column, we identified Htt and HAP40 () by mass spectrometry and immunoblotting (). Therefore, we investigated the function of HAP40 and Htt with respect to Rab5. We first tested whether HAP40, Htt, or both interact directly and specifically with Rab5. To this end, full-length HAP40 was cloned from a rat brain cDNA library and in vitro translated. Because of its large molecular mass (348 kD; ), fragments of wild-type Htt were translated to facilitate the analysis. In vitro translation yielded major products of predicted size as well as lower molecular mass bands presumably as a result of internal initiation (). Immobilized Rab5-GST fusion protein preloaded with either GDP or GTPγS were incubated with the translation products and washed, and bound proteins were eluted with glutathione analyzed by SDS-PAGE and autoradiography (). Similar to early endosome antigen 1 (EEA1), which served as a positive control, HAP40 displayed specific binding to GTPγS- versus GDP-bound Rab5 (, compare lanes 1 and 11 with lanes 2 and 12). In contrast, none of the Htt fragments exhibited significant binding to Rab5 (, lanes 3–10). Because Htt was purified on the GST-Rab5 affinity column, we next tested whether binding of Htt to Rab5 occurs indirectly and requires HAP40 as a bridge. HAP40 and Htt fragments were cotranslated in vitro (, lanes 6–9) and applied on the Rab5 columns. Indeed, the COOH-terminal part of Htt (, Htt4) was eluted together with HAP40 from GTPγS- but not GDP-bound GST-Rab5 (, lanes 7 and 8). None of the other Htt fragments displayed such binding (, lanes 1–6), which is consistent with the reported interaction map between HAP40 and Htt (). Neither the Htt fragments nor HAP40 displayed binding to Rab4, Rab6, Rab7, or Rab11 (), suggesting the interaction with Rab5 is specific. Thus, we conclude that HAP40 binds to the COOH-terminal part of Htt and links the complex to active Rab5. By applying a 10-fold excess of the COOH-terminal fragment of Htt onto the Rab5 column to reduce binding of free HAP40 to Rab5 (), we estimated the stoichiometry of the Rab5/HAP40/Htt interaction in this assay to be ∼1:1:1 (see Rab5 affinity…cloning). We began testing the functional relevance of this interaction in vivo by immunofluorescence microscopy analysis of HeLa cells. First, we verified that the anti-HAP40 antibody resulted in specific staining above background levels for detection of the endogenous antigen (). Endogenous HAP40 displayed a diffuse staining in the cytoplasm and accumulated in the nucleus, whereas endogenous Htt localized to discrete cytoplasmic structures () as reported previously (; ). Colocalization of endogenous HAP40 and Htt was hardly detectable (). Early endosomes labeled with EGFP-Rab5 displayed little colocalization with endogenous Htt (7 ± 5% overlap; = 10; ). However, the association of HAP40 with early endosomes dramatically increased upon overexpression. In cells overexpressing HAP40 (see ) but not EGFP-Rab5 alone, HAP40 significantly colocalized with endogenous Htt on EGFP-Rab5–positive early endosomes (, compare B with D; endogenous Htt and EGFP-Rab5: 43 ± 6% overlap, = 10; HAP40 and EGFP-Rab5: 31 ± 7% overlap, = 10; , compare C with D). We experimentally verified that the HAP40 fluorescence signals () were not caused by bleed-through of the Htt signals (because of extended AlexaFluor568 emission in the Cy5 channel) and that swapping these fluorescent dyes on the secondary antibodies resulted in similar distribution patterns of Htt and HAP40 (see Cell culture procedures). The endosomal colocalization of endogenous HAP40 and Htt was even more striking upon expression of the activated EGFP-Rab5Q79L mutant (Htt: 52 ± 7% overlap, = 8), which caused the characteristic swelling of early endosomes (; ). Given the difficulties detecting endogenous HAP40 and Htt on early endosomes in untreated cells () we sought to verify the changes in the localization of both proteins upon the overexpression of HAP40 biochemically. To this end, we prepared early endosomes from HeLa cells for Western blot analysis. Indeed, we found that the levels of HAP40 and Htt increased on early endosomes from the HAP40 overexpressor compared with untreated cells (). Early endosome (EEA1 and transferrin receptor [Tfr]) as well as lysosomal (lysosome-associated membrane protein (LAMP); ) and Golgi (GM130; ) markers remained unchanged, confirming equal loading and the specificity of changes through elevated HAP40 on early endosomes. To confirm the requirement of HAP40 for the recruitment of Htt onto early endosomes, we transfected HeLa cells with short interfering RNA (siRNA) duplexes against HAP40 and unrelated siRNA (against GFP) as control. The HAP40 siRNA specifically and efficiently reduced the protein levels by ∼90% (), whereas the level of EEA1 remained unchanged. When cells were cotransfected with the expression vector for EGFP-Rab5Q79L and HAP40 siRNA, Htt was no longer detectable on the enlarged endosomes (8 ± 6% overlap, = 9; ), confirming HAP40 as a prerequisite to bridge Htt to active Rab5. Collectively, these data suggest that active Rab5 and HAP40 are rate limiting for the recruitment of Htt onto early endosomes. Htt has previously been shown to bind microtubules and regulate microtubule–motor interactions (; ; ; ). Because Rab5 regulates endosome motility along microtubules (; ), we explored the role of the Htt–HAP40 complex in this process. First, by using a cell- and cytosol-free assay that recapitulates the Rab5-dependent movement of early endosomes along microtubules (), we found that Htt–HAP40 inhibited microtubule-dependent early endosome motility. Addition of the GTPγS- (containing Htt and HAP40; ) but not the GDP-loaded Rab5 column eluate reduced the motility compared with control conditions (). This inhibition was specifically rescued with anti-Htt but not unrelated antibodies. The addition of 1 μM GST-HAP40 fusion protein completely blocked the in vitro motility (). Second, we performed a biochemical early endosome–microtubule-binding assay as described previously () with some modifications to improve the quantitative assessment. An early endosome–enriched fraction was prepared from HeLa cells pulsed with rhodamine–transferrin, incubated with taxol-stabilized microtubules, ATP, and factors to be tested, and centrifuged through a sucrose cushion. The resulting pellet of microtubule-associated material was analyzed by immunoblotting () and fluorimetrically for the rhodamine–transferrin content (). In a dilution series for calibration, we verified that the amount of endosomes and fluorescence intensity correlated linearly (see Microtubule and actin spin-down assays). Western blotting revealed that the β-tubulin content was similar between samples, ruling out secondary effects on microtubule stability (, compare lanes 2–5 with lane 1). The addition of 1 μM GST-HAP40 protein decreased the amount of early endosomes in the pellet as revealed by EEA1 and Tfr (, compare lane 3 with lane 2). The GST-Rab5–GTPγS column eluate adding ∼0.3 μM HAP40 () caused a similar inhibition (, compare lane 4 with lane 2) that was rescued through the addition of antiserum against Htt (, compare lane 5 with lanes 4 and 2). To rule out the idea that the observed differences result from the bundling of microtubules causing unspecific cosedimentation of any membranous structure, we probed the pellets for nonendosomal contaminants in the fraction. Both the lysosomal (LAMP1) and Golgi marker (GM130) pelleted with similar efficiency in all samples. Collectively, these data indicate that Htt–HAP40 specifically lowers the binding of early endosomes to microtubules. Having validated the assay, we next quantified the amount of endosomes bound to microtubules fluorimetrically. Omission of either microtubules or early endosomes reduced the fluorescence signal to background levels (). As reported previously (), the association of early endosomes with microtubules was energy dependent and required active Rab5. Omitting ATP or substituting it with the nonhydrolyzable adenylyl-imidodiphosphate analogue resulted in an ∼50% reduction in microtubule binding. The requirement for ATP can be explained by the role of PI3-K in the assembly and maintenance of a functional Rab5 domain on endosomes () and in recruitment of the endosomal kinesin KIF16B (). Extraction of Rab proteins from membranes by the addition of 1 μM of recombinant Rab–GDP dissociation inhibitor (GDI; ) or treatment with RN-tre, a GTPase-activating protein for Rab5 (), caused an ∼40% reduction of endosomes in the pellet (). As for endosome motility (), Htt and HAP40 inhibited endosome–microtubule binding. The GST-Rab5–GTPγS column eluate led to an ∼25% reduction in binding (), but supplementing the reaction with antibodies against Htt restored binding to control levels. The addition of 1 μM HAP40-GST fusion protein decreased the binding by ∼60%, whereas GST alone did not have any effect. Collectively, these data underpin the ability of HAP40 and Htt to destabilize endosome–microtubule association. We next performed time-lapse video microscopy studies to correlate these biochemical findings in vitro with the regulation of early endosome dynamics in vivo. We detected a drastic reduction in motility of EGFP-Rab5–positive endosomes in HeLa cells overexpressing HAP40 () compared with cells transfected with EGFP-Rab5 alone ( and Videos 1 and 2, available at ). Whereas some residual motility activity was observed in the cell periphery, early endosomes in the perinuclear region appeared static, with frequent short-range movements almost completely impaired in long-range motility (, ). Collectively, these data suggest that the Rab5-dependent recruitment of Htt onto endosomes by HAP40 disrupts early endosome–microtubule interactions, thus leading to a reduction in organelle motility. We next asked whether alterations of endosome motility could occur in cells bearing the HD mutation. Primary fibroblasts from five healthy individuals and five unrelated HD patients were transfected with EGFP-Rab5 to compare the motility of early endosomes. The identity of the HD cell lines was confirmed by Western blotting to detect the polyQ-expanded Htt. Because the cells were derived from patients heterozygous for the HD gene, they express both normal and mutant Htt. The latter is known to display a lower mobility in SDS-PAGE (), thus causing a doublet on the blot (). Strikingly, we observed a severe reduction in early endosome motility in all HD cell lines compared with fibroblasts from healthy individuals ( and Videos 3 and 4, available at ). The similarity between this phenotype and the alterations induced upon HAP40 overexpression in HeLa cells ( and Videos 1 and 2) hinted to a common molecular basis. Interestingly, we discovered an ∼10-fold up-regulation of endogenous HAP40 levels by Western blotting in all HD cell lines compared with normal fibroblasts (). As a control, the levels of EEA1 remained unchanged. Moreover, we found that HAP40 protein levels were significantly elevated in striatal tissue (caudate, putamen, accumbens, and globus pallidus) from human postmortem brains affected by HD (adult onset grade) compared with control brains (). Our data suggest that the motility block in HD cells may be caused by elevated HAP40 levels, as mimicked by overexpression of HAP40 in HeLa cells. Consistently, endogenous Htt localized to EGFP-Rab5–labeled endosomes in fibroblasts from human HD patients as well as striatal STHdhQ cells from a HD mouse model () but not in cells lacking the mutant Htt (). This phenotype is also caused by overexpressed HAP40 in HeLa cells (). Our data do not exclude the possibility that the observed inhibition of early endosome motility may have other underlying causes. As a test to our hypothesis, we attempted to rescue the inhibition of early endosome motility by specifically ablating HAP40 from the HD fibroblasts by RNA interference (RNAi). Transfection of HAP40 siRNA efficiently reduced the HAP40 protein levels in these cells () as in HeLa cells (). Indeed, endosome motility was restored by RNAi against HAP40 ( and Video 5, available at ) but not against GFP, suggesting that the up-regulation of HAP40 is indeed the underlying mechanism of the motility defect in HD fibroblasts. Given the ability of HAP40 to reduce endosome–microtubule binding in vitro (), we investigated whether the observed motility block in HD cell lines was caused by a release of endosomes from microtubules in vivo. Immunofluorescence analysis on cells transfected with EGFP-Rab5 showed a considerable alignment of early endosomes to microtubules in healthy fibroblasts (82 ± 9% overlap, = 10) but to a much lesser extent in HD cell lines (15 ± 6% overlap, = 9; , compare F with G). In contrast, early endosomes were strikingly aligned with filamentous actin (F-actin) in all five HD cell lines (44 ± 8% overlap, = 10) but not in healthy fibroblasts (2 ± 1% overlap, = 10; , compare G with F). A similar phenotype was obtained by the overexpression of HAP40 in HeLa cells (not depicted). The alignment of early endosomes on F-actin in HD cell lines could be a secondary effect from the inhibition of endosome–microtubule interactions. To directly test the role of HAP40 in the association between early endosomes and actin filaments (), we modified the biochemical sedimentation assay used to study endosome–microtubule interactions () by replacing taxol-stabilized microtubules with freshly in vitro–polymerized F-actin. Unlike for endosome–microtubule binding, depletion of active Rab5 from endosomal membranes by treatment with either Rab-GDI or RN-tre did not decrease endosome–actin interactions (). Thus, a basal level of endosome–F-actin binding activity was independent of Rab5. However, the addition of HAP40-GST fusion protein stimulated binding (∼260%) over control levels, whereas GST alone was ineffective. Evidently, endosome binding to microtubules and F-actin is reciprocally regulated through HAP40 because concentrations (1 μM) inhibiting binding to microtubules () clearly stimulated binding to F-actin (). This effect was Rab5 dependent because when Rab-GDI or RN-tre were added together with HAP40, the stimulation was nearly abolished (). Again, immunoblotting confirmed that the amount of actin in the pellets was unaffected by any added protein and that all changes in pelletable fluorescence corresponded consistently to altered band intensities of early endosomal markers (EEA1 and TFr) but not to others (LAMP1 and GM130), indicating specific effects of Htt–HAP40 on endosome–actin binding (). Next, we investigated the role of HAP40 in Rab5 dynamics in an experimental system that is more relevant for HD using immortalized STHdhQ striatal cells (). These cells were established from embryonic normal or HD knock-in mice and either express normal (STHdhQ) or mutant Htt as a result of a CAG expansion inserted into the endogenous Htt gene (heterozygous STHdhQ and homozygous STHdhQ). Thus, they reflect the closest situation to HD patients as normal, and mutant Htt are expressed at endogenous levels. Remarkably, we again found endogenous HAP40 protein levels elevated in STHdhQ and STHdhQ compared with STHdhQ cells (), which is consistent with the data on HD fibroblasts () and brain tissue from HD patients (). Live cell imaging revealed EGFP-Rab5–positive organelles moving bidirectionally in neuronlike outgrowths as well as in the cell body of normal STHdhQ cells ( and Video 6, available at ). Conversely, STHdhQ and STHdhQ cells clearly displayed a drastic reduction of endosome dynamics ( and Videos 7 and 8), which is consistent with the observations on HD fibroblasts ( and Videos 3 and 4). Collectively, defects of Rab5 dynamics caused by pathogenic excess of HAP40 apparently occur in peripheral tissues such as fibroblasts as well as in neuronal systems and, therefore, are of potential relevance for HD. To gain some insights into possible alterations of endocytic transport caused by increased levels of HAP40, we tested the uptake of transferrin in HeLa cells overexpressing HAP40 and in fibroblasts from healthy and HD patients. shows that the uptake of transferrin was reduced by 30% in HeLa cells overexpressing HAP40 compared with mock-transfected or control cells. Consistent with this result, fibroblasts from HD patients displaying higher levels of endogenous HAP40 () displayed a similarly reduced uptake of transferrin compared with normal fibroblasts (). As for the block of endosome motility, the inhibitory effect on transferrin uptake was rescued by the depletion of HAP40 by RNAi. These data suggest that the inhibitory effects on endosome motility caused by up-regulation of HAP40 also result in defects in cargo transport through the endocytic pathway. The key finding of this study is that the complex between HAP40 and Htt is a direct downstream effector of Rab5 that regulates the dynamics of early endosomes through a switch from microtubules to F-actin. These findings provide important new insights into how the motility of early endosomes is regulated under physiological and pathological conditions. Htt has been implicated in clathrin-mediated endocytosis, regulation of the actin cytoskeleton, and microtubule-dependent transport along the endocytic pathway via interactions with its numerous binding partners (). Such multiplicity of roles implies that the activities of Htt in endocytic membrane trafficking need to be spatially and temporally coordinated. In showing for the first time that a complex between Htt and one of its binding partners, HAP40, can be recruited onto endosomes by interacting directly with Rab5, our data provide novel insights into the mechanisms governing the targeting of Htt to early endosomes and its regulatory activity on cytoskeleton-dependent dynamics. Under normal physiological conditions, early endosomes undergo frequent short-range movements on actin but also long-range bidirectional movements along microtubules (; ; ). Plus end movement of early endosomes along microtubules is propelled by KIF16B (). Overexpression of HAP40, which is rate limiting for the recruitment of Htt on the membrane, caused the detachment of early endosomes from microtubules and their preferential association with actin filaments, thus limiting both their velocity and range of movements. Therefore, the Htt–HAP40–Rab5 complex is a key regulator of the switch from one type of filaments to another. Our data are consistent with previous studies documenting alterations in microtubule-dependent motility in HD model systems (; ; ). However, these studies have exclusively implicated alterations between mutant Htt and HIP/HAP effectors. For example, wild-type Htt has been shown to interact with the dynactin subunit p150 via HAP1 and mutant Htt to disrupt the dynein motor complex in axonal transport (). In this study, we have uncovered a different mechanism based on the up-regulation of an Htt adaptor, HAP40. Compelling evidence in support of this mechanism was provided by the rescue of the motility block upon depletion of HAP40 by RNAi both in HD fibroblasts and in striatal cells. Because the COOH-terminal part of Htt is responsible for the underlying interactions with HAP40 and Rab5, this endosomal recruitment affects normal as well as mutant Htt. Consistently, wild-type (overexpression of HAP40 in HeLa cells), heterozygous (STHdhQ), and homozygous (STHdhQ) HD cells display very similar phenotypes. In this way, functional competition with the Rab5-dependent endosomal kinesin KIF16B and disruption of the dynein–dynactin complex () could account for the compromised bidirectional microtubule-based motility. Therefore, our data suggest that multiple defects may contribute to the block of vesicular transport in HD cells, which implicates various HIP/HAPs interacting with different regions of Htt. Previous findings have shown that Htt is also linked to Rab8 through FIP-2 () and that this interaction regulates cell polarization and morphogenesis. Whereas the mechanism underlying the latter process is unclear, in light of our observation, we propose that Htt is a multifunctional protein that may be recruited by different Rab GTPases through different adaptors at different intracellular locations to regulate organelle dynamics. The primary pathological cause of HD is attributed to the abnormal activity of Htt and its interacting partners in the nucleus (). The majority of HAP40 is also nuclear under normal conditions (; ), although the functional significance of this localization is unclear at present. Our data indicate that HAP40 fulfills a function in the cytoplasm. It is interesting in this respect to note that HAP40 adds to the increasing list of proteins implicated in a dual role in endocytic trafficking and nuclear signaling (). The molecular mechanism underlying the up-regulation of HAP40 unexpectedly observed in cells and brain tissue from HD patients remains to be determined. The most likely explanation is that it arises as a consequence of alterations in gene expression caused by mutated Htt in the nucleus. Messenger RNA microarray studies have revealed many transcriptional abnormalities in HD (; ; ), although no changes for HAP40 or any other Htt-interacting partners have been reported so far. Up-regulation of HAP40 at the protein level might thus serve as a new diagnostic indicator for the disease. However, it is premature to judge to what extent this abnormality and the consequent alterations in membrane dynamics contribute to the pathogenetic mechanism of HD as compared with nuclear activities of Htt and HAP40. Although endosome motility is not perturbed in HD cells to the extent of compromising cell viability, we nevertheless uncovered defects in endocytic transport, specifically a decrease in transferrin uptake. These observations may be of particular functional relevance in neurons given the importance of long-range transport (; ; ). Therefore, it is possible that alterations in endosome motility and transport may further compromise the pathological state induced by mutated Htt on the survival and function of neurons or some selected populations of neurons where endosomal transport is particularly rate limiting in vivo. Addressing this hypothesis requires a more thorough investigation of HAP40 and Htt function in the transport of different types of cargo along the endocytic/recycling as well as degradative pathway in neuronal cells rather than fibroblasts. A deeper analysis of the role of HAP40 on the molecular level will hopefully improve our understanding of both HD pathogenesis and the mechanisms coordinating organelle–cytoskeleton interactions during membrane trafficking. GST-Rab5 affinity chromatography with bovine brain cytosol or in vitro–translated proteins was performed as described previously (). HAP40 was cloned from rat brain cDNA by PCR, Htt fragments were cloned by PCR on full-length Htt cDNA in pBlueScript (a gift from M. Sherman, Boston University School of Medicine, Boston, MA), and all PCR products were cloned into pcDNA3.1 (Invitrogen) and sequenced to confirm their identity. To estimate the stoichiometry of the bound complex, GST-Rab5 was incubated in the presence of a 10-fold excess of the COOH-terminal fragment of Htt over HAP40. Band intensities obtained under these conditions were quantified in ImageJ software (National Institutes of Health; Htt4/HAP40 = ∼6:1) and corrected for the number of [S]methionine residues incorporated (Htt4/HAP40 = 22:4). The resulting ratio was ∼1:1. Assuming a 1:1 ratio of Rab protein to effector, the stoichiometry of the Rab5/HAP40/Htt interaction was estimated at ∼1:1:1. Monoclonal mouse antibodies against Htt were obtained from Chemicon and Y. Trottier (Institut de Génétique et de Biologie Moléculaire et Cellulaire, Université Louis Pasteur, Illkirch, France), and polyclonal rabbit antibody against HAP40 was obtained from Chemicon. Microtubules were revealed with mouse monoclonal anti–β-tubulin antibody (BD Biosciences), and F-actin was revealed with AlexaFluor568-conjugated phalloidin (Invitrogen). Secondary anti–mouse and rabbit IgG for immunofluorescence microscopy were conjugated to AlexaFluor568 and Cy5 (Invitrogen). Rabbit polyclonal anti-EEA1 antibody was described previously (), mouse monoclonal antibody against the cytoplasmic tail of Tfr was purchased from Invitrogen, mouse monoclonal anti-LAMP1 antibody was obtained from Becton Dickinson, and mouse monoclonal anti-GM130 antibody was obtained from Abcam. Rab-GDI and RN-tre were prepared as described previously (; ). GST and HAP40-GST fusion proteins were affinity purified from lysates according to standard procedures. HeLa and human primary fibroblast cells were grown according to standard procedures. Immortalized STHdhQ striatal cell lines from control and HD knock-in mice were cultured and differentiated as described previously (). For transient expression studies, cells were transfected using LipofectAMINE 2000 (Invitrogen) and used 12 h after transfection for immunoblot analysis or intracellular localization studies according to standard protocols. Fixed cells were analyzed using a 100×/NA 1.40 plan-Apochromat oil immersion lens (Carl Zeiss MicroImaging, Inc.) on a microscope (Axiovert S100TV; Carl Zeiss MicroImaging, Inc.). Illumination was performed with a 100-W mercury lamp with filter sets for GFP, AlexaFluor568, DAPI, and Cy5 fluorescence. Images were acquired at 1,300 × 1,000 pixels with a digital camera (SP 1.4.4; Diagnostic Instruments) controlled by the MetaVue 6.1 software package (Universal Imaging Corp.). Raw images from various color channels were assembled and colorized in Adobe Photoshop 7.0. Brightness and contrast were adjusted for visual clarity. To validate immunofluorescence microscopy studies using AlexaFluor568- and Cy5-conjugated secondary antibodies, HAP40 was overexpressed in HeLa cells and cells stained with a mixture of anti-HAP40 and -Htt antisera. Single addition of secondary antibody against mouse IgG (for Htt) conjugated with AlexaFluor568 resulted in no detectable bleed-through into the Cy5 channel at the exposure time and filter settings used for HAP40 detection with Cy5. Moreover, signals for HAP40 that were obtained with Cy5 looked very similar in the presence or absence of AlexaFluor568. Finally, when these dyes were swapped on the secondary antibodies, very similar distribution patterns for HAP40 and Htt were obtained. Collectively, the colocalization between HAP40 and Htt revealed in and is not caused by bleed-through. The percentage of colocalization between immunofluorescence signals was determined as follows: raw signals from discrete membrane structures were manually counted and inspected for overlapping with signals in other channels to calculate the percentage of colocalization. Only signals above value 70 on the eight-bit tonal scale were considered specific and used for the analysis. Typically, 10 cells ( = 10) were counted to calculate the mean colocalization and SD. For overlapping studies of EGFP-labeled early endosomes with cytoskeletal filaments, these organelles were considered colocalized to either microtubules or F-actin if at least three early endosomes in a row were clearly aligned to a filament. In areas where microtubules and F-actin converged (e.g., close to the cell edge), early endosomes could often not be assigned to either type of filament. The sum of early endosomes aligned to microtubules, F-actin, and unassignable early endosomes (100%) was used to calculate percentages of colocalization. Internalized biotinylated transferrin was detected with ruthenium-labeled antitransferrin antibodies and subsequently quantified by ECL analysis as described previously (). The primary human fibroblast cell lines GM00023, GM00024, GM00037, GM00038, and GM00041 (Apparently Healthy cell collection) and GM04281, GM04723, GM03621, GM04869, and GM04847 (Huntington Disease cell collection) were obtained from Coriell Cell Repositories. Cell lines were established from both genders from healthy, unrelated individuals by the age of 3–31 yr or from clinically affected onset HD patients by the age of 19–32 yr. Immortalized striatal cell lines established from mouse embryos (STHdhQ) were gifts from M. MacDonald (Richard B. Simches Research Center, Boston, MA; ). Human postmortem brain samples (CAP; Globus Pallidus) from healthy and HD grades 3 and 4 donors were provided by the Harvard Brain Tissue Resource Center. Gel-separated Coomassie-stained proteins were excised from the gel slab and in-gel digested with trypsin as described previously (). Tryptic peptides were sequenced by nanoelectrospray tandem mass spectrometry on hybrid quadrupole time-of-flight mass spectrometers (Q-TOF I; Micromass, Ltd. and QSTAR Pulsar I; MDS Sciex) as described previously (). Database searching was performed by Mascot software (Matrix Science, Ltd.). The in vitro motility assay was essentially performed as described previously () with the following modifications (). KHMG (110 mM KCl, 50 mM Hepes-KOH, pH 7.4, 2 mM MgCl2, and 10% glycerol) was used as an assay buffer. For preparation of the antifade solution, BRB80 buffer was substituted with KHMG, and the solution was further supplemented with 10% serum. The energy mix consisted of 75 mM creatine phosphate, 10 mM ATP, 10 mM GTP, and 20 mM MgCl2 in BRB80. For the assay, taxol-stabilized microtubules were perfused in a microscopy chamber and allowed to bind to the coverslip. Next, 10 μl of 10% nonspecific rabbit serum in antifade solution was perfused in the chamber followed by 10 μl of the assay mixture (2 μl of 5 mg/ml fluorescently labeled early endosome, 1 μl of energy mix, 6 μl of the antifade solution, and 1 μl of saturated hemoglobin solution in KHMG) and was incubated for 5 min at RT. At least three videos per sample were recorded (60 frames with 2-s intervals) and analyzed as described previously (). Cells were grown on glass coverslips, transfected with EGFP-Rab5 alone or together with siRNA against HAP40, and transferred to custom-built aluminum microscope slide chambers () just before observation. Cells were analyzed in CO-independent medium/10% FBS (Invitrogen) at 37°C using a 100×/NA 1.40 plan-Apochromat oil immersion lens (Carl Zeiss MicroImaging, Inc.) on a microscope (IX70; Olympus) placed in a temperature-controlled chamber. Illumination was performed with a 100-W Xenon lamp fitted with a monochromator to excite GFP fluorescence. Images were acquired at 512 × 512 pixels with a camera (Cascade 512B; Roper Scientific) in stream acquisition mode over 2 min in 300-ms intervals using the MetaVue 6.1 software package. Image stacks were either converted to QuickTime videos or motility events highlighted by generating Z projections of the entire stack using the ImageJ software package. Binding of early endosomes to microtubules was performed as described previously () but with some modifications. 15 μg of prepared early endosomes pulsed with rhodamine–transferrin (Sigma-Aldrich) were incubated at RT with 5 μl of energy mix (see previous section), 1 μl of nonspecific rabbit serum, and 1 μM of recombinant candidate factors (HAP40, Rab-GDI, or RN-tre) in KHMG buffer supplemented with 1 mg/ml HeLa cell cytosol. After 20 min, 16 μg of taxol-stabilized microtubules were added to a total volume of 50 μl. After 10 min more at RT, the reaction mixture was laid over 100 μl of a 35% (wt/vol) sucrose cushion. After sedimentation at 100,000 for 20 min at 22°C in a rotor (TLA 100; Beckman Coulter), the supernatant was removed, and the resulting pellet was washed twice in cytosol-free KHMG buffer. The pellet was lysed in 150 μl KHMG supplemented with 1% sodium deoxycholate, and the released rhodamine–transferrin was quantified in a fluorimeter. Excitation was at 550 nm, and emission was at 582 nm. A calibration curve revealed a linear correlation between emission and early endosome concentration. The equation of the linear regression analysis was y = 4,548.4x − 2,653.1 with R = 0.9879, where y is the emission in arbitrary units and x is the concentration of early endosomes in protein mass units. For binding of early endosomes to F-actin, globular-actin (Cytoskeleton, Inc.) was polymerized in KHMG buffer at RT for 1 h and centrifuged for 30 min at 100,000 . The resulting pellet was washed with KHMG buffer and resuspended at 10 mg/ml. The assay was then performed with 10 μg of F-actin as described for microtubules. All videos show EGFP-Rab5 dynamics, and images were acquired as described in In vivo motility of early endosomes. Video 1 shows a wild-type HeLa cell transfected with plasmid for EGFP-Rab5 expression only. Video 2 shows a HeLa cell cotransfected with plasmids for EGFP-Rab5 and HAP40. Video 3 shows a primary human fibroblast from a healthy person transfected with plasmid for EFGP-Rab5 expression only, whereas Video 4 shows the fibroblast from a HD patient. Video 5 shows a primary human fibroblast from a HD patient cotransfected with plasmid for EGFP-Rab5 transfection and siRNA duplexes against HAP40. Video 6 shows a striatal STHdhQ cell transfected with plasmid for EGFP-Rab5 expression only, Video 7 shows the same for an STHdhQ cell, and Video 8 shows this for an STHdhQ cell. Online supplemental material is available at .

Cell migration is crucial for many biological processes, including embryonic development, wound healing, and immune surveillance. Migration is a complex and highly coordinated process that requires a cell to polarize, extend protrusions in the direction of movement, form adhesions at the leading edge, translocate the cell body, and, finally, detach from the substratum at the trailing edge (; ). Directed cell migration is usually initiated in response to extracellular cues such as chemoattractants, growth factors, and the extracellular matrix. The establishment and maintenance of polarity during directed migration are mediated by feedback regulations involving integrins, phosphoinositides, cytoplasmic adaptor proteins, and Rho family guanosine triphosphatases (GTPases; ). The Rho family GTPase Cdc42 plays a crucial role in determining cell polarity during directed migration. Cdc42 is activated at the leading edge of polarized cells () and inhibition of Cdc42 activity or expression impairs directed migration assayed either in the cell culture system (; ) or in vivo (). Studies on slow moving cells such as astrocytes and fibroblasts indicate that Cdc42 controls polarized migration through two mechanisms. First, Cdc42 restricts the formation of protrusions at the front, which is mediated by a spatially specific activation of Rac at the leading edge, thereby promoting a polarized actin polymerization activity toward the direction of migration (). Second, Cdc42 is required for the reorientation of the microtubule-organizing center (MTOC) and Golgi to face the direction of migration (; ), which may contribute to polarity establishment by facilitating microtubule growth to the lamella and directed vesicle transport to the leading edge to maintain forward protrusions (). The effect of Cdc42 on MTOC and Golgi positioning is mediated through the Par6–Par3–aPKC complex (), which inactivates glycogen synthase kinase-3β to promote the capture of microtubule plus ends at the leading edge via adenomatosis polyposis coli (). Another mediator of Cdc42-induced MTOC polarization is IQGAP1, which forms a complex with two microtubule plus end–binding proteins, CLIP-170 () and adenomatosis polyposis coli (). Together, these findings indicate that Cdc42 coordinately regulates both actin and microtubule cytoskeletons via distinct pathways, thereby establishing the polarized morphology. It has been implicated that integrins act upstream of Cdc42 during directed migration. Although the function of integrin in migration is best known to be involved in the formation of cell adhesions, emerging evidence has revealed its role in cell polarization. For instance, integrin α5β1 mediates fibronectin-dependent cell polarization and protrusion through Rho family GTPases (). Integrin engagement is also essential for Cdc42 activation and polarity establishment during wound-healing migration (). Recently, α4 integrin was found to regulate cell polarity by recruiting the paxillin–GIT1 complex, in which GIT1 functions as an Arf-GAP to decrease Arf6 activity, thereby leading to Rac inhibition. As this complex is spatially restricted to the sides and the rear of the cell, Rac activation is limited to the leading edge, thus facilitating directed migration (). Despite these findings, it remains unclear whether proteins that affect integrin activity could influence cell polarity during migration. Death-associated protein kinase (DAPK) is a calmodulin-regulated and cytoskeleton-associated serine/threonine kinase (). Several lines of evidence indicate that DAPK plays an important role in tumor suppression. First, the expression of DAPK is frequently lost in various human cancer cell lines and tumor tissues, and this loss of expression correlates strongly with the recurrence and/or metastasis incidence of several human cancers (for reviews see ; ). Second, the antitumorigenic effect of DAPK was directly demonstrated in a mouse model system in which DAPK expression plays a causative role in suppressing the ability of Lewis lung carcinoma to form metastases in mice (). Third, DAPK is capable of suppressing c-myc– and E2F-induced oncogenic transformation by activating a p53-mediated apoptotic pathway (). The tumor-suppressive function of DAPK has been attributed to its effect on promoting apoptosis. DAPK is a well known proapoptotic protein and participates in a wide variety of apoptotic paradigms (; , ; ; Raveh et al., 2001; ; ; ). Recently, we found that DAPK exerts an apoptotic effect by inside-out inactivation of integrin β1, thereby suppressing the matrix survival signal and activating a p53-dependent apoptosis pathway (). Accordingly, the proapoptotic activity of DAPK is largely dependent on the existence of functional p53 protein, and several p53-deficient cell lines either escape from cell death () or undergo autophagic death in response to DAPK overexpression (). However, given the broad involvement of DAPK in tumor suppression and the frequent loss or mutation of p53 in various tumors, we postulate that DAPK elicits the second, apoptosis-unrelated mechanism to suppress tumor progression. Notably, the effects of DAPK on integrin inactivation () and actin cytoskeleton reorganization (; ) raise the possibility of DAPK involvement in motility regulation. We demonstrated that DAPK interferes with directional persistence during random migration and with cell polarization during directed migration and that these effects of DAPK are mainly mediated by its suppression of the integrin–Cdc42 signaling axis. Even in tumor cells that are resistant to DAPK-induced apoptosis, DAPK can elicit this motility-inhibitory effect and functions as a determining factor in tumor cell invasion. Together, our study identifies a novel role of DAPK in regulating cell polarity during migration, which may contribute in part to the tumor-suppressive function of DAPK. To characterize the function of DAPK in cell motility, we first examined whether DAPK affects the migration of free-moving fibroblasts. NIH3T3 cells were infected with retrovirus carrying control vector, DAPK, DAPK42A (a dominant-negative mutant), or DAPKΔCaM (a constitutively active mutant; ), together with a puromycin-resistant gene. After selection with puromycin, pools of cells that overexpressed various DAPK proteins were generated (). Because prolonged culture of cells overexpressing DAPK or DAPKΔCaM led to apoptosis (), populations of cells were used immediately after selection to test for their migration using time-lapse microscopy. We first observed that the four populations of cells exhibited different morphologies. Cells expressing DAPK or DAPKΔCaM often displayed many protrusions all around the cells, and these protrusions extended and retracted frequently during cell migration ( and Video 1, available at ). Cells carrying control vector or DAPK42A, however, often showed a long and polarized morphology () and, consequently, less protrusions () and smaller spreading areas () than the other two populations. When analyzing the trajectory of each individual cell during a 120-min migration period by tracing its centroid from the time-lapse movie, we found that cells expressing DAPK or DAPKΔCaM displayed much shorter net translocation than the control cells, whereas cells expressing DAPK42A showed longer paths (, right). Furthermore, the DAPK- or DAPKΔCaM-expressing cells made directional changes much more frequently than cells carrying control vector, whereas the DAPK42A-expressing cells migrated on a straighter path. These effects on random migration could be more clearly visualized by reproducing sample cell movement paths on window plots (). showed the rates of cell migration, the distances of net translocation during the 120-min time period, and the directional persistence of cells. These analyses revealed that each type of cell did not display a significant difference in their migration rate and that the differences in their net translocation were mainly attributed to their directional persistence properties during migration. Together, our data indicate that DAPK inhibits random migration by reducing the directional persistence. The reduced directional persistence suggests that DAPK renders cells unable to maintain a stable direction during targeted migration. Therefore, we assayed these NIH3T3 derivatives for wound-healing migration. Similar to what was observed with free-moving cells, overexpression of DAPK or DAPK ΔCaM caused a marked delay in wound closure, whereas DAPK42A accelerated wound closure (). Close examination of cells at the wound edge revealed that control cells displayed a polarized phenotype, with cell protrusions perpendicular to the wound and microtubules elongating to the tip of protrusions (; and Video 2, available at ). This polarization of wound-edge cells was also made evident by the reorientation of their MTOC and Golgi in the direction of wound (). Although the polarized phenotype was preserved in cells expressing DAPK42A, cells expressing DAPK or DAPKΔCaM manifested a drastic disruption of polarity. Although these cells still formed protrusions at the wound edge, their directions were more random, and a substantial proportion of cells displayed multiple short protrusions or protrusions parallel to the wound ( and Videos 3–5, available at ). Furthermore, their microtubules showed scattered distribution, and a significantly lower number of cells in the front row exhibited polarized MTOC and Golgi (). Time course analysis indicated that the DAPK-expressing cells exhibited a delayed kinetics of Golgi reorientation compared with control cells, whereas DAPKΔCaM-expressing cells did not show reorientation, even 7 h after wounding. In contrast, cells expressing DAPK42A modestly accelerated wound-induced Golgi reorientation (). These observations collectively identify an inhibitory role of DAPK in cell polarization during directed migration. Cdc42 activity is required for establishing polarity during directed migration (). To study the molecular mechanism through which DAPK inhibits the polarization of cells at the wound edge, we assessed the activity of Cdc42 during wound-healing migration. In cells carrying vector or DAPK42A, Cdc42 activity was drastically induced 15 min after wounding and then declined gradually. However, this wound-induced Cdc42 activation was delayed and modestly attenuated in DAPK-expressing cells and completely abrogated in DAPKΔCaM-expressing cells (). Therefore, we investigated whether the suppression of Cdc42 activation accounts for the polarity defects induced by DAPK. Indeed, expression of a constitutively active Cdc42 mutant rescued the Golgi polarization defect seen in DAPKΔCaM-expressing cells (). Thus, DAPK inhibits cell polarity by suppressing Cdc42 activation. The migration-induced Cdc42 activation and cell polarization can be abolished by a cyclic RGD peptide, demonstrating a role for integrin in the establishment of cell polarity (). Accordingly, we observed that integrin β1 was preferentially activated at the leading edge of migrating cells, as judged by comparing the distribution of total integrin β1 with that of activated β1 (). Intriguingly, DAPK was also localized at the leading edge of cells undergoing wound-healing migration (). Consistent with this DAPK distribution and a previously reported effect of DAPK on integrin inactivation (), overexpression of DAPK or DAPKΔCaM attenuated integrin downstream signaling at the leading edge, as monitored by autophosphorylation of FAK (). FAK autophosphorylation at the leading edge was rescued by pretreatment of DAPK- or DAPKΔCaM-expressing cells with the integrin β1–activating antibody TS2/16 (). Thus, these findings not only suggest the existence of polarized integrin β1 activity and signaling in cells at the wound edge, but also reveal an inhibitory role of DAPK in these polarized events. Next, we explored the mechanism by which DAPK inactivates integrin. Integrin activation can be regulated by several Ras family GTPases, such as R-Ras and Rap1, and, at least in certain circumstances, R-Ras acts upstream of Rap1 for integrin activation (). However, expression of DAPK or its mutants did not affect Rap1 activity, as monitored by a pull-down analysis (Fig. S1, available at ). Another important intracellular molecule for integrin activation is talin, which was recently found to be a common downstream effector of many signaling pathways that control integrin activation (). To investigate the role of talin in DAPK-triggered integrin inactivation, we used RNA interference to knockdown talin expression. The introduction of talin small interfering RNAs (siRNAs) to 293T cells led to a great reduction of talin expression, but not of cell surface expression of integrin β1, which was monitored by the β1-specific antibody AIIB2 (). As expected, this down-regulation of talin in 293T cells decreased the activity of integrin β1, as monitored by the monoclonal antibody B44, which specifically recognizes the active integrin β1 (). A similar reduction of B44 binding was observed in cells overexpressing DAPK or DAPKΔCaM, which is consistent with our previous study (). Conversely, DAPK42A led to an increase of B44 binding. However, in talin knockdown cells, the expression of DAPK or DAPKΔCaM could not further decrease the binding of B44, whereas DAPK42A could no longer increase B44 binding. Furthermore, activation of integrin by treatment of cells with Mn completely rescued the effect of talin knockdown on DAPK42A-expressing cells (). Collectively, these results indicate that DAPK acts upstream of talin to inhibit its integrin activation function. We next tested whether DAPK elicits an inhibitory effect on the association of talin-H with integrin β1 tail. To avoid the detection of association induced by any post-adhesion event, we prepared lysate from suspension cells. We found that in suspension cells, the expression of DAPK or DAPKΔCaM reduced the capability of coexpressed talin-H to interact with bacterially expressed GST-β1 tail. Conversely, DAPK42A enhanced this association (). To investigate whether DAPK acts directly on talin-H to prevent its interaction with integrin β1 tail, we tested the possibility of talin-H as a DAPK substrate. In vitro kinase assay revealed that DAPK did not phosphorylate talin-H, whereas myosin light chain was heavily phosphorylated by DAPK (). Thus, DAPK may use an indirect mechanism to interfere with the recruitment of talin-H to integrin β1 tail. Nevertheless, as this DAPK-induced blockage of talin-H–integrin interaction could be detected in the absence of cell adhesion, this effect likely contributes to a mechanism by which DAPK inactivates integrin. We reasoned that the DAPK-induced inhibition of integrin may represent a mechanism by which DAPK suppresses wound-healing and random migrations. Indeed, enforced activation of integrin β1 by the mouse β1–activating antibody 9EG7 rescued wound-induced Cdc42 activation in NIH3T3 cells expressing DAPK or DAPKΔCaM (). Consequently, 9EG7 restored Golgi polarization () and rescued wound-healing migration defects () in these two populations of cells. Furthermore, in random migration, 9EG7 completely reversed the DAPK-induced defects in net cell translocation and directional persistence, and the four populations of cells moved in a similar manner under 9EG7 treatment (; and Fig. S2, available at ). Notably, integrin activation converted the morphology of DAPK- and DAPKΔCaM-expressing cells to a typical polarized morphology, i.e., with a single, stable protrusion at the leading edge and a long trailing tail (Video 6). As a control, a nonactivating integrin β1 antibody MB1.2 could not rescue the random migration defects induced by DAPK or DAPKΔCaM (). Thus, we conclude that inactivation of β1 integrin accounts for a major mechanism through which DAPK suppresses cell motility and migratory polarity. Having demonstrated that DAPK effects motility inhibition in fibroblasts, we next investigated whether this function of DAPK could be recaptured in carcinoma cells that are resistant to the apoptotic effect of DAPK. As the proapoptotic activity of DAPK is largely dependent on the existence of functional p53 protein (; ), we used two human tumor cell lines that contain mutant p53, i.e., the epidermoid carcinoma cell A431 () and the breast carcinoma cell MDA-MB-231 (). Again, DAPK or its mutants were introduced to the two cell lines by retrovirus-mediated gene transfer. Overexpression of DAPK or DAPKΔCaM in either cell line could not trigger apoptosis (Fig. S3 A, available at ). In fact, even maintained in culture for 26 d, these cells were still alive and expressed exogenous DAPK proteins at similar levels as those cultured for only 3 d (). When the four populations of A431 cells were assayed for wound-healing migration, we observed a substantial delay of wound closure in cells expressing DAPK or DAPKΔCaM, which was reversed by pretreatment of cells with TS2/16. In contrast, DAPK42A accelerated wound closure ( and Fig. S3 B). The migratory capabilities of these stable lines were also tested with haptotactic migration assays. Similar to what was found in wound-healing migration, expression of DAPK or DAPKΔCaM greatly suppressed collagen-induced haptotactic migration of A431 cells, whereas DAPK42A promoted haptotaxis. Again, the effect of DAPK or DAPKΔCaM on haptotactic migration was completely overridden by TS2/16 (). Furthermore, we observed a similar inhibitory effect of DAPK or DAPKΔCaM on fibronectin-induced haptotaxis in MDA-MB-231 cells, which, again, could be rescued by TS2/16 (). These results indicate that DAPK is capable of inhibiting the migration of tumor cells that are resistant to the apoptotic effect of DAPK and suggest the existence of a second role for DAPK in tumor suppression. Because the migratory capacity of tumor cells is one of the determining factors in their abilities to invade a matrix barrier, we tested the effect of DAPK on the invasiveness of the two tumor cells using Matrigel. As shown in , the invasive potential of both cell lines was significantly reduced by overexpression of DAPK or DAPKΔCaM, and this effect was again reversed upon the addition of TS2/16. Furthermore, DAPK42A potentiated invasion of both A431 and MDA-MB-231 cells. In conclusion, our study identifies a novel function of DAPK in suppressing tumor cell migration and invasion, which is mediated by integrin inactivation but independent of apoptosis induction. To further characterize the role of DAPK in tumor invasion, we used a pair of tumor cell lines. CL1-0 is a human lung adenocarcinoma cell line that was previously used to generate the highly invasive subline CL1-5 by progressive selections through the invasion chamber (). We found that CL1-5 expressed a significantly lower level of DAPK than CL1-0 (), implying that reduction of DAPK expression conferred a selective advantage to develop a highly invasive phenotype during tumor progression. To determine whether DAPK expression level plays a role in the distinct invasiveness of the two cell lines, we overexpressed DAPK or DAPKΔCaM in CL1-5 cells (). As expected, such overexpression of DAPK or DAPKΔCaM led to a reduction in cell invasion, which was reversed by TS2/16 (). Next, we used DAPK siRNA to down-regulate the expression of endogenous DAPK in CL1-0 cells. DAPK level was significantly reduced in CL1-0 cells treated with DAPK-specific siRNA1 or siRNA2, but not with a control siRNA (). Compared with untransfected cells or cells receiving control siRNA, cells expressing either of the DAPK siRNAs displayed a marked increase in the invasion capability (). Together, these results not only demonstrate a role of endogenous DAPK in regulating tumor cell invasion, but also indicate that DAPK functions as one of the determining factors in the invasiveness of CL-1 cells. We next investigated whether down-regulation of endogenous DAPK could affect cell polarization during migration. Using a wound-scratch assay, we found that the parental CL1-0 cells or CL1-0 cells carrying control siRNA did not reorient their Golgi at 5 h after wounding (), and polarization began to emerge 7 h after wounding (not depicted). However, in cells receiving either of the DAPK-specific siRNAs, a significant population (∼60%) sensed the wound and displayed Golgi reorientation at 5 h after wounding (). This finding identifies a physiological role of DAPK in the inhibition of cell polarization during directed migration. It is well documented that DAPK elicits proapoptotic activity. However, in this study, we uncover a novel biological function of DAPK in regulating cell migration. We found that DAPK inhibits random migration by suppressing directional persistence. During directed migration, DAPK functions as a potent inhibitor of cell polarization, as made evident by its perturbation of the formation of static protrusion at the leading edge, and polarization of MTOC and Golgi. Notably, these effects of DAPK are independent of its proapoptotic activity, as they are detected in cells without any sign of apoptosis. Indeed, even in cells that are resistant to DAPK-induced apoptosis, the migration/invasion inhibitory function of DAPK is still evident. We reason that this function of DAPK may contribute in part to its tumor-suppressive activity. In support of this notion, we found that the expression level of DAPK plays a determining role in the invasive capability of the CL1 cells, as increased expression of DAPK in CL1-5 cells inhibits tumor invasiveness, whereas decreased expression in CL1-0 cells promotes it. Notably, these alterations of DAPK expression in CL1-0 and CL1-5 cells did not significantly affect cell proliferation and survival (unpublished data), and therefore the role of DAPK in suppression of invasion is best explained by its motility effect. The motility-inhibitory function of DAPK is expected to play two significant roles in suppressing tumor development and/or progression. First, as cell migration is central to tumor invasion through basement membrane and formation of metastasis in vivo, DAPK likely suppresses these processes through its effect on cell migration, thereby functioning in the late stage of tumor progression. Accordingly, DAPK has been demonstrated to suppress metastasis of Lewis lung carcinomas in a mouse model system (). Furthermore, hypermethylation of the DAPK gene and/or loss or reduction of DAPK expression have been shown to associate with metastasis and/or the advanced stages of many human cancer types (for reviews see ; ). Second, the motility-inhibitory effect of DAPK would be particularly important to suppress tumors that have escaped from DAPK-induced apoptosis, which can be achieved by mutations, leading to the inactivation of genes involved in the apoptosis core machinery, such as p53. Indeed, the apoptosis-promoting activity of DAPK is lost in many p53-defective tumor cell lines (; ), including the A431, MDA-MB-231, and CL1-5 cells used in this study. Given that inactivating mutations of p53 are frequently found in a wide range of human tumors, the migration/invasion inhibitory function of DAPK might be of greater clinical consequences than its proapoptotic role. Regardless of the relative contributions of the two mechanisms to tumor suppression, the proapoptotic and antimigratory activities of DAPK could act in a cooperative and complementary fashion to prevent malignancy during different stages of tumor development. This double safeguard mechanism would allow DAPK to act as an efficient and versatile tumor suppressor, which might explain the frequent observation of DAPK promoter hypermethylation in a wide variety of cancer types at different stages (for reviews see ; ). Importantly, both proapoptotic and antimigratory functions of DAPK are mainly medicated by integrin inactivation (). Thus, integrin appears to play a central role in the tumor-suppressive effect of DAPK. In this study, we not only demonstrate a critical role of integrin β1 in the motility-inhibitory function of DAPK but also explore the mechanism for DAPK-induced integrin inactivation. We first demonstrated that the integrin regulation function of DAPK or its mutants is completely abolished by depletion of talin, whereas activation of integrin by Mn compensates for the effect of talin down-regulation. These results indicate that DAPK acts upstream of talin in regulating integrin activation. We further provide evidence showing that DAPK elicits an inhibitory effect on talin-H association with integrin β1 tail, a well known mechanism for inside-out activation of integrin (, ; ). As this inhibitory effect of DAPK can be detected in the absence of cell adhesion, it most likely represents a mechanism, rather than a consequence, of the integrin-modulating function of DAPK. How DAPK interferes with talin-H binding to integrin is currently unclear. However, the inability of DAPK to phosphorylate talin-H in vitro implies the involvement of an indirect mechanism. Additional studies are needed to elucidate whether DAPK signaling leads to an inhibitory posttranslational modification on talin-H or an activation of a competitor for talin-H–integrin interaction. Furthermore, as talin affects the activation status of multiple integrin β subunits (), it would be interesting to determine the effect of DAPK on the activities of other integrin β subunits. Although the effects of DAPK on random and wound-healing migrations are described as the failure to maintain directional persistence and to establish cell polarity, respectively, these two events are likely to be mechanistically related. First, both events are rescued by enforced activation of integrin β1. Second, in both migration modes, DAPK- or DAPKΔCaM-expressing cells fail to establish a polarized morphology and static protrusion toward the direction of migration. In random migration, this leads to a decrease of directional persistence. In wound-healing migration, even though cells are unable to move toward different directions because of a lack of free space, this failure to maintain a polarized morphology would be expected to hamper efficient and persistent movement toward the wound, and therefore delays wound closure. Given that DAPK does not significantly alter migration speed in random migration, the effect of DAPK on wound-healing migration is predicted to be mainly resulted from the decrease of directional persistence. The inability of DAPK to significantly reduce migration speed seems to be inconsistent with its integrin inactivation function, as cell-substratum adhesion strength plays a determining role in the migration speed of fibroblasts (). However, the actin cytoskeleton–localized DAPK might regulate other cytoskeleton-residing molecules to compensate for the effect of integrin inactivation on migration speed. Notably, DAPK was shown to phosphorylate myosin light chain at Ser19 in vivo, thereby increasing actomyosin contractility (). Future studies will be aimed at evaluating the contribution of this phosphorylation event to the migration properties of DAPK-expressing cells. The effect of DAPK on migratory polarity and persistence is mainly attributed to its suppression of the integrin–Cdc42 polarity pathway. Although Cdc42 is well documented as a key polarity regulator (), less attention has been focused on the role of integrin in controlling cell polarization. Inhibition of integrin function by a cyclic RGD peptide has been shown to block wound-induced Cdc42 activation and cell polarization in cultured astrocytes (). This finding highlights a role of integrin in polarity establishment during directed migration, but the type of integrin involved in this process has not been defined. In this study, we identify integrin β1 as a crucial factor of polarity establishment during both random and directed migrations and as a mediator of DAPK-induced regulation of Cdc42 activity. Notably, a role of integrin β1 in polarized migration has been demonstrated in β1-null keratinocytes, which exhibit defects in the rapid reorientation of actin cytoskeletons toward the polarized movement (). In addition, α4β1 integrin has been shown to regulate cell polarity during directed migration through its recruitment of the paxillin–GIT1 complex in the side and rear of cells to inhibit Rac activation (). In contrast to these studies, a recent study demonstrated that adhesions mediated by αvβ3 are more static than those mediated by α β1. Consequently, αvβ3, rather than α5β1, permits the formation of a single broad lamellipod at the leading edge and persistent migration (). Even though our study does not address the role of β3 in migration, we did observe the formation of static adhesion and persistent migration by activating β1 integrin in cells overexpressing DAPK or DAPKΔCaM. One explanation for this discrepancy in β1-mediated migration behavior may be the difference in migration speed of cells used in the two studies. As the GE11 epithelial cell used in a previous study (Danen et al.) migrates at a two- to threefold slower rate than the NIH3T3 fibroblast used in this study, more static cell-matrix adhesion (e.g., β3-mediated adhesion) may be needed to provide a sufficient time for the slow-migrating GE11 cells to complete the subsequent events in a migratory cycle. Alternatively, the downstream signaling events and migratory effect of β1 may be influenced by the existence of other types of integrin. Despite certain inconsistencies observed in different cell systems, these findings collectively underscore the importance of integrin-mediated cell-matrix adhesion in the establishment and/or maintenance of cell polarity during migration via its regulation of Rho family GTPases. pRK5- and pBabepuro-based expression vectors for DAPK, DAPKΔCaM, and DAPK42A were described previously (; ). Human Cdc42V12 cDNA was obtained from T.-S. Jou (National Taiwan University, Taipei, Taiwan). The plasmid pGEX-β1 tail (residues 757–798 of the human integrin β1A cDNA) was obtained from H.-C. Chen (National Chung Hsing University, Taichung, Taiwan). The plasmid pGEX2T–talin-H (residues 1–435) was obtained from T.-L. Shen (National Taiwan University, Taipei, Taiwan), and the talin-H fragment was subcloned to pEGFPC1 vector. The antibody to DAPK was described previously (), and the human integrin β1-activating antibodies TS2/16 and AIIB2 were purified as described previously (). The mouse β1 integrin–activating antibody 9EG7 was purchased from BD Biosciences, and the antibody to phosphorylated FAK (p-FAK) was purchased from Biosource International. The integrin β1 antibodies P4C10, B44, and MB1.2 were obtained from CHEMICON International, Inc. Antibodies to Cdc42 and Rap1 were purchased from Santa Cruz Biotechnology, Inc. Antibodies to β-COP and α-tubulin were obtained from Sigma-Aldrich. Antibody to γ-tubulin was obtained from S.-C. Lee (National Taiwan University, Taipei, Taiwan). 293T, NIH3T3, A431, and MDA-MB-231 cells were maintained as described previously (; ). CL1-0 and CL1-5 cells (provided by P.-C. Yang, National Taiwan University, Taipei, Taiwan) were cultured in RPMI 1640 medium containing 10% FCS. Transfections of 293T, CL1-0, and CL1-5 cells were performed using the calcium phosphate method. Generation of recombinant retroviruses and infection of cells were performed following previously described procedures (). For random migration analysis, cells were seeded on 6-well plates at a density of 10 cells/well in the regular culture medium and placed in a temperature- and CO-controlled chamber of a microscope (Axiovert 100TV; Carl Zeiss MicroImaging, Inc.) equipped with 20, 40, and 100× objective lenses. Time-lapse recording started 30 h after plating. Images were collected at 3-min intervals over 120 min with a cooled charge-coupled device (CCD) video camera (CollSNAP fx; Roper Scientific) operated by Metamorph image analysis software (Molecular Devices). Motility parameters, including migration path, distance, rate, and directional persistence, were obtained from time-lapse movies. To track the migration path of individual cells, cells were manually traced for each frame and the geographical centers were recorded using Metamorph. The migration paths were expressed as graphs using the Excel program (Microsoft). The rates of cell migration were calculated as a ratio of the total length of migration paths and the duration of migration. Migration distances were determined as the net translocation during a 120-min period. Directional persistence was calculated as a ratio of the direct distance during a 120-min period and the total length of the migration path. For wound-healing migration assay, cells were seeded on 6-well plates at a density of 6 × 10 (for NIH3T3 cells) or 5 × 10 cells/well (for A431 cells) in culture medium. 30 h after seeding, the confluent monolayer of culture was scratched with a fine pipette tip, and migration was visualized by time-lapse imaging. The rate of wound closure was calculated by a ratio of the average distance between the two wound edges and the total duration of migration. For haptotactic migration assay, the underside of Transwell polycarbonate membrane (8-μm pore size; Costar) was coated with 15 ng/ml collagen I (for A431 cells) or 50 ng/ml fibronectin (for MDA-MB-231 cells). For the invasion assay, the membrane was coated with 0.54 μg/μl Matrigel (BD Biosciences). 10 cells resuspended in culture medium were plated onto the upper chamber, and the same medium was added to the lower chamber. Cells were incubated at 37°C for various time points. At the end point of incubation, cells on the upper side of the membrane were removed by wiping it with a cotton swab, and cells that had migrated onto the lower membrane surface were fixed by 4% formaldehyde, stained with Hoechst 33342, and counted. When transiently transfected cells were used, GFP-positive cells that had migrated to the underside of the membrane were counted under a fluorescence microscope (Axioskop; Carl Zeiss Microimaging, Inc.). Each assay was set up in triplicate, and ten random fields were analyzed for each membrane. To observe the effects of integrin β1 activation, 5 μg/ml 9EG7, 5 μg/ml MB1.2, or 2 μg/ml TS2/16 was added to culture medium before the recoding (for random and wound-healing migration) or to the upper chamber of Transwell plates. Cells were fixed with 4% formaldehyde in PBS for 20 min. For β-COP staining, cells were permeabilized with PBS containing 0.01% Triton X-100 and 0.05% SDS for 5 min, and blocked with blocking solution (0.1% saponin and 0.2% BSA in PBS) for 30 min. Cells were incubated with β-COP antibody diluted in blocking solution for 2 h, and then incubated with Texas red–conjugated secondary antibody and Hoechst 33342 for 1 h. For integrin β1 (total or active), p-FAK, α-tubulin, or γ-tubulin staining, cells were permeabilized with extraction buffer containing 50 mM NaCl, 300 mM sucrose, 10 mM Pipes, pH 6.8, 3 mM MgCl, and 0.5% Triton X-100 for 5 min, blocked with PBS supplemented with 10% goat serum, 1% BSA, and 50 mM NHCl for 1 h, and then incubated with various primary antibodies diluted in PBS containing 0.2% BSA and 5% goat serum for 1 h. Cells were then incubated with FITC- or Texas red–conjugated secondary antibody for 1 h. Cells were washed, mounted, and examined with either an epifluorescence microscope (Axioskop; Carl Zeiss MicroImaging, Inc.) with a 40 or 100× oil objective lens, or a confocal microscope (510 Meta; Carl Zeiss MicroImaging, Inc.) with a 100× oil objective lens. Fluorescent images were captured with a cooled CCD camera operated by the Image-Pro Plus Software (Media Cybernetics) or the Laser Scanning Microscope LSM510 Software. The images were arranged and labeled using Photoshop software (Adobe). Cdc42 () and Rap1 activity () were determined essentially as previously described. For Cdc42 activity, a monolayer of cells was repeatedly scratched (∼30 times in a 60-mm dish) and lysed in a buffer containing 25 mM Hepes, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10 mM MgCl, 1 mM EDTA, 10% glycerol, 1 mM PMSF, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM sodium vanadate, 2 mM sodium pyrophosphate, and 10 mM NaF. For Rap1 activity, cells were lysed in lysis buffer containing 50 mM Tris, pH 7.4, 200 mM NaCl, 1% NP-40, 2.5 mM MgCl, 10% glycerol, 1 mM PMSF, 10 μg/ml aprotinin, and 10 μg/ml leupeptin. Cell lysates were incubated with GST-PAK-CRIB (for Cdc42) or GST-RalGDS-RBD (for Rap1; provided by J.L. Bos, University Medical Centre, Utrecht, Netherlands) coupled to glutathione–Sepharose 4B beads for 1 h at 4°C. The beads were washed, resuspended in SDS-PAGE sample buffer, and analyzed by Western blot with anti-Cdc42 antibody or anti-Rap1 antibody. DAPK siRNAs, talin siRNAs (SMARTpool), and control siRNA were purchased from Dharmacon RNA Technologies. The sequences of DAPK siRNAs are as follows: DAPK siRNA1, sense: 5′-CAAGAAACGUUAGCAAAUGUU-3′ and antisense: 5′-CAUUUGCUAACGUUUCUUGUU-3′; DAPK siRNA2, sense: 5′-GGUCAAGGAUCCAAAGAAGUU-3′ and antisense: 5′-CUUCUUUGGAUCCUUGACCUU-3′. DAPK siRNAs were transfected to CL1-0 cells, together with a CD2 marker plasmid or a GFP plasmid. The transfected cells were isolated 48 h after transfection by CELLection CD2 kit (Dynal) or visualized by GFP fluorescence. Talin or control siRNA, together with various DAPK expression vectors, were cotransfected into 293T cells, and talin expression was determined at 72 h after transfection. GST pull-down analysis was performed essentially as previously described (). In brief, cells transiently cotransfected with GFP–talin-H and various DAPK constructs were cultured in suspension for 4 h and then lysed in lysis buffer containing 20 mM Tris, pH 8.0, 137 mM NaCl, 1% NP-40, 10% glycerol, 1 mM PMSF, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM sodium vanadate, 2 mM sodium pyrophosphate, and 10 mM NaF. Lysate containing 2 mg of total proteins was incubated with 250 μg GST or GST–integrin β1 tail conjugated to glutathione beads, and bound protein was analyzed by Western blot. 293T cells were cotransfected with various DAPK constructs and talin siRNA or control siRNA at a ratio of 1:1 and then treated with or without 5 mM MnCl for 30 min. The total and active integrin β1 at cell surface were monitored by AIIB2 and B44 antibodies, respectively, followed by flow cytometry analysis as described previously (). To test protein phosphorylation by DAPK, Flag-tagged DAPK was immunoprecipitated from cell lysates and then incubated in 40 μl of kinase buffer containing 50 mM Hepes, pH 7.5, 8 mM MgCl, 2 mM MnCl, 0.1 mg/ml BSA, 1 μM bovine calmodulin (Sigma-Aldrich), 0.5 mM CaCl, 50 μM ATP, and 10 μCi γ-[P]ATP, in the presence 2 μg GST-MLC or GST–talin-H at 25°C for 15 min. Protein phosphorylation was detected by autoradiography. The effect of DAPK on Rap1 activity, random and wound-healing migrations of 9EG7-treated cells, as well as the cell death and wound-healing migration assays for A431 stable transfectants are provided as supplemental figures. In addition, supplemental videos of migratory experiments are provided. Online supplemental material is available at .

Arf was first discovered, purified, and functionally defined as the protein cofactor required for cholera toxin–catalyzed ADP ribosylation of the stimulatory regulatory subunit (Gs) of adenylyl cyclase (; ) and, shortly thereafter, was shown to be a GTP-binding protein (). Use of the acronym Arf is currently preferred to ADP ribosylation factor, as only Arf1–6 shares the cofactor activity for cholera toxin and because ADP ribosylation does not appear to be involved in any aspect of the normal cellular actions of any member of the family. The use of all capital letters (e.g., ARF1) refers specifically to the human gene or protein, whereas when only the first letter is capitalized (e.g., Arf1), it may refer to the protein from more than one species, an activity, or a group of proteins. Since their discovery, they have been found to be ubiquitous regulators of membrane traffic and phospholipid metabolism in eukaryotic cells (for reviews and discussion of Arf actions see ; ; Kahn, 2004). Arfs are soluble proteins that translocate onto membranes in concert with their activation, or GTP binding. The biological actions of Arfs are thought to occur on membranes and to result from their specific interactions with a large number of effectors that include coat complexes (COPI, AP-1, and AP-3), adaptor proteins (GGA1-3 and MINT1-3/X11α-γ/APBA1-3), lipid-modifying enzymes (PLD1, phosphatidylinositol-kinase, and phosphatidylinositol-kinase), and others. Arf proteins are activated by guanosine diphosphate (GDP) to GTP exchange, which is stimulated by the Sec7 domain of Arf guanine nucleotide exchange factors, and their activity is terminated upon the hydrolysis of GTP, which is stimulated by interaction with an Arf GTPase-activating protein. Cloning and sequencing of the first Arf family member () led directly to the realization that Arfs are closely related to both the Ras and heterotrimeric G protein α subunit families of GTPases, and all are thought to have arisen from a common ancestor. The very high degree of conservation of Arf sequences in eukaryotes (74% between human and yeast) was also noted early on and has allowed the ready identification of orthologues in every examined eukaryote, including , which lack Ras and G protein α subunits (). Cloning by low stringency hybridization and chance led to the identification of additional members of the Arf family in a wide array of eukaryotic species. The number of mammalian Arfs grew to six by 1992 () and were named in their order of discovery (; ; ; ). The first confusion in the nomenclature was that the current human ARF4 was originally published with the name ARF2 (). In fact, humans appear to have lost the ARF2 orthologue, which is present in other mammals (including rats, mice, and cows). The combination of protein sequence comparisons and intron/exon boundaries of Arf genes led to further classification of the six mammalian Arfs into classes: class I (ARF1–3 are >96% identical), class II (ARF4 and ARF5 are 90% identical to each other and 80% identical to the other Arfs), and class III (ARF6 is 64–69% identical to the other Arfs). Phylogenetic analyses support the conclusion that the three classes of Arf diverged early, as flies and worms have single representatives of each of the three classes, and the number of genes/proteins in class I and II were later expanded in vertebrates. deletion in , and directly activate PLD were given the name Arf. Thus, with the chance cloning of an essential gene in that encoded a protein closely related to the Arfs (50–60% identity) but lacking in these activities, it was named (). When orthologues were found in several other species, the name was changed to Arf-like 1 () in those species (; ; ). Note that although the name Arf still denotes a protein with one or more specific functions or activities, the term Arl does not. The term Arl indicates only that the protein is structurally related to Arfs. Thus, the Arls are not a coherent group either functionally or phylogenetically. PCR amplification with degenerate oligonucleotide primers (; ) revealed the existence of a large number of mammalian cDNAs encoding closely related proteins. The next to be cloned and sequenced were (), (), (), and (). Each of the encoded proteins has a glycine at position 2, the site of -myristoylation in all Arf proteins. Note that although ARL2 and ARL3 have the NH-terminal glycine, they appear not to be substrates for -myristoyltransferases. Around this time, a protein with similar percent identities to the Arf and Arls was found, but it lacked the NH-terminal glycine, was membrane associated, and displayed distinctive nucleotide handling properties (). Thus, it was given the name Arf-related protein 1 () to distinguish it from the Arls and Arfs. We realize today that this was unfortunate, as several of the more recently identified Arls also have functions and biochemical properties that are quite divergent from Arfs. SAR1 was among the earliest members of the Arf family sequenced, and it came out of genetic screens in the yeast as a suppressor of sec12(ts) (). Its name is derived from its identification as a secretion-associated and Ras-related protein. Cloning of the mammalian orthologues revealed the presence of two closely related (90% identity) proteins/genes (). With <30% identity to Arfs or Arls, the SAR proteins are only slightly closer in sequence to Arfs than to other families of GTPases, but they also share considerable functional relatedness to Arfs in that they act through the recruitment of coat proteins or complexes to initiate vesicle budding. SARs lack the other aforementioned Arf activities. An interesting variation is found in ARD1/tripartite motif 23 (TRIM23), a 64-kD protein that possesses a ∼20-kD domain at its COOH terminus with 60% identity to Arfs (). Originally named based on the presence of the Arf domain, ARD1 is also a member of the TRIM family, from which it obtained its current name, TRIM23. A large extension is also seen in ARL13B, a protein of 428 residues that contains an Arl domain at its NH terminus (; ). Although the NH-terminal portion of TRIM23 may possess GTPase-activating protein activity toward its own Arf domain () and E3 ubiquitin ligase activity (), the COOH-terminal portion of ARL13B has no defined domains or functions to date. As the discussion above suggests, there are no shared functions or activities that justify grouping Arf, Arl, and SAR proteins into a family with a common nomenclature. Similarities in protein sequences within the Arf family were first identified by alignment and phylogenetic analyses and were shown to provide distinct signatures that allowed differentiation from Ras, G protein α subunits, and other GTPases. These include an NH-terminal extension, a glycine acceptor for myristate at position 2, an aspartate at position 26 (in contrast to the glycine 12 of Ras that carries oncogenic potential), and other residues that are very highly conserved within the family. These early observations were put on more solid functional footing when they were found to map to unique elements in their three-dimensional structures, which allow for the GDP/GTP switch to be coupled with interaction signals opposite to the nucleotide-binding site (for review see ). The prominent feature of this unique nucleotide switch is a nonconventional GDP-bound form in which the two β strands that connect the nucleotide-sensitive switch 1 and 2 regions (also called the interswitch) are retracted in the protein core and must undergo a two-residue shift to reach the active conformation (). However, the interswitch cannot do so unless the NH-terminal helical extension, which caps the interswitch and locks it in the retracted conformation, has been displaced. In the case of ARF1, biochemical studies have established that this requires the interaction of the NH terminus with membranes, thus allowing the nucleotide-binding site to detect and respond to remote protein–membrane interactions (). Like Arf proteins, each Sar has an NH-terminal amphipathic helix that functions as a structural GDP/GTP switch to anchor the GTP-bound form to membranes of the endoplasmic reticulum (; ). Furthermore, membrane insertion of this NH-terminal helix was recently shown to initiate membrane bending at the early stages of COPII coat assembly and to be subsequently required for the completion of COPII vesicle fission (). Structural analysis of ARF1 and ARF6 GDP/GTP cycles and their comparison with those of small GTP-binding proteins whose interswitch does not toggle identified three structural determinants for this movement: a helical NH-terminal extension that fastens the retracted, GDP-bound interswitch; a shorter interswitch that can retract completely; and a sequence signature (wDvGGqXXXRxxW) that provides both flexibility for the movement (GG) and hydrogen bonds for stabilization of the active conformation (R/W). These characteristics are present in all Arf and most Arl sequences, which, therefore, are predicted to have the ability to undergo the interswitch toggle to detect interactions opposite to the nucleotide-binding site, whatever their nature, and propagate them to this site (). These structural criteria for unifying Arf and Arl proteins as a family have since been supported by various structures of GDP-bound Arf and Arl proteins (Table S1, available at ). It should be noted, however, that one subgroup, ARL4, has a long interswitch that may have lost the ability to toggle, whereas structures of NH-terminally truncated ARL8A and ARL8B bound to GDP have a GTP-like conformation. This suggests that truncation of the NH terminus is sufficient in this family to destabilize the retracted interswitch or that these proteins have lost their ability to undergo the interswitch toggle. Recent work on ARL3 suggests that proteins interacting with the NH terminus could also work as the displacing factor as an alternative to membranes (; ). contains information on proposed and previous names as well as other information on the human ARF family members. EST and genomic sequencing resulted in the identification of subsequent Arf-like proteins, and these proteins/genes were often misnamed or named multiple times by different research groups. Some of these names suggest relationships that are misleading, and some are called Arfs despite (presumably) lacking any Arf activities. In many cases, no functional data are yet available for the most recently identified Arf family members. One protein has been referred to by four different names, and some proteins/genes were named by curators of databases responding to specific requests in a manner that disagreed with common usage by researchers in the field. The confusion is magnified when species differences are considered (e.g., yeast Arl3 is the orthologue of ARFRP1). The need for a generally agreed upon nomenclature for the ARF family has become acute as a result of increasing confusion and interest in their study. It is not possible today to propose a completely consistent nomenclature, as there are simply too many studies with some of the earlier discovered proteins (e.g., ARFRP1 should be an ARL). The nomenclature developed and described in this article builds on previous efforts to describe phylogenetic relationships and bring consistency to nomenclature (; ; ). It is the result of many discussions between researchers in the field and with the HUGO Genome Nomenclature Committee (HGNC) and has been widely circulated to Arf family researchers. We describe the presence in the human proteome of 29 members of the Arf family and a system for naming newly identified proteins in human or other species. The use of letter suffixes is reserved for those groups of proteins within the family that share higher percent identities and are, therefore, likely to share some level of functional redundancy. One exception to this is the ARL13A and ARL13B proteins, which have been given a common number based upon phylogenetic evidence. The consensus nomenclature for the Arf family is shown in along with previous names and unique gene/protein identifying information. Note that in three cases (, , and ), the intron/exon boundary predictions in the database are thought to be incorrect (based upon comparisons with sequences in other species), resulting in differences in the predicted protein sequences. In these cases, we use our corrected sequences for comparisons and provide the predicted protein sequences of the human proteins (see supplemental material, available at ). In addition, there is one case () in which it appears that alternative splicing yields two different proteins, one of which is truncated and predicted to be unable to bind nucleotides, so both are provided in the supplemental protein sequence material. We also identify several gene sequences that have questionable EST/mRNA support and are likely pseudogenes derived from members of the Arf family. These genes, which are annotated by the HGNC, are therefore not included as Arf family members and are listed, along with their identifiers, in . It is expected that additional pseudogenes will be found and added to this list over time. We also note some uncertainty as to whether ARL5C in is a transcribed gene, as it may lack part of the consensus GTP-binding signature depending on which predicted protein sequence is used. Finally, we note that although the large majority of Arf family members appear to have very broad and perhaps ubiquitous tissue expression patterns, a few are far more restricted in their expression. Thus, it is expected that further additions and perhaps even deletions will be needed to keep the nomenclature of this family current and as consistent as possible. To ensure that new family members are assigned unique symbols, we strongly encourage authors to consult the HGNC before publishing any new names for members of this gene/protein family. This is a confidential service provided by the HGNC that will help prevent future confusion from arising. We also suggest that curators and researchers focusing on other organisms use the information provided in this article as much as possible to simplify and clarify the nomenclature across species. Other researchers supporting the use of this nomenclature include: Bruno Antonny, Bill Balch, Vytas Bankaitis, Gary Bokoch, Juan Bonifacino, Chris Burd, Jim Casanova, Tamara Caspary, Dany Cassel, Rick Cerione, Pierre Chardin, Philippe Chavrier, Shamshad Cockcroft, Peter Cullen, Ivan de Curtis, Maria Antonella De Matteis, Julie Donaldson, Cryslin D'Souza-Schorey, John Exton, Victor Faundez, Jim Goldenring, Jean Gruenberg, Alan Hall, Fuchu He, Wangjin Hong, Victor Hsu, Mary Hunzicker-Dunn, Trevor Jackson, Cathy Jackson, Hans Joost, Toshi Katada, Fang-jen Lee, Michel Leroux, Jennifer Lippincott-Schwartz, John Logsdon, Alberto Luini, Vivek Malhotra, Ed Manser, Tobias Meyer, Paul Melancon, Joel Moss, Aki Nakano, Kazu Nakayama, Tommy Nilsson, Susanne Pfeffer, Richard Premont, Paul Randazzo, Anne Ridley, Scotty Robinson, Anne Rosenwald, Craig Roy, Hisataka Sabe, Randy Schekman, Nava Segev, Val Sheffield, Phil Stahl, Elizabeth Sztul, Chris Turner, Anne Theibert, Martha Vaughan, Kanamarlapudi Venkateswarlu, Fred Wittinghofer, Keqiang Ye, and Marino Zerial.

Cytoplasmic dynein is responsible for the distribution and transport of diverse membranous organelles and is involved in chromosome segregation and mitotic spindle organization and orientation. Recent studies have also suggested a role for dynein in the removal of metaphase checkpoint proteins from the kinetochore (; ) and in directed cell migration (). The mechanisms by which dynein interacts with a diversity of cargoes and subcellular targeting sites is incompletely understood. The dynein intermediate, light intermediate, and light chains have each been implicated in cargo binding (for review see ), as has another multisubunit complex, dynactin (). Zeste white 10 (ZW10) is a kinetochore protein that participates in the mitotic checkpoint and also serves to link dynactin and dynein to mitotic kinetochores (for review see ). -null mutants exhibit chromosome missegregation and abnormal separation of sister chromatids in the presence of colchicine (), a result confirmed by the injection of HeLa cells with anti-ZW10 antibody (). -null mutants also lack dynein at their kinetochores (). ZW10 was found to interact with the dynactin subunit dynamitin (), which, when overexpressed, had been found to displace the rest of the dynactin complex, along with cytoplasmic dynein, from kinetochores (; ). ZW10 exists as a complex with (), , and the Rad50-interacting protein-1 (RINT-1; , ; ; ; ; ; ). Despite clear evidence of a mitotic function, ZW10 is expressed uniformly throughout the cell cycle (). Furthermore, its interacting partner dynamitin participates in general dynein function (; ). ZW10 has been reported to colocalize with the ER by immunocytochemistry and to participate in ER–Golgi trafficking through a SNARE-dependent mechanism (), although the specific role of ZW10 in this process was uncertain. This study was initiated to explore potential dynein-related activities of ZW10 during interphase. We report that ZW10 associates with specific interphase structures and regulates minus end–directed Golgi, endosome, and lysosome transport, which is consistent with a general role in dynein regulation and targeting. To examine the subcellular distribution of ZW10 in interphase cells, immunofluorescence microscopy was performed. A previously characterized antibody, “anti-HZW10,” recognizes a doublet at the ZW10 position (; ), only the lower band of which is enhanced in cells expressing recombinant ZW10. An additional antibody produced against a COOH-terminal human ZW10 peptide, “anti-Cter,” was specific for ZW10 (). Using either of the antibody punctates, pericentrosomal staining was generally observed in interphase cells () that was absent in preimmune controls and was dispersed by brefeldin A or nocodazole (not depicted). Specific colocalization of ZW10 could be observed with the Golgi markers β-coatomer protein (β-COP), γ-adaptin (), 58K protein, and -acetylglucosaminyltransferase (NAGT; not depicted), and additional ZW10-positive spots could be seen throughout the cell. Specific labeling of ER, endosomes, or lysosomes could not be discerned, as is the case with other dynein pathway components. Consistent with a membrane association, ZW10 appeared primarily in the insoluble fraction in HeLa () and COS-7 cell extracts (, bottom band), but could be solubilized using a nonionic detergent (, TX100, SN). ZW10 was also enriched with Golgi membranes in subcellular fractionation studies (, top) and was separated from soluble markers, including the top 90-kD band recognized by the anti-HZW10 antibody (; ) and α-tubulin. The Golgi-associated ZW10 could be sedimented and solubilized with Triton X-100 (, bottom), which is consistent with a membrane association. The detection of ZW10 in other membrane flotation fractions () supports an association with additional organelles as well. We used multiple approaches to inhibit ZW10 function. Overexpression of the COOH-terminal ZW10 fragment HZW10-4 produced a pronounced dominant-negative phenotype (Fig. S1 A,available at ), though no effect was observed with other fragments or full-length ZW10. Mitotic index was dramatically reduced (0.5 ± 0.1 vs. 2.6 ± 0.3% for controls), even in the presence of nocodazole (0.2 ± 0.05 vs. 5.2 ± 0.1% for controls; ; ), and there was a substantial increase in multinucleate cells (23.5 ± 4.2 vs. 4.3 ± 0.6% in controls), many of which had micronuclei and/or unequally sized micronuclei (48.4%; ). HZW10-4 over-expression resulted in a marked increase in cells with a dispersed Golgi apparatus (68.4 ± 3.9 vs. 8.1 ± 3.3% for nonexpressing cells, and 12.9 ± 2.8% for high level β-galactosidase overexpressers as controls; Fig. S1, B and C). Microinjection of COS-7 cells with affinity-purified anti-Cter ZW10 peptide antibody also caused Golgi dispersal (49 ± 3%; Fig. S2 A, top, available at ), compared with cells injected with peptide-blocked antibody (13.9 ± 4.6%; Fig. S2 A, bottom). Both HZW10-4–overexpressing and antibody-injected cells exhibited a substantial loss of centrosome-centered microtubule organization (Fig. S2 B, top), a result also produced by expression of dynamitin and other dynactin polypeptides (; ). To test whether Golgi dispersal is a direct or indirect effect of altered ZW10 function, we performed quadruple labeling using anti-tubulin, anti-58K, DAPI, and the injected anti-ZW10 antibody. For cells with clear Golgi dispersal, a normally centered microtubule organization was still detectable in 26.9 ± 1.8% of the cases (Fig. S2 B, bottom), indicating that the distribution of the Golgi is affected independently of microtubule defects in our experiments. We also used RNA interference (RNAi) to inhibit ZW10 expression. ZW10 expression was greatly reduced after 3 d of ZW10 small interfering RNA (siRNA) treatment, whereas no reduction was detected after using a scrambled siRNA control (). Dynein and dynactin levels were unaffected by reduced ZW10 expression (). As for HZW10-4 overexpressers, ZW10 RNAi reduced mitotic index (2.1 ± 0.2%) versus control siRNA-treated cells (5.5 ± 0.3%; ). Cultures with reduced ZW10 also displayed a dramatic increase in Golgi dispersal, as visualized using the NAGT-GFP marker (49.3 ± 4.2% for ZW10 siRNA vs. 17.3 ± 2.3% for control siRNA; ), which is similar to results obtained by in HeLa cells. However, as observed for anti-Cter–injected cells (see previous paragraph), centrosome-centered microtubule organization was altered in many of the COS7 cells treated with ZW10 siRNA (, middle; 80% of the cells showed dispersed Golgi), but Golgi dispersal could still be observed in cells displaying a normally centered microtubule network (, bottom). We also transfected cells with a cDNA encoding a ZW10 short hairpin RNA (shRNA) corresponding to a target sequence distinct from that of the siRNA. Recipient cells again showed clear Golgi dispersal, as well as microtubule disruption (not depicted). The disruption of the Golgi apparatus by multiple means strongly supports a role for ZW10 in controlling Golgi organization. A previous study attributed similar phenotypic effects to a SNARE-related mechanism (), despite ZW10's known role in mitotic dynein function (). To test directly for a role for ZW10 in interphase dynein function, we conducted live imaging of Golgi vesicles in cells subjected to ZW10 RNAi. To ensure proper scoring of minus end– versus plus end–directed movement, we coexpressed YFP-tubulin along with the RNAi. Only cells in which a clear radial microtubule organization persisted were included in this analysis (Videos 1 and 2, available at ). As in the fixed images, Golgi elements labeled with NAGT-GFP were dispersed by ZW10 RNAi. Analysis of vesicle movements revealed an ∼70% decrease in the number of minus end–directed movements, relative to results obtained using a scrambled control (). Plus end–directed movements were also reduced, but this effect was much smaller. This result is reminiscent of recent observations of the effects of dynamitin overexpression in frog melanophores, where it was attributed to a role for dynactin in the anchoring of kinesins, as well as dynein, to vesicular organelles (; ). A pronounced increase in the percentage of stationary NAGT-GFP vesicles was also observed (). To determine whether other minus end–directed membranous structures were also affected by ZW10 RNAi, we examined cells expressing YFP-tubulin and labeled with either the endosomal marker FITC-Tf or the lysosomal marker LysoTracker red. Vesicular elements labeled with each marker were dispersed as revealed by both immunocytochemistry and live imaging (). Analysis of vesicle motility (Videos 3–6, available at ) again revealed a clear decrease in minus end–directed movements that was comparable in magnitude to that observed for Golgi elements, a similar smaller decrease in plus end–directed movements, and a substantial increase in stationary particles (). To test for a role for ZW10 in anchoring dynein to membranous organelles, we stained HeLa cells subjected to ZW10 RNAi with an antibody specific for Golgi cytoplasmic dynein heavy chain that was reported for cells overexpressing the dynactin subunit dynamitin (). Dynein staining at the Golgi apparatus was also reduced by ZW10 RNAi (). Vector-based ZW10 RNAi (, bottom) reduced Golgi-associated dynein immunofluorescence almost completely. In contrast, the dominant-inhibitory HZW10-4 cDNA had little detectable effect on Golgi dynein heavy chain staining (not depicted). This result suggests that HZW10-4 produces its inhibitory effects by a means independent of dynein targeting. In support of this possibility, HZW10-4 failed to displace dynein and dynactin from kinetochores (unpublished data), despite a pronounced effect on mitotic progression (see Inhibition of ZW10 function). HZW10-4 appears to contain part of, but not the entire, interaction region for the dynactin subunit dynamitin (). Although we observed coimmunoprecipitation of overexpressed dynamitin with full-length ZW10, we have not detected an interaction between dynamitin and the HZW10-4 fragment in this assay or by sucrose density gradient centrifugation of HZW10-4–transfected cell lysates (McKenney, R., personal communication). The fragment could be detected at kinetochores (unpublished data). However, the persistence of dynein at this site and at the Golgi apparatus argues against a role for the corresponding ZW10 domain in dynein targeting. Instead, these data suggest additional roles for ZW10 or its interacting proteins in mitotic checkpoint and dynein motor regulation. This fragment seems to differ in its phenotypic effects from an NH-terminal fragment reported to interfere with ER–Golgi trafficking (). The latter was reported to interact with the t-SNARE syntaxin-18, suggesting a role for ZW10 in membrane budding and fusion. Therefore, ZW10 conceivably participates both in syntaxin-18–mediated membrane sorting and in dynein-mediated membrane transport, though some of the earlier data could be dynein related. A key finding of that study was a ZW10-associated defect in vesicular stomatitis virus glycoprotein trafficking, a classic indicator of ER–Golgi sorting, though this process was delayed rather than blocked. Because transport of the ER–Golgi intermediate compartment vesicles is under dynein control (; ), a delay in trafficking could conceivably result from a transport, rather than a trafficking, defect. During interphase, cytoplasmic dynein associates with the cortex of migrating cells and can be seen prominently at the leading cell edge (). In view of the expanded range of dynein-related functions in which our data now implicate ZW10, we tested further for its distribution in migrating cells (). ZW10 immunoreactivity was observed in a pattern strikingly similar to that of cytoplasmic dynein, which is associated with the leading cell edge in advance of most of the microtubules in the region. This study provides the first evidence for an association of ZW10 with Golgi elements in particular and for a general role in interphase cytoplasmic dynein function. Furthermore, it strongly suggests that ZW10 serves a role during interphase related to that at the kinetochore, i.e., at least in part as an anchor for dynein and dynactin. In light of the reported interaction between ZW10 and dynamitin (), we anticipate a comparable hierarchy of interactions at the surface of Golgi elements, as well as of endosomes and lysosomes (). An important issue is how this model relates to the role proposed for the spectrin cytoskeleton in dynein targeting (; ). As seen in , the two models are not mutually exclusive. A Golgi-associated spectrin isoform, βIII spectrin, was found to bind to the Arp1 subunit of dynactin (, ), whereas ZW10 binds to the dynamitin subunit (). Thus, it is possible that dynactin has two distinct interaction partners at the Golgi surface, which could function in concert. Based on this study, ZW10 appears to be necessary for normal binding. Its contribution relative to that of the spectrin-mediated mechanism remains to be fully assessed. How ZW10 itself binds to membranes is unknown. This interaction could, conceivably, involve syntaxin-18. Immunoprecipitates of this protein were found to contain ZW10; RINT-1, which is a protein first identified in the Rad50 DNA repair pathway; and p31, another SNARE-related protein (). It is now known that ZW10 and RINT-1 exist as a complex (; ) and that both proteins, along with p31, are released from syntaxin-18 by α−SNAP in the presence of ATP (). How these results, and the lack of a direct interaction between ZW10 or RINT-1 with syntaxin-18, may relate to dynein targeting remains to be explored. Rab6 () and () have also been implicated in Golgi dynein targeting. Whether ZW10 functions in concert with these proteins or independently is unclear. It is appealing to speculate, as a final consideration, that other features of the interaction of ZW10 with kinetochore pertain to its association with cellular membranes. ZW10-interacting protein-1, for example, has been identified as an upstream interactor for ZW10 at kinetochores (; ). Whether ZW10-interacting protein-1 itself or some other yet to be identified protein serve in this capacity during interphase remains an important question for further investigation. Full-length HZW10 () was inserted into the pcDNA3.1 vector (Invitrogen) using the BamHI and XhoI restriction sites. A myc tag was inserted at the 5′ end of each HZW10 fragment (Fig. S1 A) using the same restriction sites and plasmid vector. pCNG2, encoding a GFP-,β1, 2-acetylglucosaminyltransferase I (NAGT-GFP) fusion protein, was a gift from D. Shima and G. Warren (Imperial Cancer Research Fund, London, UK). β-Galactosidase overexpression was performed using pCMVβ vector (CLONTECH Laboratories, Inc). The polyclonal antibody anti-HZW10 was previously described elsewhere (). A polyclonal antibody termed anti-Cter was raised against the 23 COOH-terminal amino acids of HZW10 (Research Genetics). This peptide, coupled to a cyanogen bromide–activated Sepharose 4B column, was used to affinity purify anti-Cter, and the antibodies were eluted with a pH2–pH5 gradient of glycine/HCl (Research Genetics). Monoclonal anti-p150 and anti-GM130 antibodies were obtained from BD Biosciences. Polyclonal Golgi-specific anti-dynein heavy chain () was a gift from V. Allan (University of Manchester, Manchester, UK). Monoclonal antibodies directed against the Golgi markers 58K, γ-adaptin, β-COP, and α-tubulin (clone DM1A) were purchased from Sigma-Aldrich. Polyclonal anti–β-gal antibody was obtained from US Biological. COS-7 and HeLa cells were grown in DMEM with 10% fetal calf serum supplemented with 100 U/ml penicillin and 100 mg/ml streptomycin. NIH3T3 cells were grown in DMEM with 10% bovine calf serum, and wound-healing assays were performed as previously described (). For microinjection, the affinity-purified antibody was dialyzed overnight into microinjection buffer (50 mM potassium glutamate and 0.5 mM MgCl, pH 7.0) and concentrated to 6.8 mg/ml. Cells were fixed 6 h after injection. Microinjection was performed using a micromanipulator (model 5171; Eppendorf) coupled to a microscope (model DMIRBE; Leica). siRNA were prepared and introduced into COS7 cells according to the specifications of the manufacturer (Dharmacon RNA Technologies). The RNA sequences used were AAGGGUGAGGUGUGCAAUAUG, for ZW10 siRNA, and AUUGUAUGCGAUCGCAGACUU, for a scrambled negative control. We also designed a pRNAT-ZW10 cDNA (pRNAT-U6.1/Neo), which encodes an shRNA corresponding to a distinct ZW10 target sequence (CGGTGAATTTACAGACTTAAA), as well as GFP (GenScript Corp.). To test for anti-Cter ZW10 peptide antibody specificity (), cells were rinsed in ice-cold PBS (137 mM NaCl, 2.7 mM KCl, 8.1 mM NaHPO, and 1.5 mM KHPO, pH 7.4) and harvested in 150 mM NaCl, 10 mM Tris, pH 7.2, 1% deoxycholate, 1% Triton X-100, and 0.1% SDS containing a protease inhibitor mixture (2 μg/ml each of aprotinin and leupeptin, 1 mM EGTA, and 1 mM 4-(2-aminoethyl)benzenesulphonyl fluoride). After incubation on ice for 20 min, the cell lysate was spun for 10 min at 13,000 , and the supernatant was subjected to SDS-PAGE and Western blotting. For immunoprecipitation, cells were extracted in 150 mM NaCl, 50 mM Tris, pH 8.0, 1% NP-40, and protease inhibitors. Membrane flotation in a discontinuous 2/1.6/1.4/1.2/0.8 M sucrose gradient was conducted as described (). 1-ml fractions were collected from the bottom of the gradient and analyzed by Western blotting. Each fraction was also diluted in 150 mM NaCl and 50 mM Tris, pH 8.0. Membranes were sedimented at 15,000 for 15 min and solubilized by adding 1% Triton X-100. Direct membrane sedimentation tests on Dounce-homogenized total cell extract were also performed at 15, 000 for 15 min. For Golgi staining, cells were rinsed in PBS and fixed for 6 min at −20°C in methanol. For costaining of Golgi and microtubules after ZW10 antibody injection, cells were rinsed in PHEM buffer (120 mM Pipes, 50 mM Hepes, 20 mM EGTA, and 4 mM magnesium acetate, pH 6.9) and fixed for 6 min at −20°C in methanol. Sequential antibody incubations were used because both GM130 and α-tubulin antibodies were monoclonal. To determine the mitotic index, cells were rinsed in PHEM buffer and fixed in 4% paraformaldehyde, 0.05% glutaraldehyde, and 0.05% Triton X-100 for 12 min. Cells were then permeabilized for 25 min in 0.5% Triton X-100 in PBS, and incubated twice for 10 min in PBS containing 10 mg/ml NaBH. 0.05% saponin was added for ZW10 Golgi staining. Images were obtained using a DMIRBE microscope equipped with a camera (ORCA 100; Hamamatsu) and Metamorph software (Universal Imaging Corp.). Z-series stacks were acquired with 0.3-μm steps; the out of focus signal was reduced using Metamorph 2D deconvolution, and the final images were obtained by total projection of the image stacks. Confocal microscopy was performed with a microscope (Diaphot 200; Nikon) coupled to a system (MRC1000; Bio-Rad Laboratories) equipped with a Kr/Ar laser (MRC1000; Bio-Rad Laboratories). For live-cell imaging of Golgi, COS7 cells were cotransfected with NAGT-GFP and tubulin-YFP constructs, along with ZW10 siRNA or control oligonucleotide treatment, for 3 d. For live imaging of endosomes, cells cotransfected with tubulin-GFP and ZW10 RNAi or control oligonucleotide were treated with TRITC-transferrin for 1 h before obtaining the time-lapse images. Cells were treated with LysoTracker red for 2 h to visualize lysosomes in live cells transfected with tubulin-GFP and ZW10 siRNA or control oligonucleotide. Time-lapse images were acquired at 37°C every 1 s for a period of 2 min (1.5 min for endosomes) using a DMIRBE microscope equipped with an incubation chamber for temperature and CO control. Figs. S1 and S2 show the effects of dominant-negative ZW10 cDNA expression and anti-ZW10 antibody injection on Golgi organization in COS7 cells. Videos show the effect of ZW10 RNAi on Golgi (Videos 1 and 2), endosome (Videos 3 and 4), and lysosome (Videos 5 and 6) motility in control oligonucleotide versus siRNA-transfected cells. Online supplemental material is available at .

Cilia are remarkable microtubule (MT)-based “nanomachines” that project from the surface of most eukaryotic cells and play diverse roles in motility and sensory reception (; ; ; ). Sensory cilia on the dendritic endings of chemosensory neurons in act as sensory antennae that control the animal's chemotactic movements in response to chemical gradients in the environment (). These cilia have distinct neuron-specific morphologies and sense different chemical stimuli (; ). For example, although the amphid channel cilia have a typical unbranched cylindrical morphology and detect hydrophilic molecules and high osmolarity, the adjacent amphid wing (AWC) cilia have a fan-like morphology and detect volatile odorants (e.g., benzaldehyde; ; ; ). How these distinct structures and sensory modalities arise is unclear, although differences in intraflagellar transport (IFT) motor function in different cilia could play a key role. IFT motors of the kinesin 2 family () assemble and maintain cilia by transporting ciliary precursors, bound to multimeric protein complexes called IFT particles, from the cell body, along the axoneme, for incorporation into ciliary structures (; ; ). In some organisms, a single heterotrimeric kinesin 2 motor transports the IFT particles to the distal tip of the cilium, and a second motor, IFT dynein, recycles kinesin 2, IFT particles, and turnover products back to the cell body (, ; , ; ; ; ; ). In amphid channel cilia, however, two kinesin 2 motors, heterotrimeric kinesin II and homodimeric OSM-3 kinesin, cooperate in a semiredundant fashion to build two distinct domains of the axoneme: the middle segment consisting of nine MT doublets surrounding a few central singlets and the distal segment consisting of MT singlets, which are extensions of the middle segment A tubules (; ; ; ; ). (We use the standard name, kinesin 2, to describe the family and the terms kinesin II and OSM-3 kinesin to discriminate the two IFT kinesin holoenzymes.) The two motors move the same IFT particles along the middle segment in a process that is required to build the middle segment, and then OSM-3 kinesin alone moves them the rest of the way along the distal segment in a process that is required to elongate the distal singlets. Therefore, in , the basic “canonical” mechanism of IFT, in which a single anterograde motor delivers precursors to the tip of the assembling cilium, is complicated by the use of two kinesin 2 motors, but what selective advantage might this confer? We address the possibility that the differential regulation of the two motors in different cilia may contribute to neuron-specific sensory ciliary diversity. We reconstructed the three-dimensional (3D) structure of the cilia from serial EM sections of wild type (WT), kinesin II (, ), OSM-3 kinesin (), and kinesin II/OSM-3 ( and ) single- and double-mutant animals. To ensure optimal preservation of cell ultrastructure, worms were high-pressure frozen and freeze substituted (). The results confirm our model in which OSM-3 and kinesin II function redundantly to build the middle segments of amphid channel cilia, whereas OSM-3 alone builds the distal singlets (), but suggest that, in the adjacent fan-like AWC cilia, the kinesin 2 motors have a different functional relationship that appears to contribute to the morphological and sensory differences between the AWC and channel cilia. shows the general organization of the head of an adult as revealed by transmission EM (TEM) of high pressure–frozen, freeze-substituted WT animals and 3D reconstructions from serial sections. We observed excellent preservation of the animal's fine structure, including obvious MT doublets and singlets of the cilia (). In the middle segments, nine doublet MTs surround a variable number of singlets, characteristic of immotile cilia (). We consistently observed that the matrix of some cilia was more intensely stained than others, possibly reflecting a larger number of electron-dense IFT particles (; ). In the reconstructions, the hexaradiate symmetry of the head is obvious (), and in each of the two amphids, 10 cilia form a bundle that extends to the tip of the channel (), whereas the fan-shaped AWC cilium does not enter the channel but instead penetrates the adjacent sheath and socket cells. We reconstructed amphid channel cilia from WT animals and from a variety of IFT motor mutants () and observed that in WT animals and single ( and ) and double () kinesin II mutants, full-length cilia, ∼7.5 μm in length, were present. In mutants lacking IFT dynein function (), the cilia are truncated and bulbous, whereas cilia in mutants lacking OSM-3 kinesin function () were only 4.5 μm long because of the specific loss of distal segments (). In OSM-3 kinesin/kinesin II double mutants ( and ), however, the entire ciliary axoneme was missing (). In previous work, we observed no fluorescent IFT particles (GFP::OSM-6) extending from the transition zone of these double mutants (). Our TEM analysis rigorously eliminates the possibility that this reflected a failure of transport of IFT particles along residual cilia in these animals (assembled, for example, by diffusion of ciliary precursors to their site of assembly) but instead is due to a complete loss of ciliary axonemes, confirming that either OSM-3 kinesin or kinesin II is required to build the middle segments of these axonemes, whereas OSM-3 alone is required to extend the distal singlets (; ). Although kinesin II and OSM-3 kinesin act in a partially redundant fashion to build the channel ciliary axonemes, they act in a completely redundant fashion to build the adjacent AWC cilia (– ). We used fluorescence microscopy () to examine the distribution of GFP-tagged translational reporters of IFT particle proteins (OSM-5 or OSM-6::GFP) in amphid channel cilia and a red fluorescent protein (RFP)–tagged transcriptional reporter of the AWC guanylyl cyclase ODR-1::RFP () to examine the morphology of AWC cilia. The channel and wing cilia in WT animals () were indistinguishable from those in single mutants in the kinesin II motor (). However, although the apparent lengths of the channel cilia in mutants were reduced from 7.5 to 5 μm because of the loss of their distal segments (), the morphology of wing cilia was indistinguishable from WT (compare ). No differences in the morphology of any single mutant and WT AWC cilia could be detected by visual inspection of scores of animals. Our efforts to quantify the dimensions of these AWC cilia yielded equivocal results. The largest measured difference was between WTs (length, 5.04 ± 1.08 μm, and width, 8.9 ± 2.16 μm; = 32) and (length, 4.56 ± 0.98 μm, and width, 8.62 ± 1.92 μm; = 33), but this small difference probably results from problems encountered in tracking fluorescence from edge to edge along these asymmetric cilia. However, clear and obvious differences were noted in kinesin II/OSM-3 kinesin double mutants (), where no fluorescence was observed distal to the transition zone, consistent with a complete loss of cilia on the AWC neuron (). We confirmed that kinesin II and OSM-3 kinesin are indeed expressed in AWC neurons because translational fusions of KAP-1::GFP () and OSM-3::GFP () colocalized with the AWC marker, ODR-1::RFP (). The results suggest that either kinesin II or OSM-3 kinesin can build the AWC cilia and that either motor, but not both, is dispensable for this. No differences in IFT dynein function were seen, based on the morphology of channel and AWC cilia (). We used serial-section TEM to confirm that the morphology defects seen in AWC cilia by light microscopy of the IFT motor mutants reflect changes in ciliary structure (). By examining multiple serial sections, we found that AWC cilia were intact in WT and in single kinesin II or OSM-3 mutants but were completely absent in double mutants (). Careful examination of WT cilia revealed mixtures of doublet and singlet MTs that lie side by side rather than being organized into a cylinder (, insets). Significantly, moving toward the distal tip of the cilium, both the singlet and doublet MTs terminated at the same point, ∼0.5 μm below the overlying ciliary membrane. Because serial sections were taken every 100 nm, the maximum length of any distal singlets must be less than this distance (i.e., 1/25 the length of the distal singlets in channel cilia), suggesting that AWC cilia lack distal singlets. Therefore, the two motors act redundantly to build the doublet and singlet MTs of AWC cilia just like in the middle segments of channel cilia, whereas OSM-3 extends the distal singlets in channel cilia but not in AWC cilia, where its function is unknown. Why channel cilia require distal singlets whereas AWC cilia do not is unclear, and in general the function of distal segments is unknown. The distal segments of cilia and flagellar axonemes in other systems are of variable length and prominence; e.g., they are 200 μm long in frog olfactory cilia (), whereas in they are lacking, but they do elongate transiently during mating (). There are speculations that distal singlets are dynamic structures that are required for some aspect of sensory signaling because the specific loss of the singlets (e.g., in mutants) leads to loss of sensory signaling. However, both the channel and wing cilia have sensory functions in , so this may not be universally true. Based on studies of amphid channel cilia (), we had predicted that OSM-3 kinesin would be deployed in a broad range of cilia to specifically elongate the distal singlets. However, the current study suggests that this is an oversimplification because AWC cilia in WTs do not have distal singlets and these cilia were morphologically indistinguishable in mutants and WTs. It will now be interesting to identify regulatory factors that act via OSM-3 to control the elongation of the distal singlets either by promoting their extension in channel cilia or by suppressing their extension in wing cilia. Such factors may be uncovered among ciliary mutants (; ; ). Sensory cilia in the head of the animal have diverse functions and provide diverse sensory modalities. Defects in osmotic avoidance are associated with the channel cilia and are present in mutants but not in or mutants (). We used chemotaxis toward benzaldehyde as a specific assay for AWC cell function (; ) and observed that chemotaxis was normal in WT animals and in mutants, but in or single and double mutants, chemotaxis was impaired by ∼50%, similar to the IFT particle mutant, (). The results suggest that these two motors are functionally redundant for AWC ciliary length determination, but once a full-length cilium is assembled, kinesin II has additional chemosensory functions that may involve the delivery of signaling molecules for incorporation into cilia, as proposed for kinesin II (). Future studies of ciliary mutants may uncover this predicted cargo of kinesin II. How kinesin II and OSM-3 redundantly assemble AWC cilia is unclear because we were unable to detect IFT in assays identical to those that demonstrate vigorous IFT along channel cilia (; ). Thus, it is possible that in AWC neurons, ciliary assembly does not depend on IFT and the kinesin 2 motors build cilia using a different mechanism. However, we favor the idea that IFT does operate in these cilia but was not seen because it occurs transiently during de novo ciliary assembly or for technical reasons such as a low level of expression of the IFT machinery. Thus, in the simple, canonical mechanism of IFT, kinesin II alone delivers precursors to the tip of assembling cilia to build the axoneme. In and perhaps elsewhere, OSM-3 is deployed as an “accessory” motor to functionally substitute for the loss of kinesin II function or, when kinesin II is active, to augment the canonical IFT pathway and, in some cases, to elongate the distal singlets. Our results suggest that modulating the activity of the two IFT kinesins in different cilia contributes to functional differences between cilia on different sensory neurons, providing a selective advantage by allowing the animal to receive and respond to a greater diversity of sensory cues. Animals were maintained and crossed using standard methods () as described previously (). Homozygosity was confirmed for and by dye filling and for and by single-worm PCR. Because and are on the same chromosome, we crossed with to produce animals with the final genotype . Amphid channel cilia were marked by introducing either or + into motor mutants, and AWC cilia in WT and mutants were imaged with oyIs44[]. ODR-1::RFP was supplied by P. Sengupta (Brandeis University, Waltham, MA). As described previously (; ), worms were anesthetized with 10 mM levamisole, mounted on agar pads, and maintained at 21°C. Fluorescent images were collected on a microscope (Olympus) equipped with a 100×, 1.35 NA objective and a spinning disc confocal head (UltraView; PerkinElmer) and further processed using MetaMorph software (Universal Imaging Corp.). Specimens were ultrarapidly frozen using a high-pressure freezer (HPM 010; Bal-Tec) and freeze substituted at −90°C in anhydrous acetone containing 1% osmium tetroxide and 0.1% uranyl acetate using an EM AFS (Leica; ). The cryofixed specimens were raised to room temperature and subsequently infiltrated and embedded in Epon-Araldite resin. Samples were cut into 100-nm-thick sections using an Ultracut E microtome (Leica). Serial sections were collected on slot grids covered with a Formvar support film, poststained with 2% uranyl acetate followed by Reynold's lead citrate, and covered by a second layer of Formvar for increased stability. Grids were visualized on a transmission electron microscope (JEM-2100FEG; JEOL) operated at 200 keV. Images of each section were recorded at ∼4 μm defocus on a 4,096 × 4,096–pixel charge-coupled device camera (Tietz; TVIPS) and at a nominal magnification of 8,000, yielding a pixel size of 1.25 nm. The images of the serial sections were manually aligned using the manual image deformation and alignment system () and modeled as surface renderings using 3DMOD to depict the length and paths of the amphid pore and each cilium (). Chemotaxis assays were performed as described previously (). 15–20 young adult animals were placed onto a 10-cm Petri dish without bacteria for 1 h and then transferred to a 10-cm Petri dish containing 1.6% Bacto agar, 5 mM potassium phosphate, pH 6.0, 1 mM CaCl, and 1 mM MgSO, along with 2 μl of the attractant, benzaldehyde (1:200 in ethanol), and 2 μl of the counterattractant, ethanol. The worms were placed at a marked origin, equidistant from the attractant and counterattractant, 2 μl of sodium azide was used to anesthetize the animals, and the numbers located at attractant and counterattractant were counted after 1 h. The chemotaxis index was defined as follows: (numbers at attractant − numbers at counterattractant) ÷ total (). Assays were repeated multiple times. A one-way analysis of variance was performed using Proc Mixed (SAS Institute), and obvious assay differences between strains were observed using Tukey's test for multiple comparisons at a significance level of 0.05. On the graphs (), standard error bars are shown. Fig. S1 shows a transmission electron micrograph of sections of each strain ∼5 μm from the tip of the head. Online supplemental material is available at .

Polarized mammalian epithelia form tight junctions between adjacent cells to regulate transepithelial permeability (gate function) and to separate the distinct apical and basolateral domains (fence function; ). The complexity of tight junction functions is reflected in the large number of associated signaling and regulatory molecules, including the polarity proteins Par-3, Par-6, and atypical PKC (aPKC; ). Tight junctions are attached to a cortical actin network through several linker proteins, and modulation of the actin cytoskeleton has a profound impact on both the assembly and functions of tight junctions. Despite intensive studies on the roles of polarity proteins in controlling polarization and cell junction formation (for review see ; ), relatively little is known about how they interact with other cellular components to orchestrate these events. We have recently shown by RNA interference (RNAi) that Par-3 is essential for the proper assembly of tight junctions in mammalian epithelial cells () partially through its interaction with Tiam1 to spatially restrict the activity of Rac. We now show that Par-3 can also regulate cofilin phosphorylation by LIM kinase 2 (LIMK2). Cofilin binds to and severs actin filaments, and it is crucial for numerous fundamental cellular processes such as migration, cytokinesis, and phagocytosis (). Cofilin activity is inhibited by a single phosphorylation on Ser3, which is mediated by LIMKs or testicular protein kinases. Dephosphorylation is executed by protein phosphatases such as Slingshot (). LIMKs contain two LIM domains and a PDZ domain () and are activated by members of the Rho family of small GTPases. Recently, cofilin-mediated actin turnover has been shown to contribute to the disassembly of the apical junctional complex, which is induced in epithelial cells by calcium depletion (). However, it has not been implicated in tight junction assembly. Given the close connections between actin reorganization and tight junction assembly, we examined the effects of reduced Par-3 expression on the phosphorylation of cofilin. The specificity of Par-3 suppression by RNAi has been confirmed previously (). Depletion of Par-3 caused a substantial increase in cofilin Ser3 phosphorylation ( and Fig. S1 A, available at ). In contrast, depletion of the tight junction protein occludin had only a small effect on phospho-cofilin levels (Fig. S1 B). Epithelial (E) cadherin knockdown (KD) also resulted in elevated levels of phospho-cofilin (Fig. S1 B), but depletion of Par-3 did not affect E-cadherin levels or disrupt adherens junctions under normal calcium conditions and only had minor effects on adherens junction assembly during calcium switch (). Moreover, when MDCK cells were subjected to calcium depletion overnight to disrupt all cell–cell junctions, the phospho-cofilin levels in Par-3–depleted cells were still higher than in control cells (, LCM), indicating that Par-3 can regulate phospho-cofilin levels independently of cell junction status and extracellular calcium levels. In addition, a double KD of Par-3 and E-cadherin led to an additive increase in phospho-cofilin compared with that achieved by a single KD of Par-3 or E-cadherin (Fig. S1 C), further suggesting that Par-3 and E-cadherin regulates cofilin phosphorylation through distinct mechanisms. Transient expression of human Par-3c, which is not recognized by the short hairpin RNA (shRNA) targeting canine Par-3, partially reversed the increase in phospho-cofilin (), supporting the argument that the increased phospho-cofilin is caused by Par-3 depletion rather than by off-target effects. Par-3c is a splice variant that lacks the aPKC-binding site (), suggesting that the ability of Par-3 to regulate cofilin phosphorylation does not involve aPKC. To determine whether regulation of phospho-cofilin by Par-3 is conserved across different cell types, we suppressed the expression of endogenous Par-3 in HeLa cells and found that cofilin phosphorylation was also enhanced after Par-3 depletion (). Similarly, depletion of endogenous Par-3 in MCF-7 breast carcinoma cells also markedly elevated the pool of phospho-cofilin (). Neuregulin (NRG)-1β treatment of MCF-7 cells activated the cofilin-specific phosphatase Slingshot () and completely blocked the increase in phospho-cofilin in Par-3–depleted cells (), indicating that the loss of Par-3 does not interfere with the activation of Slingshot. These data suggest that Par-3 is generally involved in down-regulating cofilin phosphorylation. Because the depletion of Par-3 results in phosphorylation of cofilin and its inactivation, we investigated whether reduced cofilin activity might contribute to the defects in tight junction assembly caused by Par-3 depletion. Using the transmembrane protein occludin as a marker, we observed, as reported previously, that MDCK cells lacking Par-3 had disrupted tight junctions for many hours after a calcium switch (; ). However, the normal cortical distribution of occludin was partially rescued at later stages of junction assembly by transient expression of a phosphorylation-defective, constitutively active cofilin mutant (S3A; ). Quantification of the mean length per cell of occludin at cell–cell contacts confirmed the significance of the enhanced cortical localization of occludin (). Importantly, the expression of cofilin S3A did not affect the efficiency of Par-3 KD (). As shown previously, the loss of Par-3 substantially delayed the development of transepithelial electrical resistance (TER) after calcium readdition, and this was partially reversed by the expression of cofilin S3A (; ). Silencing of cofilin expression alone did not disrupt tight junctions, however, suggesting that cofilin activity is not a limiting regulator of junction assembly (Fig. S1, E–G). Altogether, these results suggest that cofilin activity contributes to tight junction formation during epithelial cell polarization and that one of the functions of Par-3 is to prevent inappropriate cofilin phosphorylation. We have shown previously that Par-3 binds to Tiam1 to down-regulate Rac activity (). However, a double KD of Par-3 and Tiam1 had no effect on the augmented phospho-cofilin levels induced by Par-3 suppression (Fig. S1 D). Furthermore, the expression of a constitutively active fragment of Tiam1 (Tiam1 C1199; ) did not elevate phospho-cofilin levels (Fig. S1 D), supporting the idea that the elevated phospho-cofilin in Par-3 KD cells is independent of misregulated Tiam1 activity. Interestingly, cofilin S3A expression did not further improve the TER establishment in Par-3 and Tiam1 double KD cells (Fig. S1 H), suggesting that cofilin activity and the Tiam1–Rac pathway may regulate similar aspects or stages of tight junction assembly. We next asked whether Par-3 might directly or indirectly regulate the activities of LIMKs or phosphatases to influence cofilin phosphorylation. LIMKs are activated by Rho small GTPases through their downstream effectors Rho-associated kinase (ROCK) and p21-activated kinase (; ). No specific binding of Par-3 to the Slingshot phosphatase was detected (unpublished data), which is consistent with our observation that Slingshot activation is not altered by the loss of Par-3 (). However, recombinant Par-3 fragments (Par-3c-B and Par-3c-D) specifically pulled down LIMK2 but not LIMK1 from COS cell lysates (). Furthermore, the COOH terminus of Par-3 (Par-3c-E) alone coimmunoprecipitated LIMK2 just as efficiently as a longer fragment of Par-3 (Par-3c-D; ), suggesting that the COOH terminus of Par-3 is sufficient to mediate the interaction with LIMK2. In contrast, another PDZ domain–containing polarity protein, Pals1, did not coimmunoprecipitate LIMK2. When endogenous LIMK2 from MDCK cells was immunoprecipitated, an association of endogenous Par-3 with LIMK2 was detected (). These data demonstrate that Par-3 and LIMK2 interact in vivo. To further dissect the region of LIMK2 that is involved in its interaction with Par-3, deletion fragments of LIMK2 were tested for their abilities to coimmunoprecipitate the COOH terminus of Par-3 (). Deletion of the entire NH-terminal half of LIMK2 (LIMK2 ΔN) or of the two LIM domains (LIMK2 ΔLIM) severely abrogated association with the Par-3 COOH terminus, indicating that the LIM domains are required for the interaction. In contrast, LIMK2 ΔC or a kinase-dead mutant of LIMK2 was able to associate with Par-3 ( and Fig. S2 A, available at ). To examine whether Par-3 fragments can bind to the NH terminus of LIMK2 in vitro, recombinant Par-3 fragments immobilized on beads were incubated with GST–LIMK2 ΔC or with GST-Rac as a negative control ().GST–LIMK2 ΔC interacted very weakly with the Par-3c-A fragment, but it showed more robust interaction with the Par-3c-B and Par-3c-D fragments, which is in agreement with the pull-down experiments. LIMK2 can be activated by ROCK downstream of RhoA (). A specific ROCK inhibitor H-1152 greatly reduced the phospho-cofilin level in Par-3 KD MDCK cells (), supporting the involvement of LIMK2 in cofilin phosphorylation. To confirm that endogenous LIMK2 mediates cofilin phosphorylation in Par-3–depleted cells, we partially suppressed the expression of LIMK2 in MDCK cells by RNAi without affecting the total levels of Par-3 and cofilin (). Double KD of both Par-3 and LIMK2 inhibited the increase in phospho-cofilin caused by Par-3 depletion (), suggesting that LIMK2 is important for increasing phospho-cofilin levels after Par-3 depletion. Consistent with the reduction in phospho-cofilin level, the suppression of both Par-3 and LIMK2 also promoted the assembly of tight junctions after a calcium switch, as monitored by the improved localization of occludin to cell–cell borders () and by the small but reproducible enhancement in TER establishment (Fig. S2 C). RhoA activity is not altered in Par-3 KD cells (), and we could not detect any interaction between Par-3 and ROCK (Fig. S2 B). The ability of Par-3 to interact with LIMK2 suggests that Par-3 is acting directly on LIMK2. Interestingly, cells lacking Par-3 showed a small increase in the level of phosphorylation on Thr505 of LIMK2 (Fig. S3 A, available at ). Phosphorylation at this site is mediated by ROCK and leads to LIMK2 activation (). Therefore, Par-3 might be involved in negatively regulating LIMK2 activation by ROCK in vivo. We next examined whether Par-3 can mediate the inhibition of LIMK2 kinase activity in vitro. Immunoprecipitated LIMK2 specifically phosphorylated GST-cofilin, and the presence of Par-3b inhibited this activity (). Par-3b was also phosphorylated during the kinase assay, which was likely caused by endogenous aPKC associated with Par-3b (). However, the Par-3c splice variant also inhibited the kinase activity of LIMK2 (). This variant does not bind to aPKC (), and it was minimally phosphorylated compared with Par-3b. To test the specificity of the inhibition, LIMK1 was used, and its activity was not inhibited by Par-3 (Fig. S3 B). Moreover, when we replaced Par-3 with another PDZ domain protein, Pals1, no inhibition of LIMK2 activity was observed (Fig. S3 C). To determine whether Par-3 can inhibit cofilin phosphorylation in vivo, Par-3c was transiently expressed in COS cells followed by treatment with lysophosphatidic acid (LPA), which activates the Rho–ROCK–LIMK2 pathway (). Overexpression of Par-3c decreased the basal phospho-cofilin level (0 min) and inhibited the increase in cofilin phosphorylation induced by LPA. Together, our data suggest that Par-3 binding to LIMK2 mediates the inhibition of LIMK2 activity and the down-regulation of cofilin phosphorylation. LIMK1 and 2 share similar domain structures and have 50% overall amino acid identity (), but they are regulated by distinct mechanisms. LIMK1 is activated specifically by p21-activated kinase and is inhibited by association with bone morphogenetic protein type II receptor, which binds to the NH-terminal region of LIMK1 (). LIMK2 is activated specifically by ROCK and, as we show here, is inhibited by association through its NH-terminal region with Par-3. The two kinases also exhibit different tissue distributions, with LIMK1 highly expressed in the brain and LIMK2 broadly expressed in the epithelia of nonneuronal tissues (), which is consistent with a role for LIMK2 in epithelial polarization. Although Par-3 is known to be a key regulator of epithelial polarization, its molecular actions have remained elusive. We have now identified a novel function for Par-3 in the regulation of LIMK2, which phosphorylates and inactivates cofilin. Depletion of Par-3 leads to elevated cofilin phosphorylation, and the restoration of cofilin activity by either active cofilin mutant or double KD of Par-3 and LIMK2 partially rescues tight junction assembly, indicating that the suppression of cofilin phosphorylation by Par-3 plays a positive role in junction assembly. Although it is possible that other indirect mechanisms contribute to the increase in phospho-cofilin after Par-3 KD, our data strongly suggest that Par-3 is directly involved in regulating cofilin phosphorylation through its interaction with LIMK2. Together with our previous finding that Par-3 regulates Rac activity through its interaction with Tiam1 (), our current study suggests a possible role for Par-3 in modulating actin dynamics to facilitate junction assembly and polarization. On the other hand, actin organization is required for the correct localization of Bazooka (the homologue of Par-3) in embryonic epithelia (). We propose that one of the functions of Par-3 is to coordinate and fine-tune the actin remodeling processes associated with junction assembly through its abilities to interact with and regulate multiple signaling pathways that control actin reorganization. It will be interesting to examine whether this mechanism of Par-3–directed cofilin activation is used in other cellular processes that require polarization, such as cell migration. The following constructs have been described previously: myc and HA-tagged full-length Par-3b, Par-3c and Par-3c fragments (Par-3c-A, Par-3c-B, Par-3c-D, and Par-3c-E), and S-tagged Par-3c fragments (; ). Human cofilin-1 was generated by PCR from a human kidney cDNA library (CLONTECH Laboratories, Inc.) and cloned into the pKH3 and the pGEX-2T vectors. Cofilin S3A was created using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). Human Slingshot in pcDNA3.1/myc-His was a gift from T. Uemura (Kyoto University, Kyoto, Japan). Flag-tagged LIMK1 in pEF-BOS was a gift from O. Bernard (The Walter and Eliza Hall Institute of Medical Research,Victoria, Australia). HA-tagged full-length rat LIMK2 and the LIMK2 kinase-dead mutant in pcDNA3 were gifts from T. Nakamura (Osaka University, Osaka, Japan). LIMK2 deletion fragments were generated by PCR and cloned into pKH3 and pGEX-2T. ROCK KD-IA construct was a gift from S. Narumiya (Kyoto University, Kyoto, Japan). Tiam1 C1199 was provided by J.G. Collard (Netherlands Cancer Institute, Amsterdam, Netherlands). MDCK, COS-7, HeLa, and MCF-7 cells were grown in DME supplemented with 10% calf serum, 100 U ml penicillin, and 100 U ml streptomycin (Invitrogen). For transfection of MDCK cells, 2–4 μg DNA was introduced into 2–4 × 10 cells by Nucleofection (Amaxa) according to the manufacturer's recommendations. Transfection efficiency was generally >70%. For transfection with pSUPER constructs, enhanced YFP was coexpressed as a marker for transfected cells. Experiments were performed 2–3 d after transfection. COS-7 cells were transiently transfected using the calcium phosphate method. For RNAi in MDCK cells, pSUPER constructs expressing gene-specific shRNA were introduced into MDCK cells by Nucleofection. Generation of pSUPER constructs expressing shRNA targeting canine Par-3 or Tiam1 has been described previously (). The target sequences for other pSUPER constructs are as follows: canine occludin, AGAGGAATTATTGCAAGCA; canine E-cadherin, GTCTAACAGGGACAAAGAA; canine LIMK2, TGTTGACAGAGTACATCGA; canine cofilin-1 #1, GGATCAAGCATGAGTTACA; and canine cofilin-1 #4, AGATCCTGGTAGGTGATGT. Empty pSUPER vector or a pSUPER construct targeted against luciferase was used as a negative control. To silence Par-3 in HeLa and MCF-7 cells, a pool of four siRNA duplexes (SMARTpool reagent M-015602-00) was purchased from Dharmacon. As a negative control, we used a siCONTROL nontargeting siRNA. HeLa and MCF-7 cells were transfected using Oligofectamine and LipofectAMINE 2000 (Invitrogen), respectively. Cells were processed 48 or 72 h after transfection. Calcium switch experiments were performed as described previously (). To detect occludin and YFP, cells were washed and fixed in 1:1 methanol/acetone (vol/vol) for 3 min at −20°C. After washing with PBS, cells were blocked with 5% BSA in PBS for 25 min and incubated with primary and secondary antibodies. Primary antibodies were monoclonal anti-occludin (Zymed Laboratories) and rabbit anti-GFP (Invitrogen). Secondary antibodies were AlexaFluor-conjugated goat anti–mouse and goat anti–rabbit antibodies (Invitrogen). AlexaFluor594-conjugated phalloidin was obtained from Invitrogen. Coverslips were mounted on slides, and images were obtained and processed as previously described (). To quantify the mean occludin length per cell, six to seven fields of cells with similar density were randomly selected from each coverslip, and the total length of occludin at cell junctions in each field was measured using the measurement tool in Openlab (Improvision). Only fields with >70% transfected cells (as determined by YFP expression) were measured. Cell numbers were counted for each field by DAPI staining, and the mean occludin length per cell on each field was calculated (). Between 300 and 500 cells were counted for each coverslip. Each bar represents the mean value from all of the fields on each coverslip. Statistical analysis was performed using a two-tailed test. Epifluorescence images were collected using an inverted microscope (Eclipse T200; Nikon) equipped with a 60× NA 1.2 plan Achromatic water immersion lens (Nikon) and a charge-coupled device camera (Orca C4742-95-12NRB; Hamamatsu). Images were collected at a 12-bit depth and 1,024 × 1,280 pixel resolution with 1 × 1 binning using Openlab 4.0 software (Improvision). Images were converted to eight-bit TIFF files and processed in Photoshop 7.0 (Adobe) to increase the gray scale range and to reduce haze using an unsharp mask. After transfection, 6.6 × 10 MDCK II cells were plated into polycarbonate Transwell TM filters (0.4-μm pore size and 12-mm diameter; Corning Costar). 2 d later, cells were changed into low calcium medium (LCM) for 18–20 h before being switched back to normal growth medium (high calcium medium [HCM]). TER was measured with an Epithelial Voltohmmeter (EVOM; World Precision Instruments). TER values were calculated by subtracting the blank value from an empty filter with HCM and were expressed in ohm.cm. Three parallel filters from each transfection were measured for each time point. Immunoprecipitation from transiently transfected COS-7 cells was performed as previously described (). Anti-HA antibody 12CA5 or anti-myc antibody 9E10 was used. To immunoprecipitate endogenous complexes of Par-3 and LIMK2 from MDCK cells, polyclonal anti-GFP antibody (Invitrogen), polyclonal anti-LIMK2 antibody (Santa Cruz Biotechnology, Inc.), and monoclonal anti–β-catenin antibody (BD Biosciences) were used. In vitro kinase assays were performed essentially as described previously (). Constructs expressing HA-tagged proteins were individually transfected into COS cells. 40 h later, HA-tagged proteins were immunoprecipitated with an anti-HA antibody 12CA5 either alone or together after combining the lysates. After extensive washing with lysis buffer and kinase buffer, immunoprecipitates were incubated for 20 min at 30°C in 16 μl of kinase buffer containing 50 μM ATP, 5 μCi γ-[P]ATP (4,500 Ci/mmol; MP Biomedicals), and 6 μg GST-cofilin or GST-Rac1. The reactions were terminated by adding 4× SDS sample buffer followed by SDS-PAGE, Coomassie staining, and autoradiography. Fig. S1 shows the effects of Par-3, occludin, and E-cadherin KD as well as Par-3/E-cadherin double KD on phospho-cofilin levels in MDCK cells (A−C). Par-3/Tiam1 double KD does not reduce phospho-cofilin levels (D). It also shows that KD of cofilin alone does not disrupt tight junctions (E−G) and shows the effects of Tiam1 KD and cofilin S3A expression on TER establishment in Par-3–depleted MDCK cells (H). Fig. S2 shows that the kinase activity of LIMK2 is not required for its association with Par-3 (A) and that Par-3 does not associate with ROCK (B). It also shows that LIMK2 KD promotes TER development in Par-3–depleted MDCK cells (C) and the actin organization during calcium switch in control and Par-3 KD cells (D). Fig. S3 shows the elevated levels of Thr505 phosphorylation on LIMK2 in Par-3–depleted MDCK cells (A). It also shows that Par-3 inhibits LIMK2 but not LIMK1 kinase activity in vitro (B) and that Pals1 does not inhibit LIMK2 activity (C). Online supplemental material is available at .

Reversible protein phosphorylation is the major general mechanism regulating most physiological processes in eukaryotic cells. Hundreds of protein kinases have been identified to date, with significantly fewer phosphatases found to counteract their action. This discrepancy can be explained in part by mechanisms used to control phosphatase activity. Protein phosphatase 1 (PP1), a major serine/threonine protein phosphatase involved in a wide range of cellular processes, exists in the cell as an oligomeric complex. The core catalytic subunit binds a spectrum of interacting proteins, termed targeting subunits, that modulate both its intracellular localization and substrate specificity (for review see ). Most of the known targeting subunits, including those targeting PP1 to such substrates as glycogen and myosin, contain a conserved “RVXF” motif (consensus Arg/Lys-Val/Ile-Xaa-Phe/Trp; ) that mediates direct binding to PP1. Using primarily biochemical approaches, >50 PP1 targeting subunits have been identified. However, these most likely cannot account for the large number of regulatory pathways in which PP1 plays a critical role, suggesting that many targeting subunits remain to be discovered. In addition to the signaling diversity provided by targeting subunits, PP1 is also expressed in mammalian cells as three closely related isoforms, α, β/δ, and γ, which are encoded by separate genes. These isoforms are >89% identical in amino acid sequence, with minor differences primarily at their NH and COOH termini (for review see ). Although most biochemical studies have not directly addressed the significance of the different isoforms, in vivo data show that they have distinct subcellular localization patterns (; ). This implies differences in their specificity of interaction with particular targeting proteins and hence preferential incorporation into different signaling complexes. Expression of PP1 isoforms as fusions with fluorescent protein (FP) tags offers both a means to compare their targeting in living cells and a method to recover them from cell lysates and analyze the proteins with which they associate. Because of the unreliability of commercially available PP1 isoform– specific antibodies for immunostaining (; ), we have gone to great lengths to validate the use of these FP fusion proteins as markers for endogenous pools of PP1. FP-tagged PP1 isoforms are functionally active phosphatases with distinct interphase localization patterns () matching those observed by antibody staining of endogenous PP1 isoforms in fixed cells (). Importantly, these distinct localization patterns can be overridden if a single targeting subunit is exogenously overexpressed in cells (). In this case, the abundantly expressed targeting subunit recruits all PP1 isoforms into the same distribution pattern, suggesting that the expression levels of individual targeting subunits are critical for determining the normal localization patterns of PP1 isoforms. We recently reported the establishment and characterization of HeLa cell lines stably expressing FP-PP1γ (). During interphase, FP-PP1γ was found in both cytoplasmic and nucleoplasmic pools, showing a prominent accumulation within nucleoli, whereas time-lapse imaging revealed dynamic targeting to specific sites during cell division, including kinetochores and chromatin. This changing spatiotemporal pattern implicates PP1γ in multiple regulatory pathways, in agreement with previous work linking its activity to the regulation of such cellular events as transcription, chromatin remodeling, chromosome condensation and segregation, cytokinesis, and reassembly of the nuclear envelope (for review see ). We have now established a HeLa cell line stably expressing FP-PP1α. The separate HeLa and HeLa cell lines allow the contrasting localization patterns and dynamic properties of both isoforms to be analyzed throughout the cell cycle and facilitate a comparison of isoform-specific binding partners detected via affinity purification of the respective tagged proteins. Mass spectrometry–based proteomics has become a powerful tool for identifying and quantifying the components of multiprotein complexes (for review see ). More recently, several techniques have exploited the use of heavy isotope tags to compare and quantitate relative protein levels under different biological conditions (for review see ). In the case of stable isotope labeling of amino acids in cell culture (SILAC), cells are metabolically labeled through growth in medium containing a specific amino acid with either carbon or nitrogen or both substituted with the heavy isotopes C and N. By using substituted arginine or lysine, proteins are labeled specifically at sites of trypsin cleavage, which is convenient for subsequent analysis of tryptic peptides by mass spectrometry. We adapted the triple encoding SILAC approach () to identify both novel and known PP1-interacting proteins and quantify their relative distribution between PP1α and -γ in vivo. One of these proteins, Repo-Man, defines a novel pool of PP1, selectively involving the γ isoform, which loads onto chromatin at anaphase and remains bound throughout the following interphase. Several lines of evidence indicate that the Repo-Man–PP1 complex is critical for cell viability. A stable HeLa cell line (HeLa) was established that expresses PP1α fused at its NH terminus to EGFP. As observed for the previously characterized HeLa stable cell line (), this HeLa cell line is homogenous, with >95% of the cells expressing full-length catalytically active FP-PP1α at levels similar to those of endogenous PP1α (unpublished data). A comparison of the cell cycle progression of both HeLa and HeLa cells with parental HeLa cells showed that their growth rates are equivalent, and FACS analysis confirmed that the relative populations in G1, S, and G2/M are similar (). Fluorescence imaging of HeLa and HeLa cells revealed striking differences in the intrinsic localization patterns of the two isoforms (), similar to those observed previously by transient transfection of the fusion proteins () and by antibody labeling of the endogenous proteins (). Although both are found in the cytoplasm and in the nucleus throughout G1, S, and G2, nuclear FP-PP1α is found mainly in a diffuse pool and in a few as-yet-unidentified foci (, arrow) and largely excluded from nucleoli. In contrast, nuclear FP-PP1γ shows a strong accumulation within nucleoli (, arrowhead; ∼20% of the total nuclear FP-PP1γ; ). FP-PP1α and FP-PP1γ also show differences in their dynamic distribution during cell division. Time-lapse imaging revealed that although both isoforms accumulate at kinetochores in metaphase (, H and I, 0 min), FP-PP1α is predominantly excluded from other chromatin regions throughout mitosis (open arrowheads). In contrast, low levels of FP-PP1γ appear on chromatin at metaphase and there is a sudden and rapid recruitment of a larger pool (∼20% of the total cellular pool; ) onto chromatin as the cell enters anaphase (, min; ). Both isoforms also localize at the cortex and midbody in telophase (, dotted arrow; ). FP-PP1α is recruited back into the nucleus much later in telophase (, 36 min), whereas FP-PP1γ, which is already associated with chromatin, reaccumulates within nucleoli at this time (, 36 min). High-resolution images of mitotic HeLa cells more clearly demonstrate the chromatin exclusion of FP-PP1α during metaphase and anaphase (, J and K, arrowheads), as compared with FP-PP1γ and its dramatic recruitment to chromatin at anaphase (, L and M, arrowheads). FP-PP1α also accumulates at centrosomes (, arrows), which was confirmed in both interphase and mitotic cells by counterstaining with anti-tubulin antibodies (not depicted) and is in agreement with previous observations (). These data imply that PP1α and -γ preferentially interact with distinct targeting subunits throughout the cell cycle and hence play distinct roles. We thus sought to identify proteins that associate preferentially with either FP-PP1α or FP-PP1γ by immunoprecipitating the fusion proteins from interphase nuclear lysates using the same anti-GFP antibody. Although this approach was not likely to identify mitosis-specific targeting proteins (i.e., those that only bind PP1 during mitosis), we made the assumption, based on known regulatory roles for PP1 in transcription and chromatin remodeling (for review see ), that the chromatin targeting of PP1γ at anaphase was maintained into the following interphase. No clear differences were detected when immunoprecipitates were screened by SDS-PAGE separation and silver staining (). However, a far Western assay using in vitro–translated S-labeled PP1 offered a means to directly detect and compare the subset of proteins coprecipitated with FP-PP1α (, lane 2) and FP-PP1γ (lane 3), which bind directly to the phosphatase. This revealed differences in the proteins that coprecipitated with the respective isoforms (, arrowheads). Immunoprecipitation of unfused FP from the HeLa cell line, using the same anti-GFP antibody, coprecipitated little or no PP1 binding proteins, confirming the specificity of these interactions (, lane 1). A SILAC proteomic analysis was performed to identify and quantify the proteins that immunoprecipitate with anti-GFP antibodies from HeLa, HeLa, and HeLa cell lines (). Sample peptide spectra are shown for three proteins, each with a different arginine ratio (). Contaminating proteins that copurify with FP alone as well as with the FP-PP1 fusion proteins show an arginine ratio of 1:1:1 (hnRNP K; ), whereas proteins that copurify specifically with either FP-PP1α (NIPP1; ) or FP-PP1γ (Q69YH5; ) have ratios reflecting this enrichment for heavier forms of arginine. The accuracy of this mass spectrometric approach was confirmed by Western blot analyses for both NIPP1 (, inset) and Q69YH5 (, inset). Quantified proteomic data obtained for the full spectrum of proteins identified in this screen are presented in the graph in , in which the mean arginine isotope ratio of FP-PP1 to FP for all quantifiable peptides is plotted for each protein. To facilitate visual comparison, the ratio is shown as a positive value for FP-PP1α and a negative value for FP-PP1γ. It is clear that although most of the proteins are not specific PP1-interacting factors and also copurify with FP alone (ratios ∼1), several proteins show enhanced copurification with FP-PP1 (ratios >1.5), and several of these show a preference for one of the isoforms. The large fraction of copurifying proteins revealed by isotope labeling to be contaminants prompted us to add a selective elution step with an RVXF motif–containing peptide that displaces proteins bound to PP1 via this motif (data included in Table S1, available at ). However, although this reduced the number of proteins recovered by ∼50%, the resulting protein mixture still contained a large fraction of contaminants, illustrating the utility of the isotope-labeling method and the importance of including the unfused FP internal control. The PP1-interacting factors revealed by the isotope labeling as specific (ratio >1.5) are listed in and include known PP1 binding proteins as well as multiple novel factors, a subset of which contains an RVXF motif and is thus likely to bind PP1 directly. Factors lacking this motif may bind PP1 by another mechanism or may represent indirectly associated proteins that are part of larger PP1 complexes. A comprehensive list of all copurifying proteins identified, including contaminants, is provided in Table S1. We were particularly interested in proteins showing preferential interaction with PP1γ, with the aim of identifying the mechanism through which this isoform specifically localizes to chromatin. One of the novel proteins found to interact preferentially with PP1γ contained a canonical RVXF motif, suggesting it was a good candidate for further analysis. This protein, Q69YH5, whose function is unknown, copurified with PP1γ in a 3:1 ratio over PP1α (). Analysis of the sequence of Q69YH5 revealed stretches of basic residues upstream and acidic residues downstream of the RVXF motif, which have been shown to enhance interaction with PP1 (). Q69YH5 is a unique protein with homologues in vertebrates but apparently not in lower eukaryotes. The region surrounding the RVXF motif is particularly conserved among vertebrate species (; see Fig. S1, available at , for a full sequence alignment). The human gene is located at chromosome 8p21.2, and Northern blot analysis suggests that it is ubiquitously expressed (GeneNote; ). To aid the molecular characterization of Q69YH5, we raised antibodies against peptides from both the NH and COOH termini (). Both antibodies recognize a protein of the expected molecular mass (∼113 kD) for the endogenous Q69YH5 in HeLa nuclear extracts () and detect appropriately sized bands when tagged derivatives are exogenously expressed (Fig. S2, available at ). We note that the antibody raised against the NH-terminal peptide recognized an additional band in HeLa extracts of ∼65 kD, which has not been characterized further. Using the anti-peptide antisera, we confirmed that endogenous Q69YH5 interacts with PP1 in vivo. Affinity-purified anti-Q69YH5 antisera coimmunoprecipitated PP1γ from HeLa interphase nuclear lysate (, lane 3), and this coimmunoprecipitation was lost when the antibody was preincubated with the cognate peptide (, lane 4). Purification of endogenous PP1 complexes from interphase nuclear lysates using microcystin affinity chromatography also verified that endogenous Q69YH5 (, lane 2) is in a complex with PP1. Using HeLa cell fractions, we showed that endogenous Q69YH5 is predominantly nuclear (). This was confirmed by immunostaining paraformaldehyde-fixed HeLa cells with anti-Q69YH5, which revealed a widespread nucleoplasmic accumulation of the protein, largely excluding nucleoli (). This staining pattern was lost upon preincubation of the antibody with the cognate peptide (unpublished data). Interestingly, immunolocalization of Q69YH5 in mitotic cells showed a diffuse pattern in metaphase () and an enhanced accumulation on chromatin at anaphase and telophase (, F and G, arrowheads). This was similar to the distribution seen for PP1γ and encouraged us to examine the dynamic behavior of Q69YH5 in live cells. We therefore fused the Q69YH5 cDNA with EGFP and performed time-lapse imaging of FP-Q69YH5 transiently expressed in HeLa cells. Although a nucleoplasmic localization pattern similar to that observed for endogenous Q69YH5 is maintained throughout the G1, S, and G2 stages of the cell cycle, it changes dramatically at M phase (). The protein initially becomes diffuse throughout the cell as the nuclear membrane breaks down, and a faint accumulation is seen later on metaphase chromatin (, arrow). As the cell progresses to anaphase, there is a large accumulation of FP-Q69YH5 on chromatin (, box). Parallel time-lapse analysis of HeLa cells indicates that the timing of this chromatin association of FP-Q69YH5 coincides with that of FP-PP1γ (, box). These data, along with further characterization of the PP1 interaction, strongly indicate that Q69YH5 acts as a targeting subunit to recruit PP1 to chromatin at anaphase. We have therefore named Q69YH5 Repo-Man, for recruits PP1 onto mitotic chromatin at anaphase, to reflect this function. Having established that Repo-Man is recruited onto chromatin at anaphase and maintained there throughout interphase, we next addressed whether it is required for cell viability by reducing its levels using RNA interference (RNAi) knockdown. Four RNA duplexes targeted against Repo-Man were transiently transfected into HeLa cells, and the cells were monitored over time for viability, cell cycle progression, and levels of endogenous Repo-Man. As shown in , two of the Repo-Man duplexes caused a >80% reduction in protein levels by 24 h after transfection (), and this reduced protein level was maintained for a further 24 h, with levels only starting to rise again 72 h after transfection. The remaining two duplexes caused a smaller reduction in protein levels (∼30%). As a negative control, cells were transfected with a nonspecific scrambled duplex (). A duplex targeting lamin A/C was used as a positive control for the transfection, causing a sustained reduction (>80%) in lamin A/C protein levels by 48 h after transfection (). Cell cycle progression was also monitored over time by FACS analysis in cells transfected with Repo-Man duplex 1. Although there was a massive reduction in cell number over time after uptake of Repo-Man duplex versus scrambled (see the end of the paragraph), no significant differences were observed in cell cycle distribution patterns (, vs. ), suggesting that cells are dying regardless of cell cycle stage. A prominent sub-G1 peak is observed from 48 h after transfection onwards for cells treated with the Repo-Man duplex (), indicating an accumulation of apoptotic cells. This peak was not observed in cells treated with scrambled duplex, even at 72 h after transfection (). The small number of cells recovered after 72 h of Repo-Man knockdown () compared with 72 h of treatment with scrambled duplex () is consistent with a reduction in cell viability when Repo-Man levels are low (note the differences in scale in y axes). Apoptosis of Repo-Man–depleted cells was confirmed by counterstaining with Annexin V–FITC and propidium iodide 24 h after transfection. At this stage, Annexin V binds to the externalized phosphatidyl serine characteristic of apoptotic cells, but propidium iodide is excluded because membrane integrity remains intact. Cells treated with the scrambled duplex did not show any significant Annexin V staining or propidium iodide incorporation (), whereas a high proportion of cells treated with Repo-Man duplex 1 bound Annexin V, and the majority of these cells did not incorporate propidium iodide (). This indicated that membrane integrity was intact and that cells were therefore in an early stage of apoptosis. Further analysis of the mode of cell death during the 24–48-h time period was done by time-lapse imaging of cells transfected with either scrambled duplex or Repo-Man duplex 1, monitoring cells both by differential interference contrast (DIC) and by fluorescence imaging of Hoechst 33342–stained DNA for the characteristic membrane blebbing and nuclear fragmentation observed in apoptosis. The graph in sums up the scored results, with >60% of cells apoptosing in interphase after uptake of Repo-Man duplex 1. Less than 3% of these cells divide successfully over this time period, which explains why so few mitotic cells are observed in these knockdown experiments. A soft agar colony–forming assay was also performed to assess cell viability. At 24 h after transfection, cells treated with either scrambled duplex or Repo-Man duplex 1 were transferred to the soft agar matrix at a clonal dilution. After 2 wk, there was a >50-fold difference in the number of colonies formed from cells treated with the scrambled duplex (, arrows) as compared with cells treated with the Repo-Man duplex (). In summary, the data show that RNAi-induced reduction in Repo-Man levels promotes the onset of apoptosis within 24 h and severely reduces cell viability. To characterize the interaction of Repo-Man with PP1 in more detail and to address whether this activity is linked with its chromatin binding, we analyzed a Repo-Man mutant in which the two conserved hydrophobic residues within the putative PP1 binding domain were changed to alanine residues (i.e., RVXF to RAXA; ). Transient overexpression of this mutant FP–Repo-Man () showed that it had a localization similar to that of wild-type FP–Repo-Man (). Although both express at similar levels when transiently transfected into HeLa cells (not depicted), Western blotting reveals that only the wild type coprecipitates endogenous PP1γ (, lane 3), indicating that interaction with PP1 has been disrupted by mutation of this RVXF region. We further tested the interaction of wild-type and RAXA mutant FP–Repo-Man with PP1γ by titrating varying levels of wild-type and mutant plasmid into HeLa cells. Under normal conditions during interphase, PP1γ is found in various cytoplasmic and nucleoplasmic pools, along with a prominent nucleolar accumulation (), whereas Repo-Man is nucleoplasmic (). When high levels of wild-type FP–Repo-Man are overexpressed, however, the cytoplasmic and nucleolar pools of PP1γ are greatly diminished and the majority of the protein shows a similar nucleoplasmic distribution to Repo-Man as it is retargeted by the expressed protein. In contrast, overexpression of RAXA mutant FP–Repo-Man, which does not bind PP1 (), fails to relocalize PP1γ. This is best illustrated by comparing the localization of FP-PP1γ, which is exclusively nucleoplasmic at a 100:1 ratio of transfected wild-type to mutant Repo-Man expression plasmid, with the characteristic nucleolar accumulation of PP1γ seen at a 1:100 ratio of the same plasmids (). When both wild-type and mutant Repo-Man are overexpressed equally, an intermediate effect is observed. Thus, the ability of Repo-Man to influence PP1 localization in a concentration-dependent manner critically depends on the presence of a functional PP1-interaction domain. We conclude that Repo-Man, which binds directly to PP1 and influences localization, has the properties of a classic PP1-targeting subunit. During mitosis, both PP1γ and Repo-Man simultaneously relocalize to chromatin at anaphase, which is most clearly illustrated by time-lapse fluorescence imaging of HeLa cells coexpressing FP-PP1γ and FP–Repo-Man (). Using a relocalization assay involving FP-PP1α, which does not accumulate on chromatin at anaphase (), it was shown that Repo-Man is directly involved in recruitment of PP1 to chromatin. Increased levels of exogenous FP–Repo-Man caused an ectopic recruitment of PP1α to chromatin at anaphase (). Interestingly, this accumulation of Repo-Man on chromatin appears to be independent of its ability to bind PP1. In cells expressing both FP-PP1α and RAXA mutant FP–Repo-Man, the mutant Repo-Man accumulates on chromatin as normal, whereas FP-PP1α retains its typical mitotic localization pattern (). We conclude that Repo-Man is a targeting subunit that recruits PP1 to chromatin and that this recruitment occurs initially at the metaphase–anaphase transition during mitosis. Interestingly, overexpression of both wild-type and mutant Repo-Man severely reduced both the total cell population and the number of mitotic cells observed, suggesting a deleterious effect on interphase cell function and/or entry into mitosis. It is therefore difficult to assess the possible mitotic defects associated with these perturbations, although we have found some evidence for chromosome condensation and segregation defects in cells overexpressing wild-type Repo-Man (Fig. S2). As noted in the previous paragraph, the RAXA mutant of FP– Repo-Man accumulates on chromatin in a fashion similar to wild type but no longer recruits PP1. The binding of Repo-Man to chromatin is reversible, as shown by FRAP analysis (Fig. S2). If a limited number of binding sites exist, we would predict that expression of high levels of RAXA mutant Repo-Man could displace endogenous Repo-Man–PP1 complexes on chromatin. To test this, we compared the anaphase chromatin targeting of FP-PP1γ in the presence of high levels of exogenously expressed RAXA mutant FP–Repo-Man () to that observed normally (). A reduced accumulation of FP-PP1γ on chromatin was evident (, arrowheads), as compared with cells not expressing the mutant Repo-Man, and we therefore propose the displacement model shown in . However, we note that almost no mitotic cells and relatively few interphase cells were detected expressing very high levels of RAXA mutant FP–Repo-Man, indicating that it adversely affects cell viability (see the following paragraph). These data suggest that RAXA mutant Repo-Man can act as a dominant negative to displace endogenous Repo-Man–PP1γ complexes from chromatin during both anaphase and interphase. The RNAi knockdown experiments () showed that Repo-Man is important for cell viability. The dominant-negative effect of the RAXA mutant provided a means to determine whether the cellular requirement for Repo-Man was connected to its role as a PP1-interacting protein. To do this, we compared the level of cell death resulting from either mock-transfection or transfection with the dominant-negative RAXA mutant of Repo-Man. A much larger level of cell death was observed for cell populations expressing RAXA mutant FP–Repo-Man (). A more detailed inspection of the dying cells revealed that the majority of cell death occurred during interphase (, lane 3). We did not observe an obvious mitotic block, although a proportion of cells died either during or immediately after mitosis (, lanes 4 and 5). As noted in the previous paragraph, there was a severe reduction in the number of mitotic cells observed, which is in agreement with the rapid onset of apoptosis. These data strongly argue that the essential role of Repo-Man in cell viability is directly related to its PP1-targeting activity. Distinct pools of PP1 play different roles in the cell, mediated by the binding of a common catalytic subunit to different targeting proteins. In this study, we used a combination of time-lapse fluorescence imaging and a novel proteomics strategy to characterize targeting proteins that can bind differentially to the α and γ isoforms of PP1. Specifically, we observed a pool of chromatin-bound PP1γ that is first recruited at anaphase and identified the targeting protein responsible. The Repo-Man–PP1 complex is loaded onto chromatin at anaphase, where it remains during interphase and is subsequently displaced when cells enter prophase (). Further characterization of Repo-Man showed that perturbation of its PP1 binding or expression level caused a variety of cell defects, as discussed later in this section. We adopted a mass spectrometric strategy involving differential stable isotope labeling, which allowed a quantitative comparison of the relative distribution of targeting subunits between the two PP1 isoforms and the ready elimination of nonspecific contaminants. The identification of Repo-Man was facilitated by the enhanced sensitivity available using this method. For example, Repo-Man was not detected in previous studies using affinity chromatography with immobilized microcystin, a cyclic peptide toxin that binds PP1 and PP2A covalently (). However, we found that Repo-Man does copurify among the proteins binding endogenous PP1 that are enriched by microcystin chromatography from interphase nuclear lysates when we detect it using anti–Repo-Man antibodies (), suggesting it was previously missed because of its relatively low abundance compared with other PP1 binding factors. In addition to offering increased sensitivity of detection, the SILAC approach also proved useful for identifying real PP1-specific proteins among the large number of copurifying contaminants. This aids identification of proteins that may be lost upon higher stringency purification strategies. Because as high as 90% of the copurified proteins proved to be contaminants (), the success of the strategy depended on inclusion of an internal negative control (cells expressing the FP tag alone). Any putative interacting proteins not specific to PP1 contained equal amounts of C/N-labeled peptides (from the FP cell line) to heavy isotope-labeled peptides (from the FP-PP1 cell lines). Furthermore, the ratio of the C/N to C/N-labeled peptides distinguished whether the targeting subunits bound preferentially to PP1α or -γ. Although we have applied this to the study of PP1, we envisage that a similar strategy could be adopted to analyze and quantitate the differential interactions of any closely related proteins or modified versions of the same protein, with binding partners in other biological systems. Time-lapse fluorescence imaging of the respective FP-PP1 stable cell lines revealed that both FP-PP1α and FP-PP1γ localize to kinetochores in metaphase; however, PP1α appears to be predominantly excluded from other chromatin regions at this stage, in contrast to PP1γ. Furthermore, we observe a dramatic increase in the accumulation of FP-PP1γ on chromatin at anaphase, where it is retained as daughter nuclei reform. This differential binding to chromatin explains why FP-PP1α appears in daughter nuclei much later and indicates the existence of chromatin binding factors that can selectively target PP1γ. Our strategy of screening PP1 binding proteins for factors showing a preference for PP1γ thus enabled us to identify Repo-Man as a targeting subunit that mediates the regulated interaction of PP1γ with chromatin. Repo-Man is a previously uncharacterized protein first detected as being a member of a group of proteins whose mRNA expression correlated with that of known cell cycle–related proteins (). Interestingly, another of these proteins, borealin, has recently been identified as a novel chromosomal passenger protein required for stability of the bipolar mitotic spindle (). More recently, Repo-Man was found among a group of cell cycle genes up-regulated in aggressive neuroblastoma tumors (). We show that Repo-Man is a chromatin binding factor that also associates with PP1 via the canonical RVXF motif. Although the loading of Repo-Man–PP1γ complexes onto chromatin occurs dramatically at the onset of anaphase, several lines of evidence suggest that the proteins remain associated in a complex bound to chromatin during interphase. Direct biochemical and proteomic analysis of interphase nuclei showed that both endogenous and FP-tagged PP1 copurify with Repo-Man, and fluorescent resonance energy transfer (FRET) experiments confirm that PP1 and Repo-Man can interact directly during interphase (). It is important to note that although PP1γ shows a strong nucleolar accumulation during interphase, it represents only ∼20% of the total nuclear pool of PP1, with the remaining 80% in the nucleoplasm. The identity of the nucleolar targeting subunit for PP1γ remains unknown, but our present data indicate that Repo-Man, along with NIPP1, ZAP3, p99/PNUTS, and other proteins, is involved in targeting pools of PP1γ to different nucleoplasmic sites, including chromatin, during interphase. Further evidence supporting the persistence of the Repo-Man–PP1γ complex on chromatin during interphase comes from photobleaching studies, which show a similar turnover rate for nucleoplasmic Repo-Man in interphase nuclei compared with Repo-Man on mitotic chromosomes (Fig. S2). Based on the photobleaching data, it appears that the protein is released from chromatin in late prophase, remaining predominantly diffuse until its recruitment back on to chromatin first at low levels during metaphase and then more dramatically at the onset of anaphase. Western blotting analysis revealed no major differences in Repo-Man levels at different stages of the cell cycle (unpublished data), indicating that its accumulation on chromatin at anaphase results from relocalization of an existing pool and not new synthesis. It will be important to determine what controls this cell cycle–modulated association of Repo-Man with chromatin and whether it is a direct (e.g., covalent modification) or indirect (e.g., interaction with another factor) regulatory event. Localization of PP1 to chromatin is in agreement with several of its known regulatory roles, which include dephosphorylation of histones and other mitotic phosphoproteins, regulation of chromosome decondensation and nuclear membrane assembly at the M–G1 transition, and regulation of both transcription and chromatin remodeling during interphase (for review see ). An important future goal, therefore, is the identification, by a similar SILAC/proteomic approach, of new binding partners for Repo-Man, which will provide clues to the specific regulatory pathways in which this PP1 complex is involved. Mapping the specific regions of Repo-Man required for chromatin localization will also facilitate the identification of binding partners and/or the mechanism of chromatin interaction, which was shown here to be separable from its PP1 binding ability. Changes in expression levels of many cell cycle–regulated kinases and phosphatases have serious consequences for cell viability (). Similarly, we find that depletion of Repo-Man by RNAi knockdown leads to cell death by apoptosis, primarily in interphase. Overexpression of mutant Repo-Man that does not bind PP1 leads to a similar phenotype, suggesting that it is the loss of endogenous Repo-Man–PP1γ complexes from chromatin that initiates the apoptotic pathway, either directly or indirectly. There is no apparent mitotic block, and time-lapse imaging of chromatin in these cells does not provide any further clues to the disrupted pathway (Fig. S3, available at ). The very rapid onset (within 24 h) of apoptosis results in very few cells surviving into mitosis ( and ), making it difficult to assess the possible mitotic defects associated with these perturbations. For example, although analysis of large numbers of HeLa cells treated with Repo-Man duplex 1 revealed that, as in parental HeLa cells, the majority of cells die rapidly through apoptosis in interphase, only a few mitotic cells (<2% of the total population screened) could be found to confirm that FP-PP1γ no longer accumulates on chromatin at anaphase and telophase under these conditions (Fig. S3). Overexpression of wild-type Repo-Man also promotes apoptosis but with lower efficiency. In the small proportion of cells surviving into mitosis, we observed evidence for chromosome segregation defects (Fig. S2). Collectively, the knockdown and overexpression data suggest that the Repo-Man–PP1γ complex may play important roles at multiple stages of the cell cycle, including both interphase and mitosis. For example, Repo-Man may target PP1γ to dephosphorylate chromatin-associated factors, thus modulating their role in maintaining the degree of chromatin condensation throughout the cell cycle, which is critical for both cell division and viability. Additionally, the Repo-Man–PP1γ complex may play an important role in modulating chromatin accessibility or the activity of specific transcription factors during interphase. Disruption of this interphase role is the likely cause of the large-scale apoptosis observed upon knockdown of Repo-Man, whereas the chromosome segregation defects found upon overexpression of Repo-Man point to disruption of a mitotic regulatory role. Further work is therefore required to investigate the essential roles of the Repo-Man–PP1 complex throughout the cell cycle. Repo-Man was cloned from expressed sequence tags (IMAGE clones 4803608 and 4838503; Geneservice Ltd.) using oligonucleotide primers, inserted into EGFP/EYFP/ECFP-C1 and monomeric dsRed-C1 FP vectors (Invitrogen) and confirmed by restriction analysis and DNA sequencing. The Val and Phe residues of the putative PP1 binding motif (residues 393 and 395) were changed to Ala using two-step recombinant PCR mutagenesis. Polyclonal antibodies were raised in rabbits against Repo-Man peptides () and affinity purified using the cognate peptides (Eurogentec). Anti–lamin A/C, anti-PP1α, and anti-PP1γ antibodies were obtained from Santa Cruz Biotechnology, Inc., all HRP-conjugated secondary antibodies were obtained from Perbio Science, and microcystin-Sepharose was obtained from Upstate Biotechnology. All FP-PP1 constructs and the HeLa, HeLa, and HeLa stable cell lines were obtained as described previously (, ). The HeLa stable cell line was established in a similar manner. Characterization of expressed FP-PP1 and FACS analyses were performed as previously described (). Cells were cultured in glass-bottomed dishes (WILLCO; Intracel) and mounted on a wide-field fluorescence microscope (DeltaVision Spectris; Applied Precision) fitted with an environmental chamber (Solent Scientific) to maintain temperature at 37°C and a CoolMax charge-coupled device camera (Roper Scientific). Before imaging, growth medium was replaced with Phenol Red-free CO independent medium (Invitrogen). DNA was stained by incubating the cells for 30 min in medium containing 0.25 μg/ml Hoechst No. 33342 (Sigma-Aldrich). Cells were imaged using a 60× NA 1.4 Plan-Apochromat objective. For time-lapse imaging, five optical sections of 1 μM each were recorded every 3 min, with an exposure time of 0.1 s for EGFP and 0.05 s for Hoechst 33342. DIC imaging was obtained with the appropriate prism insert (Olympus). All fluorescence time-lapse images are presented as two-dimensional projections of the three-dimensional datasets. For high-resolution imaging, either 20 (interphase cells) or 40 (mitotic cells) optical sections of 0.5 μM each were collected, with exposure times of 0.5 s for EGFP. Similar settings were used to image ECFP- and EYFP-tagged proteins, using the appropriate filter sets (Chroma Technology Corp.). SoftWorX software (Applied Precision) was used for image acquisition, data deconvolution, and volume rendering. Fluorescence lifetime imaging microscopy/FRET experiments were performed on a fluorescent lifetime imaging system (Radiance 2100MP; Bio-Rad Laboratories). Cells were grown for six cell divisions in -arginine (HeLa), -arginine C N (HeLa), or -arginine C N (HeLa) labeling media before purification. Nuclei were isolated from cells using a variation of a previously described technique (). Purified nuclei were resuspended in RIPA buffer to solubilize proteins. Lysates from each cell line were mixed in a 1:1:1 ratio based on total protein concentration, and FP proteins were affinity purified using anti-GFP monoclonal antibodies (Roche) covalently coupled to protein G–Sepharose, as previously described (). For the peptide elution experiment, the beads were incubated with an RVXF peptide from ZAP3 (sequence GKKRVRWADLE) to elute PP1-interacting proteins. Immunoprecipitated proteins were separated on NuPAGE 4–12% Bis-Tris gels and excised into 12 slices. Peptides resulting from in-gel digestion were extracted from the gel pieces, desalted, concentrated on reverse phase–C18 tips, and eluted into 96-well plates for automated mass spectrometric analysis. For some experiments, the FPs from each nuclear lysate were affinity purified separately and subjected to one-dimensional SDS-PAGE, followed by silver staining or immunoblotting. For the far Western assays using S-PP1, the substrate was prepared by in vitro translating PP1γ (in a pcDNA vector) in the presence of [S]methionine using the TNT T7 quick coupled transcription/translation system (Promega). Blots were blocked overnight with 10% milk powder in PBS, washed well with PBS, and incubated with S-PP1 for 4 h. Additional PBS washes were performed before exposure to x-ray film to detect S-PP1–labeled bands. High-resolution mass spectrometric analysis was performed as described previously () using the LTQ-FT-ICR mass spectrometer (Thermo Finnigan). Protein ratios were calculated for each arginine-containing peptide as the peak area of [C]Arg divided by the peak area of [C]Arg and the peak area of [C]/[N]Arg divided by the peak area of [C]Arg for each single scan mass spectrum. Peptide ratios for all arginine-containing peptides sequenced for each protein were averaged. MS-Quant, a software program developed in house, was used to extract information from the Mascot HTML database search files (Matrix Science) and to evaluate the certainty in peptide identification and in peptide abundance ratio. The program is available as open source at . HeLa cells were transfected in 24-well plates with 1 μg of small interfering RNA (siRNA) duplexes (Dharmacon) using RNAiFect (QIAGEN) and either harvested or passaged 24 h after transfection. For the soft agar assays, cells were grown and transfected with siRNA duplexes as described in the previous section and then trypsinized and mixed at a dilution of 4e cell/ml with 0.3% agar in growth medium. This mixture was carefully pipetted onto 2 ml of base agar (0.5% agar in growth medium) in a 6-well plate at 0.5 ml/well and incubated at 37°C/5% CO for 2 wk. Table S1 contains a comprehensive list of all proteins identified in the SILAC/proteomic screen. Fig. S1 shows a Clustal alignment of Repo-Man protein sequences from several vertebrates. Fig. S2 shows Repo-Man detected by Western blotting, mitotic defects associated with its overexpression, and analysis of its turnover kinetics throughout the cell cycle. Fig. S3 shows DNA fragmentation induced by RNAi knockdown of Repo-Man and displacement of PP1γ from chromatin by overexpression of RAXA mutant Repo-Man. Video 1 shows a mitotic HeLa cell. Video 2 shows a mitotic HeLa cell. Video 3 shows FP-Q69YH5 transiently expressed in a mitotic HeLa cell stained with Hoechst 33342 to mark chromatin. Video 4 shows FP-PP1γ in a mitotic HeLa stained with Hoechst 33342 to mark chromatin. Video 5 shows a mitotic HeLa cell transiently expressing ECFP–Repo-Man. Video 6 shows a mitotic HeLa cell transiently expressing monomeric dsRed–Repo-Man. Video 7 shows a mitotic HeLa cell transiently expressing RAXA mutant monomeric dsRed–Repo-Man. Video 8 shows a field of HeLa cells mock-transfected and imaged 8–40 h after transfection. Video 9 shows a field of HeLa cells transfected with RAXA mutant FP–Repo-Man and imaged 8–40 h after transfection. Online supplemental material is available at .

Phosphoinositides are required in cellular events such as signal transduction, cytoskeletal rearrangements, and membrane trafficking. Phosphoinositides function by recruiting protein effectors to specific membrane domains. Most phosphoinositides bind to multiple effectors. This enables cells to control multiple functions simultaneously by regulating the levels of a single phosphoinositide isomer. Therefore, tight control of phosphoinositide levels is critical for the ability of a cell to orchestrate diverse signaling pathways. The most recently discovered phosphoinositide, PI3,5P (phosphatidylinositol [PI] 3,5-bisphosphate), is required for signaling pathways that invoke a response within the endomembrane system. For example, PI3,5P is important for insulin-induced trafficking of the glucose transporter GLUT4 to the plasma membrane (; ). In addition, increases in PI3,5P levels accompany a cellular response to UV irradiation () and EGF stimulation (). A similar increase in PI3,5P is observed in yeast and plant cells in response to acute increases in environmental osmolarity (; ; ). PI3,5P levels are also important for constitutive membrane traffic to the lysosome. Mammalian EGF receptor trafficking to the lysosome is impaired in cells that overexpress the PI3,5P 3-phosphatase mytotubularin (). In addition, secrete a phosphoinositide phosphatase (SopB) that decreases host PI3,5P levels () and diverts maturing phagosomes away from the lysosomal degradation pathway (; ). In , the levels of PI3,5P under normal conditions are low (; )—20-fold lower than the other detectable phosphoinositides PI3P, PI4P, and PI4,5P (). Yeast mutants that cannot produce PI3,5P have grossly enlarged vacuoles. Moreover, the vacuoles are not acidified properly despite the proper localization of the vacuolar ATPase (; ). Furthermore, some proteins that are normally transported to the lumen of the vacuole through multivesicular bodies are mislocalized to the limiting membrane of the vacuole (; ; ). Vesicular traffic out of the vacuole is also compromised in the absence of PI3,5P (; ). These observations demonstrate that PI3,5P controls multiple vacuolar functions. The low levels of PI3,5P in increase dramatically when cells are acutely exposed to hyperosmotic conditions (). Within 5 min, there is a 20-fold increase in PI3,5P levels (). These elevated levels are briefly maintained, but by 30 min, PI3,5P levels return to their normal, low levels. It is likely that the transient increase in PI3,5P provides protection against osmotic changes in the environment. For example, vacuole volume decreases dramatically as PI3,5P levels increase (; ). PI3,5P may also regulate channels that release ions, other osmolytes, and water from the vacuole into the cytoplasm (). In , Fab1p is the sole kinase that generates PI3,5P from a pool of PI3P (). Overexpression of does not lead to an increase in PI3,5P levels (). This suggests that Fab1p activity is tightly regulated. In fact, two putative activators of Fab1p have been identified: Vac7p and Vac14p. Like Δ cells, Δ and Δ cells have no detectable PI3,5P under basal conditions. Moreover, in response to hyperosmotic shock, Δ and Δ cells are defective in producing the 20-fold increase in PI3,5P levels observed in wild-type cells (; ; ). Furthermore, human Vac14 was recently shown to physically associate with mammalian Fab1 (PIKfyve), and overexpression of human Vac14 increased the PI3P 5-kinase activity of PIKfyve (). These observations support the hypothesis that Vac7p and Vac14p regulate PI3,5P levels through the activation of Fab1p. Neither Vac7p nor Vac14p have identifiable conserved domains, and it is not yet known how they regulate PI3,5P levels. Fig4p is a known PI3,5P 5-phosphatase (). Thus, our recent observation that Δ cells are defective in the hyperosmotic shock–induced increase in PI3,5P levels was surprising (). It raised the possibility that Fig4p also functions as an activator of Fab1p. Moreover, Fig4p forms a complex with Vac14p (; ; ), suggesting that Vac14p may regulate Fig4p or that Fig4p may regulate Vac14p. Therefore, we sought to determine the roles of Vac7p, Vac14p, and Fig4p in regulation of the lipid kinase activity of Fab1p and determine the potential roles of Vac14p and Vac7p in regulation of the lipid phosphatase activity of Fig4p. In this study, we identify specific mutations in two distinct regions of Fab1p that enhance its activity. The resulting hyperactive Fab1p mutants produce elevated levels of PI3,5P under basal conditions and respond normally to hyperosmotic shock when Vac7p, Vac14p, and Fig4p are present. In their absence, the Fab1p mutants are still able to produce detectable levels of PI3,5P but have lost their ability to respond to hyperosmotic shock. These hyperactive Fab1p mutants allowed us to individually test the roles of the putative Fab1p activators in the regulation of PI3,5P levels. We found that Vac7p and Vac14p independently regulate PI3,5P levels. We also show that Vac14p is absolutely required for Fig4p-mediated turnover of PI3,5P after hyperosmotic shock. Lastly, we show that Fig4p point mutants defective in PI3,5P turnover are simultaneously defective in hyperosmotic shock–induced elevation of PI3,5P levels, suggesting a direct role for Fig4p in both PI3,5P synthesis and turnover. The ability of Fig4p and Vac14p to act in both the synthesis and turnover of PI3,5P may be key to the ability of cells to acutely regulate transient changes in PI3,5P levels. In vivo, Fab1p cannot produce PI3,5P in the absence of either Vac7p or Vac14p. Furthermore, in the absence of Fig4p, there is a dramatic defect in PI3,5P elevation after hyperosmotic shock. Therefore, to analyze the roles of Vac7p, Vac14p, and Fig4p as potential activators of Fab1p, Fab1p mutants that function in the absence of these proteins were sought. These mutants would enable us to study the individual roles of Vac7p, Vac14p, and Fig4p in controlling PI3,5P levels and may also reveal regulatory regions within the Fab1p protein. A Fab1p mutant that is partially active in the absence of Vac7p had been previously identified () but had multiple mutations that were distributed throughout two thirds of the protein. Thus, we isolated new Fab1p mutants with single amino acid changes. Random mutagenesis was performed on the region of that encodes the kinase domain (<30% of the total gene; ). constructs subjected to mutagenesis were transformed into ΔΔ cells (). Mutant alleles that exhibited wild-type growth rates at 37°C and suppressed the large vacuole volume of Δ cells were sequenced (). Two of these mutant alleles, (T2250A) and (G2238W), were each the result of a single amino acid substitution of highly conserved residues in the COOH-terminal end of the kinase domain. Notably, all previously reported kinase domain mutations had a deleterious effect on Fab1p function (; ). A second type of informative mutant (e.g., ) had mutations exclusively in the region proximal to the kinase domain. Both sets of mutants also displayed reduced vacuole volume in Δ, wild-type, and, to a lesser extent, Δ cells (). The identification of these mutants supports a hypothesis that specific residues in the kinase domain and the region proximal to the kinase domain are involved in the up-regulation of Fab1p kinase activity. The Fab1p mutants identified above restored normal vacuole volume in Δ cells but had much less effect in Δ cells (). There are two possibilities as to why the mutants suppressed Δ better than Δ mutants. First, Δ cells have a greater defect in the production of PI3,5P. Therefore, the basal PI3P 5-kinase activity of these Fab1p mutants may be increased enough to restore normal PI3,5P levels and vacuole volume in Δ cells but not in Δ cells. Alternatively, the mutants may specifically bypass Vac14p function but not Vac7p function. Therefore, we performed an additional screen for Fab1p mutants that could restore normal vacuole volume in Δ cells. We randomly mutagenized to ensure that new mutations would be obtained. New mutations might specifically bypass Δ defects or may be more general and bypass Δ defects while further bypassing Δ defects. The screen yielded mutants in this latter category. The mutants contained the original kinase domain mutation plus mutations in the region proximal to the kinase domain (). These Fab1p mutants led to a decrease in vacuole volume in both Δ and Δ cells as well as wild-type and Δ cells. This suggests that specific mutations in either domain result in Fab1p molecules with elevated kinase activity. Mutations in each domain were additive. For example, the most active mutant () contained mutations in both the kinase domain and in the region proximal to the kinase domain. The kinase domain mutation (T2250A) was identical to that of . The proximal region mutated residues (E1822V/F1833L) were nearly identical to those of a proximal region mutant, (E1822K/N1832Y). mutant and a mutant () had identical suppression of Δ and Δ vacuole volume defects (not depicted). In addition, suppression of vacuole volume defects by was equal to that of and combined (). The ability of isolated Fab1p mutants to suppress vacuole volume defects of Δ and Δ cells suggests that these Fab1p mutants produce elevated levels of PI3,5P. Indeed, the (0.24 units) and (0.68 units) mutants had significantly higher PI3,5P levels than wild-type (0.04 units; ). The observation that results in three times more basal PI3,5P than likely explains why suppresses both Δ and Δ vacuole volume defects, whereas suppresses only Δ. In the presence of , basal levels of PI3,5P in Δ cells are similar to wild type (; ), whereas basal levels of PI3,5P in Δ or Δ cells are significantly lower (). These relative differences between the strains were also present when either or was expressed as the sole copy of in Δ, Δ, or Δ cells (). In each case, substitution of with resulted in a modest increase in steady-state levels of PI3,5P, whereas substitution of with resulted in a larger increase in PI3,5P levels. These findings suggest that neither mutant specifically bypasses the requirement for Vac7p, Vac14p, or Fig4p under basal conditions. Although the and alleles result in increased Fab1p PI3P 5-kinase activity, our data do not indicate whether the increased activity is a result of changes in the catalytic activity of Fab1p itself or whether the mutations affect modulators of Fab1p activity. For example, the Fab1p mutants presented in this study may have increased PI3P 5-kinase activity because of a reduction in negative regulation by Atg18p. The observation that Δ cells have extremely high levels of PI3,5P led to the proposal that Atg18p negatively regulates Fab1p (; ). Furthermore, Atg18p has also been reported to bind Fab1p in the region proximal to the kinase domain, which is the site of mutations in (). We tested whether the or mutants produced even higher levels of PI3,5P after hyperosmotic shock. Indeed, in a wild-type background, both mutants responded to hyperosmotic shock (). Unexpectedly, the mutants produced the same level of PI3,5P as wild-type . This was surprising because under basal conditions, the and alleles had elevated levels of PI3,5P compared with wild-type . Perhaps these Fab1p mutants are already partially activated under basal conditions and cannot be further activated beyond a specific maximum. Alternatively, the regulation of Fab1p under basal conditions may be different than the activation of Fab1p after hyperosmotic shock. Another alternative is that there may be regulated inhibition of Fab1p activity to maintain specific levels of PI3,5P at specific times. Along similar lines, perhaps regulated turnover of PI3,5P prevents the levels from exceeding a maximum. It should also be noted that the levels of PI4P and PI4,5P did not change significantly with the expression of or when compared with (not depicted). In the absence of Vac7p, Vac14p, or Fig4p, neither nor achieved the normal levels of PI3,5P induced by hyperosmotic shock (). Furthermore, the relative differences in hyperosmotic shock–induced levels of PI3,5P observed when or were expressed in Δ, Δ, or Δ mutants parallel those observed under basal conditions. Specifically, when these mutant alleles are expressed in Δ cells, they are most defective in the elevation of PI3,5P levels; when expressed in Δ cells, they are moderately defective; and when expressed in Δ cells, they are least defective in the elevation of PI3,5P levels. A previous study showed that mouse Fab1 (PIKfyve) PI3P 5-kinase activity is increased upon its dephosphorylation (). Based on this information, we tested the hypothesis that the allele () is more active because it prevents a possible inhibitory phosphorylation at the conserved Thr2250 residue. mutant to mimic inhibitory phosphorylation and observed that the mutant had elevated levels of PI3,5P that were almost identical to that of the mutant before and after hyperosmotic shock (unpublished data). Therefore, it is unlikely that the threonine at position 2,250 regulates the activity of Fab1p via its phosphorylation/dephosphorylation. A Δ strain containing the hyperactive mutant exhibited a small increase in PI3,5P levels after hyperosmotic shock (). Likewise, when the mutant was integrated into the chromosomal locus of a Δ strain, hyperosmotic shock induced a modest increase in PI3,5P levels (0.08–0.12 units) and a corresponding decrease in vacuole volume (). These results suggest that a second mechanism of elevating PI3,5P exists that is independent of Vac7p. This other mechanism may involve the Vac14p–Fig4p complex. To test this possibility, we simultaneously deleted Vac7p, Vac14p, and Fig4p and measured the levels of PI3,5P. The ΔΔΔ strain produced low but detectable levels of PI3,5P under basal conditions (). This is in contrast with wild-type , where PI3,5P cannot be detected in the absence of Vac7p, Vac14p, and Fig4p. Notably, hyperosmotic shock of the ΔΔΔ strain did not induce an increase in the levels of PI3,5P or a decrease in vacuole volume (). These observations strongly suggest that a hyperosmotic shock–induced increase in the levels of PI3,5P requires Vac7p, Vac14p, and/or Fig4p. To determine whether Vac7p, Vac14p, or Fig4p alone elicit an increase in the levels of PI3,5P, we individually expressed each gene in the ΔΔΔ strain. Because the localization of Vac14p and Fig4p are partially dependent on each other (; ), each gene was overexpressed to mitigate potential problems in their localization. Overexpression of in the ΔΔΔ strain had no effect on the levels of PI3,5P before or after hyperosmotic shock (). This indicates that Fig4p alone cannot elevate the levels of PI3,5P. Overexpression of resulted in a 4.5-fold increase in the basal levels of PI3,5P. This suggests that Vac14p can elevate levels of PI3,5P in the absence of Vac7p and Fig4p. After hyperosmotic shock, there was no further increase in PI3,5P levels. Overexpression of resulted in a 4.5-fold increase in the basal levels of PI3,5P. Moreover, after hyperosmotic shock, there was a further increase in PI3,5P levels. This suggests that Vac7p can elevate levels of PI3,5P in the absence of Vac14p and Fig4p. Moreover, it also suggests that the majority of the hyperosmotic shock–induced elevation of PI3,5P levels occurs via Vac7p. To test whether the potential regulators act in concert with each other, we overexpressed pairwise combinations of , , and in the ΔΔΔ strain. When and were overexpressed simultaneously, the basal levels of PI3,5P were lower than when was expressed alone (). This is likely caused by the ability of the Vac14p–Fig4p complex to turn over PI3,5P under basal conditions. Upon hyperosmotic shock, the levels of PI3,5P increased. This is in contrast to the overexpression of Vac14p or Fig4p alone where hyperosmotic shock had no effect on the levels of PI3,5P (). This finding suggests that the Vac14p–Fig4p complex plays a role in the hyperosmotic shock–induced increase in PI3,5P levels. This postulate is also supported by the observation of a Vac7p-independent hyperosmotic shock–induced increase in PI3,5P levels in Δ cells (). Analysis of a Fig4p point mutant is also consistent with a role of the Vac14p–Fig4p complex in the hyperosmotic shock–induced increase in PI3,5P levels (see the last section of Results). When and were overexpressed simultaneously, the levels of PI3,5P were the sum of the levels observed with or alone. This is consistent with the hypothesis that Vac7p and Vac14p independently elevate the levels of PI3,5P. The highest levels of PI3,5P observed under both basal and hyperosmotic shock conditions occurs when all three regulators were overexpressed simultaneously. This further supports the hypothesis that all three proteins play important roles in the elevation of PI3,5P levels. All of the aforementioned results support a model in which Vac7p and Vac14p independently increase PI3,5P levels. Under basal conditions, Vac7p and Vac14p play an equal role. After hyperosmotic shock, Vac7p is the main protein required to elevate PI3,5P levels. In addition, after hyperosmotic shock, Fig4p is important for the elevation of PI3,5P levels and requires Vac14p for this function. Analysis of Δ cells revealed that after hyperosmotic shock, there is a delay in the turnover of PI3,5P (). This delay is similar to that observed in Δ cells where the levels of PI3,5P at 30 min after hyperosmotic shock are the same or even higher than the levels achieved at 10 min. This delay in the turnover of PI3,5P can also be observed when the allele was expressed in Δ or Δ cells (). These findings clearly demonstrate that in either a Δ or Δ mutant, there is a delay in the turnover of PI3,5P. Therefore, both Vac14p and Fig4p are required for the hyperosmotic shock–induced turnover of PI3,5P. The Fig4p sequence is similar to known lipid phosphatases. Moreover, Fig4p functions in vitro as a PI3,5P 5-phosphatase (). These findings, combined with the aforementioned analysis, strongly suggests that the rapid turnover of PI3,5P after hyperosmotic shock is the result of the Fig4p-dependent conversion of PI3,5P to PI3P. The requirement for Vac14p in hyperosmotic shock–induced turnover of PI3,5P was surprising because Vac14p was first proposed to be an activator of Fab1p. More recently, Vac14p had also been shown to interact with Fig4p (; ). Here, we find that the levels of Fig4p are greatly diminished in a Δ mutant (). This suggests that Vac14p may play a role in the turnover of PI3,5P either by stabilization and/or regulation of Fig4p. The observation that hyperosmotic shock–induced PI3,5P levels eventually decrease even in the absence of Fig4p suggests that there are other pathways that degrade PI3,5P. Five other proteins (Sac1p, Sjl1p, Sjl2p, Sjl3p, and Inp54p) are predicted to have phosphoinositide 5-phosphatase activity. However, defects in PI3,5P turnover have not been observed in the corresponding knockout strains. Only when multiple genes are knocked out are elevated PI3,5P levels observed (; ; ). Furthermore, none of these phosphoinositide phosphatases localize to the vacuole membrane, which is the site of PI3,5P synthesis and function. Although it is likely that at least some of these proteins can degrade PI3,5P, none of them are likely to have a primary role in the rapid turnover of PI3,5P after hyperosmotic shock. The Δ strain produced very little PI3,5P in response to hyperosmotic shock, but these levels decreased at 30 min (). This suggests that Vac7p is not required for the rapid turnover of PI3,5P. We used the ΔΔΔ strain to further test the function of Vac14p and Fig4p in the hyperosmotic shock–induced turnover of PI3,5P. Simultaneous overexpression of , , and in the ΔΔΔ strain resulted in the hyperosmotic shock–induced turnover of PI3,5P (). Overexpression of alone resulted in no turnover of PI3,5P (). Similarly, overexpression of alone resulted in no turnover of PI3,5P. However, simultaneous overexpression of and resulted in the turnover of PI3,5P. This supports the hypothesis that the Vac14p–Fig4p complex functions in the turnover of PI3,5P after hyperosmotic shock. When was overexpressed alone in the ΔΔΔ strain, a modest decrease in PI3,5P levels was observed (). However, the amount of turnover (0.19 units) was less than half the amount (0.41 units) observed when , , and were overexpressed (). The observed turnover is likely the result of a Fig4p-independent mechanism (discussed in the previous section). Simultaneous overexpression of and had a rate of PI3,5P turnover that was similar to alone. Simultaneous overexpression of and also had a rate of PI3,5P turnover that was similar to alone. These data suggest that Fig4p is the major enzyme responsible for the rapid turnover of hyperosmotic shock–induced PI3,5P levels and that Fig4p stability and function require Vac14p. Perturbation in the localization of Vac14p in a Δ strain (; ) makes it difficult to directly assess a role for Fig4p in the hyperosmotic shock–induced elevation and turnover of PI3,5P. Therefore, we used a Fig4p point mutant in which the localization of Vac14p is normal (). point mutant () is defective in PI3,5P 5-phosphatase activity in vitro (). mutant. mutant are elevated (). This observation is consistent with a defect in PI3,5P 5-phosphatase activity. appears to retain some lipid phosphatase activity. cells have degraded 0.5 units of PI3,5P compared with 0.0 units in Δ cells () and 1.3 units in cells. In an attempt to produce a Fig4p mutant with no phosphatase activity, we generated a second Fig4p mutant, . Based on a similar mutation in Sac1p (), the mutant was predicted to have no lipid phosphatase activity. Moreover, the D469 residue is conserved in all Sac1 domain–containing proteins, and the residue lies within the conserved CXRT catalytic site of lipid () and protein () phosphatases. mutant (unpublished data). Both mutants had elevated levels of PI3,5P under basal conditions, a decreased rate of hyperosmotic shock–induced turnover of PI3,5P, and a defect in the elevation of PI3,5P in response to hyperosmotic shock (∼65% of wild type). Notably, the extent of the partial defect observed in PI3,5P turnover parallels the extent of the partial defect in PI3,5P elevation. These observations suggest that the Sac1 phosphatase domain of Fig4p is important for both an increase in PI3,5P levels and their turnover. #text Strains are listed in . Strains were grown at 24°C in either YPD (yeast extract/peptone/glucose) or synthetic complete (SC) minimal media. All yeast plasmids were derived from the pRS400 series of vectors (). PCR reactions used PfuUltra HF (Stratagene). Restriction enzymes and buffers were obtained from New England Biolabs, Inc. For pRS415- , a 660-bp BsaBI–NcoI fragment from pET-15b- () was cloned into BsaBI–NcoI-gapped pRS415-. The mutant was generated by site-directed mutagenesis. A 542-bp ClaI fragment from pRS415- was cloned into pUC18. The GAC codon (aspartic acid) was changed to AAC (asparagine) using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). The ClaI fragment was cloned into a pUC18 plasmid containing the SacI fragment of . This mutated SacI fragment was then cloned into pRS415 to generate pRS415-. The mutation was confirmed with DNA sequencing. Log phase cells were labeled with FM4-64 as described previously (). Salt treatment of labeled cells was performed by resuspending cells in media containing 0.45 M NaCl, which elicits the same changes in vacuole volume as 0.9 M NaCl. A large change in refractive index when using 0.9 M NaCl makes it difficult to achieve focused images on the microscope. Cells were viewed at room temperature on a microscope (Axioplan II; Carl Zeiss MicroImaging, Inc.) with a 100× NA 1.40 ∞/0.17 plan Apochromat objective using a TRITC filter. Images were captured with a camera (RT-Spot; Diagnostic Instruments) and analyzed with MetaMorph software (Universal Imaging). Experiments were performed as described previously (). Log phase cells were lysed in cytosol buffer, and the extracts were spun at 300 to sediment unbroken cells. The supernatant was spun at 13,000 for 10 min at 4°C. The supernatant (S13) was moved to a fresh tube, and the pellet (P13) was resuspended in cytosol buffer in an equal volume to the supernatant. Fractions were subjected to SDS-PAGE and transferred to nitrocellulose for Western blot analysis. Blots were probed with rabbit anti-Fig4p sera or goat anti-Vac14p sera (). HRP-conjugated secondary antibodies (Invitrogen) followed by SuperSignal Chemiluminescence (Pierce Chemical Co.) was used to visualize immunoreactive proteins. was PCR amplified from pRS416-, forward primer CTCAAACCGAAGAACGTCCG, reverse primer GGCACCCCAGGCTTTACAC (in pRS416), and standard Taq polymerase conditions (Boehringer). This produced a 3.5-kbp product that included 2,053 bp (encoding aa 1,595–2,278). The PCR product was cotransformed into a ΔΔ strain with NheI–NotI-gapped pRS416- and transformants selected on SC-uracil plates. Transformants were replica plated onto SC-uracil plates containing 100 μM of the iron chelator bathophenanthroline disulfonic acid (Fluka). Wild-type colonies have normal growth rates and are white, whereas Δ, Δ, or Δ cells are red. After a 36-h incubation at 37°C, large white colonies were selected as Δ bypass candidates. Their ability to bypass the large vacuole morphology of Δ cells was verified with FM4-64 labeling. Candidate plasmids were isolated and retransformed into ΔΔ cells to confirm that a plasmid encoded a hyperactive Fab1p mutant. The mutagenized regions were sequenced. The same approach was used to screen for mutants that bypass Δ, except pRS416- was used as the starting PCR template. The kanamycin cassette integrated along with the mutations was PCR amplified from the yeast haploid knockout collection (Open Biosystems) using primers that had 45 bp of flanking sequence that would anneal to either the 3′ end of or the 3′ untranslated region of . This PCR product was transformed into wild-type yeast along with BsrGI-digested (partial) pRS416- or pRS416- and transformants selected on YPD supplemented with gentamycin. Plasmid preps from candidate transformants were performed, and the pRS416- cassettes were verified by restriction digest and DNA sequence. Positive plasmids were digested with SwaI–NotI, and the resulting 3.5-kbp fragment was transformed into Δ (LWY2054) and transformants selected on YPD supplemented with gentamycin. Proper integration of the mutants was verified with PCR. Generation of the ΔΔΔ mutant (LWY7197) was performed by mating ΔΔ (LWY6538) with Δ (LWY7179). Diploids were sporulated, and genotypes of the resultant tetrads were determined by the presence or absence of expected nutritional markers and PCR. PI was extracted and analyzed as described previously (). In brief, total PI was extracted from [H]inositol-labeled cells and deacylated with methylamine. The inositol head groups were then separated with HPLC, and radioactive counts were measured. Values for PI3,5P are reported as the percentage of total PI extracted from each strain to normalize for [H]inositol incorporation, the number of cells, and the amount of extracted inositol injected into HPLC.

The endosomal system coordinates intracellular protein trafficking between several subcellular compartments, including the Golgi, plasma membrane, and lysosome. A variety of sorting reactions confer distinct outcomes onto endosomal membrane proteins: recycling to an earlier compartment, residence within a compartment, or degradation in the lysosome. Delivery of endosomal membrane proteins to the lumen of the hydrolytic lysosome is the result of a sorting event that occurs during the formation of multivesicular bodies (MVBs; for reviews see ; ; ). MVB formation occurs when the limiting membrane of the endosome invaginates and buds into its own lumen, bringing with it a subset of the proteins residing therein, including endocytosed, activated cell surface receptors. Heterotypic fusion of the MVB with the lysosome results in the delivery of the cargo-containing intralumenal MVB vesicles to the degradative activities sequestered within the lysosomal lumen. The sorting of proteins to the intralumenal vesicles of an MVB is not limited to endocytic cargoes. A subset of cargo directly sent from the Golgi apparatus is actively sorted into these vesicles without transiting the cell surface, typically referred to as biosynthetic MVB cargo (for review see ). In addition to facilitating delivery of proteins to the lysosome for degradation or processing, the MVB machinery participates in the budding of retroviruses from the host cell (for review see ). Thus, understanding the mechanisms governing MVB sorting is important with regards to processes as diverse as the modulation of cell surface–receptor signaling and the budding of retroviruses. For cellular homeostasis to be maintained, entry into the degradative MVB pathway must be tightly regulated, and the machinery mediating this sorting event is highly conserved between yeast and man (for reviews see ; ; ). Recent progress toward our understanding of this process has been greatly facilitated by genetic and biochemical studies in yeast. Posttranslational modification of endosomal proteins with ubiquitin serves as a signal for their inclusion into intralumenal vesicles during MVB formation (for review see ). Endosomal membrane proteins that have been covalently modified with ubiquitin are sorted into the MVB pathway via the action of the class E vacuolar protein sorting (Vps) proteins. The majority of the class E Vps proteins are subunits of protein complexes called ESCRTs (endosomal sorting complexes required for transport; for reviews see ; ; ). Class E mutants display the phenotype of missorting MVB cargoes to the limiting membrane of the vacuole and the class E compartment rather than the vacuolar lumen, indicating a requisite role for these factors in the function of the MVB pathway (). The biochemical activities of the ESCRTs and their accessory proteins have begun to be characterized. At present, the best defined are the cargo-recognition activities of Vps27/Hrs and ESCRT-I, both of which bind ubiquitinated cargoes and direct them into the MVB pathway in a coordinated manner (; ; ). Genetic evidence places ESCRT-II function downstream of ESCRT-I, where it plays a critical role in the assembly of ESCRT-III (). Appropriate ESCRT-III assembly is required for the recruitment of additional factors involved in the deubiquitination of MVB cargoes before their inclusion into intralumenal vesicles (e.g., Doa4 and Bro1; ; ). Additionally, ESCRT-III recruits the AAA-ATPase (ATPase associated with a variety of cellular activities) Vps4 to the MVB sorting site for dissociation of the ESCRTs at a step late in the sorting reaction (). Loss of Vps4 ATPase activity results in an accumulation of ESCRT proteins on endosomal compartments and concurrent loss of MVB sorting (, ). Although ESCRT-III has been shown to recruit Vps4 to the site of MVB sorting, no regulators of its ATPase activity have been described to date. Vps4 is the only class E Vps protein that has a readily measurable enzymatic activity, the hydrolysis of ATP. In the absence of ATP, Vps4 exist as a dimer. ATP binding by Vps4 results in its homooligomerization into a complex of ∼440 kD. ATP hydrolysis by the Vps4 oligomer, required for the release of the ESCRTs from the endosomal membrane, returns Vps4 to its inactive dimeric state (). Spatial and temporal regulation of this Vps4 cycle is critical for the function of the MVB sorting reaction, as this represents a potential mechanism by which to modulate flux through the pathway. We provide evidence that Vta1 is a positive regulator of Vps4 in vivo. This finding is supported by biochemical analyses indicating Vta1 directly binds to Vps4 and stimulates its ATPase activity in vitro. Additionally, we define a 30-residue COOH-terminal sequence highly conserved in yeast Vta1, mammalian SBP1 (SKD1 binding protein 1), and plant LIP5 as a new region (Vta1/SBP1/LIP5 [VSL]) necessary and partially sufficient for stimulation of Vps4 ATPase activity. These findings implicate Vta1 as a novel regulatory component of the MVB machinery that contributes to the temporal and spatial activation of Vps4. Conservation of the VSL region suggests that this is an evolutionarily conserved mechanism by which Vta1, SBP1, and LIP5 regulate the MVB sorting reaction. The () ORF (YLR181c) was identified as required for proper function of the MVB pathway during the course of genetic analyses using the Research Genetics (BY4742) deletion collection ( and Fig. S1, available at ; unpublished data). Consistent with this observation, two recent reports have implicated Vta1 in the function of the MVB pathway, but the function of this protein was not elucidated (; ). To better define the role Vta1 plays in the MVB pathway, endosomal MVB cargoes originating from either the Golgi or plasma membrane were analyzed in cells lacking Vta1 () as compared with wild-type cells or cells lacking the class E protein Vps4 (; and Fig. S1). Generation of deletion strain in our standard genetic background (SEY6210) revealed a significantly weaker phenotype than was observed in the BY4742 background. For this reason, MVB sorting phenotypes were examined in SEY6210, BY4742, and SF838-9D genetic backgrounds to document any variation in the phenotype. The proper sorting of GFP-tagged MVB cargo proteins is scored by their localization within the lumenal space of the vacuole, as opposed to its limiting membrane. Using the biosynthetic MVB cargo GFP-tagged carboxypeptidase S (CPS) to visualize this process, the vast majority of the fusion protein was localized to the vacuole lumen in wild-type cells (). In contrast to cells lacking Vps4, the MVB sorting pathway was only partially defective in cells lacking Vta1, as indicated by the presence of GFP-CPS in the lumen as well as the limiting membrane of the vacuole. In addition, cells lacked the characteristic class E compartment that was readily observed in cells ( and Fig. S1 A, arrowheads). One possible explanation for the defects seen in CPS sorting would be a defect in its ubiquitin modification. This was addressed by immunoprecipitation of CPS from wild-type and cells, followed by anti-ubiquitin Western blotting. This analysis revealed no defect in ubiquitin modification of CPS in cells (unpublished data). The trafficking of an endocytic MVB cargo (Ste3-GFP) or an ubiquitin-independent biosynthetic MVB cargo (Sna3-GFP) was also examined. As expected, both Ste3- and Sna3-GFP delivery to the vacuole lumen was perturbed in a strain (Fig. S1). Depending on the genetic background, the severity of the phenotype observed with loss of Vta1 function varied; cells in the SEY6210 background displayed a phenotype largely indistinguishable from wild type, whereas the other backgrounds displayed more severe missorting phenotypes (Fig. S1 A). Additional tests were performed to address hallmarks of the class E phenotype, including secretion of carboxypeptidase Y and the processing of CPS to its mature form; these analyses again revealed a distinct difference between and cells, with cells exhibiting wild-type to intermediate phenotypes (Fig. S1). These results indicate that endosomal function in general, and MVB sorting in particular, is not dramatically perturbed in the mutant in all genetic backgrounds, in contrast to previously characterized class E mutants. This suggests that Vta1 is likely not a core component of the class E Vps machinery but rather represents a potential modulator of ESCRT function. In vitro, Vta1 has been demonstrated to bind the AAA-ATPase Vps4 (). Similarly, the murine homologue of Vta1, SBP1, has also been shown to bind the murine homologue of Vps4, SKD1 (). These observations, together with the Vta1 localization data presented in , led us to examine whether Vta1 impacted the Vps4-dependent MVB sorting reaction in vivo. To uncover a synthetic genetic interaction, was combined with a previously described temperature-sensitive allele of Vps4 () and the trafficking of Ste2-GFP was analyzed in the double mutant as well as in and single-mutant cells at permissive temperature. Ste2-GFP was delivered to the lumen of the vacuole in both and cells (), an indication of efficient function of the MVB pathway. double mutant accumulated Ste2-GFP in exaggerated endosomal compartments with no obvious vacuolar signal, indicative of defective MVB sorting. The synthetic genetic interaction suggests that Vta1 acts with Vps4 in the MVB pathway in vivo. The endosomal association of ESCRT-III subunits (Vps20, Snf7, Vps2, and Vps24) is exquisitely sensitive to the activity of Vps4. In wild-type cells, only 30% (or less) of these subunits are endosome associated because of the ESCRT-III disassembly activity of Vps4. In cells lacking Vps4, >90% of ESCRT-III subunits are localized to endosomal membranes (). To determine whether Vta1 and Vps4 act together at a common point in the MVB pathway, the impact of Vta1 function on ESCRT-III membrane association was examined. cells, and the resulting fractions were probed for the ESCRT-III subunits Snf7 and Vps24 (). Consistent with previous results, Snf7 and Vps24 were largely soluble in wild-type lysates and largely membrane associated in the lysates. In contrast, cells exhibit an intermediate phenotype, with Snf7 and Vps24 equally distributed between the soluble and membrane fractions. Interestingly, the genetic backgrounds that displayed more severe MVB sorting defects also revealed increased membrane association of ESCRT-III subunits with deletion of (Fig. S1 B). The double-mutant fractionation profile is indistinguishable from the profile observed for the cells (). These findings are consistent with Vta1 stimulating ESCRT-III release via Vps4 and suggest that Vps4 function is partially compromised when Vta1 function is lost. To analyze Vta1 localization, a Vta1-GFP fusion protein expressed from an integrated gene was analyzed by fluorescence microscopy. Identification of phosphatidylinositol-3-phosphate (PtdInsP)–positive endosomal structures was facilitated by coexpressing a DsRed-FYVE (PtdInsP binding domain defined by Fab1/YDR313c/Vps27/EEA1) chimera (). Extensive colocalization of Vta1-GFP with the DsRed-FYVE reporter was observed in wild-type cells, with an additional cytoplasmic signal (). This colocalization indicates that Vta1 associates primarily with endosomal membranes, consistent with previous interpretations () and supportive of a role for Vta1 in MVB protein sorting. All known class E Vps proteins accumulate on the class E compartment in cells, as Vps4 is required to dissociate the ESCRT complexes (, ; ,). Surprisingly, although cells displayed aberrant class E structures (visualized with the DsRed-FYVE chimera), Vta1-GFP did not display obvious colocalization with these structures (). Instead, Vta1-GFP displayed an increased cytoplasmic distribution and membrane associations that were distinct from the class E compartment. To more quantitatively address this altered localization, subcellular fractionation was performed with cells expressing a functional, chromosomally integrated HA-tagged form of Vta1. Fractions were probed with anti-HA or marker antibodies, and immunoreactive species were quantitated. In wild-type cells, ∼40% of the Vta1 is present in the membrane fraction (). In contrast, Vta1-HA membrane association was decreased to ∼10% in the strain (). Thus, in contrast to previously described ESCRT subunits, Vps4 appears to be required for Vta1 association with the class E compartment. At present, the nature of the structures with which Vta1-GFP associates in the cells is unclear. However, as seen in wild-type cells, a portion of the Vta1-GFP does not colocalize with DsRed-FYVE and is associated with unique structures. It would appear that loss of Vps4 function causes a shift in distribution toward these PtdInsP-negative structures. One interpretation of these observations is that Vta1 may associate with the endosomes undergoing MVB formation via Vps4 itself. To test this idea, the ATP hydrolysis–defective form of Vps4 (Vps4), previously shown to accumulate on endosomal membranes (), was used to examine the impact on Vta1-GFP endosomal localization. Visualization of cells expressing Vps4 revealed extensive colocalization of Vta1-GFP with DsRed-FYVE–positive structures (), consistent with a role for Vps4 itself in the recruitment of Vta1 to endosomal structures. Similarly, subcellular fractionation revealed that Vta1-HA membrane association was increased to ∼65% in the context of the Vps4 mutant (). These data indicate that Vps4 plays a role in appropriate Vta1 localization to the MVB sorting reaction and suggest that the recruitment of Vta1 occurs through a direct association with Vps4. In contrast, Vta1 is not required for Vps4 association with the MVB sorting machinery because Vps4 localization is indistinguishable in and cells (). Vta1 is clearly not essential for Vps4 function because cells do not phenocopy cells in all respects. These data implicate Vta1 as a positive modulator of Vps4 in vivo. To elucidate the mechanism by which Vta1 impacts Vps4 function, biochemical characterization of Vta1 was initiated. Consistent with published data, we found that both Vta1 from yeast cell extract () and bacterially expressed Vta1 elute from a gel filtration column in the range of 300 kD, in spite of a predicted molecular mass of ∼40 kD (). Together, these observations suggest that Vta1 forms a homomeric complex. The subtle difference in size observed between bacterially produced Vta1 and yeast Vta1 may be due to the sizes of the epitopes used in each case (His for bacteria vs. 3× HA for yeast); alternatively, Vta1 homooligomers may be associated with other factors in yeast extracts. Molecular mass determination by gel filtration can be misleading, as the shape of the molecule affects the measurement. Analytical ultracentrifugation was therefore used to determine the precise molecular mass of the His-Vta1 homooligomer. Varied concentrations of Vta1 protein were subjected to equilibrium centrifugation, and the resulting data indicated a native molecular mass for Vta1 of 82 kD, suggesting that Vta1 forms a dimer (theoretical molecular mass of His-Vta1 is 40.9 kD). To ensure that during the analytical centrifugation experiment Vta1 did not change its oligomeric state, gel filtration was performed after equilibrium centrifugation. The Vta1 sample again eluted in a mass range of 300 kD, demonstrating that the Vta1 complex remained stable during centrifugation (unpublished data). Thus, the discrepancy between the gel filtration data and the ultracentrifugation data indicate that the Vta1 dimer has a very large Stoke's radius and is likely to be rod shaped. To examine the direct association of Vps4 and Vta1, purified, bacterially expressed proteins were subjected to gel filtration chromatography using a Superose 6 column in the presence of either ATP or ADP (). Vps4 bound to ATP forms a homooligomeric complex of ∼500 kD. The ATPase activity of wild-type Vps4 prevents resolution of this complex by gel filtration, requiring the use of the ATPase-defective mutant Vps4 (; ). Consistent with previous results, in the presence of ADP, Vps4 fractionates predominantly as a dimer. Dimeric Vta1 complex migrates at a slightly smaller apparent mass than the 440-kD marker and the oligomeric Vps4 complex; once again, the presumptive rodlike structure of Vta1 is likely the explanation for this unexpected fractionation pattern of the Vta1 dimer. The addition of equimolar amounts of Vta1 to the Vps4 plus ATP sample promoted formation of a Vta1–Vps4(ATP) complex that eluted at nearly 1 MD (); moreover, Coomassie blue staining of the 1-MD complex fractions revealed an apparent Vta1/Vps4 ratio of ∼1:1 (, inset; silver staining yielded identical results [not depicted]). Further addition of Vta1 (threefold molar excess) did not substantially affect the amount of complex formed, suggesting that the ∼1:1 stoichiometry represents a Vta1-saturated complex. Formation of this 1-MD complex was dependent on ATP because the Vta1 + Vps4 ADP sample did not form this high–molecular mass species (). This is in apparent contrast to previously reported results wherein an interaction between Vps4 and Vta1 was found to be independent of the nucleotide-bound state of Vps4 (). However, although we detect a weak association between Vta1 and Vps4 in GST pull-down assays in the absence of ATP, the near stoichiometric association of GST-Vta1 and Vps4 required ATP ( and Fig. S2, available at ). Although ATP is required for stoichiometric binding, the Vps4 NH-terminal region is dispensable for Vta1–Vps4 complex formation (). These observations suggested that in the presence of ATP Vta1 binds to either the AAA or COOH-terminal region Vps4, consistent with published two-hybrid data (). One interpretation of the data presented in is that Vta1 might function as an assembly factor that supports the oligomerization of ATP-bound Vps4 dimers into the active, higher order oligomeric form of Vps4. If this were true, addition of Vta1 might help assemble multimeric Vps4-ATP when it is too dilute to form complexes by itself. To test this model, we used gel filtration to analyze the size of ATP-bound Vps4 at low concentration (10-fold lower than in ) in the presence or absence of Vta1 (). Western blot analysis of the resulting fractions using anti-Vps4 antiserum revealed that Vps4 elutes from the column in the mass range of the Vps4 dimer (∼100 kD) at this low-protein concentration. In contrast, in the presence of Vta1, Vps4 is found in fractions corresponding to the 1-MD mass range, indicating that the binding of Vta1 to Vps4 stabilizes the higher oligomeric form of the ATPase. Oligomerization of Vps4 is required for the hydrolysis of the bound ATP (). Vta1 induces oligomerization of Vps4 and therefore would be predicted to increase the ATP hydrolysis rate of Vps4. To test this hypothesis, the ATPase activity of Vps4 was measured (ADP/min/Vps4 molecule) in the presence and absence of Vta1. The addition of Vta1 to Vps4 (100 nM or 1.5 μM) stimulated Vps4 ATPase activity in a concentration-dependent manner, with half-maximal stimulation at ∼2 μM Vta1 (1.6–2.8 μM; ); however, Vps4 ATPase activity stimulation was not observed with the addition of either BSA (not depicted) or certain truncated forms of Vta1 (, , and ), and Vta1 exhibited no ATPase activity by itself (not depicted). Consistent with previous results, Vps4 ATPase activity was strongly dependent on Vps4 protein concentration, with 100 nM Vps4 exhibiting activity of 5 ADP/min/Vps4 and 1.5 μM Vps4 exhibiting activity of 30 ADP/min/Vps4 (; ). The protein concentration–dependent activity of Vps4 shown in can be described by the Michaelis-Menten equation with a K of 1.1 μM, similar to the previously reported K of 0.6 μM (). The addition of either 1 or 6 μM Vta1 resulted in a decrease of the Vps4 K to 118 and 69 nM, respectively. In addition, the V increased from the previously reported 45 to 71 ADP/min/Vps4 with 6 μM Vta1. The lower Vps4 K values observed in the presence of Vta1 are consistent with the model that Vta1 promotes the oligomerization of Vps4 dimers. However, the increase in the V of Vps4 suggests that Vta1 not only promotes the oligomerization of Vps4 but may also affect other aspects of the hydrolysis kinetics. To further dissect the mechanism of Vta1 action, a structure–function analysis was initiated. The COILS program () identifies three potential coiled-coil regions in Vta1—residues 34–63, residues 231–263, and the COOH-terminal 30 amino acids (). To analyze the involvement of these regions in Vta1 activity, we generated alleles lacking residues 1–68 (Vta1), 37–68 (Vta1), 224–330 (Vta1), or 290–330 (Vta1). These alleles, as well as wild-type Vta1, were expressed in yeast as COOH-terminal fusions to PrA (IgG binding units of protein A from ). Whereas PrA-Vta1 could complement the CPS sorting defects observed in the BY4742 strain, none of the Vta1 truncations could restore function ( and not depicted), although equivalent expression was observed (). This result indicates that both the COOH- and NH-terminal coiled-coil domains are important for Vta1 function in vivo. To examine homodimerization, PrA-Vta1 alleles were expressed in yeast harboring a -GFP allele at the genomic locus. Cell lysates were generated under nondenaturing conditions, PrA-Vta1 fusions were affinity purified with IgG Sepharose and resolved by SDS-PAGE, and Vta1-GFP and PrA-Vta1 were detected by Western blotting with anti-GFP monoclonal antibody. As expected, Vta1-GFP copurified with full-length PrA-Vta1, consistent with the formation of a Vta1 homodimer (). Deletion of the last 106 or 40 residues of Vta1 was sufficient to abolish the association with Vta1-GFP. Consistent with this observation, purified Vta1 and Vta1 behaved as apparent monomers when subjected to Superose 6 gel filtration analyses ( and not depicted). Together, these results imply that the COOH-terminal 40 amino acids of Vta1 are required for homooligomerization. In contrast, the PrA-Vta1 and PrA-Vta1 proteins were competent to bind Vta1-GFP; thus, the NH-terminal coiled-coil domain is not required for Vta1 self-assembly. Vta1 has been reported to associate with the class E Vps proteins Vps60/Mos10 and Vps20 (; ). Although the gene name () implies that Vta1 binds Vps20, we have been unable to detect any interaction between Vps20 and Vta1 under a variety of conditions (unpublished data). In contrast, the association of full-length Vta1 with Vps60 was readily detectable with this PrA-based copurification approach (). Although Vta1 was unable to homodimerize, the PrA fusion protein retained the ability to bind Vps60 (). In contrast, deletion of NH-terminal residues 1–68 or 37–68 abolished the Vta1–Vps60 interaction. These findings implicate the NH-terminal coiled-coil region as mediating the Vta1 association with Vps60. To examine domains of Vta1 required for association with Vps4, the various forms of Vta1 were expressed in bacteria as NH-terminal GST fusions. GST pull-down assays were performed with purified Vps4 in the presence of 1 mM ATP and the various forms of Vta1. Vps4–Vta1 associations were analyzed by Coomassie blue staining and Western blotting (). Whereas full-length GST-Vta1 was capable of binding Vps4, deletion of the last 40 residues of Vta1 abolished interaction with Vps4. This finding indicates that COOH-terminal portion of Vta1 is required for both Vta1 dimerization and Vps4 binding, although the absence of Vps4 binding could be a secondary consequence of a defect in Vta1 dimerization. In contrast to Vta1, the NH-terminal deletions (Vta1 and Vta1) are still competent to bind Vps4 (). To further address the relevance of the Vta1–Vps4 association, the GST-Vta1 and GST-Vta1 proteins were examined for their abilities to stimulate Vps4 ATPase activity. To minimize the amount of fusion protein required, we used a Vps4 ATPase assay that included glutathione beads and the conjugated proteins to supply the reactions with the desired amount of Vta1 protein. Using this assay with Vta1 as well as GST-Vta1, we were able to demonstrate that the GST tag and the glutathione beads do not interfere with Vta1 enhancement of Vps4 ATPase activity (). The GST-Vta1 and GST-Vta1 fusion proteins were able to stimulate the ATPase activity of Vps4 to levels comparable to those of wild-type Vta1. In contrast, deletion of the COOH-terminal 40 residues abolished the ability of Vta1 to stimulate Vps4. These findings indicate that the Vta1 NH-terminal coiled-coil region is required for interaction with Vps60 but not for homodimerization, nor binding and activation of Vps4. In contrast, the COOH-terminal region (in particular, residues 290–330) is required for Vta1 homodimerization, Vps4 binding, and Vps4 activation but is dispensable for Vps60 binding. These results suggest that Vta1 binds Vps60 and -4 independently. Consistent with this model, we have observed the association of PrA-Vta1 and Vps60-HA in lysates generated from strains (unpublished data), and we have reconstituted the Vta1–Vps4 complex in vitro without Vps60 (). However, given that Vta1 truncations that fail to bind Vps60 also fail to complement (), we cannot exclude the possibility that the Vps60–Vta1 interaction modulates Vta1 stimulation of Vps4 in vivo. The Vta1 COOH-terminal 40 residues are required for Vta1 homodimerization and Vps4 binding and stimulation. Although no defined domains have been identified in Vta1, this portion of Vta1 is highly conserved in homologous proteins in yeast, plants, and mammals. In particular, the last 30 residues of Vta1, mammalian SBP1, and plant LIP5 () exhibit very significant sequence conservation. Based on these representative members, we have named this sequence motif the VSL region. The spacing of leucine and small hydrophobic residues in conjunction with the requirement of this VSL region in Vta1 dimerization suggests that the VSL region may form a coiled-coil structure, although this has not yet been confirmed experimentally. Other residues of note include the conserved charged residues at positions 2, 5, 14, 15, and 25 and the conserved aromatic residues at positions 6 and 13 (). To examine the role of this VSL region in Vps4 binding and stimulation, site-directed mutagenesis was used to alter anticipated surface residues, including Lysines 299, 302, and 322 and Serine 306. Mutants were then analyzed for function both in vivo and in vitro and revealed a range of defects. PrA-tagged forms of each mutant were expressed in cells harboring Vta1-GFP to analyze the ability of these proteins to dimerize; all four mutants were able to copurify Vta1-GFP to levels indistinguishable from wild type (unpublished data), indicating that dimerization is unaffected by these mutations. GST-tagged forms of each mutant were expressed in bacteria and purified for use in Vps4 association and activation assays. Although homodimerization was unaffected, mutation of Lysines 299 and 302 and, to a lesser extent, Serine 306 disrupted binding to Vps4 in the GST pull-down assay (). In contrast, Lysine 322 was largely dispensable for the Vta1–Vps4 interaction. In the Vps4 activity assay (), GST-Vta1 exhibited Vps4 ATPase stimulation comparable to wild type; GST-Vta1 displayed a modest reduction in Vps4 stimulation, consistent with the partial defect in Vps4 binding; and the GST-Vta1 and GST-Vta1 mutants were most compromised for Vps4 stimulation. Thus, Vps4 stimulation deficits correlated with Vps4 binding. Analysis of in vivo function also corresponded with these in vitro phenotypes. Vta1 and Vta1 exhibited the least Vps4 stimulation in vitro and similarly displayed the most severe GFP-CPS sorting and ESCRT-III recycling defects when expressed in cells (). Vta1 displayed more subtle defects in vivo, and Vta1 was indistinguishable from wild type. Together, these data demonstrate a correlation between the abilities of Vta1 to bind Vps4 via the VSL region, to stimulate Vps4 ATPase activity, and to effect efficient function of the MVB pathway. Moreover, these results demonstrate that the VSL region is directly involved in Vps4 binding in addition to being required for Vta1 dimerization. To determine whether the VSL region is sufficient for Vps4 binding and activation, the last 40 residues of Vta1 (290–330) were fused to the COOH terminus of GST (GST-VSL). This bacterially expressed protein was then used in GST pull-down assays as previously described for GST-Vta1. As can be seen in , GST-VSL is capable of binding Vps4, similarly to full-length Vta1, whereas GST-Vta1 cannot. This result indicates that the VSL region is both necessary and sufficient for Vta1 association with Vps4. To address whether the VSL region is capable of stimulating Vps4 activity, GST-VSL was used in Vps4 ATPase assays. Whereas GST or GST-Vta1 addition did not stimulate the ATPase activity of Vps4, the addition of GST-VSL significantly stimulated Vps4 activity at this concentration from 3 to ∼10 ADP/min/Vps4 (P = 0.03; unpublished data). We also performed ATPase assays with the purified 40-mer VSL peptide (VSL sans GST) and found that the Vta1 VSL region displayed concentration-dependent Vps4 stimulation (). Although this activation was less pronounced than stimulation by full-length Vta1, this result indicates that the Vta1 VSL region is both necessary and partly sufficient for Vps4 activation. Murine SBP1 has been demonstrated to bind the Vps4 orthologue SKD1 (). Therefore, human SBP1 (hSBP1) and the VSL region of hSBP1 (VSL) were also examined for Vps4 binding and stimulation to determine whether the human VSL region exhibited comparable activity. Full-length hSBP1 or the last 41 residues of SBP1 (267–307) were fused to the COOH terminus of GST (GST-SBP1 and -VSL) and used in pull-down assays with Vps4. GST-VSL and -SBP1 were capable of binding Vps4 (). Moreover, examination of Vps4 ATPase stimulation indicated that SBP1 and the VSL region were also capable of enhancing yeast Vps4 ATPase activity (P = 0.0002 and P < 0.0001; ) to an extent similar to the VSL region. This result suggests a conserved mechanism by which the VSL region can stimulate Vps4. Vps4 is an AAA-ATPase whose activity is required for proper function of the MVB sorting pathway (for review see ). As a family, the AAA-ATPases function in a variety of cellular processes through a similar mechanism of action. AAA-ATPases use the energy from ATP hydrolysis to induce conformational changes in other proteins or protein complexes, ultimately causing unfolding or dissociation of the substrate protein (for review see ). The function of Vps4 would appear to be the removal or disassembly of the ESCRT machinery at a late step within the MVB sorting reaction (for review see ). Loss of Vps4 function results in accumulation of ESCRT machinery on the endosomal membrane and concurrent dysfunction of the MVB sorting pathway (, ). The mammalian homologue of Vps4, SKD1, is believed to play an equivalent role in mammalian cells (; ). The pivotal role for Vps4 in the later steps of MVB cargo sorting highlights the importance of understanding how its ATPase activity is regulated during this complicated sorting event. In the present study, we have demonstrated that Vta1 binds directly to Vps4 through a conserved region, resulting in stimulation of Vps4 ATPase activity and potentiating the recycling of ESCRT proteins. The mechanism by which Vta1 stimulates Vps4 function appears to be evolutionarily conserved, underscoring its importance in the context of the MVB sorting reaction. We have demonstrated that Vta1 plays an important role in stimulating Vps4 activity for maximal function in vivo. Our data indicate that Vps4 is still recruited to the endosome in cells and that the basal Vps4 activity is sufficient to maintain some level of MVB sorting. The positive regulatory role of Vta1 might explain the phenotypic variations observed upon deletion of in different genetic backgrounds (; ). It is possible that in alternative genetic backgrounds, the basal Vps4 activity is not sufficient to escape the class E phenotype. Additionally, changes in environmental/growth conditions may increase flux through the endosomal system, which could place a bigger burden on the MVB pathway, specifically Vps4. If Vps4 function was already compromised by loss of Vta1 function, this defect could be amplified under increased endosomal flux. We also observed that there is some level of cargo-specific MVB sorting defects occurring in cells. Together, these data suggest that Vta1 functions as a positive regulator of Vps4 function, thereby impacting the function of the MVB pathway. The ORF was amplified from yeast genomic DNA and cloned into the BamHI and SalI sites of the pET28b expression vector (Novagen) to generate pET28Vta1. pET28Vta1, pET28Vta1, pET28Vta1, and pET28Vta1 were constructed by PCR amplifying the relevant regions of and cloning into the BamHI and SalI sites of pET28b. hSBP1 was amplified from a human breast cDNA library (a gift from C. Mendelson, University of Texas Southwestern, Dallas, TX) and cloned into the BamHI and SalI sites of pGST-parallel1 (pGST–hSBP1). VSL-region coding sequences of Vta1 and SBP1 were synthesized and cloned into the BamHI and SalI sites of pGST-parallel1 (pGST-VSL and pGST-VSL). All cloned PCR products and oligonucleotides were sequenced to exclude unexpected mutations. Inserts from pET28Vta1 constructs were then subcloned into the BamHI and SalI sites of pGST-parallel1 and pPrA416, the yeast expression vector pGPD416 containing an NH-terminal Protein A tag (this study). The following yeast strains were used in this study: SEY6210 (MATα []), SEY6210.1 (SEY6210; MATa []), MBY4 (SEY6210; []), JPY48 (SEY6210; [this study]), JPY47 (SEY6210.1; [this study]), JPY50 (MBY4; [this study]), JPY42 (SEY6210; [this study]), JPY43 (MBY4; [this study]), JPY46 (SEY6210; [this study]), and JPY45 (MBY4; [this study]). BY4742 strains were obtained from Open Biosystems. SF838-9D strains were a gift from R. Piper (University of Iowa, Iowa City, IA). Protein expression was performed in the HMS174 DE3 bacterial strain at 25°C for 14–20 h with 0.5 mM IPTG. His-fusion proteins were purified by Ni-affinity chromatography (5 ml HiTrap Chelating FF or Ni-NTA resin), treated with thrombin (optional), incubated with ATP to dissociate chaperones, subjected to anion exchange chromatography (5 ml HiTrap Q FF or Bioscale Q2 column), and resolved on a Superose 6 10/30 column (optional). GST fusion proteins were purified by glutathione-affinity chromatography (glutathione Sepharose 4B resin) including treatment with ATP to dissociate chaperones and elution of the Vta1 protein or peptide by cleavage with HisTEV; HisTEV was then removed with Ni-NTA resin. Alternatively, GST fusion proteins were used in pull-down assays and ATPase assays as fusion proteins still bound to the glutathione Sepharose 4B resin. Vps4 was purified as previously described (). GST pull-down experiments were performed as previously described (), with the following modifications: ATPase buffer (see ATPase assay) plus 0.05% Tween 20 was used with varying concentrations of ATP as needed, and Vps4 was used at 6 μg/ml. Visualization of bound material was accomplished by Coomassie staining or Western blotting with anti-Vps4 antisera. Protein A purification was performed as in , with slight modifications. 5 OD units of cells were spheroplasted, lysed in PBS plus 0.05% Tween 20 (PBST), cleared, and incubated with IgG Sepharose 6 Fast Flow in spin microcolumns for 60 min. Bound material was visualized with monoclonal HA.11 (Covance) or monoclonal anti-GFP (CLONTECH Laboratories, Inc.). Gel filtration analysis was performed as previously described (). Analysis of CPS and carboxypeptidase Y transport to the vacuole by pulse-chase immunoprecipitation was performed as previously described (). Vta1 equilibrium centrifugation was performed in an Optima XL-1 (Beckman Coulter) centrifuge at concentrations of 0.5 and 2 mg/ml in PBS. Subcellular fractionation was performed as previously described (). Microscopy was performed on living cells using a fluorescence microscope (Nikon) fitted with FITC and rhodamine filters and a digital camera (Coolsnap HQ; Photometrix), and images were deconvolved using Delta Vision software (Applied Precision, Inc.). Measurement of Vps4 ATPase activity was performed as previously described (). To ensure accurate calculations of activity, experiments were performed in a manner such that substrate (ATP) was not limiting and product inhibition was not observed. α-[P]ATP was combined with cold nucleotide to yield a 10-mM ATP × Ci/mole mix. Vps4 (100 nM–1.5 μM) was combined with Vta1 proteins (100 nM–30 μM) in ATPase reaction buffer (0.1 M KOAc, 20 mM Hepes, and 5 mM MgOAc, pH 7.5) in a total of 18 μl at 30°C. Reactions were initiated by the addition of ATP to 1 mM. 1-μl samples were removed at various time points after ATP addition (5, 10, 15, and 20 min for low Vps4 concentrations and 40 s, 1 min 20 s, and 2 min for high Vps4 concentrations) and resolved by thin-layer chromatography using precoated PEI Cellulose TLC glass plates (Merck) and developing buffer (0.75M KPO, pH 3.5). Plates were dried and exposed to phosphoimager screens for 4–12 h. Screens were processed using the Storm 840 system (GE Healthcare), and ADP and ATP signal was quantitated using ImageQuant software package (GE Healthcare). For analysis of GST fusion proteins, samples (lysate and glutathione beads) were washed extensively with PBST and ATPase buffer. Residual buffer was aspirated with a 30-gauge needle, and 500 nM Vps4 was added in a total of 18 μl ATPase reaction buffer. ATP addition, time-point collection, and sample processing was then performed as described for the untagged proteins. Data was analyzed with Excel (Microsoft) to determine ATP hydrolysis rates and Prism 4 (GraphPad) to determine kinetic and statistical parameters. Fig. S1 shows variation of phenotypes in distinct genetic backgrounds. Table S1 show that the phenotype is distinct from the phenotype. Fig. S2 shows that ATP is required for stoichiometric Vta1–Vps4 interaction in vitro. Online supplemental material is available at .

Polyglutamine (polyQ) disorders such as Huntington's disease (HD) are caused by a dominantly heritable expansion mutation of a triplet repeat in the coding region of the gene. The expression of this mutant protein leads to the onset of a slow, progressive disorder that invariably leads to death. Thus far, neither an effective treatment nor viable targets for drug design are available. A prevalent feature of HD and other polyQ diseases is the accumulation and aggregation of the mutant protein. These changes lead to the formation of cytoplasmic and nuclear inclusion bodies, the appearance of which indisputably signifies the inability of the cell to properly dispose of the mutant protein. Indeed, overexpression of expanded polyQ proteins has been shown to alter proteasome () and lysosome function (). Over the past several years, animal models of HD (; ; ) and spinocerebellar ataxia type 1 (; ) revealed that cells have the capacity to clear these products if the continuous production of the mutant transgene is halted. Invariably, clearance of the protein is accompanied by reversal of the disease-like symptoms in the mice. In light of these findings, it is critical to determine the pathway responsible for alleviating this protein accumulation to define targets to fight these diseases. To determine the pathway responsible for the clearance of mutant huntingtin (htt), we conducted a two-tiered functional genetic screen. We first used gene arrays to quickly assess the transcriptional changes induced by pathogenic polyQ lengths. Although these changes alone can be somewhat informative, it is difficult to determine the functional relevance of these changes. Thus, we next targeted transcripts of genes that were “increased” with chemically synthesized small interfering RNAs (siRNAs) to determine which of these proteins were required for mutant htt clearance. Those specific proteins revealed by the second screen then became the focus of further investigation. Of the 56 up-regulated transcripts, 23 were required for mutant htt clearance. Interestingly, the pattern of genes revealed that activation of insulin receptor substrate 2 (IRS-2), a scaffolding protein that mediates the signaling cascades of growth factors such as insulin and insulin-like growth factor 1 (IGF-1; ), leads to a macroautophagy-mediated clearance of the accumulated polyQ proteins. Clearance is present despite the activation of Akt, mammalian target of rapamycin (mTOR), and p70 S6 kinase. This is surprising because activation of mTOR is an inhibitor of the classic, starvation-induced macroautophagy (). The significance of this is twofold: first, that macroautophagy in the presence of accumulated proteins can also occur in an mTOR-independent manner; and second, that this represents another important pathway through which factors such as insulin and IGF-1 may exert beneficial effects. To determine if the clearance of mutant protein can be observed in stable cell lines, we designed a series of functional cell-based assays that were similar to the HD94 mouse model (). Cell lines inducibly express exon1 of htt (exon1htt) carrying a polyQ expansion of 25, 65, or 103 residues. Inducibility is conferred using the tet-off system (). To monitor the state of the proteins, and to ensure that aggregation was mediated primarily by the polyQ repeat, the COOH termini were fused to monomeric enhanced CFP (mCFP; ). To ensure that our siRNA-based screen can be conducted as efficiently as possible, we first focused on HeLa-based cell lines (). siRNA transfection efficiency in these cells reaches >80% (; ; unpublished data). To confirm our findings, however, we also used a neuronal background with Neuro2a cell lines (N2a; ), which have been previously used to characterize different cellular aspects of HD (; , ). The cell lines demonstrated a polyQ length–dependent increase of intracellular, predominantly cytoplasmic, inclusions (). After transient transfection, we observed acute, polyQ length–dependent cell death, as in other transiently transfected cell–based studies (; ; ; ). After stable transfection, however, this was no longer observed across three independent cell lines during the duration of our experiment (Fig. S1, available at ). It is likely that the expression levels achieved were not high enough or that the time of expression was not long enough to elicit acute cell death (). Thus, this assay is well suited to identify regulators of protein degradation because the absence of cell death will allow a clearer interpretation of the hits. Nonetheless, neuronal cell death is a critical aspect of HD and other polyQ diseases, and therefore should always be considered when integrating these findings in the context of the disease. Next, we determined if the abolition of mutant exon1htt expression would lead to protein clearance. Inhibition of polyQ expression with 100 ng/ml doxycycline (dox) led to clearance of both the soluble and aggregated protein (). Similar to primary cultures derived from the HD94 mice (), within 6 d the inclusions cleared ( and see ). Thus, both nonneuronal and neuronal cell lines are capable of clearing the mutant forms of exon1htt. These findings indicate that the elimination of accumulated mutant exon1htt is very slow. Furthermore, because the amount of time required for clearance is similar across cell types, including primary neurons, the process underlying the elimination of this protein may be a general cellular event. To identify genes that are altered because of expression of mutant exon1htt, we tested our hypothesis that stable expression of 65Q or 103Q leads to transcriptional changes that reflect sequestration and elimination of inclusions. To examine these changes in an unbiased global manner, we determined the genetic profile of the cell lines using Affymetrix gene arrays. Comparisons between exon1htt-65QmCFP (65Q) and 25QmCFP clones (25Q) revealed a total of 70 transcripts increased and 89 transcripts decreased. Comparisons between 103QmCFP (103Q) and 25Q revealed 132 increased and 96 decreased (). Common across both pathogenic glutamine lengths was the 56 increased and 42 decreased (Table S1, available at ). Interestingly, increases were seen in only one proteasomal subunit (PSMD8), but in two lysosome-associated membrane proteins (LAMP1 and 2). The profile also revealed changes in vesicle trafficking, signaling proteins, metabolic proteins, and hypothetical proteins. Eight EST transcripts were excluded from the siRNA screen. For the remaining 48 transcripts, three to four siRNA sequences per gene were individually transfected into two different 65Q and 103Q cell lines. 48 h after transfection, cells were exposed to 100 ng/ml dox for another 48 h to shut down production of new protein and permit 50% of clearance to occur. Cells were fixed and analyzed for the number of inclusions per cell using InCell Analyzer (INCA) 3000 software and calculated for an accumulation index, as described in Materials and methods. An example is shown in . Loss of function of 23 out of the 48 transcripts led to a complete or partial inhibition of clearance, of 9 led to cell death, and of 16 led to no change (). As predicted, genes involved in protein degradation were the most prevalent, including the lysosomal membrane proteins LAMP1 (, probes 3H5, 3H6, 3H7, and 3H8) and LAMP2 (, probes A7, A8, A9, and A10). LAMP2 has previously been shown as essential for lysosomal function in LAMP2 knockout mice. LAMP1, on the other hand, was not required; however, it was surmised that its function was redundant to LAMP2 (; ). The relatively acute loss of function offered by siRNA-mediated silencing may prevent the ability of such compensation to occur. Putative knockdown of the only proteasomal subunit that was altered in the gene arrays, PSMD8, led to cell death, and thus its role in clearance could not be elucidated. Unexpectedly, knockdown of IRS-2 led to an inhibition of aggregate clearance ( and ). A scaffolding protein that transmits the phosphatidylinositol 3-kinase (PtdIns3K) signaling of growth factors like IGF-1 and cytokines, IRS-2 knockout mice have also revealed an important role in the brain (). Western blot analysis confirmed that siRNAs effectively down-regulates the protein (Fig. S2, available at ). shows representative images taken from the INCA 3000 of one of the two 103QmCFP clones in the absence of dox, in the presence of dox, and in the presence of dox after transfection with siRNA number 8H06, which is one of the three siRNAs against IRS-2. Quantification using the INCA 3000 software of these images shows that, after transfection of 08H06, the clearance normally observed by abolishing transgene expression is inhibited (). Because elimination of IRS-2 inhibited clearance, IRS-2 activation may stimulate clearance. Therefore, we tested a series of ligands known to activate IRS-2 in the following cells: insulin (), IGF-1 (), and interleukin-4 (IL-4; ). All three cells demonstrate a dose-dependent clearance of accumulated polyQ proteins (). Unlike our normal clearance paradigm using dox, this clearance occurred despite maintaining continuous expression of exon1htt. As shown in , silencing IRS-2 led to a complete inhibition of the enhanced clearance triggered by IGF-1. Western blot analysis also found that IGF-1–mediated clearance is inhibited by small interfereing IRS2 (siIRS-2; Fig. S2). Silencing another IRS family member, IRS-1 (), had no effect ( and Fig. S1). Although this result is compelling, we cannot completely eliminate its role because up-regulation of IRS-2 may mask an effect. Similar dependence on IRS-2, but not -1, was observed with the enhanced protein degradation triggered by insulin and IL-4 (unpublished data). We next attempted to gain insight into the intracellular pathway by which IRS-2 activation triggers protein clearance. It has been previously shown that activation of IRS-2 can lead to the production of PtdInsphosphate by turning on the class I PtdIns3K (). However, it has also been shown that IRS-2 phosphorylation may lead to the production of PtdInsphosphate (PI3P; ; ; ). These latter lipid products are predominantly formed by a class III PtdIns3K, called hVps34 (; ). We first determined whether the presence of PI3P leads to clearance. Synthetic dipalmitoyl-PI3P has been effectively delivered into the cell using liposomes (; ). Administration of these liposomes to the exon1htt-65Q or 103QmCFP cell lines for 3 d led to a significant decrease in the number of inclusions per cell ( and not depicted). Coadministration of PI3P with IGF-1 or insulin did not significantly enhance this effect, suggesting that they lie in the same pathway (unpublished data). siRNA against hVps34 effectively eliminated the exon1htt clearance caused by IRS-2 activation (). Interestingly, similar to IRS-2, the loss of hVps34 also eliminated the clearance revealed by eliminating transgene expression with dox (unpublished data). The requirement of hVps34, as well as LAMP1 and 2 (), suggests that macroautophagy is being triggered. Macroautophagy, which we will henceforth refer to as autophagy, is a means by which long-lived cytoplasmic proteins are degraded by the lysosome via engulfment and encasement by a multilamellar structure and fusion to the lysosome (). Indeed, several studies have previously indicated that the induction of autophagy can lead to the clearance of aggregated polyQ proteins (; ), especially cytoplasmic ones (). It has been previously shown that the inhibition of autophagy elicited by 3-methyladenine (3-MA) is caused by the inhibition of hVps34 (; ; ). hVps34 acts in a multiprotein complex that includes p150 and the mammalian orthologue of Apg6, Beclin1 (; ; ). As seen in , silencing of Beclin1 abolished the clearance stimulated by IRS-2 activation. To determine if this effect is downstream of class I PtdIns3K activation, we next knocked down the serine/threonine kinase Akt. Akt is a survival kinase that was previously shown to be protective in several diseases, such as amyotrophic lateral sclerosis (ALS) and HD (; ; ; ). Although this siRNA against Akt efficiently knocks down the protein (Fig. S2; ), no effect was seen on the IRS-2–mediated clearance (). It has previously been shown that Akt phosphorylation of serine 421 on htt leads to fewer aggregates (; ). Our cell lines express only exon1htt, and thus this residue is not expressed. Perhaps monomeric htt is degraded in an Akt- and proteasome-dependent manner. It is plausible that once an inclusion forms, a different method of protein degradation is required. To further ascertain that IRS-2 activation potentiates autophagy-mediated elimination of the accumulated protein, we examined the effect of established inhibitors of lysosomal degradation and macroautophagy on clearance. As shown in , lysosomal inhibitors inhibit the ability of the cell to eliminate the inclusions. 3-MA and wortmannin also eliminated clearance. These findings further demonstrate the importance of lysosomal function in exon1htt clearance and imply a role in autophagy. Next, we examined if the inclusions colocalized with autophagosomes or lysosomes. Administration of 65 nM of Lysotracker, a red fluorescent dye that accumulates in acidic organelles, showed that mCFP-positive inclusions were in acidified compartments (). Using the autophagosome marker LC3 (; ) and serial Z-sections, we found mCFP-positive inclusions surrounded by an LC3-positive structure ( and ). Although most autophagosomes are known to be much smaller than the inclusion, under certain conditions, such as bacterial invasion, large autophagosomes have been shown to occur (). Furthermore, in a cellular model of another polyQ disease, spinal and bulbar muscular atrophy, found polyQ inclusion bodies bound by double-membrane structures that were also quite large. Quantifications revealed that ∼10% of the inclusion-positive cells had inclusions within LC3-positive vacuoles (). Consistent with the enhanced clearance, activation of IRS-2 using IGF-1 significantly increased the frequency of colocalization. siRNA-mediated silencing of IRS-2 and Beclin1 abolished this increase. Silencing Akt or IRS-1, again, had no effect. Growth factors like insulin are classic signaling molecules that inform the cell of the presence of nutrients. From yeast to mammalian cells, these factors generate PtdInsphosphate. This leads to a signaling cascade that activates Akt, mTOR, and the translation activator p70S6K (). A known consequence of this pathway is an inhibition of macroautophagy, which is a catabolic process. This is contrary to our results, in which the same receptor activates degradation of mutant exon1htt inclusions by autophagy. To ensure that our findings are not caused by a disturbance of normal signaling processes, we reexamined the effect of insulin and IGF-1 on both the proper phosphorylation of known downstream kinases and on autophagy caused by amino acid deprivation. Whole cell lysates were collected from 65QmCFP- or 103QmCFP-expressing cells after exposure to IGF-1 or insulin for 30 min (). Phosphospecific antibodies against Akt and p70S6K revealed that, indeed, the targets were rapidly activated in these clones. Therefore, we concluded that the classical autophagic response is intact in our cell lines. Recent work by , however, found that long-term activation of p70S6K can also induce autophagy. To ensure that mTOR signaling is intact in the cell lines, phosphospecific antibodies against mTOR were also used. Phosphorylated mTOR was readily detectable in the presence of mutant exon1htt expression, as well as in the presence or absence of mutant htt expression ( and not depicted). Amino acid deprivation and rapamycin decreased mTOR phosphorylation in these cells, whereas insulin and IGF-1 inhibited the effect of amino acid withdrawal. Thus, the IGF-1 and insulin lead to the predictive phosphorylation of p70S6K and mTOR. We next examined proteolysis of long-lived proteins. Proteolysis was measured by [C]valine-labeled long-lived proteins. 103QmCFP cell lines were under dox suppression for 2 wk to ensure no transgene expression because the presence of these proteins led to high levels of baseline protein degradation, despite the presence of full serum. Insulin and IGF-1 significantly attenuated the amount of proteolysis, whereas PI3P induced proteolysis despite the presence of complete media. Furthermore, the administration of 3-MA also diminished degradation (). Thus, the autophagy-mediated clearance of inclusions by IGF-1 and Ins occurs despite proper signaling by mTOR. IRS-2 is expressed in all insulin-responsive organs, including the brain. We examined if IRS-2 could activate macroautophagy in response to protein accumulation in neuronal cell lines (). Similar to the HeLa cells, elimination of novel polyQ protein production led to clearance over a period of 6 d. We stimulated IRS-2 using IGF-1 to determine if a similar mechanism was at play. IRS-2 activation using IGF-1 also led to an autophagy-mediated clearance of polyQ proteins in cells of a neuronal lineage. The clearance was accompanied by an increased colocalization of mCFP-positive inclusions in YFP-LC3 autophagosomes. Again, knockdown of Beclin1 inhibited the colocalization of inclusions in the autophagosomes. Using a unique two-tiered functional genetic screen, this study revealed an unexpected means by which autophagy-mediated clearance of accumulated mutant protein can be activated. We found that the activation of IRS-2 led to macroautophagy-induced clearance of the accumulated polyQ proteins (). The activation was dependent on class III PtdIns3K activation and occurred despite activation of Akt, mTOR, and p70S6K. These findings highlight several points. The first is that activation of IRS-2 can lead to the clearance of accumulated mutant exon1htt. IRS-2 is widely expressed and, together with IRS-1, mediates the signaling of insulin and IGF-1 in most tissue (, ). found that loss of IRS-2 function in knockout mice led to an accumulation of neurofibrillary tangles containing phosphorylated tau in the absence of changes in the kinase glycogen synthase kinase 3β. In light of our findings, it is possible that the loss of IRS-2 exposed a potential role for these proteins to mediate clearance of these complex proteins. It would be interesting to determine if protein accumulation is occurring in other tissues that are IRS-2 deficient. Activators of IRS-2, such as insulin and IGF-1, have both been shown to strongly promote neuronal survival through stimulation of Akt. Consequently, its efficacy has been tested as such in other neurodegenerative diseases, such as ALS. For example, retroviral delivery of IGF-1 in a mouse model of ALS led to amelioration of the phenotype, together with a diminishing of the accumulated mutant SOD1 (). A placebo-controlled trial in American ALS patients found that the progression of functional impairment significantly slowed in the treated patients, with no adverse side effects (). The outcome suggested an IGF-1 dose-dependent treatment effect. For HD, have examined the neuroprotective effect of Akt stimulation by transiently transfecting mutant htt into primary neurons. Similar to the findings in ALS, they found that IGF-1 administration led to both a decrease in the number of aggregates formed and a decrease in cell death. The mechanism through which IGF-1 elicited both effects was believed to be directly downstream of Akt (, ). We were able to determine that the autophagocytic clearance can also occur independent of Akt because its knockdown did not eliminate clearance. Monomeric and aggregated proteins may be degraded differently. These findings indicate that the protection conferred by the insulin signaling pathway in diseases with protein accumulation may be twofold; the classical neuroprotective pathway triggered by Akt and the enhanced clearance stimulated by hVps34 activation. Because cytoplasmic inclusions are more readily degraded by autophagy, it is now critical to determine to what extent these inclusions contribute to pathology. Another point highlighted by this study is that macroautophagy is indeed capable of degrading large inclusions and is stimulated under conditions previously deemed inhibitory. Activation of IRS-2 in the HeLa cell lines was achieved by using insulin, IGF-1, or IL-4. Signaling through these receptors, however, is also known to inhibit mTOR-mediated autophagy through the activation of class I PtdIns3K and mTOR. Nonetheless, there is evidence that autophagy may occur despite mTOR activation. For example, transgenic expression of the autophagosomal marker LC3 demonstrates that in certain tissue autophagosomes constitutively form, even in the absence of starvation (). Consistent with these findings, immortalized cells of certain lineages also have a higher basal activity of macroautophagy (). More recently, found that proteasome inhibition was a potent activator of autophagy, although the role of mTOR in the response was not explored. In this study, our findings indicate that macroautophagy is the mechanism by which the aggregates are cleared, despite mTOR activation. This could be achieved if the autophagy regulators downstream of mTOR, such as hVps34, could be activated. Therefore, the activity observed in our results may represent a competition between the inhibitory effects of activating class I versus III kinases. Moreover, clearly demonstrated in the fat body that the regulation of autophagy may occur differently from what was previously believed. They found that constitutive activation of p70S6K was indeed required for autophagy, rather than inhibitory as previously described (). The presence of expanded polyQ proteins may offset the balance between p70S6K and mTOR, thus, allowing for IRS-2 activation to perpetuate a signal to drive autophagy. Another possibility is that inclusion formation itself triggers autophagy () in an mTOR-dependent fashion, and IRS-2 activation perpetuates this response downstream of mTOR regulation. Indeed, a previous study showed that inclusions can sequester mTOR and, thus, could activate autophagy (). However, in our cell lines, we did not find mTOR sequestration. Nonetheless, differentiation between mTOR dependence and independence is difficult because the activation of IRS-2 is upstream of both the inhibitory and stimulatory signaling. To test this hypothesis, we must first understand how mTOR negatively regulates autophagy in mammalian cells. This is not yet fully defined. In yeast, TOR has been shown to negatively regulate the activation of the Apg13–Apg1 complex, a kinase required for autophagy, but it is not certain if this is attributable to direct phosphorylation by TOR (). Furthermore, in mammalian cells, the orthologues of neither Apg13 nor Apg1 have yet to be identified. mTOR-dependent autophagy has mainly been confirmed in mammalian systems through the use of chemical inhibitors such as rapamycin. Chemical compounds that directly activate mTOR are not available. In conclusion, by starting with a nonhypothesis-driven approach, we were able to discover that the regulation underlying degradation of accumulated proteins differs from the regulation underlying conditions of starvation. Furthermore, this work demonstrates that we can no longer assume that regulatory mechanisms studied under nonpathogenic conditions are static; cells compensate when confronted with toxic conditions. Accordingly, we have found that the insulin-signaling pathway may be an important avenue through which this might be achieved. Indeed, activation of IRS-2 has been an attractive target for the treatment of type II diabetes, and thus this line of research may also benefit other disorders. In any case, it is clear from these studies that more information regarding the importance of lysosomal degradation pathways such as autophagy in neuronal systems is absolutely crucial. Controlling these pathways that degrade mutant htt will allow us to finally begin to treat this terrible disease. N2a cells were purchased from American Type Culture Collection. Insulin, IGF-1, IL-4, and dox were purchased from Sigma-Aldrich. Antibodies were purchased from Upstate Biotechnology (anti–IRS-2, anti–phospho Akt, and anti–phospho S6 kinase), BD Biosciences (Akt, S6 kinase, and IRS-1), Cell Signaling Technology (anti–phospho-mTOR and anti-mTOR), and Roche (anti-GFP). htt-exon1 (CAGCAA) constructs were obtained from A. Kazantsev (Massachusets General Hospital, Charlestown, MA). pYFP-LC3 was obtained from T. Yoshimori (National Institute of Genetics, Mishima, Shizuoka, Japan). N2a were selected to be tTA-positive by transfection with P-tTA-IRES-neo and selection with 800 μg/ml G418. PolyQ cell lines were created by cotransfecting Hela and N2a with tetO-htt (25Q, 65Q, or 103Q) exon1-mCFP and P-hygro (CLONTECH Laboratories, Inc.) and then selected with hygromycin using 800 and 200 μg/ml, respectively. 100 ng/ml dox was also maintained in the culture media during selection to maintain suppression of transgene expression. HeLa cells were maintained in DME with 10% FCS, whereas N2a cells were maintained in 50% DME/50% Optimem in 10% FCS. Cells were plated in a 100-mm dish, harvested using 100 μL TRIzol reagent (Invitrogen) and isolated per the manufacturer's instructions. RNA was resuspended and further purified using the RNAeasy kit (QIAGEN). RNA was labeled and hybridized onto human U133A chips at the Genome Core Facility at Memorial Sloan Kettering Cancer Center (MSKCC). Array results were analyzed using GeneSpring 2.0 (Agilent Technologies) and an Affymetrix software package 5.0. siRNAs were designed using an algorithm designed by . siRNA were created at either Integrated DNA Technologies or the Functional Proteonomics Project at MSKCC. Scramble siRNA sequences were purchased from Dharmacon. A final concentration of 10 or 20 nM of siRNAs was used for silencing. Cells were transfected using OligofectAMINE per the manufacturer's instructions. 7.5 × 10 cells were plated in 96-well ViewPlates (Packard Instrument Co.) and transfected the next day. 48 h after transfection, cells were split across two wells and treated with dox. Cells were examined 48 h later. Transfection of plasmid DNA was accomplished using LipofectAMINE per the manufacturer's instructions. Compounds were administered 48 h after transfection unless otherwise noted. For the high throughput screen, images were collected on the INCA 3000 (GE Healthcare). Cells in 96-well ViewPlates were fixed for 10 min with 4% PFA, and the nuclei were stained using 1 μM Hoechst 333342 for 30 min. After scanning the plates, images were analyzed using the granularity analysis module on the accompanying software. In brief, the granularity analysis quantifies the number of inclusions (grains) within a cell, using a two-color strategy to identify individual cells and to analyze associated grains. After recognizing the objects, in this case the Hoechst-positive nuclei, the algorithm next identifies, using a specified size range (in pixels) and fluorescent intensity gradient, the grain in the proximity of the object. Both the Ngrains (number of qualifying grains per cell) and the fraction of fluorescent within the qualifying grains gave similar results. To calculate the accumulation index used in , values were first normalized as a percentage of control for scramble siRNA–transfected cells treated with 2 d of 100 ng/ml dox, and set to 0 by subtracting 100%. This would permit comparison across all of the experiments conducted and a quick assessment of the direction of the change—an increase in accumulation (>0) or an increase in clearance (<0). A gene was considered required for clearance when the absolute value of accumulation index was greater than two standard deviations of the Scramble siRNA + 2 d dox. Each siRNA was transfected on 8 wells per 96-well plate, and each experiment was repeated five to eight times. Cell viability was determined through several measures. First, we examined the number of cells per well scanned. If a drug or siRNA led to significantly fewer cells, they were initially considered toxic. If this toxicity consistently appeared across experiments, we next confirmed toxicity using an inclusion/exclusion assay of cell death, known as a LIVE/DEAD assay (Invitrogen). Cells were grown on glass (HeLa) or poly--lysine–coated (N2a) coverslips in 24-well plates. Cells were fixed for 10 min with 4% PFA. Nuclei were stained with Hoechst 33342 for 30 min, and membranes were stained using Alexa Fluor 633–labeled cholera toxin subunit B obtained from Invitrogen. Images were acquired using a confocal microscope (TCS SP2; Leica) at 63× magnification, along with the accompanying software package. Data acquisition was performed using National Institutes of Health Image 4.0. Analysis of variance (ANOVA) and post hoc analyses were conducted using Statview 5.0 (SAS Institute, Inc.). Synthetic dipalymitoyl-PI3Ps (Matreya) were dried together with phosphatidylserine at a 1:1 concentration under argon and vacuum and resuspended in 25 mM Hepes, 100 mM KCl, and 1 mM EDTA to a total lipid concentration of 800 μM. Liposomes were freeze-thawed, then manually extruded through two 50-nm polycarbonate membranes. PI3P/phosphatidylserine liposomes were administered to cells at 20 μM for 3 d, fixed, and assessed for clearance. This protocol was adapted from previously reported protocols (; ). Cells were incubated for 18 h at 37°C with 0.5 μCi/ml L-[C]valine–supplemented media. Cells were rinsed with HBSS to remove unincorporated radioisotopes and then chased in fresh media overnight to allow degradation of short-lived proteins. Cells were rinsed in HBSS + 10 mM Hepes and incubated for 4 h with either full media ± rapamycin, 20 mM PI3P liposomes, or HBSS + 10 mM Hepes ± 3-MA,20 μM PI3P liposomes, 100 nM insulin, or 100 nM IGF-1. Cells were scraped and, using TCA, protein was precipitated from both the incubation media and the cells. Proteolysis was assessed as the acid-soluble radioactivity divided by the radioactivity maintained in the precipitate. Fig. S1 shows cell lines that were monitored for cell death using the LIVE/DEAD assay. Fig. S2 shows whole cell lysates from transfected cells that were run on SDS-PAGE gels, transferred to PVDF membranes, and probed with the antibodies as noted. Online supplemental material is available at .

The microtubule motor cytoplasmic dynein and its activator dynactin, which mediate minus end–directed movement, have important roles in both interphase and dividing cells. In interphase cells, the dynein–dynactin complex is essential for vesicle and organelle transport, such as ER-to-Golgi vesicular trafficking (for review see ). The dynein–dynactin motor complex also transports RNA particles (), aggresomes (), and virus particles along microtubules (). During cell division, dynein and dynactin play a critical role in both nuclear envelope breakdown and spindle formation (for review see ). Consistent with these multiple cellular roles, dynein and dynactin function are required in higher eukaryotes. Loss of dynein or dynactin is lethal in (), and mice homozygous for loss of cytoplasmic dynein heavy chain die early in embryogenesis (). Cells cultured from dynein heavy chain–null embryos show fragmented Golgi and a dispersal of endosomes and lysosomes throughout the cytoplasm (). Neurons appear to be particularly susceptible to defects in dynein–dynactin complex function. The dominant-negative mutation in Glued, which encodes a truncated form of the p150 subunit of dynactin, shows defects that are most profound in neurons (). Two -ethyl--nitrosurea–induced point mutations in cytoplasmic dynein heavy chain cause slowly progressive motor neuron disease in mice (). () and () mice each carry missense mutations in a highly conserved domain of cytoplasmic dynein that mediates subunit interactions. When homozygous, these mutations are lethal; heterozygous mice exhibit progressive loss of motor neurons, leading to muscle weakness and atrophy (). A similar phenotype is observed in transgenic mice with a targeted disruption of dynactin in motor neurons (). In humans, a G59S missense mutation has been identified in the gene encoding p150 (DCTN1) in a kindred with slowly progressive motor neuron disease (). Affected patients develop adult-onset vocal fold paralysis, facial weakness, and distal-limb muscle weakness and atrophy. Clinical, electrophysiological, and pathological investigations have confirmed the selective loss of motor neurons in this disorder (). p150 is the dynactin subunit responsible for binding to dynein and microtubules (; ). The G59S substitution occurs in the highly conserved NH-terminal cytoskeleton-associated protein, glycine-rich (CAP-Gly) domain, which interacts directly with microtubules () and the microtubule plus-end protein EB1 (). In this study, we examined the biochemical and cellular effects of the G59S substitution in p150. Our data suggest that the G59S mutation leads to both decreased microtubule binding and enhanced dynein and dynactin aggregation, suggesting that both loss of function and toxic gain of function contribute to the motor neuron degeneration observed in affected patients. The G59S mutation is located within the highly conserved CAP-Gly domain of the p150 polypeptide, a domain that mediates the binding of dynactin to microtubules. We compared the microtubule binding affinities of wild-type and G59S p150 peptides (). The CAP-Gly domain of wild-type p150, which spans residues 1–107, bound weakly to microtubules (unpublished data). This 1–107 peptide lacks the serine-rich region of p150 (111–191), which may be required for efficient microtubule binding by CAP-Gly proteins (). of 1.1 ± 0.2 μM. of 2.6 ± 0.5 μM, indicating a modest decrease in affinity. More striking, however, was the observation that even at saturating microtubule concentrations, only half of the mutant protein was able to bind to microtubules in this assay (). Similar results were observed in experiments with full-length wild-type and G59S p150 (unpublished data). We performed sequential microtubule binding experiments, in which the unbound fraction of G59S p150 (1–333) protein was incubated for a second time with a saturating concentration of microtubules (25 μM), and observed that ∼60% of the protein pelleted with microtubules (Fig. S1 A, available at ). These data suggest that there may be a rapid equilibrium between two populations of the mutant polypeptide, one that can bind and one that cannot. Mixing of wild-type and G59S p150 at a 1:1 ratio resulted in 60% of protein pelleting with 25 μM microtubules (Fig. S1 B). These data suggest that mutant protein does not significantly inhibit the binding of wild-type polypeptide to microtubules. We next investigated the effects of the mutation on the binding of p150 to microtubules in cells. We used transient transfection assays to compare the distribution of wild-type and G59S p150 in COS7 cells as well as MN1 cells, motor neuron–like cells that extend neurites (; ). Although endogenous dynactin generally has a punctate cellular localization, with decoration of dynamic microtubule plus ends, overexpression of p150 results in the decoration of the microtubule cytoskeleton (). As shown in , 24–48 h after transfection of GFP-tagged full-length constructs of wild-type p150, there was decoration of microtubules, as assessed by colocalization with tubulin. In contrast, GFP-tagged full-length G59S p150 was distributed diffusely in the cytoplasm and showed no colocalization with tubulin (). Similar results were obtained using GFP-tagged NH-terminal 1–333 constructs of wild-type and G59S p150, as well as untagged full-length wild-type and G59S p150 constructs (unpublished data). We performed microtubule binding experiments using protein extract from COS7 cells that had been transfected with GFP-tagged, full-length p150. Almost all of the exogenous polypeptide from wild-type p150–transfected cells pelleted with taxol-stabilized microtubules. However, only approximately half of the protein from G59S p150–transfected cells pelleted with microtubules (unpublished data). This observation confirms our in vitro data that only a portion of the G59S p150 protein population may be available for microtubule binding. The NH-terminal CAP-Gly domain of p150 binds to EB1 (). Crystallographic studies demonstrate that the COOH terminus of EB1 contacts p150 in a hydrophobic cleft of the CAP-Gly domain (). We therefore examined the binding of G59S p150 to EB1 using affinity chromatography. The wild-type peptide bound to the EB1 column and was retained until elution with high ionic strength buffer, but the G59S peptide had decreased retention on the column, indicating reduced affinity for EB1 (). Previous studies have shown that p150 tracks dynamically with growing microtubule ends together with EB1 (). To investigate the effect of the G59S mutation on the localization of p150 to microtubule plus ends, we transfected COS7 cells with GFP-labeled wild-type or G59S p150. We selected for cells with low levels of expression, as microtubule plus-end tracking behavior is not evident at higher expression levels because of the decoration and bundling of microtubules induced by high levels of exogenous p150. Wild-type p150 tracked dynamically with growing microtubule ends ( and Video 1, available at ), whereas the G59S construct showed no microtubule association, even at tips ( and Video 2). In cells with higher levels of expression of the G59S construct, we noted apparent aggregates of misfolded protein, but these aggregates showed no directed movement (Video 2). To study the cellular effects of the G59S mutation in the p150 subunit of dynactin, we established fibroblast and lymphoblast cell lines from two symptomatic individuals known to be heterozygous for the G59S missense allele. Control fibroblast cell lines were obtained from two age-matched control individuals, and a control lymphoblast cell line was derived from an age-matched subject. In these lines, we examined whether the G59S mutation alters the expression of dynein and dynactin. In both lymphoblasts and fibroblasts, quantitative RT-PCR analysis of RNA levels showed no difference in p150 transcript levels between cell lines heterozygous for the G59S mutation and control cell lines (). Western blots of protein extract from patient cell lines showed up-regulation of levels p150, but not of dynein or other dynactin subunits, compared with control cell lines (). To determine whether the wild-type and mutant proteins are both expressed in cells cultured from patients heterozygous for the G59S mutation, we performed quantitative Western blotting using both a monoclonal antibody to the microtubule binding region of p150, which binds the wild-type protein with a much higher affinity than the mutant protein, and a polyclonal antibody to p150, which recognizes both forms equally well (). Analysis of patient cells indicated that the total level of p150 expression (as determined using the polyclonal antibody) is 147 ± 7% the level observed in control cells (). Western blots with the monoclonal antibody demonstrated that patient cells express 82 ± 4% of the wild-type p150 that control cells express (). Thus, we estimate that the mutant protein makes up ∼44% of the total p150 population in patient cells. To examine the structural integrity of the dynactin complex, we fractionated cell extracts from the patient-derived and control fibroblast cell lines by sucrose density gradient centrifugation. Intact dynactin was observed to sediment at ∼19S in both the patient and control samples, consistent with the large size of the multimeric complex. No significant pool of unincorporated p150 subunits was observed in the lower S value fractions from either the patient or control cells (), suggesting that expression of the mutant polypeptide does not significantly disrupt dynactin structure and that the mutant polypeptide is incorporated into dynactin in these cells. Incorporation of the mutant polypeptide into dynactin might be expected to disrupt dynactin localization in patient-derived cells; however, we observed no change in the cellular localization of dynactin in fibroblasts derived from patients compared with control fibroblasts ( and Fig. S2 A, available at ). Dynactin was present diffusely in the cytoplasm in a fine, punctate pattern, with no visible dynactin aggregates. We also noted no change in the cellular localization of cytoplasmic dynein, which was also found in a punctate cellular distribution, partially overlapping with dynactin staining in both patient and control cells (), or EB1, which was localized specifically to microtubule tips (). We examined the effects of the G59S mutation on the integrity of the Golgi and the assembly of the mitotic spindle in the patient-derived fibroblasts. Disruption of dynactin by dynamitin overexpression has been shown to disrupt the Golgi in interphase cells () and the mitotic spindle in dividing cells (). However, no gross morphological defects in the organization of the Golgi or the mitotic spindle were evident in patient-derived heterozygous cells under normal growth conditions ( and Fig. S2 B). In addition, no consistent defects in the growth rate were observed in the patient fibroblasts (unpublished data). To test the patient fibroblasts for dynactin function, we looked at several dynein/dynactin-dependent processes. Dynactin, as well as dynein and the dynein-interacting protein LIS1, are necessary for directed fibroblast migration (). However, wounded monolayers of patient cells recovered at the same rate as control cells (unpublished data). Aggresome formation has also been shown to be dynein dependent (). To test the effect of the mutation on aggresome formation, an androgen receptor containing an expanded polyglutamine repeat that induces inclusion formation () was expressed in patient fibroblasts. These fibroblasts formed inclusions at a rate indistinguishable from control cells (unpublished data). Although a single wild-type copy of the gene for p150 may be sufficient to mediate dynein-dependent processes under normal conditions, conditions of cellular stress may reveal latent effects of the G59S mutation. Nocodazole, a microtubule-depolymerizing drug, causes dispersal of the Golgi. During recovery from nocodazole treatment, microtubules reassemble and the Golgi fragments coalesce near the centrosome in a dynein/dynactin-dependent manner (). have shown a slowing in the recovery of the Golgi after nocodazole treatment in fibroblasts cultured from homozygous mice. Therefore, we assayed the cytoskeletal and organelle recovery rates in heterozygous G59S and control fibroblasts after nocodazole washout. Microtubules were depolymerized and the Golgi body dispersed after 1 h of nocodazole treatment. 1 h after drug washout, microtubules had reassembled in both control and patient-derived cells; however, Golgi complex morphology was significantly different in patient cells. In control cells, 75 ± 2% of cells had an intact Golgi complex, 22 ± 3% of cells had a partially disrupted Golgi complex, and 3 ± 1% of cells had completely disrupted Golgi complex (). In contrast, in patient-derived cells only 46 ± 8% of cells had intact Golgi complexes, whereas 44 ± 5% of cells showed partial disruption and 11 ± 6% of cells showed complete disruption of the Golgi. Golgi reassembly after 24 h was essentially normal in patient-derived fibroblasts (unpublished data), indicating that expression of mutant dynactin slows but does not block the minus end–directed transport of Golgi elements toward the microtubule organizing center. We also observed that the localization of EB1 to microtubule plus-end tips was altered in patient cells during nocodazole recovery. After microtubule depolymerization with nocodazole, EB1 demonstrated diffuse cytoplasmic staining. After 30 min of recovery in conditioned growth media, EB1 was localized specifically to the plus ends of microtubules in control cells, forming comet tails that were 1.20 ± 0.06 μm long (). In patient-derived cells, EB1 was not limited to microtubule tips but was also observed to localize along microtubules (). EB1 tail length increased significantly in patient-derived cells, often to >5 μm, although overlap of adjacent microtubules prevented exact measurements of the elongated EB1 tails. These data suggest a defect in the specific localization of EB1 to microtubule plus ends. To compare these data to a loss of function of dynactin, we used RNA interference to knockdown p150 expression levels in HeLa cells by 70–90% (). This knockdown caused dispersal of the Golgi throughout the cell body (). In addition, we observed an increase in the length of EB1 comet tails from 1.08 ± 0.05 μm in mock-transfected cells to 1.28 ± 0.07 μm in cells transfected with small interfering RNA (). The lengthening of EB1 comet tails is similar to what was observed in patient fibroblasts recovering from nocodazole treatment and correlates with a loss of dynactin function. In the microtubule binding assays described in , we observed the binding of only half of the mutant p150 polypeptide to microtubules, suggesting that some portion of the mutant protein population is unavailable for binding to microtubules. To investigate this further, we expressed differentially tagged (T7 and His) truncated forms of wild-type and G59S p150 in vitro and performed immunoprecipitation with an antibody to the T7 tag. Although our constructs, which include amino acids 1–333, span part of the first coiled-coil domain of p150 hypothesized to mediate dimerization (), we observed no association of the T7- and His-tagged wild-type polypeptides (). However, we did observe coimmunoprecipitation of the differentially tagged NH-terminal G59S constructs. These data suggest that the G59S polypeptide, but not the wild type, has a tendency to self-associate. There was no coimmunoprecipitation after incubation of differentially tagged wild-type and G59S p150 (unpublished data), indicating that the wild-type and G59S proteins do not interact under these conditions. We next investigated whether aberrant biochemical species of the G59S p150 protein could be isolated from protein extracts of cells overexpressing this protein. COS7 cells were transfected with full-length wild-type or G59S GFP-tagged p150. 24 h after transfection, the extract from these cells was fractionated over a sucrose gradient and analyzed by SDS-PAGE gel electrophoresis and Western blot. In cells transfected with wild-type p150, the peak concentration of dynamitin and endogenous p150 was at 19S (). The exogenous p150 protein (as determined by the increase in molecular weight that is due to the GFP tag) was present at 19S, as well as at less dense fractions. This indicates that some exogenous protein is incorporated into the dynactin complex but some remains unincorporated in lower molecular weight fractions, most likely because its expression is in excess of the other subunits of dynactin. In contrast, extracts from cells transfected with GFP-tagged G59S p150 demonstrated higher molecular weight species in fractions 2–4. This suggests the presence of aggregated forms of G59S p150 with a molecular weight well above that of endogenous dynactin (). Endogenous p150 and dynamitin are not present in these fractions, indicating that they do not copurify with the aggregated protein. The aggregated protein remains soluble, as we did not observe the formation of detergent-insoluble aggregates (unpublished data). As shown in , G59S p150 was cytoplasmically dispersed in COS7 cells 24–48 h after transfection, whereas wild-type p150 decorated microtubules. At longer time points, however, we noted a centripetal localization of the proteins. Wild-type p150 became preferentially localized along microtubules in the perinuclear region (). In contrast, G59S p150 localized to inclusions surrounding the nucleus, which may correspond to the aggresomes of misfolded protein described by . These structures were also observed in very highly expressing cells at earlier time points, but their frequency increased with time after transfection (Fig. S3, available at ). In some MN1 cells transfected with GFP-tagged G59S p150, single or multiple inclusions were evident most often in the cell body () and rarely in neurites. They were similar in appearance to those observed in motor neurons from the brainstem of an affected patient (), and their frequency increased with time after transfection (Fig. S3). Inclusions stained positive for dynein intermediate chain (DIC), the Golgi marker GM130, and the 20S proteasome but not kinesin heavy chain, microtubules, neurofilaments, vimentin, microtubule-associated protein 2, Cu/Zn superoxide dismutase (SOD1), and survival of motor neurons, (unpublished data). Thus, in both neuronal and nonneuronal cells, mutation of the glycine 59 appears to decrease microtubule binding by the p150 CAP-Gly domain and leads to aggregation and inclusion formation by the mutant protein. To look at the ultrastructure of the inclusions, transfected COS7 cells and MN1 cells were observed by EM. MN1 cells transfected with GFP-tagged full-length G59S p150 and labeled with immunogold showed granular, nonfibrillar inclusions of mutant protein (). Analysis of nonimmunogold-labeled, glutaraldehyde-fixed COS7 cells demonstrated that the inclusions were not membrane bound (). These micrographs show inclusions that look remarkably like the dynein- and dynactin-containing inclusions seen in patient neurons by immunohistochemistry (). In these ultrastructural studies, mitochondria frequently surrounded or were contained within the G59S p150 inclusions (). To examine the possibility that mitochondria localization was altered by the inclusions, COS7 cells were transfected with wild-type or G59S p150 and stained with an antibody to mitochondrial chaperone Hsp60. Mitochondria were partially relocalized in the area of the aggregates (). Quantification of the cross-sectional area of the cells that contained mitochondria demonstrated that mitochondria in cells transfected with G59S p150 were less widely distributed than in cells transfected with wild-type protein (). It may be that mitochondria cannot be transported to the cell periphery because of aberrant interaction with the aggregated G59S p150. Alternatively, it is possible that loss of dynein/dynactin transport causes mitochondrial mislocalization, as expression of dynamitin has also been shown to cause an inward collapse of the mitochondrial array (). The expression of the G59S polypeptide led to an increase in cell death in MN1 cells, as determined by propidium iodide (PI) exclusion. Cells were transfected with GFP-tagged wild-type p150, G59S p150, or GFP alone. The MN1 cells transfected with G59S p150 demonstrated a significantly higher percentage of cell death than cells transfected with wild-type p150 or GFP alone (). Furthermore, the percentage of cell death increased with time after transfection, corresponding to an increase in the percentage of cells containing inclusions visible by immunofluorescence (Fig. S3). Embryonic rat motor neurons expressing G59S p150 also demonstrated an increase in cell death compared with motor neurons expressing exogenous wild-type p150 in a time-dependent manner (Kalb, R.G., personal communication). Neuronal cells may be uniquely sensitive to the G59S polypeptide, as the expression of G59S p150 does not increase cell death in COS7 cells (). Overexpression of the chaperone Hsp70 has been reported to suppress protein aggregate formation and prevent cell death in several protein misfolding disease models (). 56 ± 6% of COS7 cells expressing the G59S p150 protein for 2 d contained visible inclusions (). However, cells expressing both Hsp70 and G59S p150 exhibited a disperse localization of both exogenous proteins and only 17 ± 4% of transfected cells contained visible inclusions (). Hsp70 containing the T13G mutation cannot undergo the conformational change necessary for chaperone activity (). In cells cotransfected with G59S p150 and T13G Hsp70, the proportion of cells containing inclusions was not significantly different from that of cells transfected with G59S p150 alone (). A Western blot of the cell protein lysates showed that levels of G59S p150 were decreased when wild-type, but not T13G, Hsp70 was coexpressed (). The chaperone function of Hsp70 may aid proper folding of G59S p150, thereby avoiding the formation of inclusions and allowing effective degradation of the mutant protein by the ubiquitin–proteasome pathway. Transfection of G59S p150 into MN1 cells led to an increase in cell death compared with cells transfected with wild-type p150 (). However, coexpression of G59S p150 and wild-type Hsp70 reduced the percentage of MN1 cell death to levels similar to those of cells transfected with wild-type p150 (). This protection was not observed when MN1 cells were cotransfected with G59S p150 and either empty vector or T13G Hsp70 (). These data demonstrate that expression of active Hsp70 reduces the amount of G59S p150 aggregates, decreases the amount of p150 expressed, and protects MN1 cells from the toxicity associated with expression of the mutant p150. A key question in the analysis of many neurodegenerative diseases is the cell-type specificity observed: why would a mutation in a ubiquitously expressed protein preferentially affect a single cell type? This question is particularly critical in the study of motor neuron diseases, such as Amyotrophic Lateral Sclerosis, in which multiple mutations in a ubiquitously expressed protein, SOD1, result in motor neuron–specific degeneration and cell death. Several mechanisms have been proposed, including neuron-specific aggregation and defects in axonal transport (for review see ). We focus on the cellular effects of a point mutation in the p150 subunit of dynactin. Dynactin is ubiquitously expressed in vertebrates, interacting with cytoplasmic dynein to serve as the major motor for microtubule minus end–directed transport in the cell. The dynein–dynactin complex is required for a range of cellular functions, including mitotic spindle assembly, ER-to-Golgi trafficking, and endosome and lysosome motility. Although complete loss of dynactin function is therefore likely to affect all cell types, patients expressing the G59S mutation in the p150 subunit of dynactin develop an autosomal-dominant, slowly progressive degeneration specific to motor neurons (). The G59S missense mutation results in a subtle impairment of dynactin function. A subtle loss of function in a protein required for retrograde axonal transport may be sufficient to induce a slow degeneration of motor neurons. Mice with a targeted disruption in dynactin function or with point mutations in cytoplasmic dynein heavy chain exhibit a slowly progressive loss of motor neurons, resulting in muscle atrophy (; ). The G59S mutation also results in a toxic gain of function, as the G59S polypeptide is prone to aggregate. have shown that evolutionarily conserved glycines inhibit aggregation because of their low propensity to form β structure. The G59S substitution alters a highly conserved glycine residue within the NH-terminal CAP-Gly domain of the protein and is predicted to result in steric crowding and misfolding of this domain (). Comparisons of aggregate formation in both neuronal and nonneuronal cells overexpressing the G59S mutation suggest that motor neurons are uniquely vulnerable to aggregate formation, leading to enhanced cell death. One explanation for this observation is that motor neurons may not express adequate levels of chaperones to cope with the high levels of misfolded protein. Recent studies have shown that motor neurons are not able to up-regulate Hsp70 in response to cellular stress and that they are particularly vulnerable to depletion of Hsp70 (). Consistent with this mechanism, overexpression of Hsp70 led to decreased aggregations of the G59S polypeptide and decreased cell death. The mechanism by which aggregate formation leads to cell death remains to be determined. However, EM analysis of the aggregates demonstrates the presence of trapped organelles, including mitochondria. The aggregates may either actively trap organelles or passively disrupt microtubule-based transport via “organelle jams” (for review see ). The sequestration of cytoplasmic dynein in these aggregates, as observed in both transfected cells and patient motor neurons (), would further disrupt axonal transport. This disruption in transport is likely to be most deleterious to motor neurons because of their overall size and extended axons. Based on our observations, we propose the following model to explain the motor neuron–specific phenotype observed in patients expressing the G59S mutation in dynactin (). The mutation leads to a decreased efficiency in minus end–directed transport. This subtle loss of function does not significantly perturb nonneuronal cells but may be sufficient to affect the overall efficiency of retrograde axonal transport in neurons. However, the mutation also results in a gain of function, as the G59S polypeptide has an enhanced propensity to misfold. Aggregation of the misfolded protein is concentration dependent, and the p150 polypeptide is highly expressed in motor neurons (; unpublished data). Further, motor neurons may be specifically vulnerable to misfolding and aggregation of the G59S polypeptide because of insufficient expression of molecular chaperones. Finally, both the sequestration of active motors and the trapping of organelles by the p150 aggregates will further exacerbate the inhibition of axonal transport. Together, our data provide the foundation for a testable model for the cellular mechanisms leading to the motor neuron–specific degeneration observed in patients expressing the G59S mutation, involving both loss of function and toxic gain of function. We anticipate that these studies will provide further insight into the mechanisms by which a mutation in an essential cellular protein can result in specific degeneration of motor neurons in vivo. Wild-type and G59S p150 were expressed and labeled with [S]methionine using the TNT T7 Quick system (Promega), clarified by centrifugation at 39,000 for 30 min, incubated for 30 min at 20°C with increasing concentrations of microtubules polymerized from purified tubulin (Cytoskeleton, Inc.), and stabilized with paclitaxel (Cytoskeleton, Inc.). Microtubule bound and unbound proteins were separated by centrifugation at 39,000 for 20 min and analyzed by SDS-PAGE and fluorography. Results were quantitated by densitometry using NIH ImageJ. Prism Software (GraphPad) was used to fit the binding data to the one-site ligand binding equation y = B × x/( + x). Affinity matrices were prepared by cross-linking recombinant EB1 to activated CH Sepharose 4B (GE Healthcare) beads at 4 mg/ml ligand. In vitro–expressed p150 was incubated with the EB1-bound beads for 30 min at room temperature. These mixtures were loaded onto a column and washed extensively with 50 mM Tris and 25 mM KCl, pH 7.4, with 0.1% Triton X-100. Retained proteins were eluted with 2 M NaCl. Fractions were analyzed by SDS-PAGE and Western blot. Fibroblast cell lines were derived from forearm punch skin biopsies from two symptomatic patients with the G59S mutation and from an age-matched, unaffected control sibling (). An additional age-matched control fibroblast line (AG02222) was obtained from the Coriell Cell Repository. Lymphoblast cell lines were derived from blood samples from an affected family member carrying the G59S mutation using standard techniques. Cell lines were established in the Cytogenetics Laboratory at Georgetown University. Fibroblast cell lines were maintained in Hams F-10 culture media supplemented with 15% fetal bovine serum; lymphoblast cells were maintained in RPMI media with 10% fetal bovine serum. COS7 cells (American Type Culture Collection) were maintained as described previously (). HeLa-M cells (a gift from A. Peden, Genentech, South San Francisco, CA) were maintained in DME with 10% fetal bovine serum. For immunofluorescence assays, cells were grown to ∼75% confluence and then fixed in −20°C methanol and processed for immunocytochemistry. Monoclonal antibodies to tubulin (DM1A from Sigma-Aldrich), hemagglutinin (Sigma-Aldrich), kinesin heavy chain (Chemicon), neurofilament (NE14 from Sigma-Aldrich), vimentin (Sigma-Aldrich), microtubule-associated protein 2 (Sigma-Aldrich), Cu/Zn SOD1 (StressGen Biotechnologies), survival of motor neurons (BD Biosciences), Golgi protein GM130 (BD Biosciences), cytoplasmic DIC (Chemicon), EB1 (BD Biosciences), TGN46 (Serotech), Hsp60 and -70 (StressGen Biotechnologies), and dynactin subunits p150 and dynamitin (BD Biosciences) were purchased commercially. Affinity-purified polyclonal antibodies to p150, Arp1, and DIC have been described previously (; ; ). Immunostaining was visualized with Alexa 350–, 488–, and 594–conjugated secondary antibodies (Invitrogen). Images were acquired on a microscope (DMIRBE; Leica) with a 63× or 100× Plan Apo objective using OpenLab software (Improvision) and a charged-coupled device camera (Orca ER; Hamamatsu). MN1 cells were maintained as described previously () and fixed in 4% paraformaldehyde at room temperature for 10 min, permeabilized with 0.1% Triton X-100 in PBS for 10 min, and incubated with monoclonal antibodies to α-tubulin (clone 2.1 from Sigma-Aldrich) followed by Texas red–conjugated secondary antibodies (Jackson ImmunoResearch Laboratories). Deconvoluted images were acquired with a microscope (Olympus) using DeltaVision software (Applied Precision) on a Silicon Graphics workstation. Transient transfection assays were performed using Fugene (Roche) and plasmids encoding either wild-type or G59S full-length human p150 or truncated constructs of the wild-type and G59S polypeptide spanning residues 1–333, both fused to GFP and untagged. Hemagglutinin-tagged Hsp70 constructs were a gift from Y. Argon (University of Pennsylvania, Philadelphia, PA). For nocodazole recovery assays, patient and control fibroblasts were treated with nocodazole at 5 μg/ml for 1 h and then allowed to recover for 0, 30, or 60 min or 24 h in conditioned culture media at 37°C in 5% CO before fixation. Live cell time-lapse recordings were performed on transiently transfected COS7 cells expressing either full-length wild-type or G59S p150 or NH-terminal residues 1–333 of wild-type or G59S p150, all fused to GFP. Cells on glass coverslips were sealed in an imaging chamber (FCS2; Bioptechs) and maintained at 37°C in culture media. Sequential time-lapse fluorescent images were acquired at 12-s intervals. HeLa-M cells were transfected using Oligofectamine (Invitrogen) with 100 nM of a mixture of four RNA duplexes targeting different regions of human DCTNI (SMARTpool siRNA reagent [Dharmacon]; available from GenBank/EMBL/DDBJ under accession no. ): 5′-gaagaucgagagacaguu-3′, 5′-cgagcucacuacugacuua-3′, 5′-caugagcgcuccuuggauu-3′, and 5′-ggagcgcuguaucguaaga-3′. Cells were methanol fixed after 72 h and processed for immunocytochemistry or resuspended in denaturing sample buffer and processed for Western blot analysis. RNA was extracted from cells using TRIzol (Invitrogen) and purified with the RNAeasy clean-up kit according to the manufacturer's protocol (QIAGEN), and cDNAs were generated using the High Capacity cDNA Achieve kit (Applied Biosystems). Quantitative PCR reactions were run in triplicate using the ABI Prism 7900 sequence detection system (Applied Biosystems). A forward primer (5′-gaagggcatggcatctttgtg-3′), a reverse primer (5′-gaagcagaagaatcaggtgtctct-3′), and a fluorescent probe (5′-FAM-ccagtcccagatccag-BHQ1-3′) were designed to amplify p150 transcripts. Endogenous controls were simultaneously amplified using commercially available primers (Applied Biosystems). The reactions were performed in triplicate and averaged, and p150 Ct values (cycle number when signal reaches a threshold above background) were corrected for endogenous control Ct values using the ΔΔCt method per the Applied Biosystems User Bulletin 2. Cells from 3–6 flasks of patient fibroblasts and 3–6 flasks of control fibroblasts were washed in PBS, harvested, and homogenized in 20 mM Tris-HCl, pH 7.4, 2 mM EGTA, and 1 mM EDTA with protease inhibitors (leupeptin, peptstatin A, -p-tosyl--arginine methyl ester, and PMSF). Triton X-100 was added to 0.4%, and the homogenate was clarified by low-speed centrifugation. The resulting supernatant fraction was layered over a 5–25% linear sucrose density gradient and centrifuged at 126,000 for 16 h. The gradients were eluted in 0.5-ml fractions, which were resolved by SDS-PAGE, and analyzed by Western blot. For aggregation assays, His- and T7-tagged constructs of wild-type and G59S p150 were coexpressed in vitro and incubated for 2 h at 30°C. The reactions were then incubated sequentially with protein A beads to preclear the extracts, followed by protein A beads with bound monoclonal antibody to T7 (Novagen). After a 1-h incubation with antibody bound beads, the beads were isolated by centrifugation; washed four times with 50 mM Tris, pH 7.3, 50 mM NaCl, and 0.1% Triton X-100; and eluted by boiling in denaturing gel sample buffer. The immunoprecipitates were analyzed by Western blots probed with antibodies to the His and T7 tags. A FACS-based survival assay was used to measure cell death (). MN1 or COS7 cells were transfected with pEGFP wild-type and G59S p150 constructs in 6-well plates. Each transfection condition was performed in triplicate. After 24, 48, or 72 h, cells were harvested with trypsin, gently pelleted with centrifugation, and resuspended in 1 ml of PBS. Cells were stained with 2 μg/ml PI (Sigma-Aldrich) and gently vortexed. For each sample, 50,000 nongated events were acquired using a FACSCalibur instrument (BD Biosciences) and Cell Quest software (Becton Dickinson). GFP fluorescence was collected in the FL-1 channel, and PI fluorescence was collected in the FL-3 channel in dot and density blot formats. Results were expressed as a percentage of PI-positive cells (cell death) divided by the total number of GFP-positive cells (transfected cells). MN1 and COS7 cells were plated on permanox chambered slides (LabTek; Nunc) and transfected with pEGFP wild-type and G59S p150 constructs or wild-type and G59S p150 (untagged) constructs. 48 h after transfection, one set of cells (for immunogold labeling) was fixed with 4% paraformaldehyde in PBS for 1 h and another set was fixed in 4% glutaraldehyde in cacodylate buffer. Paraformaldehyde-fixed cells were washed three times with PBS and then blocked and permeabilized with 0.1% saponin for 1 h. The p150-GFP–transfected cells were then incubated with mouse polyclonal GFP antibody (Invitrogen), and p150-transfected cells were incubated in dynactin polyclonal antibody (UP235) followed by anti-mouse or anti-rabbit Nanogold (Nanoprobes). Slides were subjected to staining and silver enhancement as described previously (). After dehydration, embedding, and sectioning, samples were examined with an electron microscope (1200EX; JEOL). Fig. S1 shows sequential and mixed microtubule binding of G59S p150. Fig. S2 shows that the localization of p150 and the formation of spindles are normal in cells heterozygous for the G59S mutation in p150. Fig. S3 shows that the percentage of cells containing G59S p150 inclusions increases with time after transfection. Video 1 shows microtubule plus-end tracking of GFP-labeled wild-type p150 in transfected COS7 cells. Video 2 demonstrates a loss of microtubule tip localization and plus-end tracking of GFP-labeled G59S p150 in transfected COS7 cells. Online supplemental material is available at .

Intermediate filament (IF) proteins are assembled into either homopolymer or heteropolymer 10-nm-diam cytoskeletal filaments in a complex multistep process (). Pairs of protein chains interact in parallel and in register to form an α-helical coiled–coil dimer, which is the basic building block of IF. Little is known about the mechanisms responsible for dimer formation. However, it is known that in vitro dimers assemble into antiparallel tetramers that associate laterally to form unit-length filaments (ULFs). The ULFs are ∼60-nm-long and contain ∼32 protein chains. These anneal end-to-end to form ∼10-nm-diam IFs (). Of these various IF structures, only small amounts of tetramer have been detected in soluble fractions of lysed cells (; ). Type III IF proteins such as vimentin and peripherin exist in several states within cells, including nonfilamentous particles (). Particles form short IF, or squiggles, which form the long IF that comprise the cytoskeletal networks of interphase cells. Although the composition of particles is unknown, they most likely contain essential IF building blocks such as dimers or ULFs. The assembly of particles into IF networks has been studied in spreading fibroblasts and differentiating nerve cells (). During spreading, ∼70% of the particles move rapidly along MT in a kinesin- and dynein-dependent manner (). As spreading progresses, many particles are converted into polymerized IF. Similar particles are seen in other IF systems, including type IV neurofilaments () and the types I and II keratins in epithelial cells (). Although it is obvious that nonfilamentous particles are precursors in the assembly pathway of cytoskeletal IF, little is known about their formation in cells. We show that in rat pheochromocytoma cells (PC12), a significant fraction of peripherin particles are assembled cotranslationally in a process that we have termed dynamic cotranslation. Evidence for dynamic cotranslation is derived from RNA FISH and the simultaneous live imaging of both peripherin mRNA and its protein product. Individual peripherin mRNA particles (messenger RNPs [mRNPs]) possess numerous copies of peripherin mRNA, suggesting a mechanism involving the coordinated synthesis of coiled–coil dimers, the building blocks of IF. The results provide important and novel insights into the linkages among the motile properties and targeting of mRNAs, their translational control, and the dynamic properties and assembly of IF proteins. The dynamic properties of IF particles were analyzed in GFP-peripherin–expressing PC12 cells by FRAP. The cells used in these experiments were optimized for particle formation by short-term (2–4 h) NGF treatment. At this time, there was a dramatic increase in peripherin expression in the form of particles (). Within 1 min after whole-cell photobleaching, fluorescent particles were visible ( and Video 1, available at ), whereas filaments could not be detected for 5–10 min (not depicted; ). This rapid recovery suggested that some particles might be engaging in de novo synthesis of peripherin. We tested this possibility by photobleaching GFP-peripherin–expressing cells in the presence of cycloheximide to inhibit protein synthesis. Under these conditions, the fluorescence recovery of particles was not detected even after 10–15 min (Video 2). We also sought to determine whether the same particles seen before photobleaching were the ones that recovered their fluorescence. However, because ∼70% of the peripherin particles in PC12 cells engage in rapid MT-dependent movements (), it was necessary to treat cells with the MT inhibitor nocodazole to inhibit motility. Transfected cells that had been differentiated in NGF for 2 h were treated with nocodazole for 30 min. Under these conditions, individual particles could be identified both before and at time intervals after photobleaching. As a control in these experiments, only half of the cell was photobleached to provide an internal standard to ensure that particles were indeed immobilized ( and Video 3, available at ). Some particles recovered their fluorescence within 1 min after photobleaching, and, by 10 min, ∼30% of the immobilized particles ( = 200) in the photobleached half seen before photobleaching were fluorescent. No further increase in the number of fluorescent particles was detected for periods up to 30 min (not depicted). To determine whether peripherin particles were associated with peripherin mRNA, GFP-peripherin–expressing PC12 cells were fixed and processed for FISH within 2–4 h after the addition of NGF (). Biotinylated antisense probes were used to detect peripherin mRNA. Analysis of ∼1,800 randomly selected peripherin particles in the thin, peripheral regions of 20 cells revealed that ∼29% were closely associated with peripherin mRNAs, which were also seen as discrete particles ( [green, peripherin particles; red, peripherin mRNA]; for split image of , see Fig. S1, A and B, available at ). A close association was defined as a minimum of 1/4 overlap between diameters of the red and green signals (Fig. S2 shows a more detailed description of the association between peripherin particles and mRNA). To ensure that this association was not random, statistical tests were performed using a modification of a previously published procedure (; see Materials and methods). The expected number of peripherin particles that could associate with the peripherin mRNA based on chance alone was calculated using this same 1/4 overlap criterion for each of the 20 cells analyzed. This calculation included the determination of the number and size of individual peripherin mRNA and protein particles, as well as the total area of the cytoplasmic region analyzed. The resulting data were subjected to a two-tailed test with Bonferroni correction to determine if the differences between the actual and expected values were statistically significant (P < 0.006; see Materials and methods). We determined that the probability of the association seen between peripherin particles and mRNA, based on chance alone, was less than one in 3.7 × 10 (P ≅ 3.7e − 8). The association of endogenous peripherin particles with peripherin mRNA in untransfected cells was also determined by a combination of immunofluorescence and FISH. Approximately 1,000 particles were analyzed as above, and the analysis revealed that ∼30% were associated with peripherin mRNA, reflecting the similar association seen in transfected cells (Fig. S1, G and H). The protein synthesis inhibitor puromycin, which acts by releasing nascent protein chains from ribosomes (; ), reduced the association between peripherin protein particles and mRNA from ∼29% to levels that were attributable to random overlap (; for split image of see Fig. S1, C and D; = 15 cells; P ≅ 0.52). As an additional control for the specificity of the interaction between peripherin particles and peripherin mRNA, we analyzed the association between peripherin particles and β-actin mRNA. Unlike the overlapping cytoplasmic distributions between peripherin particles and mRNA seen in , the majority of β-actin mRNA and peripherin particles appeared to reside in distinct regions of the cell (; for split image of see Fig. S1, E and F). No significant association could be detected based on the same statistical analyses performed for peripherin mRNA and particle association (; = 10 cells; P ≅ 0.04). As mentioned in the previous paragraphs, ∼70% of peripherin particles were motile and ∼30% were stationary (). Therefore, we attempted to determine whether moving peripherin particles were associated with mRNA. Motile particles in GFP-peripherin–transfected PC12 cells that were grown on locator coverslips were observed by time-lapse imaging for 10–20 s (Video 4, available at ), and the cells were then immediately fixed on the microscope stage. After fixation, the coverslip was processed for FISH and the appropriate microscope field was relocated (). In 15 different experiments involving 15 different cells, 40 moving particles were followed and none of these was associated with peripherin mRNA (, arrowheads). In contrast, particles in the same field of view, which were not moving during the observation period, were associated with mRNA (, asterisks). This result suggested that only stationary particles were engaged in protein synthesis (Video 5 shows footage of a different cell). To determine the motile behavior of both peripherin particles and mRNAs, a system was developed that permitted the simultaneous imaging of both a specific mRNA and its translation product in live cells. A construct containing the enhanced CFP (ECFP)–peripherin coding sequence (CDS), followed by 24 MS2–binding repeats (R24; ) and the peripherin 3′untranslated region (UTR), was introduced into PC12 cells along with a plasmid encoding YFP-MS2 (a bacteriophage MS2 coat protein fused to YFP). When the ECFP-peripherin-R24-3′UTR construct is transcribed, the R24 sequence in the mRNA forms 24 stem loops, each consisting of an identical sequence of 19 nucleotides. The YFP-MS2 protein binds to these stem-loop structures, thereby providing a visible reporter for locating and tracking the movements of these peripherin mRNAs. The translation product of this YFP-MS2–tagged peripherin mRNA is CFP-peripherin (). Live imaging of PC12 cells expressing both constructs contained CFP-peripherin particles and YFP-peripherin mRNA, which also appeared as distinct particles or mRNPs (). Time-lapse imaging revealed rapid and independent movements of both peripherin mRNPs and protein particles. These were dispersed throughout the cytoplasm, including early outgrowing neurites (Video 6, available at ; one frame of this movie is shown in ). Approximately 80% of the peripherin mRNPs were motile, with an average speed of 0.42 ± 0.15 μm/s ( = 200 mRNPs in 5 cells), which is ∼1.5 times faster than the average rate of movement of peripherin particles (). In no case could we detect mRNPs associated with protein particles moving together ( = 54 cells). In time-lapse movies, stationary mRNPs could be found in close association with stationary protein particles. In some cases, the mRNPs were seen to move rapidly away, leaving the peripherin particle behind ( = 20 mRNPs in 10 cells; and Video 7). The reverse situation, in which a protein particle moved away from a stationary mRNP, was never observed. Furthermore, treatment of cells with 10 μg/ml of the protein synthesis inhibitor puromycin for 30–60 min increased the fraction of motile mRNPs to ∼99% ( = 1,286 mRNPs in 6 cells). These findings lend further support to the idea that motile mRNPs are translationally silent. To evaluate the copy number of peripherin mRNA molecules in each mRNP, we used a quantitative FISH method, which is capable of detecting single RNA molecules (; see Materials and methods). The results of this analysis ( = 7 cells) demonstrated that peripherin mRNPs contain one or more mRNAs (). The distribution of the mRNA clusters or mRNPs shows that ∼50% of the mRNPs contain one mRNA and the other ∼50% contain two or more peripherin mRNAs (). However, the ∼50% of mRNPs that contain one mRNA account for only ∼20% of the total peripherin mRNA. Therefore, ∼80% of the total peripherin mRNA exists in clusters of two or more mRNAs (). Type III IF proteins such as peripherin are known to self-assemble into higher order structures (). This capacity for self-assembly may play a role in the clustering of peripherin mRNAs caused by the interactions between nascent protein chains. mRNA clustering may also be attributable to regulatory regions, such as the 3′UTR, within the mRNA molecules. An examination of the possible role of the nascent peripherin protein chains in mRNA clustering was initiated by inserting a stop codon (TAA) between the CFP and peripherin CDSs in the ECFP-peripherin-R24-3′UTR construct to inhibit peripherin synthesis. To verify the action of the stop codon, we transfected rat embryonic fibroblasts (Rat2) that do not express endogenous peripherin with the ECFP-TAA-peripherin-R24-3′UTR construct. Controls consisted of Rat2 cells transfected with ECFP-peripherin-R24-3′UTR. After a 24-h transfection, whole cell lysates were separated by SDS-PAGE and prepared for immunoblotting (). In cells expressing the control plasmid, both peripherin and GFP antibodies (known to react with CFP) recognized a product corresponding to the predicted size of the CFP-peripherin fusion protein, ∼81 kD. In cells expressing the construct with the TAA insertion, there was no detectable peripherin by immunoblotting and the GFP antibody instead recognized an ∼27-kD product, which is the predicted size of CFP (). Immunofluorescence analyses of cells transfected with ECFP-TAA-peripherin-R24-3′UTR showed no peripherin staining, and the CFP signal was diffuse and evenly distributed throughout the cytoplasm (unpublished data). In contrast, in cells transfected with the control plasmid, both the peripherin antibody staining and CFP fluorescence were filamentous and colocalized with the endogenous Rat2 vimentin network (unpublished data), suggesting that the fused peripherin protein was, as expected, coassembling with this type III IF protein (). The quantitative FISH method was used to analyze PC12 cells transfected with the ECFP-TAA-peripherin-R24-3′UTR construct for 24 h followed by NGF for 4 h. The results showed an ∼30% increase in the number of mRNPs containing single mRNAs (, red bars; = 8 cells) compared with controls that were transfected with the ECFP-peripherin-R24-3′UTR construct (, blue bars). However, calculation of the total amount of peripherin mRNA in clusters of two or more still accounted for ∼60% of the total cellular peripherin mRNA (). These data suggest that mRNA clusters can form in the absence of peripherin translation and that the clustering is mediated to a certain extent by the mRNA itself. To more specifically define the region of peripherin mRNA that might be involved in mRNA clustering, we tested the role of the 3′UTR by deleting this sequence from the original ECFP-peripherin-R24-3′UTR construct. Cells transfected with this construct and processed for FISH were quantitatively analyzed as described in the previous paragraph. The results were very similar to the stop codon experiment, showing an ∼30% increase in single mRNAs (, yellow bars; = 20 cells). As with the stop codon construct, ∼60% of the total peripherin mRNA still formed clusters (). To further examine the mechanisms underlying the peripherin mRNA clustering, PC12 cells expressing the ECFP-peripherin-R24-3′UTR construct were treated with nocodazole both before and during NGF treatment (see Materials and methods). Under these conditions, the absence of a fully polymerized MT system resulted in an ∼54% increase in the number of mRNPs containing single mRNAs (, green bars; = 12 cells; Fig. S3, available at ) and an ∼30% decrease in the fraction of total mRNA in clusters (). The close associations found between some of the mRNPs and peripherin particles suggested a cause and effect relationship, indicating that the protein particle formation resulted from translation of the colocalized mRNA. Therefore, we sought to observe the generation of a protein particle in vivo. Cells were double transfected with ECFP-peripherin-R24-3′UTR and YFP-MS2 and treated with NGF for 4 h, at which time nocodazole was added to depolymerize MT (30 min; see Materials and methods). This procedure inhibited the rapid movements of the majority of both peripherin mRNPs and protein particles. This afforded us the opportunity to observe individual mRNPs over a period of time consistent with the synthesis of protein by time-lapse imaging. In several cases, an mRNP was seen closely associated with a weak, diffuse CFP signal. Within 30–180 s, the CFP fluorescence intensity increased significantly, ultimately appearing indistinguishable from other IF particles ( and Video 8, available at ). Our findings suggest that ∼30% of peripherin particles are cotranslationally assembled in close association with their mRNPs. Therefore, it was expected that ribosomes would be associated with these mRNP–peripherin particle complexes. An antibody directed against the S6 ribosomal subunit was used to determine whether ribosomes were associated with these complexes in cells expressing both YFP-MS2 and ECFP-peripherin-R24-3′UTR after treatment with NGF. Approximately 70% of the peripherin particles that associated with peripherin mRNPs also associated with ribosomes (). 10 cells were analyzed and the statistical significance of the associations was determined using the method described for peripherin particle and mRNA association, but modified for triple colocalizations (see Materials and methods). It should be noted that only ∼20% of peripherin mRNPs were associated with ribosomes and that the majority were associated with neither ribosomes nor peripherin particles. Keratin IF are assembled from heterodimers containing a type I and II protein (). We analyzed HeLa cells expressing keratin 8 and 18 to determine if keratin mRNAs form clusters and whether these clusters contain both types of mRNA. In fully spread cells, keratin networks are more stable when compared with Type III IF networks such as peripherin (). Therefore, early spreading cells (after trypsinization and replating), in which the keratin network was more dynamic and undergoing active assembly (, ; unpublished data), were used for analysis. To detect the two types of keratin mRNAs in these spreading cells, a biotinylated oligonucleotide probe specific for K18 mRNA and a digoxigenin-conjugated probe specific for K8 mRNA were used for double FISH analysis (). Approximately 15% of the two types of mRNA signals appeared to colocalize or overlap extensively (; = 8 cells; ≅ 900 mRNPs; P ≅ 0.0004). The existence of hetero-mRNP clusters provides further evidence that IF mRNA clustering is a general phenomenon. The peripherin system of IFs in PC12 cells has provided a unique opportunity to observe the dynamic properties of a specific species of mRNA, along with its translation product. This is attributable to the finding that newly synthesized peripherin accumulates in close proximity to its mRNPs, which is demonstrated by the findings that ∼30% of GFP-peripherin particles rapidly recover their fluorescence after photobleaching in a cycloheximide-sensitive fashion. This rapid recovery of fluorescence shows that the folding and maturation time for GFP-peripherin is on the order of minutes. Although the folding and maturation process of GFP takes ∼1 h in bacteria and yeast (), it has recently been shown that GFP cDNA injected into the nuclei of mammalian cells is expressed and fluoresces in the cytoplasm within 40 min (). In addition, the refolding of extensively denatured GFP in vitro takes a few minutes, as determined by chromophore fluorescence (). Furthermore, our FISH analyses demonstrated that a similar percentage (∼29%) of peripherin particles was closely associated with peripherin mRNA and that this association was inhibited by puromycin. The loss of this association in the presence of puromycin shows that the relationship between peripherin mRNAs and protein particles depends on translation. The interactions between peripherin particles and mRNA were directly visualized in living cells using the dual-transfection method developed for simultaneous imaging of both peripherin mRNA and its protein product. Time-lapse imaging shows active movement of both peripherin mRNPs and protein particles throughout the cytoplasm. However, the movements were independent of one another, and motile peripherin mRNPs were never seen associated with peripherin protein particles. Various mRNA species are known to move along both MT and microfilaments, powered by molecular motors such as kinesin, dynein, and myosin (for review see ). Specific examples of kinesin-dependent mRNA movement include the mRNA-encoding myelin basic protein in oligodendrocytes () and tau mRNA in axons of neuronal P19 cells (). In addition, the localization of several zygotically transcribed mRNAs in embryos have been shown to be dependent on cytoplasmic dynein (for review see ). Although it has been hypothesized that motile mRNAs are not capable of translational activity, this has never been observed in vivo. Our dual peripherin mRNA- and protein-imaging method provides direct evidence that moving mRNPs are translationally inactive. In those instances where stationary peripherin mRNPs were found in close association with peripherin particles, the mRNP invariably moved rapidly away, leaving the protein particle behind. The reverse situation was never observed. In further support of these observations, the translational inhibitor puromycin increased the fraction of motile mRNPs to ∼99%. Visualizing the early stages of protein synthesis in association with peripherin mRNPs turned out to be extremely difficult because of the rapid motility of the majority of both mRNPs and protein particles along MT tracks. To observe the relatively small number of stationary mRNPs engaging in protein synthesis cells were treated with nocodazole. This disassembled MT and therefore stopped the movement of the vast majority of peripherin mRNPs not engaged in protein synthesis, as well as the particles not associated with mRNA. Under these conditions, we were able to observe peripherin particles forming in close association with those translation factories that had presumably formed before the disassembly of MT. In addition, ∼70% of the mRNPs associated with particles were also associated with ribosomes. The ∼30% that remained may represent mRNPs that have ceased the translational process and have already dissociated from ribosomes ( and Video 9, available at ). Most likely, newly synthesized peripherin particles will at some point engage molecular motors and initiate motility or alternatively, remain in place, ultimately assembling into short IFs that appear to link in tandem to form long IFs ( and Video 9; ). On the other hand, the departing mRNP may move to a new site of synthesis, associate with ribosomes, and begin another round of translation ( and Video 9). Quantitative analyses of peripherin mRNPs demonstrate that ∼80% of total peripherin mRNA resides in clusters of two or more. Analyses of cells expressing the ECFP-TAA-Peripherin-R24-3′UTR construct, from which peripherin protein is not synthesized, demonstrate that ∼60% of the total mRNA remains in clusters of two or more. The reduction from ∼80 to ∼60% is mostly attributable to an ∼30% increase in mRNPs containing single mRNAs and an ∼60% decrease in the largest mRNPs containing 6–30 mRNAs, suggesting that interactions between the nascent peripherin protein chains may contribute to the formation of large mRNA clusters. There was no significant change in the number of mRNPs containing two to five mRNAs under these experimental conditions. In the absence of a 3′UTR, there was a similar decrease in clustered peripherin mRNPs with the most dramatic decrease again in the large 6–30 mRNA-containing clusters. There was also an ∼30% increase in mRNPs containing single mRNAs. The MS2-binding repeat region of the mRNA is most likely not involved in the mRNA clustering, as a previous work using the same repeat region conjugated to actin mRNA showed that the vast majority of the mRNA existed as single molecules (). Treatment of cells with nocodazole resulted in the most dramatic decrease in peripherin mRNA clustering, suggesting that the mRNAs are not clustered by simple diffusion, but require a regulated mechanism involving MT and most likely MT-based motors. The partial inhibition of clustering in nocodazole-treated cells may be because of an actin microfilament–based clustering mechanism (). It has been suggested that the nonfilamentous peripherin particles contain dimers, which are the essential building blocks of IF, and perhaps higher order structures such as tetramers and ULFs (). Dimer assembly requires the formation of coiled–coil interactions between the α-helical–rich central rod domains of two parallel and precisely in register IF protein chains (). The α-helical coiled–coil structure is one of the primary subunit oligomerization motifs in proteins, originally discovered in keratin IF by x-ray diffraction (). The mechanism responsible for coiled–coil dimer assembly remains unknown. The finding that the majority of peripherin mRNA is located in clusters of two or more suggests that the formation of the coiled–coil interactions between pairs of peripherin protein chains takes place cotranslationally, involving the coordinated synthesis of closely associated mRNAs. Interestingly, during muscle differentiation, myosin assembly intermediates containing myosin heavy chain coiled–coil dimers also appear as globular “foci,” similar in appearance to peripherin particles. These foci are also thought to form cotranslationally (). The double FISH experiments showing the subpopulation of hetero-mRNPs containing both K8 and K18 mRNAs also support the possibility that the cotranslational assembly of IF dimers is a general phenomenon. Our study provides evidence for the clustering of IF mRNAs and its physiological significance. Other mRNAs, such as β-actin, have been found to exist as single mRNAs, using the same quantitative FISH method that we have described (). Furthermore, whereas peripherin mRNPs are distributed throughout the cell, actin mRNPs are concentrated in certain regions such as the lamellipodia of moving cells. This targeting of actin mRNPs is most likely related to the maintenance of high concentrations of monomeric soluble actin required for the extensive actin polymerization that takes place in lamellipodia (). In contrast, IF mRNPs can move to many sites distributed throughout the cell to initiate the synthesis and formation of insoluble, nonfilamentous precursors such as peripherin particles. This provides a cell with the capacity to control the de novo assembly of fully polymerized IF anywhere in its cytoplasm. This temporal and spatial control of IF polymerization may allow cells to locally regulate the distribution and various functions of IF, such as those involved in determining the mechanical properties of the cytoplasm and in signal transduction (). The direct observation of a specific mRNP and its protein product in the same living cell provides a powerful tool for determining the mechanisms and the order of events involved in targeted protein synthesis. Collectively, the results of this work reveal a process called dynamic cotranslation, which has profound implications for translational control in vertebrate cells. Stock cultures of PC12 cells were maintained in complete medium (DME containing 10% calf serum and 1 mM sodium pyruvate) at 37°C. For studies of peripherin particles, cells from stock cultures were removed with trypsin-EDTA (Invitrogen), replated onto laminin-coated (Roche) coverslips, and cultured in differentiation medium (DM; DME containing 5% calf serum, 1 mM sodium pyruvate, and 30 ng/ml of NGF [Roche] for 2–4 h; ). Primary antibodies used in this study included rabbit anti-peripherin (), mouse anti-biotin (Jackson ImmunoResearch Laboratories), sheep anti-digoxigenin (Boehringer), mouse anti-S6 (a gift from D.L. Spector, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, and from R. Traut, University of California, Davis, Davis, CA), mouse anti-GFP (Roche), and mouse anti-vimentin (clone V9; Sigma-Aldrich). FITC-, rhodamine-, and Cy5-conjugated goat anti–rabbit and anti–mouse and donkey anti–sheep IgG (Jackson ImmunoResearch Laboratories) antibodies were used as secondary antibodies for indirect immunofluorescence. For immunoblotting, peroxidase-conjugated goat anti–rabbit and anti–mouse secondary antibodies were used (Jackson ImmunoResearch Laboratories). For immunoblotting experiments, 70% confluent 100-mm dishes of transfected Rat2 cells were prepared for SDS-PAGE analysis and transferred to nitrocellulose for immunoblotting (). All antibody incubations were performed in PBS containing 5% nonfat dry milk (Sigma-Aldrich). For anti-biotin labeling after FISH, PC12 cells grown on laminin-coated coverslips were rinsed in PBS and fixed with 3.5% formaldehyde (Tousimis) at room temperature for 5 min. HeLa cells were spread for 2 h on uncoated coverslips before fixation with formaldehyde. After formaldehyde fixation, cells were permeabilized with 0.05% Triton X-100 for 5 min. This fixation and permeabilization method was also used for the immunolabeling of Rat2 cells. For S6 labeling, PC12 cells were permeabilized for 15 s in IF lysis buffer (PBS containing 0.6 M KCl, 5 mM EDTA, 5 mM EGTA, and protease inhibitors [1 mM PMSF, 1 mM tosyl--arginine methyl ester, and 1 mg/ml each of leupeptin, pepstatin, and aprotinin; Sigma-Aldrich]) containing 0.1% Triton X-100 () and fixed for 5 min in 3.5% formaldehyde (Tousimis). Cells were then washed with PBS and processed for indirect immunofluorescence, as previously described (; ). After staining, coverslips were washed in PBS and mounted on glass slides in gelvatol containing 100 mg/ml 1,4-diazabicyclo[2.2.2]octane (Sigma-Aldrich; ). Images of fixed, stained preparations were taken with a microscope (LSM 510; Carl Zeiss MicroImaging, Inc.; ). For the detection of peripherin mRNA in PC12 cells, either a 2′-methylated RNA oligonucleotide complementary to the 3′UTR region of rat peripherin mRNA (position 1,473–1,492) or a probe complementary to the coding region of peripherin mRNA (position 288–320) was used. Each probe was labeled at its 5′ end with either biotin or Cy3. A biotin-labeled RNA oligonucleotide probe complementary to rat β-actin mRNA (position 402–421) was used in control experiments. For quantitative FISH analyses, cells were simultaneously hybridized with a Cy5-labeled probe to a region near the MS2-binding sites () and the Cy3-peripherin probe. For keratin double FISH experiments, a probe antisense to position 338–366 in human keratin 8 mRNA CDS and a probe antisense to position 227–261 of human keratin 18 mRNA CDS were used. The 5′ end of the K8 probe was labeled with digoxigenin and the K18 was labeled with biotin. All probes were obtained from Integrated DNA Technologies. PC12 cells were transfected with either GFP-peripherin, ECFP-peripherin-R24-3′UTR, ECFP-TAA-peripherin-R24-3′UTR, or ECFP-peripherin-R24 for 24 h and maintained in DM on laminin-coated coverslips for 2–4 h, followed by fixation with 4% paraformaldehyde (in 1× PBS, pH 7.4) for 15 min. Cells were then rinsed in 1× PBS three times for 10 min each, followed by permeabilization with 0.5% Triton X-100 for 5 min on ice. Cells were washed twice for 10 min in PBS and then in 2× SSC for 10 min. Cells were hybridized in hybridization buffer (final concentration of 2× SSC, 10% dextran sulfate, 50% formamide, 1 mg/ml tRNA mix, and 1 ng/μl of RNA probe) overnight at 37°C. The hybridization was performed on a coverslip sandwiched between two pieces of parafilm that were sealed at their edges to prevent dehydration. For the keratin double FISH experiment, HeLa cells were trypsinized and replated for 2 h and then processed in the same manner as the PC12 cells. In some experiments, untransfected PC12 cells were first processed for immunofluorescence using peripherin antibodies to detect endogenous peripherin particles, followed by a second fixation and hybridization with peripherin mRNA probes. The next day, after hybridization, both transfected and nontransfected cells were washed twice in 2× SSC (50% formamide) for 30 min and then in 1× SSC for 15 min. Cells were then processed for indirect immunofluorescence using either mouse anti-biotin, sheep anti-digoxigenin, or both, followed by goat anti–mouse-FITC and donkey anti–sheep-rhodamine IgG for secondary detection. As a negative control for hybridization, cells were treated with RNase A (Sigma-Aldrich) before hybridization. Single molecule mRNA identification and quantification was performed as previously described () using the Cy3-labeled probe to the peripherin message. This probe has only one binding site on the transcript and can therefore serve for quantification. Only cells hybridized with both the Cy3-peripherin and Cy5-MS2 probes were selected for quantitation, as these are the ones expressing the exogenous construct. The number of probes per particle was determined by calculating the total fluorescent intensity (TFI) of individual RNPs and dividing that value by the TFI per probe as follows: TFI per probe was calculated for the Cy3-peripherin probe by collecting images from serial probe dilutions. 5 μl of each probe dilution (ranging from 1.5 to 1.5 × 10 ng/μl) were placed between a coverslip and a slide, onto which 170-nm blue fluorescent beads had previously been dried. Using the beads as markers, the distance between the coverslip and slide was measured (in micrometers) and the center plane of the dilution was located. A range of interest (256 × 256 pixels) that excluded beads was identified, and a single image was captured at the center plane, using an exposure time identical to that used to capture cell images. This procedure was repeated three times for each dilution, each at a different location on the coverslip. The TFI per probe was then obtained by plotting the integrated fluorescence in the total imaged volume against the known number of molecules in that volume. The slope of the resulting curve represents the TFI per one fluorescent probe molecule and is further used in the calculations. In parallel, TFI per 3D RNA particle in the FISH images was calculated from deconvolved images with a script written for IPLab. To compensate for the deconvolution, this value was then divided by the number of planes in the point-spread function used by EPR. Each RNA now had an assigned TFI, which was then divided by the TFI/probe molecule calculated from the dilution curve, resulting in the number of RNA molecules in an imaged particle. Color map image volumes were generated, using a program written for IPLab, to visualize the number and spatial position of clusters of mRNA. The program identified mRNA clusters from the acquired 3D images, which were deconvolved using a threshold and routine segmentation algorithms. The total intensity of each object was calculated and used to assign a color code for each mRNA cluster that corresponded to the number of mRNAs contained in the cluster. Signals below those calculated for single mRNAs were excluded in the color map presentations. The single FISH images (using either peripherin or actin probes) of GFP-peripherin–transfected cells were subjected to statistical tests using a modification of a previously published procedure (). These tests were performed to make certain that the associations between peripherin particles and peripherin mRNPs were not random. One cytoplasmic region from each cell ( = 20 cells for peripherin FISH; = 15 cells for peripherin FISH in puromycin; and = 10 cells for actin FISH) was used for the statistical analysis of images captured by confocal microscopy. For the purposes of our calculations, we assumed that the particles and mRNA signals were circular and that any particles that showed >1/4 overlap (in radii) with an mRNA signal were defined as “associations.” The following formula was used to calculate the expected number of peripherin particles (E) that would coincide (>1/4 overlap) with either peripherin or actin mRNA based on chance alone:where N is the total number of peripherin particles in a cytoplasmic region, N is the number of mRNPs (either peripherin- or actin-specific) in a cytoplasmic region, r is the average radius of a peripherin particle, r is the average radius of an mRNP, and A is the total area of the cytoplasmic region analyzed. Measurements (radii and area) were performed using image analysis software (LSM510; Carl Zeiss MicroImaging, Inc.). The number of associations actually observed was also determined for each image. This was defined as the number of peripherin particles that were observed to associate (by a minimum of 1/4 overlap) with peripherin or actin mRNA ( = 1,776, 750, and 500 randomly picked peripherin particles were analyzed for associations with mRNA in the peripherin FISH, peripherin FISH in puromycin, and actin FISH experiments, respectively). Finally, two-tailed tests with Bonferroni correction were used to determine if the differences between the actual and expected values were statistically significant. P < 0.006 was defined as significant. When the stringency of the association criteria was increased to a minimum of 1/2 overlap, the following formula was used to calculate the expected number of associations based on chance alone:When the stringency of the association criteria was further increased to complete overlaps, the following formula was used to calculate the expected number of associations based on chance alone:When the stringency of the association criteria was decreased to touching, the following formula was used to calculate the expected number of associations based on chance alone:To test the significance of triple associations between YFP-MS2, CFP-peripherin, and ribosomal staining, the following formula was used:where E is the total number of ribosomes expected to be associated with both peripherin particles and mRNPs by chance alone, E is the expected number of peripherin particles that would associate with mRNPs by chance alone, N is the total number of ribosomes in a cytoplasmic region, r is the average radius of a peripherin particle-mRNP doublet, and A is the total area of the cytoplasmic region analyzed. To determine the actual number of triplet associations observed. 10 cells were analyzed (167 peripherin particles that were associated with mRNPs were analyzed) and only extensive or complete overlaps were considered colocalizations. For the keratin double FISH analysis, a complete overlap between the two mRNP signals was considered as a colocalization. To determine the expected number of colocalizations based on chance alone for these associations, the following formula was used:where E is the total number of K18 mRNPs expected to be associated with K8 mRNPs by chance alone, N is the total number of K18 mRNPs in the cytoplasmic region analyzed, N is the total number of K8 mRNPs in the cytoplasmic region analyzed, r is the average radius of K8 mRNP signals, and A is the total area of the cytoplasmic region analyzed. To determine the actual number of complete colocalizations between K18 and K8 mRNPs, eight cells (∼1,000 mRNPs for each type of mRNA) were analyzed. The EGFP-C1-peripherin plasmid was constructed by amplifying rat peripherin cDNA (∼1.4 kb; provided by L. Parysek, University of Cincinnati, Cincinnati, OH) using PCR with primers that insert BamHI sites at the 5′ and 3′ ends. The resulting BamHI–BamHI fragment was subcloned into the BamHI site of pEGFP-C1 (CLONTECH Laboratories, Inc.). The ECFP-peripherin-R24-3′UTR plasmid was constructed by modifying the EGFP-C1-peripherin plasmid. 24 MS2-binding repeat sequences (∼1.3 kb) were fused to the 3′ end of the ECFP-peripherin CDS, followed by the 3′UTR (∼0.3 kb) of peripherin mRNA. The ECFP-TAA-peripherin-R24-3′UTR construct was created by inserting a TAA between the CFP and peripherin CDSs of the ECFP-peperipherin-R24-3′UTR construct. For the 3′UTR deletion construct (ECFP-peripherin-R24), this sequence was removed from the ECFP-peripherin-R24-3′UTR construct. ). Constructs were introduced into either PC12 or Rat2 cells by electroporation (). After electroporation, PC12 cells were maintained in complete medium for 24 h. After 24 h, the transfected cells were trypsinized, replated onto laminin-coated coverslips, and maintained in DM for 2–4 h (which was optimal for visualizing peripherin particle synthesis) before imaging/analysis. Immunoblot analysis of GFP-peripherin–transfected PC12 cells shows that GFP-peripherin composes ∼25% of the total peripherin in an average transfected cell and that the exogenous peripherin incorporates into the endogenous IF network (). Furthermore, transfected cells display a GFP-IF pattern that is indistinguishable from the endogenous IF in nontransfected cells prepared for indirect immunofluorescence (). Time-lapse observations were made using a LSM510 confocal microscope as previously described (). Images were captured at 5–10-s intervals at a resolution of 512 × 512 dots per inch with an average scan time of 5 s. Images were collected for 20-s to 30-min time periods. FRAP analyses were performed on GFP-peripherin–transfected PC12 cells using the LSM510 microscope as previously described (). For some experiments, the entire cell was photobleached and in other cases only half of the cell was bleached. In some experiments, differentiated cells (in DM for 2–4 h) were treated with 10 μg/ml nocodazole (Sigma-Aldrich) for 30 min or 10 μg/ml cycloheximide (Calbiochem) for 60 min before FRAP analysis. The average rate of translocation of peripherin mRNPs ( = 200) in five cells was obtained by monitoring the total distance traveled by each RNA particle during the entire image capture sequence and dividing it by the capture time, as previously described (). To analyze the relationship between translation and mRNP motility, double transfected cells that were differentiated for ∼4 h in NGF were treated with 10 μg/ml puromycin (Sigma-Aldrich) for 30–60 min. Six cells and 1,286 mRNPs were analyzed for motility in puromycin. Each video ranged from 3–10 min and rapid movement or disappearance of an mRNP out of the focal plane was also categorized as moving. , , and , and Fig. S1, using Photoshop software (Adobe) to decrease background noise. Fig. S1 (A–F) shows split images of overlays shown in . G and H are low and high magnification overlay images showing the distribution of endogenous peripherin particles and mRNA. Fig. S2 is a comparison of the average percentage of peripherin particles observed to associate with peripherin mRNA and the expected percentage based on chance alone, using different association criteria. Fig. S3 is a color-coded map showing peripherin mRNP distributions in a cell differentiated in the presence of nocodazole. A single deconvolved image from the primary image stack is also included. Videos 1–9 are of stills shown in , , , , and . Videos referred to in and another example of the experiment described in are shown as well. Online supplemental material is available at .

Directional cell migration is essential for development, chemotaxis, and wound healing and requires cell polarization in combination with a motility stimulus. Key to the cells' capacity to migrate is the dynamics of the microtubular and actin cytoskeleton. Whereas microtubules primarily play a role in dictating and sustaining cell polarity (; ; ; ), forward movement results from actin polymerization at the front and actomyosin-based contraction at the rear of the cell (). In parallel, the actin and microtubule cytoskeleton also orchestrates cell substrate adhesion to further promote directional migration (; ; ). Cytoskeletal dynamics and cell adhesion are controlled by small GTPases of the Rho family, in particular RhoA, Rac1, and Cdc42 (; ). These GTPases act as master switches, cycling between an inactive, guanosine diphosphate (GDP)–bound state and an active, GTP-bound state. This cycling is regulated through guanine nucleotide exchange factors (GEFs) that exchange bound GDP for GTP (), GTPase-activating proteins (GAPs) that promote the intrinsic GTP hydrolysis, and guanine nucleotide dissociating inhibitors (GDIs) that bind Rho-like GTPases through their COOH-terminal lipid anchor () and retain the GTPases in the cytosol (; del Poz et al., 2002). Activation of Rho GTPases is associated with membrane translocation (; ) and release from the cytosolic GDI protein. At the plasma membrane, Rho GTPases activate a wide range of effector proteins that regulate adhesion and migration as well as gene transcription (). It has been suggested that integrins indirectly regulate the recruitment of Rac1 to the membrane and promote the dissociation of RhoGDI, which would allow localized Rac1 activation and signaling (). In line with this, Rac1 association with the membrane was found to be independent of interactions with effector proteins (), suggesting that other regulatory proteins mediate membrane targeting. Membrane localization of Rho GTPases is mediated by the COOH-terminal lipid moiety. However, it is unlikely that membrane association through this lipid anchor mediates spatially restricted GTPase targeting and signaling. More likely, GTPases are specifically targeted to membrane-associated signaling complexes via protein–protein interactions (; ). Supporting this notion, the Rho-like GTPase Rac1 is activated at the leading edge of migrating cells, which is in line with its role in cell motility (; ). The related GTPase Cdc42 is also activated in cellular protrusions, albeit in a narrower, peripheral area compared with Rac1 (; ). This differentially localized GTPase activation will result in coordinated spatially confined signaling, which is required for directional cell migration. Our laboratory has previously shown that the hypervariable COOH-terminal domain of Rac1 mediates its association with the DAG kinase–phosphatidylinositol-4-phosphate 5-kinase complex as well as with adaptor proteins such as CrkII and Nck (; ). These protein interactions suggest that Rac1 and, most likely, Rho GTPases in general use their hypervariable COOH terminus to target to specific protein complexes. This is in agreement with studies showing that the COOH termini of small GTPases mediate both subcellular targeting () and signaling specificity (). In this study, we show that β-Pix (p21-activated kinase [Pak]–interacting exchange factor) can interact with the hypervariable domain of Rac1. β-Pix is a Rac/Cdc42GEF that also acts as a signaling organizer by binding to the ArfGAP paxillin kinase linker (PKL)/GIT as well as to the Rac effector Pak1 (, ; ). We demonstrate that Rac1, through the proline stretch in its COOH terminus, binds directly to the SH3 domain of β-Pix. This interaction mediates Rac1 targeting to membrane ruffles and to focal adhesions (FAs) as well as Rac1 activation and Rac1-mediated cell spreading. Finally, we show that Rac1 and Pak1 compete for binding to β-Pix and that Pak1 controls the Rac1–β-Pix interaction and adhesion-induced Rac1 activation. These data provide a model for Rac1 targeting through its activator β-Pix, which drives integrin-mediated localized Rac1 activation in polarized migrating cells. Recently, we showed that cell-permeable versions of the COOH termini of Rac1, RhoA, and Cdc42 act as selective inhibitors of the respective GTPases in a variety of cell types, presumably by interfering with the proper targeting of these GTPases (). We noticed that the COOH terminus of Rac1, which binds the SH3 domain of Crk (), shows homology to a SH3 domain–binding motif in Pak1 and SPIN90. Interestingly, both Pak1 and SPIN90 interact with the SH3 domain of Rac1/Cdc42 GEF β-Pix (; ), suggesting that Rac1 might also interact with β-Pix. Streptavidin-based pull-down assays with a series of biotinylated RhoGTPase COOH-terminal peptides using lysates from human embryonal kidney 293 (HEK293) or MDCKII cells confirmed that endogenous β-Pix specifically binds the COOH terminus of Rac1 but not of Rac2, Cdc42, or RhoA (). In contrast to β-Pix, the Rac GEFs Tiam1 () or Vav2 (not depicted) did not bind the COOH-terminal domain of Rac1. These data show that the Rac1 COOH terminus binds specifically to β-Pix. To demonstrate that β-Pix also binds full-length Rac1, we expressed either β-Pix or Rac1 in COS-7 cells and performed reciprocal pull-down assays with bacterially purified GST, GST-Rac1, or GST–β-Pix (). These experiments showed that β-Pix interacts with the full-length Rac1 protein and, conversely, that bacterially purified full-length Rac1 binds β-Pix. To test whether endogenous Rac1 binds to endogenous β-Pix, we performed pull-down assays with the biotinylated Crib domain of Pak1 (). This domain binds activated Rac from stimulated cells and was used instead of specific anti-Rac1 antibodies because these are directed against the Rac1 COOH terminus. These experiments were performed with cell lysates derived from cells in suspension or adherent to collagen to induce Rac1 activation (). We found that endogenous β-Pix coprecipitated with the Crib domain of Pak1 in lysates from the adherent cells (i.e., where Rac1 was activated) but not in lysates from the cells in suspension (). The latter finding also excludes the possibility that β-Pix interacts with the Pak-Crib domain directly. In addition, pull-down assays with GST-Rhotekin or GST–Wiskott-Aldrich syndrome protein, which bind active Rho or Cdc42, respectively, did not show any interaction with β-Pix (not depicted). To further characterize the Rac1–β-Pix interaction, we performed pull-down assays with purified GST-Rac1, GST-Rac1ΔC (a Rac1 COOH-terminal deletion mutant; ), and purified β-Pix in the absence or presence of the Rac1 COOH-terminal peptide or, as a control, the Pak-Crib peptide. Indeed, GST-Rac1 but not GST-Rac1ΔC binds to β-Pix (). Moreover, the binding between Rac1 and β-Pix was blocked by the Rac1 COOH-terminal peptide but not by a control peptide or the Pak1-Crib peptide (). These data show that the interaction between full-length Rac1 and β-Pix is direct and is mediated via the COOH terminus of Rac1 but not via its effector loop. The SH3 domain of β-Pix and a proline stretch in Pak1 mediate the Pak–β-Pix association (; ). To determine whether the homologous proline stretch in the COOH terminus of Rac1 is required for the interaction with β-Pix, we generated wild-type (WT) and activated Rac1 mutants in which the proline residues at positions 179, 180, and 181 were mutated to alanines (Rac1 P-A). Pull-down experiments with GST–β-Pix in lysates from COS-7 cells expressing either Rac1, Rac1 P-A, V12Rac1, or V12Rac1 P-A show that the P-A mutants no longer interact with β-Pix. This confirms that the proline stretch in the COOH terminus of Rac1 is required for the interaction with β-Pix and indicates that the interaction between Rac1 and β-Pix is nucleotide independent (). Subsequent pull-down experiments with WT GST-Rac1 in the presence of purified β-Pix or of a β-Pix SH3 domain mutant, β-PixW43K (), showed that β-Pix can only bind Rac1 when the SH3 domain of β-Pix is intact (). Moreover, we observed that the isolated β-Pix SH3 domain is sufficient for binding to purified Rac1 (). Together, these findings indicate that the direct interaction of Rac1 with β-Pix is mediated via the proline stretch in the COOH terminus of Rac1 and the SH3 domain of β-Pix. Moreover, the Rac1 sequence is specific for the β-Pix SH3 domain because the Rac2 or Cdc42 COOH-terminal peptides, which harbor potential SH3 domain–binding motifs, do not bind β-Pix (). To further show that the interaction between β-Pix and Rac1 is independent of the bound nucleotide, we performed pull-down assays with GST or with GST-Rac1 bound to GDP or GTP. We used COS-7 cell lysates expressing an exchange-deficient β-Pix mutant () to exclude a potential effect of β-Pix activity on the nucleotide-bound state of Rac1. shows that this β-Pix mutant binds both the GDP- and the GTP-bound forms of Rac1 to a similar extent, which shows that the Rac1–β-Pix interaction is independent of the Dbl homology (DH) domain of β-Pix and occurs with both inactive and activated Rac1. To explore the intracellular localization of the Rac1–β-Pix complex, we transfected HEK293 cells with HA-tagged Rac1 and seeded these cells on fibronectin for 1 h. Immunostainings for endogenous β-Pix and HA-tagged Rac1 showed that in the basal section of the cell, Rac1 and β-Pix colocalize to FAs (; ) particularly at the periphery of the cell (, basal and zoom). In the middle section of the cell, the two proteins colocalize markedly at the periphery but not in the center and colocalize most prominently in membrane ruffles (, middle). Rac1 is usually localized primarily in the cytosol as a result of its association to RhoGDI. However, activated Rac1 translocates to the plasma membrane in an integrin-dependent fashion (). In polarized cells, Rac1 activity has been found in the leading edge, which is in line with its function in actin polymerization, cell protrusion, and motility (; ). Given the fact that β-Pix binds Rac1 and, thus, may drive local Rac1 activation, we analyzed β-Pix localization in polarized cells. For these experiments, we cultured NIH3T3 cells on fibronectin-coated coverslips for 18 h. Cells were fixed and immunostained for endogenous β-Pix and for paxillin, a marker for FAs. We found that although paxillin is found in FAs both in the leading edge as well as in the contracting rear of polarized cells, β-Pix is concentrated in FAs in the leading edge and is virtually absent from the paxillin-positive FAs in the cell body or in the rear of the cell (). These data suggest that the differential localization of β-Pix in polarized cells drives localized activation of Rac1. To establish the functional relevance of the Rac1–β-Pix interaction, we generated HEK293 and MDCKII cells that express β-Pix at levels approximately twofold more than endogenous and incubated these cells briefly with the cell-permeable version of the Rac1 COOH-terminal peptide that blocks the Rac1–β-Pix interaction. Rac1 activity was subsequently determined using the Pak-Crib pull-down assay (). We found that the exogenous expression of β-Pix resulted in higher levels of active Rac1 in these cells, which was reduced in cells that were incubated with the Rac1 COOH-terminal peptide (). Importantly, the Rac1 COOH-terminal peptide does not interfere with the isolation of activated Rac1 by the Pak1-Crib peptide (). Furthermore, GFP–β-Pix–induced membrane ruffling in HEK293 cells was almost completely inhibited by the Rac1 COOH-terminal peptide but not by the control peptide or by a Rac1 P-A COOH-terminal peptide (), which cannot bind β-Pix (not depicted). Moreover, expression of the β-PixW43K mutant inhibited Rac1 activation compared with the expression of WT β-Pix, again indicating that the SH3 domain of β-Pix is indispensable for β-Pix–mediated Rac1 activation. In addition, when β-Pix–expressing cells were kept in suspension, Rac1 was not activated, suggesting that adhesion to an extracellular matrix is required for β-Pix–mediated Rac1 activation (unpublished data). To further test this, we reduced the levels of β-Pix in mouse embryonal fibroblasts (MEFs) using short inhibitory RNA (siRNA) expression and seeded the cells on fibronectin for 15 min, after which the Rac1 activity was assayed. Indeed, the cells with reduced β-Pix expression also showed reduced Rac1 activity after adhesion (), confirming that β-Pix mediates adhesion-induced Rac1 activation. Cell adhesion induces the activation of integrins and recruitment of paxillin (), which can associate with the GIT–β-Pix–Pak complex (), possibly driving localized Rac1 activation by β-Pix (,). Our observation that β-Pix binds specifically to Rac1 via its COOH terminus () suggests that β-Pix mediates the targeting of Rac1 to the membrane. To test this, we used a nonprenylated Rac1 mutant (V12Rac1C189S; ) that no longer localizes to the plasma membrane but accumulates in the nucleus (). However, when this Rac1 mutant is cotransfected with β-Pix, membrane localization of Rac1 is restored (). Furthermore, the membrane recruitment of V12Rac1C189S by β-Pix is blocked by the cell-permeable Rac1 COOH-terminal peptide (). This shows that the Rac1 COOH terminus by itself is not sufficient for the targeting of full-length nonprenylated Rac1 to the plasma membrane but that the interaction with β-Pix is both necessary and sufficient for the membrane targeting of Rac1, even in the absence of a lipid anchor. In the cytosol, Rac1 is tightly bound to RhoGDI, and, therefore, it is likely that RhoGDI is either released from Rac1 before Rac1 targeting to the membrane or that Rac1 and RhoGDI are targeted as a complex. Based on structural analysis, it was suggested that the COOH terminus of Rac1 is not required for the interaction of Rac1 with RhoGDI (; ). This, in turn, suggests that Rac1 can bind β-Pix through its COOH terminus even when complexed to RhoGDI. In line with this, GST–β-Pix pull-down experiments showed that RhoGDI indeed complexes with β-Pix. This interaction is blocked by the COOH-terminal Rac1 peptide (), indicating that the β-Pix–RhoGDI interaction is mediated via Rac1. Of note, the COOH-terminal Rac1 peptide does not bind RhoGDI and does not interfere with the Rac1–RhoGDI interaction (; unpublished data). These data suggest that Rac1 can be targeted to the membrane by β-Pix while in complex with RhoGDI. To further establish the Rac1 targeting to a β-Pix–GIT–paxillin complex, we performed GST-Rac1 pull-down experiments with lysates of COS-7 cells and found that endogenous paxillin associates with Rac1 (). This interaction is inhibited by the Rac1 COOH-terminal peptide but not by the Rac2 COOH-terminal peptide or by a control peptide (). Furthermore, we observed that GST-Rac1ΔC no longer associates with endogenous paxillin or PKL/GIT (). These data indicate that FA targeting of Rac1 is mediated by the β-Pix–GIT–paxillin complex and is dependent on the Rac1 COOH terminus. In line with this result, we found that V12Rac1 localizes to paxillin-containing FAs in spreading HEK293 cells, which is in contrast to the V12Rac1 P-A mutant (). Interestingly, the distribution of FAs in spreading cells that express V12Rac1 P-A appeared disorganized compared with the cells expressing V12Rac1. Moreover, the V12Rac1 P-A mutant showed no discrete localization in these cells, suggesting that the targeting of Rac1 to FAs requires their proper organization as well as an intact Rac1 COOH terminus. To assess whether the spreading of V12Rac1- or V12Rac1 P-A–expressing cells is quantitatively different, we seeded HEK293 cells expressing either empty vector, V12Rac1, or V12Rac1 P-A on fibronectin or collagen and monitored their spreading by means of electrical cell substrate impedance sensing (ECIS; ). This technique allows real-time analysis of cell spreading on gold electrodes coated with extracellular matrix proteins. V12Rac1-transfected cells spread two- to threefold more efficiently compared with the V12Rac1 P-A mutant or empty vector–transfected cells (). This was confirmed by phase-contrast imaging () as well as by staining for F-actin (). Together, these data show that Rac1 binding to β-Pix is important for Rac1 targeting to FAs and for efficient Rac1-mediated cell spreading on extracellular matrix proteins. β-Pix interacts via its SH3 domain not only with Rac1 () but also with Pak1 (; ). This suggests that Rac1 and Pak1 may compete for β-Pix binding, which could represent a means of regulating Rac1 binding to β-Pix. Therefore, association of the Rac1 COOH-terminal peptide with purified GST–β-Pix was analyzed in the presence or absence of purified GST-Pak1. The data show that GST–β-Pix directly binds to the Rac1 COOH-terminal peptide (, third lane) and that GST-Pak1 inhibits this interaction (, fourth lane). It is noteworthy that the Rac1 peptide does not interact with Pak1 (not depicted). Furthermore, we found that Pak1ΔCrib, a Pak1 mutant that cannot bind active Rac1 but can associate to the SH3 domain of β-Pix, blocks the interaction between full-length Rac1 and β-Pix (). These data show that Rac1 and Pak1 compete for binding to β-Pix. The competitive binding of Rac1 and Pak1 to β-Pix suggests that in the absence of Pak1, more β-Pix–Rac1 complexes could be formed, resulting in higher levels of activated Rac1. To test this, we used immortalized MEFs that were derived from Pak1-null (Pak1) and WT (Pak) mice and tested the Rac1 activity in these cells after seeding on fibronectin. We found that there is clearly more activated Rac1 present in the Pak1 cells compared with the WT cells (). The low but detectable levels of GTP-Cdc42 are similar in the Pak1 cells versus the WT cells (). Subsequent pull-down assays with the COOH-terminal Rac1 peptide showed that more β-Pix associates to the Rac1 COOH terminus in the Pak1 cells compared with the WT cells (). This indicates that in the absence of Pak1, there is more β-Pix available for Rac1 binding, resulting in increased Rac1 activation. The increased levels of activated Rac1 in the Pak1 cells suggest that these cells will spread more efficiently than the WT cells (). We tested this by seeding the WT and Pak1 cells on fibronectin-coated ECIS electrodes. In agreement with the aforementioned hypothesis, we found that the Pak1 cells spread more efficiently than the WT cells (). To confirm this, we seeded cells on fibronectin () or collagen (not depicted) and visualized the cells' morphology by phase-contrast microscopy. The Pak1 cells did indeed spread more efficiently, forming clearly detectable lamellipodia, whereas most WT cells were round with small membrane extensions (). Furthermore, reconstitution experiments with Pak1 WT or Pak1 P191A and P192A, Pak1 mutants that do not bind β-Pix (), showed that the enhanced spreading of Pak1 cells can be restored with WT Pak1 but not by the Pix-binding mutant (). These data suggest that Pak1 controls cell spreading by regulating the β-Pix–Rac1 interaction and, as a result, β-Pix–mediated Rac1 activation. To further confirm this, we reduced the expression of β-Pix by means of siRNA in WT and Pak1 cells and analyzed their spreading on fibronectin-coated ECIS electrodes. The enhanced spreading of Pak1 cells was reduced when β-Pix expression is knocked down (). Similarly, reducing the expression of β-Pix in WT cells also impaired their spreading on fibronectin (). This indicates that β-Pix plays a pivotal role in the enhanced spreading of the Pak1 and WT cells. The interaction between Pak1 and β-Pix is negatively regulated by the autophosphorylation of Pak1 (; ). Because activated Cdc42 promotes Pak1 autophosphorylation and activation (), Cdc42 could induce the dissociation of Pak1 from β-Pix to allow Rac1–β-Pix binding and targeting of Rac1 to the membrane. In line with this, we observed that V12Cdc42 induced the translocation of V12Rac1C189S to the plasma membrane. This confirms that Cdc42 activation promotes Rac1 membrane targeting (), which is consistent with the observation that the expression of V12Cdc42 results in the activation of Rac1 (). In this study, we describe a novel mode of interaction between a Rho-like GTPase and its GEF (i.e., Rac1 and β-Pix). We show that Rac1 interacts specifically and directly through its hypervariable COOH-terminal domain with the SH3 domain of β-Pix ( and ). The data further show that Rac1 and its effector Pak1 act as competitors for β-Pix binding. Importantly, this Rac1–β-Pix interaction is required for Rac1 targeting to the membrane and to FAs, which supports the notion that β-Pix plays an important role in the integrin-mediated recruitment and localized activation of Rac1. Rho-like GTPases signal at membranes, where the relevant GEFs as well as their downstream targets are localized. A recent in vivo study has demonstrated that the COOH termini of Rac1 and Rac2 dictate both subcellular localization and differential signaling (). Earlier studies had already shown that the COOH termini of a large number of small GTPases mediate differential intracellular localization, which is suggestive of targeting through specific protein interactions (; ). Our finding that the COOH terminus of Rac1 mediates specific protein–protein interactions that regulate its intracellular targeting and signaling may therefore apply to many, if not all, small GTPases. Indeed, recent studies in our laboratory show that the COOH termini of the different Rho-GTPases have their own, only partially overlapping repertoire of specific interactors (unpublished data). As for R-Ras and Nck () or Cdc42/Rac1 and CAPRI (), the interaction between Rac1 and β-Pix is nucleotide independent. In line with this finding, we found no competition of the Rac1–β-Pix interaction by the Crib domain of Pak1, which binds activated Rac1. Moreover, we found that Rac1 also associates with a GEF-deficient form of β-Pix. In addition, the Rac P-A mutant, which does not bind β-Pix, was not targeted to paxillin-containing FAs and was no longer able to promote adhesion-induced cell spreading. Finally, Rac1 activation by β-Pix was abrogated when their interaction was blocked by the Rac1 COOH-terminal peptide or when the β-Pix W43K mutant was used. Together, these results indicate that binding of the Rac1 COOH terminus to the SH3 domain of β-Pix is essential for Rac1 targeting and for subsequent localized activation. In resting cells, Rho GTPases are associated with RhoGDI, which retains the GTPases in their inactive state in the cytosol (). The fact that we could isolate a GDI–Rac–β-Pix complex suggests that GDI may be recruited along with Rac1 to the plasma membrane and to FAs, where it will dissociate from Rac1 to allow β-Pix–mediated nucleotide exchange. RhoGDI is likely to sterically hinder nucleotide exchange because it binds the switch I and II regions of Rac1, which are also required for interaction with the DH–pleckstrin homology domain of the Rac1 GEF Tiam1 and likely for other GEFs as well (; ). In addition, association to GDI blocks Rac1 effector binding (), preventing signaling before GDI dissociation and activation. The Rac1 effector Pak1 has been implicated in Rac1 activation, as it can phosphorylate RhoGDI, leading to its dissociation from Rac1 (). However, our finding that Rac1 activity is increased in Pak1 cells suggests that alternative means or kinases exist to dissociate Rac1 from RhoGDI. The notion that β-Pix interacts with Pak1 through its SH3 domain (; ) suggests that Rac1 and Pak1 act as competitive binding partners for β-Pix. Indeed, we observed that purified WT Pak1 as well as a Pak1 mutant that does not bind activated Rac1 interferes with the β-Pix–Rac1 interaction in vitro. In line with this, we found more β-Pix binding to the COOH terminus of Rac1 and, consequently, more Rac1 activation in Pak1 cells. Interestingly, the Pak1 cells show an increase in cell spreading, suggesting that Pak1 is not essential for Rac1-induced cell spreading. It is unknown how the competition of Rac1 and Pak1 for binding to the SH3 domain of β-Pix is regulated. However, activation and autophosphorylation of Pak1 blocks its binding to β-Pix (; ) and leads to its dissociation from FAs (). This indicates that Pak activation would allow the association of Rac1 to β-Pix. Pak autophosphorylation can be induced by activated Cdc42 (), suggesting that the Pak1–β-Pix complex mediates Cdc42-induced Rac1 activation (). Activated Cdc42, which localizes to the leading edge of motile cells (), has also been shown to recruit β-Pix to peripheral focal contacts (). In agreement with these findings, we found that V12RacC189S is recruited from the nucleus to the membrane both by β-Pix as well as by activated Cdc42. β-Pix is found at FAs and associates through PKL with the FA protein paxillin. We found β-Pix in particular at FAs in the leading edge of polarized cells, which corresponds with earlier findings on the localized activation of Rac1 in the front of migrating cells (). Intriguingly, we observed that paxillin-positive FAs in the center and at the rear of the cell were virtually devoid of β-Pix. This suggests that leading edge FAs have a preference for recruiting β-Pix. These findings are in agreement with earlier results showing the preferential localization of activated Pak1 in lamellipodia and FAs in the leading edge of motile fibroblasts (). Furthermore, we show that reducing the levels of β-Pix results in less adhesion-induced Rac1 activation and cell spreading, indicating that β-Pix is pivotal for integrin-mediated Rac1 activation (), specifically at the leading edge of cells. recently showed that phosphorylation of α4 integrins in the leading edge of polarized cells modulates their association with paxillin. As a result, paxillin and the ArfGAP PKL/GIT become preferentially associated with α4 integrins in the back of the cell. The ArfGAP was shown to be essential for inhibiting Arf6 and Rac activation in the lateral side and the rear of the cell. These findings are in agreement with our data on the preferential localization of β-Pix in FAs at the leading edge, albeit that other paxillin-binding integrins may be important in these cells because the α4 integrin is expressed at low levels in fibroblasts. Together, these differential integrin-dependent signaling cascades may explain the polarized activation of Rac1 in the leading edge of migrating cells. In conclusion, our data support a model () in which integrin engagement leads to the activation of Cdc42 in the leading edge of motile cells. Activated Cdc42 will subsequently activate FA-associated Pak1, which is bound to β-Pix. Activated Pak dissociates from β-Pix, allowing Rac1 recruitment to FAs through the β-Pix SH3 domain. This is then followed by Rac1 activation, which will trigger actin polymerization, membrane protrusion, and directional cell migration. Myc–β-Pix WT and W43K () were cloned in pEGFP-C1 (CLONTECH Laboratories, Inc.) using BamHI and EcoRI restriction sites and in GST-6P-1 (27–4597-01; GE Healthcare) using BamHI and NotI. The exchange-deficient β-Pix mutant was a gift from E. Manser (Institute of Molecular and Cell Biology, Singapore). Rac1-HA (Guthrie Healthcare System), V12Rac1C189S (gift from M.A. Schwartz, University of Virginia, Charlottesville, VA; ), V12Cdc42-myc (), GST-Rac1 and GST-Rac1ΔC (gift from R. Ahmadian, European Molecular Biology Laboratory, Heidelberg, Germany; ), GST-Pak1 (), and GST-Pak1ΔCrib () and Rac1 P-A mutants were generated with the QuikChange Site-Directed Mutagenesis Kit (200518; Stratagene). The sequences for β-Pix siRNA was GUCCCAGGAUGAGUGGUUU (Eurogentec) and for control siRNA was AUCAUUAGCAUCAAACGUC (Eurogentec). Synthetic peptides were synthesized on a peptide synthesizer (Syro II; MultiSynTech) using Fmoc solid phase chemistry and encoded a protein transduction domain (Biotin-YARAAARQARAG; ) followed by the 10 amino acids preceding the CAAX motif of Rac1, Rac2, Rac3, RhoA, RhoB, and Cdc42. The sequence of the Rac1 P-A peptide is CAAAVKKRKRK. The HEK293, MDCKII, and COS-7 cell lines were maintained in Iscove's Modified Dulbecco's Medium (IMDM; BioWhittaker) containing 10% heat-inactivated FCS (Invitrogen) at 37°C and 5% CO. Cells were passaged by trypsinization. HEK293 and COS-7 cells were transiently transfected with FuGENE (Roche). For this, we mixed 2 μg DNA and 6 μl FuGENE in 100 μl IMDM that was incubated for 15 min at RT. Subsequently, 2 ml IMDM with FCS containing 500,000 cells were added to the DNA–FuGENE mixture, which was then incubated in a six-well plate for 6 h. Transfections of siRNAs were performed with LipofectAMINE 2000 (Invitrogen). In short, 25 μl siRNA (100 μM) and 25 μl LipofectAMINE 2000 were separately diluted in 1,250 μl IMDM, after which the two solutions were mixed and incubated for 30 min at RT. Subsequently, 10 ml of cell suspension (500,000 cells/ml in IMDM/10% FCS) was added, and cells were plated in a 10-cm dish. After 6 h, the medium was replaced with fresh IMDM/10% FCS. HEK293 and MDCKII cells with stable expression of β-Pix, V12Rac1, and V12Rac1 P-A were generated by retroviral transduction. Transfection and production of amphotropic retroviruses were described previously (). MEFs were isolated from day 13.5 WT or Pak1 embryos. WT and Pak1 MEFs were immortalized with SV40 T antigen and were maintained in DME supplemented with 10% FBS. To assay the binding of β-Pix, 10 × 10 HEK293 or COS-7 cells were seeded 1 d before the experiment. Cells were lysed in lysis buffer A (50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 10 mM MgCl2, 10% glycerol, 1% NP-40, and 0.1% SDS) and centrifuged for 10 min at 14,000 rpm and 4°C. The supernatant was then incubated with the indicated COOH-terminal peptides (5 μg) in the presence of streptavidin-coated beads (Sigma-Aldrich) at 4°C for 1 h while rotating. Beads were washed five times in lysis buffer B (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl2, 10% glycerol, and 1% NP-40) and resuspended in 25 μl SDS sample buffer. β-Pix association was determined by Western blot analysis. To determine Rac activity, 5 × 10 HEK293 cells were seeded on fibronectin-coated dishes for 1 h. After a 15-min incubation with the different COOH-terminal peptides (200 μg/ml), GTP-bound Rac1 was isolated with biotinylated Pak1-Crib peptide (). Rac1 binding was detected by Western blotting. 24 h after transfection, cells were seeded for the indicated time on either fibronectin- or collagen-coated glass coverslips and fixed by 3.7% formaldehyde (Merck) in PBS for 5 min and permeabilized with 0.5% Triton X-100 in PBS. Immunostainings were performed at 37°C for 1 h with the indicated antibodies. Fluorescent imaging was performed with a confocal laser scanning microscope (Axiovert 100 M; Carl Zeiss MicroImaging, Inc.) using a 63×/NA 1.40 oil lens (Carl Zeiss MicroImaging, Inc.). Image acquisition was performed with LSM 510 software (Carl Zeiss MicroImaging, Inc.). For phase-contrast microscopy, cells were seeded on fibronectin- or collagen-coated dishes and fixed at the indicated time points. Images were obtained with a microscope (model DMIL; Leica), L20/0.30 PH1 lens, a camera (DC 300; Leica), and IM500 software (Leica). For ECIS-based adhesion experiments, gold ECIS electrodes were coated with either fibronectin or collagen in 0.9% NaCl for 1 h at 37°C. Next, HEK293 or MEF cells were seeded at a concentration of 400,000 cells per well in 400 μl IMDM with 10% FCS. ECIS was subsequently monitored for up to 2 h with the ECIS equipment (Applied Biophysics).

Cell locomotion is critical for normal embryonic development, cell-mediated immunity, and wound healing, as well as for pathological conditions such as inflammation and tumor metastasis. The process of cell motility depends on the ability of a cell to extend a leading edge in a polarized fashion and to adhere to an underlying substratum. Integrin receptors for extracellular matrix are concentrated at sites of substratum attachment, providing the structural support necessary for lamellar protrusion. Recent work has revealed that substratum adhesion sites display functional heterogeneity, with the small adhesions near the leading edge of a cell representing the sites of greatest mechanical force transmission to the substratum (). Heterogeneity in the molecular composition of adhesion sites is likely to account for their functional diversity (; ). In mammalian cells, one component of substratum adhesions that is associated with mature focal adhesions, but not with Rac-induced nascent adhesions that are concentrated at the leading edge, is zyxin (). Zyxin is a LIM domain protein that has been implicated in the regulation of actin assembly, cell division control, and nuclear–cytoplasmic communication (; ; ). A role for zyxin in actin assembly was proposed when it was recognized that a series of proline-rich repeats in the ActA protein of the intracellular pathogen are also present in zyxin (; ). The FPPPP sequences present in ActA stimulate actin assembly on the pathogen's surface and promote bacterial motility within the cytoplasm of a eukaryotic host. A common mechanism by which zyxin and ActA are thought to promote actin polymerization involves recruitment of Ena/VASP proteins, which are proposed to act by recruiting profilin-actin and by controlling the access of capping protein to the barbed ends of actin filaments (; ). Imaging of zyxin in living cells has revealed that zyxin accumulation at adhesion sites is highly dynamic. Zyxin levels are low near the leading edge of a migrating cell, where high tension and propulsive forces are generated (). Moreover, the accumulation of zyxin at substratum adhesion sites is correlated with cessation of membrane protrusion () and with a dramatic reduction in traction forces (). Because of these intriguing observations that suggest a link between the presence of zyxin and the properties of substratum adhesion sites, we sought to assess the consequences of eliminating zyxin function for cell adhesion, motility, and actin organization. Murine zyxin is a 564-aa LIM protein () whose binding partner repertoire had suggested a role in actin assembly, cell cycle control, cell motility, and cell signaling. To directly assess the contributions of zyxin to cell function, we have isolated fibroblasts from mice in which the gene is disrupted by homologous recombination (). Neither full-length nor truncated zyxin protein is present in the zyxin−/− cells (). Loss of gene function can lead to compensatory up-regulation of other family members. Two proteins that are closely related to zyxin, the lipoma preferred partner (LPP; ) and thyroid receptor-interacting protein 6 (TRIP6; ), are also expressed in fibroblasts. We detected no alteration in the level of expression of either LPP or TRIP6 in the zyxin−/− cells (), which is consistent with the view that cells do not compensate for the elimination of zyxin by modulating the expression of these highly related proteins. Some focal adhesion constituents depend on the presence of their binding partners for their stability (). Therefore, we examined whether the levels of zyxin-binding partners were compromised when zyxin was absent and detected no differences in the levels of several zyxin-binding partners (). Likewise, other focal adhesion proteins including vinculin, talin, integrin-linked kinase (ILK), FAK, paxillin, and src were detected at similar levels in wild-type and zyxin−/− cells (). Zyxin displays mitosis-dependent phosphorylation and has been shown to interact directly with the h-warts/LATS1 tumor suppressor, a serine/threonine kinase implicated in cell cycle control (). Disturbance of the LATS1–zyxin interaction by use of a dominant-interfering LATS1 truncation product (LATS1 aa 136–700) was shown to interfere with zyxin localization during mitosis and delay mitotic progression, leading to the suggestion that LATS1-dependent recruitment of zyxin to the mitotic apparatus is required for regulated exit from mitosis (). With the availability of zyxin−/− cells, we were able to test directly whether zyxin is essential for normal mitotic progression. We detected no change in LATS1 protein levels in the zyxin−/− cells (); therefore, zyxin is not essential for LATS1 stability. To look for specific changes in the cell cycle caused by loss of zyxin, flow cytometric analysis of cell cycle progression in exponentially growing wild-type and zyxin−/− cells was used (). No statistically significant differences in cell cycle progression () or the percentage of binucleate cells (not depicted) were observed. These results illustrate that, at least in fibroblasts, mitotic progression and cytokinesis can proceed normally in the absence of zyxin. Many of zyxin's binding partners and biochemical activities suggest that it may play a role in integrin- and actin-linked processes. Therefore, we focused the balance of our analysis on the possible role of zyxin in motility, adhesion, and cytoarchitecture. To determine if loss of zyxin altered cell motility, the behavior of zyxin−/− cells was evaluated using a monolayer wound assay. In blind studies, zyxin−/− fibroblasts were consistently observed to reach the midline of the wound sooner than the wild-type cells (). Quantitative analysis revealed that the zyxin−/− cells displayed a mean migration velocity of 32 μm/h compared with the wild-type migration rate of 18 μm/h (). The enhanced migration of zyxin−/− fibroblasts suggests that zyxin acts as a negative regulator of cell motility. Integrin-dependent adhesion is essential for communicating intracellular contractility to the underlying extracellular matrix and is necessary to drive motility. To explore the contribution of zyxin to cell motility in greater detail, we examined the haptotactic migration of wild-type and zyxin−/− cells toward a variety of integrin ligands using a Boyden chamber transwell migration assay (). We observed a striking difference in haptotactic motility when comparing wild-type and zyxin−/− cells. Except in the cases of 20 μg/ml fibronectin or vitronectin, the zyxin-null cells displayed statistically significant enhancement of migration, relative to wild-type cells. Matrix proteins have differential, concentration-dependent capacities to stimulate cell migration. As can be seen in , migration of wild-type fibroblasts exhibited sensitivity to fibronectin concentration, whereas zyxin-null fibroblasts migrated efficiently independent of fibronectin concentration. The migration of the zyxin-null cells in the absence of matrix cues was statistically indistinguishable from the maximal migration of wild-type cells toward fibronectin. Zyxin-null cells also displayed enhanced migratory potential relative to their wild-type counterparts in a Matrigel invasion assay (). Introduction of a GFP-tagged version of zyxin into the null cells resulted in suppression of cell migration (). GFP-zyxin localized as expected at focal adhesions in the zyxin−/− cells () and was expressed at levels approximating that of zyxin in wild-type cells (). GFP, which did not affect motility, displayed diffuse cellular localization (not depicted). The successful rescue of the null cells by reintroduction of a zyxin transgene illustrates that the migration phenotype observed for the zyxin−/− cells is directly attributable to loss of zyxin. We next tested whether altered cell–substratum adhesion might account for the enhanced motility of the zyxin−/− cells. Zyxin-null fibroblasts were more adherent than wild-type fibroblasts when plated on a variety of extracellular matrix proteins (), and reexpression of GFP-zyxin in null cells resulted in reduced adhesion (). One possible explanation for the enhanced cell adhesion observed in the zyxin−/− cells could be elevated integrin expression. However, immunoblot analysis revealed no difference in integrin levels or subunit expression (). The behavior of zyxin-null fibroblasts on integrin ligands suggested that cells lacking zyxin might somehow be primed for integrin-dependent adhesion and migration. One molecular hallmark of the integrin activation state is enhanced tyrosine phosphorylation of proteins such as FAK, which is a tyrosine kinase that colocalizes with zyxin at focal adhesions. Therefore, we compared the subcellular distributions and molecular mass profiles of tyrosine-phosphorylated proteins in wild-type and zyxin-null cells. Both wild-type and zyxin-null cells displayed phosphotyrosine (pY) associated with focal adhesions (), and Western immunoblot analysis of pY-containing proteins revealed no apparent difference in the region where FAK migrates (). However, the wild-type cells displayed a 75-kD tyrosine-phosphorylated protein (pY75) that showed reduced abundance in zyxin-null cells. In contrast, we observed an 80-kD tyrosine-phosphorylated protein (pY80) in the zyxin-null cells that was not evident in the wild-type cells. Using two-dimensional (2D) gel electrophoresis of wild-type and zyxin-null cell lysates, followed by Western immunoblot with antiphosphotyrosine antibody, both pY75 and pY80 were resolved as a cluster of isoelectric point variants ranging from ∼6.5 to 8.0 (). Mass spectrometric (LC-MS/MS) sequence analysis of the protein found four pY75 2D gel spots derived from wild-type cells and four pY80 2D gel spots derived from zyxin-null cells, such as those in , and identified all eight tyrosine-phosphorylated spots as caldesmon (). Caldesmon is an actomyosin regulator that influences stress fiber formation, cell contractility, and cell motility (). There are two major isoforms of caldesmon derived by alternative splicing (): nonmuscle or low molecular weight caldesmon (l-caldesmon) and smooth muscle or high molecular weight caldesmon (h-caldesmon; ). Our sequence results are consistent with recovery of l-caldesmon. With the exception of sequence encoded by exon 1′, which was only obtained for pY75 isolated from wild-type cells, sequence from all other classical l-caldesmon exons (; ) were represented and were nearly identical for pY75 and pY80 (). In wild-type mouse fibroblasts, a caldesmon-specific antibody detected a prominent 75-kD immunoreactive band, as well as a minor 80-kD immunoreactive band, which is consistent with l-caldesmon, whereas zyxin-null fibroblasts displayed only the slower migrating 80-kD species (). A similar quantitative shift to a slower mobility species occurred for the 140-kD h-caldesmon in the zyxin-null fibroblasts, as seen in a longer exposure (, right). This shift in caldesmon mobility was detected in multiple zyxin-null primary fibroblast isolates, as well as in immortalized zyxin-null fibroblast lines (unpublished data). Moreover, the caldesmon mobility shift was observed in cells derived from zyxin knock-out mice generated from two independently targeted embryonic stem cell lines (not depicted) and in a variety of tissues, including lung, bladder, and spleen, which were isolated directly from the zyxin-null mice (). The slower migrating caldesmon isoform is evident only in tissues derived from homozygous mutant animals, suggesting that complete loss of zyxin is required to promote the change in caldesmon. Phosphorylation can affect a protein's electrophoretic mobility and, in the case of caldesmon, is known to regulate its ability to control contractility and integrin-dependent motility (). We tested directly whether l-caldesmon from wild-type and zyxin-null cells was differentially phosphorylated on Ser497 and Ser527, two ERK-dependent phosphorylation sites implicated in regulation of caldesmon function (). Both the 75- and 80-kD caldesmon species were recognized with the phosphospecific antibody (). For a more global assessment of whether differential phosphorylation might be responsible for the caldesmon mobility difference we have observed, we treated cell lysates from wild-type and zyxin-null cells with phosphatase and assayed caldesmon and control paxillin protein mobility by SDS-PAGE (). Although the phosphatase treatment resulted in the collapse of paxillin bands indicative of dephosphorylation, no impact on caldesmon mobility was observed, suggesting that phosphorylation is unlikely to account for the altered caldesmon mobility seen in the zyxin-null cells. To directly test whether caldesmon is subject to differential posttranslational modification that affects its mobility, we transfected wild-type and zyxin-null cells with a his-caldesmon cDNA expression construct and monitored mobility of the expressed protein with Western immunoblot. The mobility of the his-tagged caldesmon was indistinguishable in wild-type and zyxin-null cells, whereas the endogenous caldesmon exhibited the mobility shift (). Collectively, our results suggest that differential posttranslational modification of caldesmon is not likely to be responsible for the altered caldesmon mobility we observe in the zyxin-null cells. Re-expression of zyxin does not restore caldesmon to its wild-type mobility (), which illustrates that the phenotypes observed in the zyxin-null cells are linked to loss of zyxin and are not attributable to alteration of caldesmon. Zyxin is concentrated in the focal adhesions of wild-type fibroblasts (); no immunoreactivity is detected in zyxin-null cells (). Elimination of zyxin did not lead to changes in focal adhesion morphology or number, as assessed by qualitative inspection of vinculin localization (). The distributions of several other focal adhesion markers, including FAK and paxillin, also appeared unperturbed in the zyxin-null cells (unpublished data). In contrast, we detected a dramatic diminution in the localization of the zyxin-binding partners, Mena and VASP, at focal adhesions (; and Fig. S1, available at ) even though their cellular levels were unchanged (). In double-labeling experiments, it was evident that Mena colocalized with vinculin at focal adhesions of wild-type cells () and was depleted from vinculin-rich focal adhesions in zyxin-null cells (). Thus, the altered Mena localization was not attributable to loss of focal adhesion structures. Expression of GFP-zyxin in the null cells restored localization of Mena to vinculin-rich focal adhesions (). Similar results were obtained for VASP (unpublished data). These observations are consistent with a major role for zyxin in localizing Ena/VASP family members at focal adhesions. Residual Mena and VASP accumulation at focal adhesions in the zyxin−/− cells likely reflects the presence of other focal adhesion proteins that can dock Ena/VASP family members. For example, both vinculin and LPP are present in focal adhesions and have the capacity to bind Ena/VASP proteins (; ). Interestingly, LPP was detected more prominently at focal adhesions of zyxin-null cells, as compared with cocultured wild-type cells (). The mislocalization of Ena/VASP proteins and the alteration of caldesmon observed in the zyxin-null cells raised the possibility that actin filament assembly or organization might be disturbed when zyxin function is compromised. Likewise, previous work involving expression of a dominant-negative zyxin deletion fragment (), RNA interference (), and exposure of cells to mechanical stimuli () suggested that zyxin might be required for actin cytoskeletal assembly or maintenance. To assess this possibility, we compared the actin cytoskeletons of wild-type and zyxin-null cells using phalloidin staining (). The presence of actin filament arrays in the zyxin-null fibroblasts illustrated that zyxin was not absolutely essential for establishment and maintenance of actin stress fibers. To explore whether we might detect an effect of eliminating zyxin when cells were challenged to reorganize actin arrays, we compared the response of wild-type and zyxin-null fibroblasts with the membrane-permeable drug jasplakinolide which stabilizes actin filaments and decreases the critical concentration of monomers required for polymerization (). Jasplakinolide-treated wild-type fibroblasts developed robust actin filament arrays detected by phalloidin staining, but jasplakinolide-treated zyxin-null fibroblasts did not exhibit such robust stress fiber arrays (). To quantitate a stress fiber thickness index (SFTI), we used an “erosion” spatial filtering approach to calculate decay constants of brightness curves (; see Materials and methods). This analysis confirmed that there was no statistically significant difference in phalloidin-stained actin filaments between untreated wild-type and zyxin-null fibroblasts (). However, jasplakinolide-induced actin stress fibers in wild-type fibroblasts were thicker than the fibers in zyxin-null fibroblasts (). When GFP-zyxin was expressed in the zyxin-null cells, the ability of the cells to generate robust actin filament arrays in response to jasplakinolide was restored. Zyxin-null cells expressing GFP-zyxin () showed more robust actin stress fibers () after exposure to jasplakinolide than adjacent cells that were not expressing the transgene. Rescue of the zyxin-null cells' ability to build robust stress fibers by expression of GFP-zyxin was confirmed by quantitative SFTI analysis (). To begin to assess how zyxin might contribute to the cellular response to jasplakinolide, we examined the behavior of zyxin in jasplakinolide-treated cells. As can be seen in , zyxin underwent a dramatic redistribution in response to exposure of cells to jasplakinolide. Zyxin was lost from focal adhesions and accumulated along the actin stress fibers (). Under those conditions, zyxin was colocalized with caldesmon along actin filaments (). In untreated fibroblasts, zyxin was colocalized with vinculin at focal adhesions (). In contrast to zyxin, which left focal adhesions in response to jasplakinolide, vinculin was retained at focal adhesions (), illustrating the specificity of zyxin's redistribution and also illustrating that the focal adhesions were not disassembling in response to jasplakinolide. In contrast with vinculin, which was retained at focal adhesions in cells exposed to jasplakinolide, Ena/VASP proteins redistributed from focal adhesions to actin filaments concomitant with zyxin redistribution (). This VASP redistribution to actin stress fibers depended on the presence of zyxin (). VASP accumulation in lamellar extensions was zyxin independent (). c a l a d h e s i o n s s e r v e a s s i t e s o f s u b s t r a t u m a t t a c h m e n t , s i g n a l t r a n s d u c t i o n , a n d a c t i n f i l a m e n t a n c h o r a g e . W e h a v e d e m o n s t r a t e d t h a t e l i m i n a t i o n o f t h e L I M p r o t e i n z y x i n b y t a r g e t e d g e n e d i s r u p t i o n a f f e c t s c e l l a d h e s i o n , m i g r a t i o n , a n d s t r e s s f i b e r r e m o d e l i n g . O u r a n a l y s i s i d e n t i f i e s s e v e r a l m o l e c u l a r a l t e r a t i o n s i n t h e z y x i n - n u l l c e l l s t h a t p r o v i d e i n s i g h t i n t o t h e m e c h a n i s m b y w h i c h l o s s o f z y x i n a f f e c t s t h e s e p r o c e s s e s . Commercial antibody sources included ICN Biomedicals (α-actinin), Sigma-Aldrich (talin and vinculin), BD Biosciences (caldesmon, ILK, paxillin, and integrins α3, αV, β1, β3), Upstate Cell Signaling (FAK 4–47, pY-4G10, and src), Santa Cruz Biotechnology, Inc. (FAK C-20 and pY-20), CHEMICON International, Inc. (integrin α5), Cell Signaling Technology (p-Tyr-100), and Qbiogene (VASP). Zyxin (B71, B72, and 1D1), TRIP6 (B65; ), and phospho-Ser527/789-caldesmon () antibodies were previously described. Other investigators supplied anti-LPP MP2 (M. Petit and J.M. Van de Ven, University of Leuven, Leuven, Belgium; ), anti-Mena 2197 and anti-VASP 2010 (F. Gertler, Massachusetts Institute of Technology, Boston, MA; ), and anti-warts kinase/LATS C2 (T. Hirota and H. Saya, Kumamoto University, Hanjo, Japan; ). Primary mouse embryonic fibroblasts (MEFs) were derived from day 14 embryos of wild-type or zyxin-null mice (backcrossed nine generations into the inbred C57BL/6 line [Charles River Laboratories] to ensure genetic homogeneity, except at the zyxin locus; ). Primary MEFs were expanded and frozen as passage 2 stocks. For experiments, the cells were thawed and used within a week (passages 4–6) as early-passage MEFs. To generate immortalized cell lines, cells were cultured from tissue (torso) explants of newborn mice. After spontaneous immortalization (generally passages 8–12), viable cell lines were isolated using cloning cylinders, expanded, and frozen for future use. All cells were grown in high glucose DME supplemented with pyruvate, glutamine, penicillin/streptomycin (Invitrogen), 10% FBS (HyClone) at 37°C, and 5% CO. Cells were treated with 100 nM of jasplakinolide or DMSO control for 2 h in DME supplemented with 10% FBS (). Unless specified in the figure legend, experiments used both primary and immortalized cell lines, with similar results obtained for both. Immortalized cells were used for rescue experiments. An enhanced GFP-mouse zyxin DNA was inserted into the retroviral vector LINXv-myc () and Phoenix-Eco cells (American Type Culture Collection) were used to generate retrovirus. Zyxin-null fibroblast cell lines were infected and sorted for GFP to select enhanced GFP-zyxin-myc–expressing cells. Caldesmon-6histidine DNA was inserted into pBABEpuro and transiently transfected with Lipofectamine 2000 (Invitrogen). Propidium iodide staining of DNA and flow cytometry followed established procedures () using a FACScan instrument and CellQuest software (Becton Dickinson). Adhesion assays were performed essentially as previously described (), with 10 cells seeded for 25 min into ECM-coated wells, and were performed four times with two pairs of independently derived wild-type and zyxin−/− cell lines. Monolayer wound assay was performed as previously described (), with confluent MEFs in optical 35-mm plates on a 37°C heated stage apparatus (Bioptechs) monitored by time-lapse (every 15 min) phase-contrast microscopy with an inverted microscope (TE300; Nikon), a 10× DIC-L objective, 0.3 NA, a 5-Mhz digital camera (MicroMax; Princeton Instruments), and QED software (Media Cybernetics). Haptotactic cell migrations on Boyden chamber transwells (24-well cell culture inserts of polyethylene terephthalate membranes with 8-μm pores; Becton Dickinson) moved from serum-free DME to DME supplemented with 10% FBS over 6 h (). BioCoat Matrigel invasion chambers with 8-μm pores (BD Biosciences) in 24-well plates were seeded with 15,000 or 30,000 cells for 4- or 24-h migration. Migratory cells in five fields per well (20× phase objective) were counted for three individual wells per condition. Cell staining followed established procedures, using the primary antibodies described in Antibodies and Alexa Fluor secondary antibodies and phalloidin (Invitrogen). Cell images were captured at room temperature using a fluorescent microscope (AxioPhot; Carl Zeiss MicroImaging Inc.) with a Plan-Apochromat 63×, 1.40 NA, oil objective and 100×, 1.30 NA, and 40×, 1.30 NA, objectives, a MicroMax digital camera (Princeton Instruments), and Openlab imaging software (Improvision), and were incorporated into figures using Photoshop 8.0 and Illustrator 11.0 software (Adobe). Cell and tissue lysates were prepared, electrophoresed, and immunoblotted as previously described (). 2D gel analysis used the Protean IEF Cell system with immobilized pH gradient gel strips (pH 3–10; Bio-Rad Laboratories) followed by denaturing 10% acrylamide gels. For MS analysis, proteins were stained with Colloidal blue (Invitrogen), excised from the gel, washed twice in 50% acetonitrile/HO, and then frozen. Peptide digestion, purification, tandem array-liquid chromatography LC/MS-MS, and sequence analysis were performed at the University of Utah Mass Spectrometry Core Facility (Salt Lake City, UT) and at the Harvard Microchemistry Facility (Cambridge, MA) by microcapillary reverse-phase HPLC nanoelectrospray tandem mass spectrometry (μLC MS/MS) on a quadrupole ion trap mass spectrometer (Finnigan LCQ DECA XP; Thermo Electron Corporation). SFTI was quantitated based on erosion spatial filtering (). An erosion filter is equal to the ninth rank filter, which sorts 9 pixels in a 3 × 3 pixel matrix by their pixel brightness and replaces the center pixel with the darkest value found in the matrix. This “erodes” contours of objects in the digital image. Decay curves of averaged brightness during filtration cycles reflect thickness of the original lines. Nonlinear curve fit was applied to compute decay constants of the brightness curves and were termed SFTI. This method was superior to other methods based on line profiles because it reflected thickness of lines in areas, rather than at points. Data analysis included analysis of variance between groups and unpaired tests using GraphPad Prism software (GraphPad Software). A P value of <0.05 was considered statistically significant. Additional examples of the Mena and VASP mislocalization in the zyxin-null fibroblasts are available in Fig. S1. Online supplemental material is available at .

xref #text xref italic #text Many years after the description of GCGs, another RNA granule was observed in the cytoplasm of tomato cells subjected to heat shock. So-called heat stress granules (SGs) contain mRNAs encoding most cellular proteins but exclude mRNAs encoding heat shock proteins (). Compositionally similar SGs appear in the cytoplasm of mammalian cells exposed to environmental stress (e.g., heat, oxidative conditions, UV irradiation, and hypoxia). In response to stress, eukaryotic cells reprogram mRNA metabolism to repair stress-induced damage and adapt to changed conditions. During this process, the translation of mRNAs encoding “housekeeping” proteins is aborted, whereas the translation of mRNAs encoding molecular chaperones and enzymes involved in damage repair is enhanced. Selective recruitment of specific mRNA transcripts into SGs is thought to regulate their stability and translation (). Stalled 48S preinitiation complexes are the core constituents of SGs, which include small but not large ribosomal subunits as well as the early translation initiation factors eIF2, eIF3, eIF4E, and eIF4G (; ). In addition, SGs contain PABP1 (), the p54/Rck helicase (), the 5′–3′ exonuclease XRN1 (), and many RNA-binding proteins that regulate mRNA structure and function, including HuR, Staufen, Smaug, TTP, Fragile X mental retardation protein, G3BP, CPEB, and SMN (see for specific references). SGs also contain putative scaffold proteins such as Fas-activated serine/threonine phosphoprotein (). Like tomato heat SGs, mammalian SGs selectively exclude mRNAs encoding stress-induced heat shock proteins (). However, SGs are not stable repositories of untranslated mRNA, as drugs that stabilize (e.g., cycloheximide) or destabilize (e.g., puromycin) polysomes inhibit or promote SG assembly, respectively, which is indicative of a dynamic equilibrium between these structures (). Moreover, FRAP analysis reveals that many SG-associated RNA-binding proteins (e.g., TIA-1, TIAR, TTP, G3BP, and PABP) rapidly shuttle in and out of SGs despite the large size (several micrometers) and apparent solidity of these cytoplasmic domains (, ). Given that these proteins also regulate mRNA translation and decay, their rapid flux through SGs supports the notion that SGs are triage centers that sort, remodel, and export specific mRNA transcripts for reinitiation, decay, or storage. At the same time, SGs contain components with no obvious link to RNA metabolism, notably TRAF2 (), which is recruited to SGs via its binding to eIF4G and may link SGs to apoptosis. Unlike other RNA granules, SGs are not seen in cells growing under favorable conditions but are rapidly induced (15–30 min) in response to environmental stress. The phosphorylation of translation initiation factor eIF2α by a family of stress-activated kinases (e.g., protein kinase R [PKR], PKR-like ER kinase, GCN2, and heme-regulated inhibitor) is critical for the assembly of SGs. Phosphorylation of eIF2α reduces the availability of the ternary complex, thereby blocking translation initiation and promoting polysome disassembly. The following data establish the central importance of phospho-eIF2α in SG assembly: the expression of a recombinant phosphomimetic mutant of eIF2α (S51D) is sufficient to induce the assembly of SGs (); the expression of a recombinant nonphosphorylatable mutant of eIF2α (S51A) prevents the assembly of SGs (); and mutant mouse embryonic fibroblasts expressing only the nonphosphorylatable form of eIF2α (S51A) do not assemble SGs in response to heat, arsenite-induced oxidative stress, or FCCP-induced metabolic stress, but they can assemble SGs when transfected with recombinant eIF2α (S51D; ). Thus, phospho-eIF2α is essential for SG assembly. Several proteins act downstream of phospho-eIF2α to potentiate or inhibit SG assembly. Self-aggregation of either TIA proteins or G3BP promotes SG assembly. Stress-induced aggregation of TIA proteins is mediated by a glutamine-rich prionlike domain that is regulated by HSP70 (). Similarly, self-aggregation of the RNA-binding protein G3BP (RasGAP-SH3–binding protein) promotes SG assembly, a process that is regulated by phosphorylation at serine 149 (). Although not present in SGs, the mitochondria-associated apoptosis-inducing factor inhibits SG assembly by shifting the cellular redox potential (). Finally, various mitochondrial poisons induce SGs, suggesting a requirement for ATP in either SG assembly or disassembly (, ). Thus, SG assembly/disassembly is regulated by numerous signaling pathways acting downstream of phospho-eIF2α. SGs are also detectable in tissues from stressed animals. In chickens treated with the ototoxic antibiotic gentamycin, the appearance of SGs in cochlear cells () occurs several hours before the onset of apoptosis. In another study, whole-animal radiotherapy induces SG assembly within individual tumor cells (), in which the radiation-induced translation of hypoxia inducible factor–1α-regulated transcripts is delayed pending SG disassembly during recovery, suggesting that the expression of these transcripts is inhibited by their retention in SGs. Similar results have been described using an animal model of stroke, in which SGs may regulate protein translation in neurons during ischemia (). These studies indicate that SGs are not in vitro artifacts of cell culture but are an integral part of the organism response to stress. The processing body (PB) is a distinct cytoplasmic RNA granule that contains components of the 5′–3′ mRNA decay machinery, the nonsense-mediated decay pathway, and the RNA-induced silencing complex. The focal distribution of mammalian 5′–3′ exonuclease XRN1 () provided the first intimation of the spatial regulation of mRNA decay (). Subsequent studies demonstrating the focal distribution of other RNA decay factors such as the decapping enzymes DCP1/DCP2 and the Lsm proteins (; , ; ; , ) supported a model of spatially discrete mRNA decay foci. Functional studies in yeast confirmed this model by demonstrating that mRNAs containing 5′–3′ exonuclease-resistant oligo-G tracts accumulate at these foci, leading to their designation as PBs (). Similarly, knockdown of XRN1 leads to the accumulation of poly(A) mRNA at PBs in mammalian cells (). Whereas SGs are heterogeneous in size and shape, PBs are uniform spheroid particles that increase in size and number in response to stress (; ; ). The components of PBs have been defined in complementary studies in lower and higher eukaryotic cells (; , ; ; , ). PBs contain enzymes that are required for each phase of the general mRNA decay pathway, including a deadenylase (CCR4), a decapping enzyme complex (including DCP1/2, Hedls, hEdc3, and p54/RCK; ; ), and an exonuclease (XRN1). PBs also contain an Lsm1–7 heptamer that regulates various aspects of RNP assembly () and components of the nonsense-mediated decay pathway (e.g., SMG5, SMG7, and UPF1; ; ). In mammalian cells, PBs contain components of the RNA-induced silencing complex (e.g., argonaute and microRNA [miRNA]; ; ), the eIF4E-binding protein 4-ET, and GW182, an RNA-binding protein required for miRNA-dependent silencing (; ; ). The translation initiation factor eIF4E and the translational silencer/RNA helicase p54/RCK are also found in PBs, suggesting that they are sites of translational control as well as mRNA decay. Finally, PBs contain several RNA-binding proteins associated with mRNA translation/decay, including TTP, BRF1, CPEB, 4-ET, and Smaug (). Many important functional studies of PBs were performed in , which lack SGs (). Yeast PBs exhibit many properties common to mammalian SGs: both increase in size and number in response to glucose deprivation, osmotic stress, and UV irradiation (); both contain mRNA in equilibrium with polysomes (); and Dhh1p/Pat1p-enforced polysome disassembly causes PB assembly in yeast (), whereas puromycin-enforced polysome disassembly causes SG assembly in mammalian cells (). Studies in fission yeast indicate that assembles SG-like structures containing eIF3, eIF4A, and RNA in response to heat shock (), making speculation as to the possible evolutionary relationship between SGs and PBs premature. In mammalian cells, SG formation requires eIF2α phosphorylation, whereas stress-induced PB formation does not (). The signaling pathways that regulate PB assembly remain to be determined. Recent work in mammalian cells defines a PB core complex (), which is likely to be a key player in PB assembly. Many metazoan PB components (hedls, GW182, and eIF4-ET) have no yeast counterparts, yet knocking down any one of these inhibits PB assembly and impairs PB function. For example, knockdown of 4E-T or GW182 concurrently inhibits adenine/uridine-rich element–mediated (; ) and short inhibitory RNA–mediated (; ; ) mRNA decay. Mammalian PBs are sites of miRNA-mediated translational silencing (which has not been reported to occur in yeast). Reporter transcripts targeted by miRNAs accumulate at PBs in a miRNA-dependent manner (; ), and mutations that displace argonaute from PBs abrogate the translational silencing of its target reporter mRNAs (). Collectively, these results indicate that PBs house multiple mRNA decay and silencing processes. Although PBs and SGs are distinct structures, they share many components () and interact with one another in stressed mammalian cells. Real-time fluorescence imaging shows that PBs are highly motile, whereas SGs are relatively fixed, although SGs exhibit fission, fusion, and occasional dispersal, which is consistent with their proposed role in sorting and export of their contents (). When a motile PB encounters an SG, it is immobilized in an apparent “docking” process. After spending some minutes tethered to the SG, it can disengage and resume its cytosolic survey. As destabilizing factors such as TTP and BRF1 promote interactions between SGs and PBs, mRNAs marked by these proteins may be delivered from SGs to PBs for decay. Thus, SGs act as intermediates between polysomes and PBs and may sort and modulate the increased flow of untranslated mRNA that accompanies stress. The close relationship between SGs and PBs in mammalian cells reiterates the dynamic link between mRNA translation and mRNA decay in yeast (). #text All RNA granules harbor translationally silenced mRNA. GCGs and neuronal granules harbor highly specific mRNA cargo, whereas SGs and PBs are less discriminating. SGs contain the majority of polyadenylated mRNAs that are subject to stress-induced translational arrest but exclude mRNAs encoding stress-induced proteins such as HSP70. PBs contain mRNAs subject to general, nonsense-, and adenine/uridine-rich element–mediated decay as well as mRNAs targeted by miRNAs and short inhibitory RNAs. lists representative proteins from different functional classes that are components of RNA granules. Although the content of each type of RNA granule is distinct, many proteins are found in more than one type of granule. Besides varying in mRNA selectivity, different RNA granules contain or exclude ribosomes. Neuronal granules appear to contain both large and small ribosomal subunits, which are packaged in an inert fashion that prevents translation. SGs contain only small ribosomal subunits derived from disassembled polysomes, whose mRNA is being remodeled and sorted for export to other sites. Finally, PBs lack ribosomal subunits altogether, nor do they contain translational initiation factors (other than eIF4E). Thus, the ribosomal composition of these granules correlates with their functions: neuronal granules house pretranslational mRNA, SGs harbor mRNA from translationally terminating RNA (e.g., disassembling polysomes), and PBs contain mRNA selected for decay. Despite recent progress, we know almost nothing about the signaling pathways and molecular mechanisms governing formation and disassembly of mRNA granules. We lack a complete list of components for a single type of RNA granule, even within a single species or cell type. In all likelihood, yeast genetics will continue to point the way toward understanding PB dynamics and regulation, but because many relevant PB components (e.g., GW182, hedls, and argonaute) are not found in yeast, studies of higher eukaryotes are also essential. A deeper understanding of the smallest and most ancient RNA granule may reveal principles that are applicable to the other granules described in this study. To borrow on an analogy, mRNA molecules lead complicated lives () and complicated deaths. Each mRNA transcript exhibits changes in wardrobe and location that correlate with its functional state throughout its lifetime. Protein coats are added cotranscriptionally and are then modified during splicing and transport to the cytoplasm. Cytoplasmic transcripts may be bundled into a germ cell or neuronal granule for maturation or transport to the worksite, dressed in a different protein uniform, and released for productive translation. If the cellular economy sours (stress), the transcript may be temporarily retired to an SG. If the cellular economy recovers, the transcript can redress and resume production at a polysome. If the cellular economy remains depressed, the transcript can be delivered to a PB to die. Thus, the localization of the transcript is linked to changes in wardrobe (protein clothing) and functional state (translating, silenced, or decaying). Whether the protein coat determines or reflects the localization and functional state of a mRNA transcript appears to be a “chicken or egg” question at present. Nevertheless, our understanding of the complicated lives of mRNA requires further elucidation of the key molecular events that dictate its wardrobe, location, and functional state.

Programmed cell death (PCD), which is often carried out by a morphologically distinct process called apoptosis, is essential for both proper animal development and cellular homeostasis (for review see ). The caspase family of cysteine proteases are the main effectors of apoptosis. These proteases specifically target a number of cellular proteins for cleavage, leading to the disassembly of cells undergoing apoptosis (for reviews see ; ). In most cases, caspases are produced as inactive zymogens that become active in response to apoptotic stimuli. In , the loss of function of a single caspase CED-3 or its adaptor CED-4, which is required for CED-3 activation, results in a complete block in developmental PCD (). The mammalian caspase family can be divided into initiator and effector caspases. Initiator caspases resemble CED-3 and are characterized by the presence of protein–protein interaction motifs such as a caspase recruitment domain (CARD), e.g., caspase-2 and -9, or a pair of death effector domains, e.g., caspase-8 and -10 (; ). Initiator caspases undergo adaptor-assisted self-activation, whereas the effector caspases, lacking a CARD or death effector domains, require proteolytic processing by an initiator caspase to become active. In mammals, the activation of the initiator caspase, caspase-9, which is mediated via its adaptor protein Apaf-1, is necessary for stress-induced cellular apoptotic responses (; ; ). However, Apaf-1 and caspase-9 knockout mice, or cells derived from knockout animals, show limited phenotypic abnormalities, and caspase activity and apoptosis is still seen in many tissues from the knockout animals (; ). Therefore, it is likely that although the caspase-9–Apaf-1 pathway is required for specific PCD, caspase activation and apoptosis pathways not requiring these conserved proteins also exist in mammals. has one Apaf-1 homologue, ARK (DARK/dApaf-1/Hac-1), which is required for the activation of DRONC, the only CARD-containing orthologue of CED-3/caspase-9 in (; ; ; ; for reviews see ; ). Apaf-1 activation has an obligate requirement for cytochrome released from mitochondria for apoptosome assembly, whereas an ARK apoptosome-like complex can assemble in the absence of cytochrome (). The DIAP-1 ( inhibitor of apoptosis protein-1) directly binds DRONC, preventing its activation by blocking the ARK–DRONC interaction (). The REAPER, HID, and GRIM proteins antagonize DIAP-1 function to facilitate DRONC activation (). Genetic and cell culture data suggest that DRONC is required for most developmental and stress-induced cell death (; ; ; ; ). Interestingly, in animals lacking DRONC, some embryonic PCD and larval midgut histolysis occur normally, indicating that DRONC is not essential for all PCD (; ). Previous studies with hypomorphic alleles and cell-based RNA interference analyses suggest that ARK is required for PCD (; ; ; ). However, the hypomorphs are viable and show restricted phenotypic abnormalities, making it difficult to fully assess the function of ARK in PCD. In this paper, we describe the analysis of two independent mutants that are strong alleles of and demonstrate that ARK is essential for normal development, most developmental PCD, and stress-mediated apoptosis. However, similar to DRONC, some PCD is ARK independent, suggesting that the ARK–DRONC pathway controls most, but not all, PCD in the fly. alleles were obtained in a screen conducted using mitotic recombination for mutations, which results in an increased relative representation of mutant over wild-type (WT) tissue. In these mutants, the mutant clones were larger than the corresponding WT twin spots. The screen of the right arm of chromosome 2 identified mutations in the locus that have been previously described (). Four alleles of were also obtained from the same screen, which were all lethal at the pupal stage of development as homozygotes or in trans to each other. Sequencing revealed point mutations or deletions in the coding sequence of the gene in each of the mutant chromosomes (). had a G to A mutation, resulting in the truncation of the protein after residue 206; had a C to T mutation, causing protein truncation after residue 660; and had a deletion after residue 592, generating a frameshift mutation, whereas possessed a T to G mutation, causing protein truncation after residue 1,357. The mutation in is predicted to affect both of the reported alternately spliced transcripts of the gene (). Because all mutants were lethal at a similar stage, only and were analyzed in our studies. Similar to Apaf-1, ARK consists of a CARD, a nucleotide-binding NB-ARC domain, and multiple WD40 repeats (). mutation truncates the protein in the NB-ARC (CED-4 domain), whereas leads to a protein lacking most of the WD40 repeats (). Both mutants are lethal and have very similar phenotypes, suggesting that they are strong loss-of-function alleles. The phenotypes also indicate that both the NB-ARC and the WD40 domains are essential for ARK function. Unlike the published hypomorphs (; ; ), all homozygous and animals die as pupae. Despite the similar overall phenotypes for and alleles, development of mutants to pupation was significantly delayed when compared with WT or alleles (), suggesting that may be a stronger allele than . Consequently, the survival of -null animals to early pupae stage was lower than that of the heterozygotes (). Although larvae and pupae from both mutants appeared grossly normal externally, some larval tissues derived from late third instar animals showed hyperplasia. For example, consistent with previous observations (; ), the larval central nervous system (CNS) was enlarged in both mutants (). This was particularly evident in the ventral ganglion that appeared to be elongated and contained longer nerve fibers. In ∼40% of and most of the animals, the wing discs were enlarged (). In a small number of both mutants, the eye discs were also enlarged (). mutant embryos contain extra cells, and the removal of maternal abolishes most cell death during embryogenesis (; ). -deficient embryos also show an enlargement of the CNS, which is presumably caused by reduced PCD (). By staining embryos with anti–embryonic lethal abnormal visual protein (ELAV) antibody to visualize neurons in the CNS and peripheral nervous system, we found extra neurons in chordotonal cell clusters in mutant embryos (). There were up to three extra cells per cluster in most mutant embryos analyzed (). Staining of embryos with BP102 antibody, which recognizes CNS axons, showed gross abnormalities in many mutant animals, with animals often showing more dramatic features (). We consistently observed stronger staining of CNS axons in mutant embryos compared with WT animals, which could result from more densely packed axons. In many mutant animals, the ventral nerve cord appeared to be improperly compacted and the spacing between longitudinal axonal tracts was enlarged (). This could be attributable to additional cells in the mutants caused by reduced PCD. To investigate whether mutants have fewer cells undergoing PCD, we analyzed various larval organs for apoptosis using acridine orange (AO) staining. In and brain lobes, AO staining was almost absent when compared with WT, which showed many AO-positive cells (). Furthermore, AO staining in wing discs was greatly reduced in both mutants (). Larval eye discs also displayed a dramatic decrease in the number of dying cells when compared with WT (). As the only predicted function of ARK is to activate caspases, we analyzed caspase activity in prepupal lysates. Caspase activity, which was measured using two different substrates, was lower in both mutants (). Although reduced, some caspase activity was detectable in mutant animals at different stages of larval development (unpublished data), suggesting that ARK-independent caspase activation may occur in some tissues and/or cell types. We conclude that ARK is essential for most PCD in the embryo and larval tissues. These data also suggest that ARK and DRONC function in the same PCD pathway. and are transcriptionally up-regulated in salivary glands after the late prepupal ecdysone pulse, and mutants show a delay in salivary gland PCD (, ; ; ; for reviews see ; ). To examine a role for ARK in larval salivary gland removal, we analyzed PCD in this tissue in mutants (). Larval salivary gland removal was markedly delayed in both and animals. Histological analysis indicated that both mutants had persistent or partially degraded salivary glands with an intact lumen at 20 h relative to puparium formation (RPF) at a time when, in the WT animals, salivary glands had been completely removed (). We could also see intact salivary glands at 30 h RPF in all mutant animals. The persistent salivary glands were highly vacuolated and appeared to be similar to persistent glands in mutant animals and prehistolyzed WT glands (; ). Adult structures, such as wings, were forming in both mutants as in the WT controls, indicating continuing pupal development (). Thus, the persistence of the salivary glands cannot be attributed to a global delay in development. No TUNEL-positive nuclei were observed in mutant salivary glands at the time when WT glands were TUNEL positive (), indicating that DNA fragmentation, which requires caspase activation, does not occur in the absence of ARK in salivary glands. These results indicate that ARK is required for caspase-dependent removal of the larval salivary glands. Along with , is up-regulated in the midgut after the late third instar larval ecdysone pulse, which triggers PCD in this organ (; ; ). However, our data demonstrate that in contrast to salivary glands, midgut removal was largely normal in both mutants (). Larval midguts are eliminated in the few hours after puparium formation. Removal of this tissue involves shortening of the gastric caecae with a concomitant reduction in midgut length, leading to the formation of the adult epithelia around the regressing larval structure (). By 4 h RPF, the proventriculus is markedly reduced in size, whereas the gastric caecum and midgut no longer resemble the larval form (). Both WT and mutant animals showed shortening of the gastric caecae by 1 h RPF and complete removal by 4 h RPF (). Histological analysis showed that by 6 h RPF both WT and mutants contained condensed larval midguts within detached adult epithelial lumen (). Within 1 h RPF, TUNEL staining, similar to that seen in WT midguts, was evident in midguts from both mutants (), indicating that DNA fragmentation occurs in the absence of ARK. We further analyzed caspase-3–like activity in prepupal (+1 h RPF) WT, , and midguts using the active caspase-3 monoclonal antibody. Caspase-3–like activity was present in both mutant and control prepupal midguts (). Midguts from both mutants showed that caspase activities were generally comparable to WT, as measured by substrate cleavage, except for animals that showed lower activity on VDVAD-AMC (). These data indicate that caspase activation occurs normally in the midgut of mutants and that ARK is not essential for midgut PCD. These results are consistent with our observations of mutants () and indicate that the ARK–DRONC pathway of caspase activation is not essential for midgut PCD and that an alternative mechanism of caspase activation is likely to be functioning in this tissue. A previous study suggests that transcription is induced by irradiation (). In mutants, stress-mediated cell death, including radiation-induced apoptosis, is completely abolished (; ; ). To investigate the role of ARK in DNA damage-induced PCD, we irradiated larvae with γ rays and analyzed the effect on apoptosis. As expected, the basal number of AO-positive cells in tissues from both mutants were very low (or absent; ). After irradiation, WT larvae showed large increases in apoptosis in brain lobes, eye discs, and wing discs, as observed by increased AO staining (). However, we did not see any increase in AO staining in any tissue from and animals (), suggesting that these organs in mutants were resistant to apoptosis that was induced by γ irradiation. These data, combined with previous studies (; ; ; ), suggest that the ARK–DRONC pathway is essential for mediating radiation-induced apoptosis. The in vivo data presented using two independent mutant alleles suggest that ARK plays an essential function in most developmental PCD in flies and that removal of results in profound developmental defects and lethality. Our results, combined with previously published data (; ; ; ), indicate that ARK is required for most PCD that occurs during embryonic and larval development, for the efficient removal of larval salivary glands during larval/pupal metamorphosis, and for stress-mediated apoptosis. Thus, it seems that ARK has a more global function in PCD than Apaf-1, its mammalian counterpart. However, consistent with Apaf-1 knockout data (), the results also suggest that some caspase activation and cell death can still occur in the absence of ARK. As mutants essentially phenocopy the loss-of- function, our data argue that these proteins act in a common pathway. Previous experiments using RNA interference have shown that ARK is required for DRONC activation (). These results suggest that the primary function of ARK is to facilitate DRONC activation. The observation that metamorphic midgut cell death occurs normally, whereas salivary gland PCD is significantly delayed, suggests that the midgut may provide a model system for studying novel caspase activation and cell death pathways that are independent of the evolutionarily conserved canonical pathway. Ethylmethanesulfonate-generated alleles were identified in a genetic screen for genes that confer a survival advantage to mutant tissue (). Meiotic recombination mapping was used to initially map the mutant locus to a 1-Mb interval around 53C and 54B. Lethal complementation analysis was then performed with small deficiencies from this region. The mutant locus failed to complement ; Df(2R)P803-Delta15, , which is a deletion spanning 14 annotated genes, but complemented ; Df(2R)ED1, which contains intact coding sequence for , but lacks the remaining 13 genes of the ; Df(2R)P803-Delta15, deletion. This revealed that mutations in the gene were likely causing the lethality of the mutant alleles we isolated. For experimental analysis we used stocks balanced over Cyo Kr-GFP to allow identification of homozygous animals. Lethality tests were performed at 25°C. Embryos deposited over 4 h were counted, and development to early pupae was monitored for 12 d. Developmental delay was analyzed by scoring the emergence of early pupae over the indicated times. Survival rates of homozygous animals were calculated following Mendelian principles using the observed number of heterozygous animals at early pupal stage to determine the expected homozygous complement. Larvae were staged by gut clearance after feeding on food supplemented with 5% Bromophenol blue. TUNEL of embryos was performed essentially as previously described (). Dissected larval tissues were fixed for 20 min in 4% formaldehyde in PBS/Tween-20, washed in PBS/Tween-20, permeabilized by incubation in 100 mM sodium citrate/0.1% Triton X-100 at 65°C for 30 min, and then TUNEL assayed using a kit (Roche). Tissues were mounted in 80% glycerol with 4 μg/ml Hoechst for confocal analysis. For AO staining, larval tissues were dissected in 1.6 μM AO/PBS, incubated for 10 min, washed in PBS, and analyzed by confocal microscopy (see Microscopy and image capture). Active caspase-3 staining was performed essentially as previously described (). Anti-active caspase-3 (Cell Signaling Technology) and anti-lamin DmO ADL67.10 (Developmental Studies Hybridoma Bank) antibodies were used at 1:50 and 1:400, respectively. For ELAV and BP102 staining, embryos were fixed in 4% formaldehyde and blocked in 10% goat/sheep sera. Anti-ELAV and BP102 antibodies (both from Developmental Studies Hybridoma Bank) were used at 2 μg/ml. Alexa Fluor 488– and 568–coupled secondary antibodies (Invitrogen) were used at 1:500. Staged animals were fixed in 85% ethanol/4% formaldehyde/5% acetic acid/1% glutaraldehyde; then they were paraffin embedded, sectioned, stained, and analyzed by light microscopy (see next section). Images in D (top), 3 A, 4 A, and 4 B were obtained using a microscope (model SZ40; Olympus) with a 110AL 2× objective and captured using a camera (model DP11; Olympus). Images in D (middle and bottom) and 3 B were obtained using a microscope (model BX51; Olympus) with UPlanApo objectives, fitted with a camera (model DP70; Olympus) and processed with Olysia Bioreport software (Olympus). Images in (A–D), 3 C, 4 (C and D), and 5 were captured using a confocal microscope (Radiance 2100; Bio-Rad Laboratories) equipped with three lasers, an Argon ion 488 nm (14 mW); a Green HeNe 543 nm (1.5 mW); and a Red Diode 637 nm (5 mW), and an inverted microscope (model IX70; Olympus) with UApo objectives. The dual-labeled cells/tissues were imaged with two separate channels (photomultiplier tubes) in a sequential setting. All image acquisitions were performed at room temperature. Images were compiled using Photoshop 6.0 (Adobe). 20–50 μg of animal or tissue lysates were used for caspase assays following established protocols (, ). Cleavage of the caspase substrates VDVAD-AMC (a preferred DRONC substrate) and DEVD-AMC (a substrate for effector caspases such as DRICE and DCP-1) was used for determining enzyme activities.

Son of Sevenless 1 protein (sos1), by stimulating the substitution of GDP for GTP, functions as a guanine nucleotide exchange factor (GEF) for the small GTPase ras. The proline-rich COOH-terminal region of sos1 (C-sos1) binds to the src homology (SH) 3 domains of the adaptor protein growth factor receptor bound protein (grb2). The grb2–sos1 complex, via the central SH2 domain of grb2, interacts with phosphotyrosine residues of activated receptor tyrosine kinases (RTKs) on the cytoplasmic side of the plasma membrane. The translocation of the grb2–sos1 complex from the cytosol to the membrane upon RTK activation allows the presentation of sos1 to ras, leading to the exchange of GDP for GTP and ras activation (). Although the grb2–sos1 complex functions exclusively as a ras activator, sos1 can also function as a GEF that is specific to the GTPase rac1. These two distinct catalytic functions of sos1 are mutually exclusive and reciprocally related. When associated with the actin binding protein eps8 and the Abl-interacting protein e3b1/Abi1 (Abl interactor-1) in a heterotrimeric complex, sos1 displays rac1-specific GEF activity (). Tyrosine phosphorylation of sos1 by Abl tyrosine kinase promotes its rac1-specific GEF activity without compromising its ras-specific GEF activity (). This implies that although Abl-induced phosphorylation modulates the rac1-specific GEF activity of sos1 within the sos1–eps8–e3b1 complex, it does not lead to the dissociation of sos1 from grb2. The mechanisms that govern the dissociation of sos1 from grb2 and the pool of free sos1 available for the formation of the sos1–eps8–e3b1 complex remain unknown. ShcA proteins, consisting of p46-, p52-, and p66shc, interact with the SH2 domain of grb2 upon tyrosine phosphorylation and can serve as coupling molecules between RTKs and grb2–sos1 complexes (; ). All three ShcA isoforms have a COOH-terminal SH2 domain (that binds to phosphorylated RTKs), a central collagen homology (CH) 1 domain, and a phosphotyrosine binding domain. P66shc has an additional NH-terminal proline-rich CH2 domain that is not found in the other two isoforms. The CH2 domain imparts functional diversity to p66shc. Unlike p46- and p52shc, p66shc inhibits rather than activates ras (). In addition, p66shc is unique in the ShcA family in its ability to control intracellular oxidant levels (). However, the molecular mechanism through which p66shc regulates reactive oxygen species (ROS) levels are not fully characterized. Because of the role of p66shc in inhibiting ras, the reciprocal relationship between the ras and rac1 GEF activities of sos1, and the importance of rac1 in regulating ROS production (), we were curious as to whether p66shc, by virtue of its CH2 domain, governs ROS levels by regulating the activity of rac1. To determine whether p66shc can activate rac1, we compared rac1 activity in mouse embryonic fibroblast (MEF) cell lines derived from mice with targeted knockout of the p66shc gene (p66shc −/− MEF) with that of their wild-type littermates (p66shc +/+ MEF). With transfection of sos1, eps8, and e3b1, active GTP-rac1 was significantly higher in p66shc +/+ than in −/− MEF (). Moreover, rac1 activity was rescued in p66shc −/− MEF that were reconstituted with full-length wild-type p66shc (p66shcWT) but not in cells expressing the CH2 domain of p66shc (p66shcCH2; ). A difference in active rac1 levels between p66shc −/− and +/+ cells was also apparent under basal conditions in which ectopic constructs were not transfected (unpublished data). These findings show that full-length p66shc activates rac1, whereas its NH-terminal CH2 domain in isolation does not. We then asked if p66shcWT stimulates the rac1-specific GEF activity of sos1. With transfection of sos1, eps8, and e3b1, the GEF activity of immunoprecipitated sos1 was twofold greater in lysates from p66shc +/+ than −/− MEF (). Expression of p66shcWT, but not -CH2, in p66shc −/− MEF, restored the rac1-specific GEF activity of sos1 to that in p66shc +/+ MEF. These findings suggest that p66shc regulates the rac1-specific GEF activity of sos1. Most SH3 domains bind to proline-rich sequences containing a core XPxXP element (where P = proline, X = hydrophobic residue, and x = any amino acid), with the prolines in the peptide core making direct contact with the hydrophobic pocket of the SH3 domain. The CH2 domain of p66shc has one putative SH3 binding core element encompassing residues 46–50 (LPPLP). We determined the importance of this motif in p66shc-stimulated rac1 activity. A mutant of p66shc in which prolines 47 and 50 were changed to alanine (p66shcP47A/P50A) was generated. In contrast to p66shcWT, expression of p66shcP47A/P50A in p66shc −/− MEF did not rescue rac1 activity () or rac1-specific GEF activity of sos1 (), suggesting that proline-mediated interactions play an important role in mediating these functions of p66shc. Sos1 functions as a ras-specific GEF when bound to grb2 and as a rac1-specific GEF when it is part of the sos1–eps8–e3b1 complex. We therefore examined the role of p66shc in regulating the formation of these sos1-containing complexes. The amount of sos1 associated with eps8 was significantly greater in p66shc +/+ than in −/− MEF (). Conversely, the amount of sos1 bound to grb2 was appreciably less in p66shc +/+ MEF than in −/− cells (). A similar inverse relationship between p66shc expression and grb2–sos1 binding was observed in human embryonic kidney (HEK) 293 cells (). Reconstitution of p66shcWT in p66shc −/− MEF increased sos1 associated with eps8 () while decreasing sos1 bound to grb2 (). In contrast, p66shcCH2 and -P47A/P50A led to little or no change in the amounts of grb2 and eps8 bound sos1 (). Thus, full-length p66shc, via prolines 47 and 50, promotes dissociation of sos1 from grb2 in vivo and increases the formation of the sos1–eps8–e3b1 complex. Proteins of the shcA family are known to associate with grb2 via phosphotyrosine–SH2 interactions (; ). To determine whether such interactions are important to p66shc-stimulated dissociation of sos1 from grb2, tyrosines 349, 350, and 427—the residues on p66shc that when phosphorylated mediate the recruitment of grb2—were mutated, and the capacity of this triple mutant (p66shcY3A) to displace sos1 from grb2 in vivo was examined. When compared with p66shcWT, p66shcY3A expression led to a decrease in displacement of sos1 from grb2 (). In addition, p66shcY3A bound less avidly to grb2 than p66shcWT (). In contrast, the in vivo binding affinity of p66shcP47A/P50A for grb2 was not appreciably diminished when compared with that of p66shcWT (). This indicates that phosphorylation of the targeted tyrosine residues is in part responsible for the binding of p66shc to grb2 and, more important, the consequent displacement of sos1 from grb2. We were intrigued by the possibility that the same features of the CH2 domain that confer upon p66shc its ability to govern rac1 activity may also determine its ability to regulate oxidative stress. We therefore assessed the importance of the CH2 domain and prolines 47 and 50 in this domain to p66shc-induced intracellular ROS generation. Expression of p66shcWT in p66shc −/− MEF resulted in a significant increase in HO, whereas expression of p66shcP47A/P50A or -CH2 had no appreciable effect (). Comparison of HO levels between p66shc +/+ and −/− MEF showed significantly lower levels in the latter, which could be rescued by expression of p66shcWT (). In addition, expression of dominant-inhibitory rac1 (rac1N17) suppressed HO levels to a considerably larger degree in p66shc +/+ MEF than in p66shc −/− cells () and abrogated p66shc-induced rescue of HO levels in p66shc −/− MEF (). Collectively, these findings suggest that rac1-dependent mechanisms play a larger role in regulating HO production in p66shc +/+ than in −/− MEF and that p66shc-mediated rescue of HO levels in p66shc −/− cells is dependent on endogenous rac1 activity. We also looked at the role of e3b1 in the increase in ROS levels induced by p66shc. P66shcWT overexpressed in a HeLa cell line with constitutive short hairpin RNA–induced down-regulation of e3b1 led to no significant increase in HO, whereas in a control HeLa cell line, overexpression of p66shcWT increased HO levels (). Moreover, as observed in p66shc −/− MEF, expression of p66shcP47A/P50A or -CH2 in control HeLa cells did not significantly increase HO (). These findings suggest that the expression of e3b1 is critical for a p66shc-induced increase in HO. Grb2 has two SH3 domains. Because the influence of p66shc on the binding partners of sos1 was dependent on putative SH3 binding proline residues in the CH2 domain, we investigated to determine whether grb2 binds to the CH2 domain via a proline–SH3 interaction. Full-length grb2 associated with p66shcCH2 in vitro (). In comparison, the binding of p66shcCH2 with mutations at prolines 47 and 50 (p66shcCH2P47A/P50A) to grb2 was much weaker, confirming the importance of these proline residues to this association. Moreover, comparison of the NH- and COOH-terminal SH3 (N- and C-SH3, respectively) domains of grb2 revealed that p66shcCH2 preferentially bound to the C-SH3 domain, suggesting sequence-specific requirements for the binding of p66shcCH2 to SH3 domains. In vitro binding assays using full-length grb2, C-sos1, and p66shcCH2 showed a reciprocal relationship between C-sos1 and p66shcCH2 with respect to binding to grb2 (). Semiquantitative analysis revealed that C-sos1 bound with much greater affinity to grb2 than did p66shcCH2 (compare ). Moreover, when compared with p66shcCH2, p66shcCH2P47A/P50A was a much weaker competitor of C-sos1 (). These findings suggest that C-sos1 and p66shcCH2 compete for binding to grb2, with C-sos1 having a significantly higher in vitro binding affinity than p66shcCH2, and that integrity of prolines 47 and 50 within p66shcCH2 is necessary for it to effectively compete with sos1. In vivo, with overexpression of grb2 and p66shcCH2, there was a weak association between the two proteins (). Differential binding affinities of the CH2 domain to the N- and C-SH3 domains of grb2 suggest that p66shc may promote but not be solely responsible for the dissociation of sos1 from grb2. Both the N- and C-SH3 domains of grb2 interact with sos1 (). Therefore, binding of the CH2 domain to the C-SH3 domain of grb2 would be expected to weaken but not entirely abrogate the sos1–grb2 interaction. It is noteworthy that our analysis of the CH2–grb2 interaction was conducted in recombinantly expressed proteins that are not posttranslationally modified. It is therefore reasonable to hypothesize that posttranslational modifications of the CH2 domain and/or grb2 may also play an important part in further modulating their interaction. Furthermore, the CH2 domain that was mutated at prolines 47 and 50 did retain some ability to bind to grb2 in vitro, suggesting that other residues may also be important for its interaction with grb2. Although the disparity in active rac1 between the p66shc +/+ and −/− cells was evident under basal conditions, this difference was much more pronounced with the overexpression of sos1, eps8, and e3b1. This might reflect very low levels of endogenous expression of these proteins (particularly eps8 and e3b1) observed in these cell lines (unpublished data), consistent with a predicted mechanism for limiting the formation of the sos1–e3b1–eps8 complex, and regulation of rac1 activity (). Notably, although endogenous expression of e3b1 and eps8 was low in both cell lines, p66shc −/− cells had appreciably higher levels of both e3b1 and eps8 when compared with their p66shc +/+ counterparts (unpublished data). This inverse relationship between active rac1 and e3b1/eps8 hints at the possibility of a compensatory feedback mechanism between rac1 and the endogenous proteins that regulate its activity. Overall, our results suggest that the NH-terminal proline-rich CH2 domain of p66shc can bind to the C-SH3 domain of grb2 via a low-affinity proline–SH3 interaction, which, by displacing sos1 from grb2, increases the rac1-specific GEF activity of sos1. These events increase intracellular rac1 activity and HO production. Importantly, this weak proline–SH3 interaction is functionally relevant, only within the context of full-length p66shc. We propose a model in which p66shc binds to grb2 primarily via a well-characterized phosphotyrosine–SH2 interaction. Once p66shc is bound to grb2, its CH2 domain, by virtue of molecular proximity and possibly changes in its conformation, can interact efficiently with the C-SH3 domain of grb2, displacing sos1 from grb2 (). This model predicts that in addition to determining the fraction of cellular sos1 bound to grb2, p66shc may also influence the binding of other proline-rich proteins to grb2 and thereby modulate a variety of cellular functions that are governed by such interactions. COS7 cells were obtained from American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium supplemented with 10% serum. Spontaneously immortalized p66shc +/+ and −/− MEF and the cDNA for p66shcWT (gifts from T. Finkel and S. Nemoto, National Heart, Lung, and Blood Institute, Bethesda, MD) have been previously described (). Mammalian expression plasmids for grb2 and sos1 were provided by D. Bar-Sagi (State University of New York, Stony Brook, NY). A HeLa cell line expressing a short hairpin RNA sequence to e3b1 and the corresponding control HeLa cell line have been previously described (). Point and deletion mutations in p66shcWT were introduced using standard methods (QuickChange; Stratagene). All mutations were verified by sequencing. Cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. SDS-PAGE and immunoblotting was performed by standard methods with the following antibodies: rac1 (Upstate Biotechnology), sos1 (Upstate Biotechnology), grb2 (Santa Cruz Biotechnology, Inc.), e3b1 (), eps8 (Santa Cruz Biotechnology, Inc.), shc (Santa Cruz Biotechnology, Inc.), GST (Santa Cruz Biotechnology, Inc.), Xpress (Invitrogen), and (His) (QIAGEN). The magnitude of GTP bound endogenous rac1 was determined with a commercial assay (Upstate Biotechnology) that affinity precipitates GTP-rac1 in cell lysates using the Sepharose-conjugated rac1 binding domain of p21-activated kinase 1. Precipitates were then immunoblotted with rac1 antibody. Rac1-specific GEF activity of sos1 was assessed as previously described (). In brief, cells were lysed in lysis buffer (20 mM Tris-HCl, 100 mM NaCl, 1 mM MgCl, 200 mM sucrose, 0.1 mM EDTA, 0.1 mM DTT, 0.5 mM PMSF, 10 g/ml leupeptin, and 10 g/ml aprotinin). Homogenates were clarified by centrifugation at 100,000 for 20 min at 4°C, and sos1 was immunoprecipitated with an anti-sos1 antibody. GST-rac1 (50 pmol) was incubated with 100 pmol of [P]GTP (8,000 Ci/mmol; PerkinElmer). Exchange reaction was performed by incubating immunoprecipitated sos1 complex with [P]GTP-GST-rac1 in the presence of 10 mM MgCl and 2 mM GTPγS. At required time points, a fixed volume of sample was removed and reaction was stopped using cold stop buffer (25 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 25 mM MgCl). Samples were then incubated with glutathione–Sepharose beads for 30 min at RT, followed by washing the beads extensively with PBS (0.901 mM CaCl, 0.493 mM MgCl, 2.67 mM KCl, 1.47 mM KHPO, 137.93 mM NaCl, and 8.06 mM NaHPO). Stop buffer was added, and samples were heated to 95°C for 5 min. Samples were spun, and [P]GTP in the supernatant was measured by scintillation counting. Coimmunoprecipitation was typically performed by incubating 3 μg of antibody and 50 μl of slurry of protein A–Sepharose beads (GE Healthcare) at 4°C overnight. Antibody–Sepharose complex was washed three times with PBS and incubated with 1.5 mg of cell lysate in lysis buffer (25 mM Hepes, pH 7.5, 200 mM NaCl, 1% IgePal CA-630, 10% glycerol, 25 mM NaF, 10 mM MgCl, 1 mM EDTA, 1 mM sodium orthovanadate, 10 μg/ml leupeptin, 10 μg/ml aprotinin, and 0.1% SDS) for 4 h at 4°C. Immunocomplexes were washed extensively with lysis buffer, boiled for 5 min, and analyzed by SDS-PAGE followed by Western blotting. An equivalent amount of nonimmune IgG was used as a control for immunoprecipitations. The CH2 domain was cloned into a (His)-tag prokaryotic expression vector, and a (His)-tagged protein was induced with 1 mM IPTG in DH5α bacteria (Stratagene) and purified using TALON affinity columns (BD Biosciences). Grb2 and the N- and C-SH3 domains of grb2 were purchased as GST fusion proteins (Santa Cruz Biotechnology, Inc.). For in vitro binding of CH2 to grb2, 1 μg of (His)-CH2 was immobilized on 50 μl of Ni-NTA beads (QIAGEN) followed by incubation with 500 ng of GST-tagged full-length grb2, N-SH3, C-SH3, or GST for 4 h at 4°C. (His)-CH2 was eluted with 150 mM imidazole. GST-tagged grb2 inputs and eluted proteins were analyzed by Western blotting using HRP-conjugated anti-(His) and anti-GST antibodies. For in vitro displacement of C-sos1 from grb2, grb2–C-sos1 complexes were first established by incubating 500 ng of GST-tagged full-length grb2 and 250 ng of GST-tagged C-sos1 for 4 h at 4°C. Increasing amounts (250, 500, and 1,000 ng) of (His)-tagged wild-type CH2 or CH2(P47A/P50A) were then added to the grb2–C-sos1 complex and incubated for another 2 h at 4°C. Similarly, for displacement of CH2 from grb2, grb2–CH2 complexes were first established by incubating 750 ng of (His)-tagged CH2 with 500 ng of GST-tagged full-length grb2, followed by the addition of increasing amounts (50, 100, 200, and 400 ng) of GST–C-sos1. GST was used to equalize the protein amount. Grb2 was then immunoprecipitated, and immune complexes and input proteins were probed with anti-GST and anti-(His) antibodies. Two methods were used to quantify cellular oxidant (HO) levels. Intracellular HO was detected and quantified by dichlorofluorescein diacetate fluorescence (Invitrogen) as previously described (). In brief, cells were washed with Krebs-Ringer buffer and loaded with 5 μg/ml dichlorofluorescein diacetate at 37°C in the dark for 5 min. After washing, cells were harvested, resuspended in Krebs-Ringer buffer, and analyzed by flow cytometry (BD Biosciences) using excitation and emission filters of 485 and 535 nm, respectively. HO was also quantified in cell media using the Amplex red assay (Invitrogen), as previously described (). Results were reproduced at least twice, and representative experiments are shown. Values are expressed as mean ± SEM, and analysis was done using a test.

A longstanding question in the study of the cellular response to DNA damage is how the complex structural environment of chromatin is altered in the presence of DNA lesions (; ). The basic repeating unit of chromatin, the nucleosome, consists of DNA wrapped around an octamer of core histones, which is composed of two molecules each of the histones H2A, H2B, H3, and H4. Nucleosomal DNA is dynamically packaged to varying degrees, resulting in different levels of chromatin compaction ranging from the 10-nm fiber to higher order structures such as the condensed mitotic chromosomes. Prior biochemical evidence suggests that chromatin structure is remodeled in the presence of double-strand breaks (DSBs; ; ; ). Similarly, the exposure of cells to UV irradiation appears to relax bulk chromatin structure within the entire nucleus (; ; ). In addition to these global changes, local perturbations in chromatin, such as the exposure of preexisting methylated residues in core histones in the vicinity of the break (; ) as well as DNA damage–induced histone modifications, may provide docking sites for DNA damage response proteins (). For example, dynamic changes in histone acetylation (; ; ; ; ; ) and phosphorylation (; ; ) on chromatin flanking DSBs may control the accessibility of damaged regions of DNA to repair/signaling proteins, including chromatin remodeling complexes. The best characterized DNA damage–induced histone modification is the phosphorylation of the core histone variant H2AX (γ-H2AX; ; ), which extends to large chromatin domains flanking each DSB (; ; ). Although the presence of H2AX is not required for the initial recognition of DSBs (), γ-H2AX organizes the dynamic assembly of multiprotein complexes into cytologically visible nuclear foci (; ; ). The physical structure and composition of γ-H2AX–containing chromatin domains is unknown, but the phosphorylation of H2AX has been linked to an increased chromatin relaxation in (), whereas in contrast, mouse H2AX is required for chromatin condensation and transcriptional silencing of the male sex chromosomes during spermatogenesis (). In addition, several chromatin remodeling complexes, including INO80 (; ) and NuA4/Tip60 (; ; ), as well as key structural components (such as cohesin; ; ) assemble on chromatin in a γ-H2AX–dependent manner. Together, these observations indicate that the phosphorylation of H2AX may directly or indirectly modulate chromatin architecture in the vicinity of a DSB, which is a hypothesis tested in this study. DNA damage–induced chromatin remodeling may account for the movement of DSB-containing chromatin domains, which is indicated by the congregation of multiple DSBs into DNA repair centers in () and by the clustering of γ-H2AX foci within tracks of DSBs in mammalian cells (). To directly monitor the mobility of chromatin containing DSBs in vivo, we expressed histone H2B tagged with a photoactivatable version of GFP (PAGFP; ) in wild-type (WT) and H2AX mouse embryo fibroblasts (MEFs; ). All of the core histones, including H2B, are tightly bound to DNA and are immobile over relatively long time periods and, thereby, provide excellent markers for chromatin in living cells (; ; ). By using the 364-nm emission from a UV laser on a confocal microscope, we simultaneously introduced localized DNA DSBs within the nucleus of cells and photoactivated H2B-PAGFP. As expected, the introduction of DSBs, monitored by the formation of γ-H2AX and the recruitment of Nbs1 in fixed cells, was dependent on sensitizing the cells with the Hoechst 33342 DNA-binding dye (WT + Hoechst and H2AX + Hoechst; ), whereas H2B-PAGFP was photoactivated regardless of the presence of the dye (). To evaluate the dynamics of chromatin domains containing DSBs in living cells, we photoactivated and induced DNA damage in specific subregions in the nucleus with distinctive patterns of circles and lines that allowed us to observe and quantify any potential dynamic movement or change in shape (; Figs. S1 and S2; and Videos 1 and 2; available at ). After compensating for cell migration and rotation, we measured the center of intensity mass for each lased region and calculated the mean squared displacement over a 10-min time period in samples in which DSBs were either introduced or not introduced ( and Fig. S1). In both WT and H2AX cells (+Hoechst), the overall pattern and position of photoactivated regions did not change significantly (mean squared displacement of <0.5 μm) even though the chromatin domains exhibited some small-scale constrained dynamics. Moreover, we did not detect significant movement of chromatin regions containing DSBs over longer time periods, including up to 4 h after irradiation ( and Fig. S1). Thus, the positioning of chromatin domains containing DSBs generated by UV laser microirradiation is stable over time. To examine the mobility of γ-irradiation–induced foci (IRIF) immediately after they form, we introduced DNA damage by γ-irradiation and monitored the spatial position of GFP-53BP1 (). Like endogenous 53BP1, GFP-53BP1 forms foci within minutes at sites of DSBs as determined by staining with γ-H2AX (Fig. S1; ). After their initial appearance, GFP-53BP1 IRIF exhibited limited diffusional motion with a mean squared displacement of 0.9 μm over the 50-min time period monitored (3 × 10 μm/s). This is similar to the range of movement of individual loci observed in undamaged human cells (). Although IRIF did not congregate within common sites in the nucleus, they did exhibit local small-scale dynamics, with IRIF within close proximity to each other frequently interacting but subsequently separating to their own nuclear space (, Fig. S1, and Video 3, available at ). This behavior occurred in a cell cycle–independent manner. In cells ( = 10) that we followed through mitosis and into early G1 and subsequently γ irradiated, we found that 53BP1 foci did not exhibit significant mobility compared with other irradiated interphase cells, nor did the foci appear to exhibit an increased propensity to associate in distinct subregions of the nucleus (Fig. S1). Similarly, early interphase cells exposed to localized UV laser microirradiation did not exhibit increased chromatin mobility compared with other interphase cells (not depicted). Collectively, our results indicate that DSBs generated by UV laser or γ irradiation do not exhibit large-scale mobility or strong cohesiveness. Despite the lack of large-scale genomic repositioning, simple visual inspection revealed that the photoactivated and DSB-containing chromatin regions underwent local expansion of photoactivated chromatin outside of the initial damaged zone (, Fig. S2, and Video 1). This expansion occurred immediately upon the introduction of DSBs and continued until 180 s after exposure to UV microirradiation, corresponding to a time interval in which DNA repair factors such as Nbs1 () and ataxia telangiectasia mutated (ATM; see ) are recruited to sites of DSBs. The persistence of photoactivated H2B in defined DNA damaged regions 4 h after UV laser microirradiation indicates that minimal amounts of H2B are released from chromatin containing DSBs (Fig. S1). Furthermore, we observed a similar expansion of H3-GFP in chromatin containing DSBs, which was confined to areas immediately surrounding regions exposed to UV laser microirradiation (). Importantly, these changes in chromatin structure were dependent on the presence of DSBs because they did not occur in photoactivated regions in cells lacking Hoescht sensitizing dye (Fig. S2 and Video 2). The change in area occupied by photoactivated chromatin corresponded to a 30% increase relative to the initial damaged area and occurred similarly in WT, H2AX, and ATM cells (). Thus, DSBs induce a local chromatin expansion that does not depend on the presence of either H2AX or ATM. To determine whether a linear stretch of transcriptionally active chromatin reacts in a similar manner, we targeted DSBs to tandem gene arrays containing a series of promoter binding sites for the gluccocorticoid receptor (GR) in a cell line expressing the GFP-tagged GR (GFP-GR; ). Upon incubation with the steroid hormone dexamethasone, GFP-GR translocates to the nucleus and binds to the gene array, allowing for visual detection of the array inside the nucleus. We observed that localized DNA damage caused a rapid dispersion of the GFP-GR signal assembled at the gene array (). By examining a cell with two gene arrays and introducing DSBs in only one of the arrays, we determined that the introduction of DSBs in a subregion of the nucleus did not cause a global change in chromatin structure because the nondamaged gene array did not exhibit any significant changes in morphology (). Moreover, there was no dispersion of GFP-GR signal when cells were UV microirradiated in the absence of Hoechst (Fig. S3, available at ). Thus, chromatin changes that occur in the damaged gene array are localized to that subregion of the nucleus, whereas the chromatin configuration in nonlased or nondamaged gene arrays did not change over the same period. The observed change in chromatin gene array density may be caused by DNA damage–induced chromatin opening such that the bound GFP-GR signal is no longer concentrated sufficiently to detect above background nucleoplasmic levels of GFP-GR fluorescence. A nonmutually exclusive possibility is that GFP-GR may dissociate from the array. To discriminate between these possibilities, we performed DNA FISH with gene array–specific probes and measured the relative area occupied by the nondamaged (pre-UV) and damaged regions within the same nuclei (). As summarized in , the FISH signal (180 s after UV) occupied a greater area than that occupied by GFP-GR before exposure to UV microirradiation and a greater area than control FISH signals in cells not exposed to UV microirradiation (). The finding that damaged gene arrays occupied a greater area than before UV laser microirradiation indicates that specific chromatin regions that are transcriptionally competent undergo further remodeling upon the introduction of DSBs. We conclude that the changes observed in the gene arrays in living cells are the result, at least in part, of chromatin decondensation. The laser microirradiation experiments were performed at the lowest doses (∼0.86 nJ/pixel) that yielded a consistent γ-H2AX signal. By measuring the whole cell γ-H2AX response to γ irradiation, we established a standard curve that was used to calibrate the extent of DNA damage induced by the UV laser (Fig. S4, available at ). By this method, we estimate that the level of γ irradiation required to induce the same density of H2AX phosphorylation throughout the nucleus as laser microirradiation within the circular regions () is ∼2.5 Gy. Assuming that 1 Gy produces 35 DSBs, this equates to 85 DSBs localized to a specific subregion of the nucleus, corresponding to a density of ∼1 DSB for every 0.95 Mb DNA. This degree of DNA damage would not be expected to extensively fragment the chromatin; otherwise, H2B-PAGFP would have been mobile, and the signal would have eventually dispersed. Instead, we found that the H2B-PAGFP signal remained within the locally irradiated region and that the integrated intensity of fluorescence did not change over time. The most likely explanation for the expansion is that DSBs induce a local decondensation of chromatin. Examining the damaged regions in greater detail by electron microscopy further supported this interpretation. To monitor chromatin changes associated with DSBs at high resolution, we used correlative fluorescence and energy-filtering transmission EM (EFTEM; ; ; ). In brief, electron spectroscopic imaging (ESI) with an energy-filtering transmission electron microscope generates element-specific maps of the specimen. Nucleic acid–containing structures are both phosphorus and nitrogen rich, whereas protein-based structures are nitrogen rich and phosphorus poor. By obtaining phosphorus and nitrogen image maps and examining the phosphorus and nitrogen content, nucleic acid–containing structures such as chromatin can be distinguished from primarily protein-based structures in intact nuclei (; ). Combining fluorescence microscopy with ESI, we were able to monitor specific subregions of the nucleus in living cells and examine the same regions at high resolution by EM. WT and H2AX cells exposed to γ irradiation or UV laser microirradiation were analyzed using this correlative ESI method. After introducing DSBs with the UV laser, WT and H2AX cells were fluorescently labeled with anti−γ-H2AX or -Nbs1 antibodies and embedded, and ultrathin sections of the cells were imaged using the light microscope to collect a fluorescence map of the chromatin regions containing DSBs. The same sections were then imaged using the ESI mode of EFTEM, and nitrogen (, green) and phosphorus (, red) elemental maps were generated. The fluorescence image of the γ-H2AX or Nbs1 signal was resampled and overlaid onto a low magnification composite ESI-generated net phosphorus image map (, grayscale; white signal on black background). Subsequently, specific regions were examined at high magnification that were sufficient to resolve individual chromatin fibers. In , WT cells exposed to UV laser microirradiation and containing DSBs were labeled against γ-H2AX to mark DSBs, which are shown by green fluorescence track overlaid onto the low magnification ESI net phosphorus image in the top left panel. The ESI image is shown again in the top right panel of but with three yellow asterisks denoting the right-most boundary of the track exposed to UV laser microirradiation (the two gray lines traversing the laser track are EM sectioning artifacts). In this low magnification image, a general decrease in phosphorus density (lower signal intensity) was found specifically in the γ-H2AX–containing domains compared with the neighboring undamaged regions (higher signal intensity). This was confirmed by line scans through the UV microirradiated regions (Fig. S5, available at ). To determine whether this change affected chromatin structure specifically, we examined at higher magnification an area in which DSBs were introduced by the UV laser in only a subregion of heterochromatin (, bottom; arrows; the yellow box in the top panel denotes the region imaged at higher magnification by ESI). We found the regions containing DSBs displayed 10–30 nm chromatin fibers and appeared less condensed (, yellow arrow) relative to neighboring heterochromatin that was not exposed to laser-directed damage (, white arrow). In addition, the chromatin fibers in the DNA-damaged regions had significant accompanying nitrogen-only content (, green) surrounding the chromatin fibers compared with the adjacent region not containing DSBs, which likely represents the accumulation of DNA damage response proteins. To determine whether H2AX is required for chromatin opening at DSBs, we examined the ultrastructure of chromatin containing DSBs in H2AX MEFs. Because Nbs1 is recruited to DSBs independently of H2AX (; ), UV laser–irradiated H2AX cells were immunolabeled against Nbs1 to mark sites of DSBs (). Introducing DSBs specifically in regions of heterochromatin (Hoechst-rich regions in mouse cells) allowed us to directly compare the difference in chromatin opening within the damaged regions (, yellow arrows) to an equivalent heterochromatin region not containing DSBs (, red arrow). We found that DSBs induced a decondensation of heterochromatin characterized by the same 10–30-nm chromatin fibers observed in the damaged WT cells. Moreover, there was significantly more protein surrounding these decondensed chromatin fibers relative to the undamaged ∼2-μm diameter condensed heterochromatin blocks. Thus, consistent with the local chromatin expansion observed in live H2AX cells (), the initial remodeling of damaged chromatin determined at high resolution occurs independently of H2AX. We then compared the chromatin topology observed by UV laser–induced DNA damage to that introduced by γ irradiation, which results in a lower density of DSBs. WT MEFs were incubated 16 h after 8 Gy irradiation and were subsequently fixed and labeled for γ-H2AX (). Individual IRIF were easily discernable by immunofluorescence microscopy (, left). Immunogold labeling with silver enhancement was used to locate the IRIF regions within the nucleus in the high magnification ESI images. The gold particles are shown in the net nitrogen and net phosphorus overlay image as white dots (, middle). This cluster of white dots represents one IRIF observed by immunofluorescence microscopy (, arrow), which likely corresponds to a single DSB. Decondensed chromatin fibers on the order of 10–30 nm and nitrogen-rich/phosphorus-poor regions were characteristic of the focus (, right), which is consistent with the nature of the chromatin observed in the UV laser–irradiated regions in the aforementioned WT and H2AX cells. Thus, a similar relaxation in higher order chromatin structure is observed both within IRIF and in UV laser–irradiated regions. To quantify the density of chromatin fibers in DNA-damaged areas, we calculated the range of phosphorus/nitrogen intensity ratios that were representative of chromatin from nonirradiated control cells (not depicted). The phosphorus/nitrogen ratios measured for nondamaged chromatin was used as a minimum threshold for display purposes, and any pixel within subsequent combined net phosphorus/nitrogen overlay images that contained a phosphorus/nitrogen ratio within this range (representative of chromatin) was displayed as white (). In both WT and H2AX cells, the damaged regions were more open, with chromatin fibers shown as white and phosphorus-poor and nitrogen-rich areas interdispersed among the chromatin fibers. By measuring the number of pixels containing the phosphorus/nitrogen ratio representative of chromatin compared with the total number of pixels within both damaged and nondamaged regions of the image (for example, see yellow outlines in ), chromatin density coefficients (see Materials and methods) were calculated. In both UV laser–irradiated WT and H2AX cells, the chromatin density coefficient decreased by ∼0.4 (on a scale of 0–1.0) for the regions containing DSBs compared with the nondamaged regions (), indicating that the density of chromatin fibers in the vicinity of DSBs is reduced by ∼40%. In addition, in the γ-irradiated WT cells, the chromatin density coefficient in the IRIF focus was 40% lower than in the region immediately outside of the focus (). Thus, the ultrastructure of chromatin within IRIF and UV laser–irradiated regions represents a consistent change that is related to the presence of DSBs. Given that the decrease in chromatin density was confined to a region surrounding DSBs, regardless of the method used to introduce DSBs, we were interested in determining how these changes related to the activation of DNA damage signaling. Specifically, it has been documented that alterations in chromatin structure rapidly induce the global nuclear activation of the ATM kinase (). A marker of ATM activation is the autophosphorylation at residue Ser 1981 (). To determine the spatio-temporal dynamics of ATM phosphorylation relative to the DNA damage–induced chromatin decondensation, HeLa cells expressing H2B-PAGFP were exposed to UV laser microirradiation in specific subregions of the nucleus, fixed with PFA at different time points after UV, and costained for ATM-1981P and γ-H2AX. Cells were relocated after staining based on the presence of the photoactivated H2B-PAGFP, and confocal images were collected using identical imaging parameters. As shown in , the phosphorylated ATM was initially confined to the laser-irradiated region, and both ATM-1981P and γ-H2AX accumulated at the DSB sites during the first minute. Similarly, the chromatin expansion was detectable within 20 s after UV microirradiation and continued until ∼180 s after irradiation ( and not depicted). At 5 min after UV exposure, the phosphorylated form of ATM began to spread out from the damaged zone to the surrounding nucleoplasm, whereas the γ-H2AX signal was restricted to the original UV-exposed region (). Background-corrected mean fluorescence intensity ratios for phosphorylated ATM signals were calculated from three different nuclear areas: inside the UV laser–exposed region (, in), outside the UV laser–exposed region but in the nucleus containing the DSBs (, out), and from neighboring nuclei not exposed to UV laser microirradiation (, non). These measurements confirmed that the phosphorylated form of ATM was present first at the damaged area containing DSBs. For example, 15 s after UV, there was a threefold greater intensity of phosphorylated ATM in the DSB region compared with the surrounding nucleoplasm of the same nucleus not containing DSBs (, in/out ratio) or compared with the background nondamaged nucleoplasm of a control cell (, in/non ratio). At this early time point, there were equally low levels of phosphorylated ATM in the surrounding nucleoplasm of the UV-treated cell and the control non–UV-treated cell (, out/non ratio), indicating that initially, ATM had not been phosphorylated away from the damage region. However, after the ATM-1981P signal accumulated at the damaged region, it began to spread to the surrounding nucleoplasm, which is indicated by the steady increase in the out/non ratio over time (). Based on the finding that the initial local concentration of phosphorylated ATM is established with the same spatio-temporal dynamics as the changes in chromatin structure induced by DSBs (), our observations support the hypothesis that local alterations in chromatin structure are associated with the activation of ATM (). DNA damage response proteins are thought to move in a passive, energy-independent manner, scanning the nuclear volume for high-affinity binding sites (; ). To determine whether DSB-induced chromatin expansion occurs via an energy-dependent mechanism, we treated cells with the metabolic inhibitors 2-deoxyglucose and sodium azide (). In contrast to normal cells, when energy-depleted cells were subjected to DSBs, a minimal amount of H2B-PAGFP expanded to outside the initial damaged region after 180 s ( and ). Thus, DSB-induced chromatin expansion is dependent on ATP. To determine whether DNA damage response proteins can still detect DSBs under energy-depleting conditions, we examined the recruitment of Nbs1 and 53BP1 to UV laser–irradiated regions and IRIF, respectively. Although γ-H2AX formed weakly in ATP-depleted cells (not depicted; but also see ), Nbs1 () and 53BP1 () failed to accumulate at sites of DSBs. Moreover, combined irradiation followed by GFP photobleaching experiments confirmed that ATP depletion did not significantly alter the mobility of Nbs1 and 53BP1 ( and not depicted). Thus, both 53BP1 and Nbs1 diffuse through the nuclear space but do not accumulate on chromatin containing DSBs in energy-depleted cells possibly because of a lack of available high-affinity binding sites or the impediment of ATP-dependent chromatin remodeling complexes. Alternatively, PI-3 kinase–related phosphorylation events may be required, although we have demonstrated chromatin decondensation in ATM MEFs. Regardless, initial changes in chromatin that occur as a result of the presence of DSBs are dependent on processes that metabolize ATP. The movement of interphase chromatin in mammalian cells is confined to submicrometer regions for periods of >1 h (). Although the mechanism that limits chromosome movement is unclear, the tethering of chromatin to nuclear substructures may contribute to the spatial constraints on diffusion (). Using real-time fluorescence microscopy, we have been able to unambiguously track the movement of chromatin containing DSBs in living cells. We find that the disruption of chromatin architecture by DSBs does not influence the large-scale mobility of chromatin, providing evidence against the hypothesis that DSBs in mammalian cells generally move in a directed manner over large distances to congregate in common repair centers. Even when neighboring IRIF interact transiently, we have shown that they do so without exhibiting strong cohesiveness. This indicates that the time-dependent increase in the size of individual foci is unlikely to be caused by the coalescence of multiple DSBs. Rather, the spreading of H2AX phosphorylation over a large chromatin domain may provide additional docking sites for a multitude of DNA damage response proteins that accumulate distal to the lesion. Because the nucleus in is relatively smaller and homologous recombination is the primary repair mechanism, the probability for interaction between DSBs in yeast may be proportionately greater than in mammalian cells (). Although DNA damage generated by UV laser or γ irradiation does not alter the nuclear volume explored by chromosomes, chromatin containing DSBs undergoes a local expansion in the vicinity of the break. As determined by correlative fluorescence and EFTEM, these nuclear regions are characterized by lower order fibers that are morphologically reminiscent of euchromatin regions found in nondamaged nuclei. Although the mechanism that determines the boundary of this DSB-induced chromatin opening or that limits the expanse of H2AX phosphorylation is unclear, neighboring heterochromatin may provide a physical block to the spreading of the damaged chromatin domain. Because a DSB interrupts the continuous chromosome fiber, it is possible that the unfolding of the damaged chromatin region is established by the break itself, which could relieve some of the torsional stress imposed during the packaging of DNA. However, we have found that the DSB-induced chromatin alteration is dependent on ATP. Therefore, it is more likely that primary DSB sensors with ATP-dependent chromatin unwinding activities, which associate with DSBs in an H2AX-independent manner, play an active role in the initial alteration of chromatin folding. Such damage sensors would thereby link the recognition of a DSB to chromatin decondensation and the local activation of the DNA damage response, which is indicated by the initial accumulation of phosphorylated ATM within close proximity to regions containing DSBs. H2AX is dispensable for the initial recruitment of DSB response proteins to DSBs and the signaling of DNA damage (). As shown in this study, the decondensation of chromatin at sites of DSBs also occurs independently of H2AX. Nevertheless, we have found that after the initial recruitment of DNA damage response proteins, they fail to stably associate with DSBs in H2AX cells and “fall off” of chromatin (). Furthermore, H2AX dephosphorylation turns off DNA damage signaling, thereby triggering exit from cell cycle arrest (). Therefore, we speculate that γ-H2AX, although dispensable for the initial remodeling of chromatin, may be essential for maintaining the decondensed accessible state of chromatin. At late time points after DNA damage, the attenuation of γ-H2AX may allow chromatin to revert to a configuration that no longer permits access to DNA damage response proteins, thereby interrupting damage signaling even in cases in which DNA damage has not been adequately repaired (; ). This hypothesis can now be tested with the aforementioned methodology, which permits direct assessment of the physical state of damaged chromatin. A confocal microscope (LSM510 META; Carl Zeiss MicroImaging, Inc.) equipped with continuous wave multiline (351 and 364 nm) argon ion UV and multiline (458, 477, 488, and 514 nm) argon ion visible lasers along with a 40× C-Apochromat NA 1.2 water immersion lens (Carl Zeiss MicroImaging, Inc.) was used primarily for fluorescence and live cell imaging experiments. A heated stage (Carl Zeiss MicroImaging, Inc.) with an objective lens heater (Bioptechs) was used to keep the cells at the appropriate temperature (37°C) and growth conditions during imaging. ESI was performed using an energy-filtering transmission electron microscope (Tecnai 200keV or CM120keV; Philips) equipped with an imaging filter electromagnetic spectrometer (Gatan) and Digital Micrograph software (Gatan). Net phosphorus and nitrogen elemental maps were obtained by collecting pre- and post-edge images specific for either phosphorus or nitrogen. The jump ratio method was used to calculate the resulting net elemental maps (). Correlative light and ESI microscopy were performed with the following steps: imaging samples using the LSM510 confocal microscope, immunolabeling the samples when applicable, embedding in a Quetol 651 epoxy-based resin (Electron Microscopy Sciences) suitable for EFTEM, sectioning the samples to be picked up on copper finder grids, imaging the sectioned samples using an upright wide-field epifluorescence microscope (Axioplan2; Carl Zeiss MicroImaging, Inc.) equipped with a 12-bit CCD camera (model ORCA ER; Hamamatsu), and collecting element-specific images using the ESI mode of an energy-filtering transmission electron microscope. The 364-nm emission of the UV laser described above was used to photoactivate GFP and to induce DNA damage. Incubation of cells with 7.5 μg/ml of the DNA-binding dye Hoechst 33342 (Sigma-Aldrich) was necessary for the UV laser to induce DSBs. However, the Hoechst dye was not necessary to photoconvert the PAGFP to the activated form using the UV laser. Cells were incubated in phenol red–free DME containing 10% FBS (Invitrogen) with or without Hoechst dye while imaging experiments were performed. The bleaching algorithm provided with the LSM510 software (Carl Zeiss MicroImaging, Inc.) was used to perform GFP photobleaching, PAGFP photoactivation, and to introduce DNA damage at specific sites within nuclei. The UV laser intensity was set to 0.86 nJ/pixel (measured at the specimen; see next section) for introducing DSBs and photoactivating PAGFP. We generally observed an ∼25-fold increase in PAGFP signal intensity upon photoconversion. Multiple populations of cells were irradiated either with γ irradiation over a range of doses (0–20 Gy) or exposed to different levels of UV laser microirradiation (50% laser output with 5, 10, 20, or 50% Acuosto optical tunable filter–modulated transmission; equivalent to 0.29, 0.48, 0.86, and 2.05 nJ/pixel, respectively, measured at the specimen plane) in specific subregions of the nucleus, fixed and immunolabeled against γ-H2AX, and imaged using consistent imaging parameters on the LSM 510 confocal microscope (described in Microscopy). The resultant confocal fluorescence images were corrected for background, and the integrated fluorescence intensity was measured for the entire nucleus in the case of γ-irradiated cells or for each individual subnuclear region in the case of UV laser–irradiated cells. A standard curve of relative integrated intensity to the amount of γ irradiation (Gy) was plotted (Fig. S4). A linear best-fit curve was calculated, and the resultant integrated intensity values were measured from the UV laser–irradiated samples fitted to the curve. The equivalent number of Gy was obtained from the equation of the best-fit curve. The 0.86 nJ/pixel level of UV laser microirradiation was used consistently for the experiments performed, which corresponds to the introduction of one DSB for every 941 kb of DNA. The density of DSBs within the UV laser–irradiated regions was measured as follows: the density of DNA within the nucleus was calculated by taking the volume of the cell nucleus to be 523.60 μm based on a 10-μm diameter sphere and the amount of DNA present in a nonreplicated nucleus to be 6 × 10 bp DNA, therefore resulting in 1.14 × 10 bp DNA/um. The volume of the subregions used and measured in the test sample was 15.2 μm, thereby providing ∼1.73 × 10 bp DNA exposed to UV laser microirradiation. From the γ irradiation (Gy) standard curve, the DNA within this subnuclear region was exposed to 5.25 Gy of irradiation, which corresponds to ∼184 DSBs given that 1 Gy introduces ∼35 DSBs in whole-cell γ-irradiated cells. This provides a density of DSBs in the UV laser–irradiated subregions of one DSB for every 941 kb DNA. The UV laser–irradiated regions in the H2B-PAGFP–expressing cells were approximately half (7.08 μm) that used in the aforementioned test cells, which, based on the same UV laser microirradiation conditions, corresponded to 2.45 Gy of γ irradiation in the subregions of H2B-PAGFP–expressing cells or ∼85 DSBs in 81 Mb DNA. WT, H2AX, and ATM MEFs as well as HeLa and mouse 3617 cells were used for imaging experiments. The H2AX and ATM cell lines were generated from embryonic day 13.5 embryos derived from intercrosses between H2AX or ATM mice as previously described (). The mouse 3617 mammary tumor virus (MMTV) gene tandem array–containing cell line has been described previously (). Cells were grown overnight in the presence of tetracycline to allow expression of the GFP-tagged GR protein. The following day and 30 min before exposing cells to UV laser microirradiation, 100 nM dexamethasone was added to the culture medium to induce translocation of GFP-GR to the nucleus and subsequent binding to promoter sequences in the MMTV gene array. For DNA FISH, MMTV gene array–containing cells were grown on gridded coverslips (Bellco Glass, Inc.). After UV laser microirradiation, cells were fixed in 4% PFA, and DNA FISH was performed as described previously (). The H2B-PAGFP gene construct was modified from the H2B-GFP and PAGFP constructs described previously (; ), which were supplied by T. Misteli (National Cancer Institute [NCI]) and G. Patterson (National Institute of Child Health and Human Development [NICHD]), respectively. GFP-53BP1 was generously provided by T. Halazonetis (Wistar Institute, Philadelphia, PA; ). H3-GFP was provided by H. Kimura (Kyoto University, Kyoto, Japan; ). NBS1-GFP was generated by cloning GFP in frame with mouse Nbs1. All cell transfections were conducted using FuGENE6 (Roche) with cells growing on either gridded on nongridded chambered coverglass (MatTek Corp.). Cells were transfected with either GFP-53BP1, Nbs1-GFP, H3-GFP, or H2B-PAGFP and exposed to UV laser microirradiation or 1–10 Gy of γ irradiation. Antibodies used in the immunofluorescence labeling protocol included anti-Nbs1 (1:1,000; ), anti-γH2AX (1:1,000; Upstate Biotechnology), and anti-ATM1981p (1:300; Rockland Immunochemicals). ATP depletion was achieved by incubating cells for 30 min before imaging in media composed of PBS supplemented with 10% FBS, 10 mM 2-deoxyglucose, 10 mM sodium azide (), and 7.5 μg/ml Hoechst dye, and ATP depletion was confirmed by the morphology and staining of mitochondria with 50 nM Mitotracker red CMXRos (Invitrogen). The Medical Imaging, Processing, and Visualization (MIPAV) software package (Matthew McAuliffe and coworkers, Center for Information Technology [CIT]/National Institutes of Health [NIH]) was used for area, center of intensity mass, and mean squared displacement measurements. Image series registration was performed using an optimized automatic 2.5D registration algorithm based on bilinear interpolation and least squares cost function. Region of interest statistics were generated using the statistics generator function in MIPAV. Area and volume measurements were also made using Imaris image processing and analysis software (Bitplane AG). Background-subtracted mean fluorescence intensities, line scan profiles, and particle tracking were also measured using MetaMorph image processing and analysis software (Universal Imaging Corp.). For H2B-PAGFP time series images, the intensity histogram was stretched for each time point to optimize viewing of the morphology of photoactivated regions for display purposes. Measurements were made on background-corrected raw data before or after image registration, with threshold levels set at an intensity value 10% greater than background and regions demarcated using the levelset volume of interest tool in the MIPAV software. Maximum intensity projections and phosphorus/nitrogen ratio (chromatin density) coefficients were generated and measured using the colocalization module of the LSM510 software. Chromatin density coefficients were calculated as follows: the phosphorus and nitrogen signal intensities representative of chromatin (taken from images of control nondamaged cells) were used as minimum signal thresholds for subsequent phosphorous/nitrogen overlay image sets. Nuclear pore complexes were used as a protein-only internal control representative of legitimate high nitrogen signal and low phosphorus signal above background. Within specific regions of the image sets, either in the irradiated areas (damaged) or adjacent to the irradiated areas (nondamaged), the number of pixels containing the phosphorus and nitrogen signal intensities representative of chromatin were added and divided by the total number of pixels within the region, which yielded the chromatin density coefficient. This method is similar to the procedure used to calculate colocalization coefficients (). Regions included in measuring the phosphorous/nitrogen chromatin density coefficients were chosen based on whether the region was inside the γ-H2AX or Nbs1 fluorescence map or contained immunogold particles, in the case of the γ-irradiated sample. Equivalent areas adjacent and outside the irradiated regions were chosen for calculating the chromatin density coefficients of nondamaged chromatin except for the H2AX sample, where a distant region of heterochromatin was included. Fig. S1 shows the mobility of chromatin regions containing DNA DSBs. Fig. S2 shows the mobility and expansion of chromatin in WT living cells. Fig. S3 shows the changes in specific chromatin structure upon UV laser microirradiation. Fig. S4 shows a comparison of γ irradiation (Gy) with UV laser microirradiation. Fig. S5 shows a line scan through a net phosphorus ESI micrograph. Video 1 shows the mobility of photoactivated chromatin containing DNA DSBs. Video 2 shows the mobility of photoactivated chromatin without DNA DSBs. Video 3 shows the mobility of GFP-53BP1–containing IRIF in living cells. Online supplemental material is available at .

xref #text To examine mammalian separase function in vivo, we performed targeted inactivation of this gene () in mice. Using homologous recombination followed by loxP-mediated deletion (), we deleted a 1.4-kb genomic DNA fragment containing exon 6 of the separase gene from embryonic stem (ES) cells. RT-PCR analysis of separase gene transcripts detected a reduction in wild-type separase expression to approximately half of normal levels. This analysis also discovered an additional transcript specifically in ES cells bearing the exon-deleted mutant allele (). Sequencing of the amplified fragment identified that this aberrant transcript was generated by the splicing of exons 5–7, which encoded a frame-shift mutation at codon 452. This transcript was present at ∼30% of the levels seen for the wild-type transcript, likely because of non-sense–mediated decay. Therefore, we concluded that functional separase expression was inactivated in cells bearing the exon-deleted mutant allele, which was designated (). Mutant mice generated from these ES cells that were heterozygous for the allele were intact and fertile. Homozygous mutants, however, could not be obtained by the breeding of heterozygotes, indicating that separase deficiency was embryonically lethal (). In utero analyses discovered the death of homozygous mutant embryos before embryonic day (E) 8.5, prompting us to investigate early development in separase-deficient animals. female mice were crossed to male mice, and blastocysts were obtained by uterine washes at E3.5. Genotype analysis identified homozygous blastocysts present at Mendelian ratios. Homozygous blastocysts could be identified by gross appearance under a dissection microscope (); the mutants were smaller than heterozygous or wild-type blastocysts () that were obtained from the same female. Culture of blastocysts in vitro for 1 d revealed that blastocysts homozygous for () were easily distinguishable from wild-type or heterozygous blastocysts by their smaller numbers of cells with abnormally large nuclei (). The total number of cells in blastocysts homozygous for was only 10% of those seen in heterozygotes. In contrast, the mean diameter of mutant cell nuclei, detected by Cytox green staining, was approximately twice that of heterozygotes (, compare F with E; quantitative data not depicted). Immunostaining of homozygous blastocysts with antibodies against pericentrin, which is a component of centrosomes (, red), revealed an increased number of centrosomes per cell (red pericentrin signals of the merged images in ). After an additional 3-d culture, neither expansion of the inner cell mass nor spreading of trophoblasts on the dish surface could be observed for homozygous blastocysts (). Heterozygous blastocysts cultured for 3 d () exhibited a normal inner cell mass (arrow) with observable spreading of the growing trophobalsts (arrowheads). DNA staining of homozygous embryos with Cytox green could not detect an increased incidence of apoptotic cell death (unpublished data), indicating that separase-deficient embryos suffered from cell cycle arrest or retarded growth at the blastocyst stage that resulted in death at an early embryonic stage. To investigate the growth defects of separase-deficient cells, we established mutant mice carrying a conditional allele of mutant separase (). These animals were generated by the insertion of a pair of loxP sequences into introns 5 and 6 (). RT-PCR analysis of separase gene transcripts detected separase expression in ES cells that was equal to the expression observed in wild-type cells, suggesting that the allele is functionally intact (). Mutant animals homozygous for , which were healthy and fertile, were crossed to mice. MEFs, which were prepared from the resulting embryos at E14.5, were infected with a recombinant adenovirus bearing the Cre gene (AxCre; ) to inactivate separase expression. We analyzed the growth profile of separase-deficient MEFs by quantitating cell numbers at the specified time points after infection (). MEFs exhibited a growth capacity identical to that of wild-type MEFs after infection with AxCre at a high multiplicity of infection (MOI; , dashed lines). AxCre infection, however, significantly retarded the growth of MEFs (, continuous lines). This effect was not observed after mock infection. This growth inhibitory effect was amplified by infections at higher MOIs, suggesting that the observed growth retardation resulted from separase deficiency. As we could not detect an increased incidence of apoptosis in the growth-retarded cells (unpublished data), we hypothesized that separase deficiency inhibited the observed increases in MEF cell numbers. To understand the mechanisms underlying this inhibition, we performed a cell cycle analysis of cultures using laser scanning cytometry (LSC; ). After infection with AxCre, MEFs maintained normal ploidy (2C and 4C; , bottom left), whereas cultures of MEFs revealed an accumulation of cells with an abnormally high ploidy, usually 8C and 16C after 2 and 4 d, respectively (, right). The increase in DNA content was not observed in mock-infected cells (, top left). As the total DNA content in MOI 200 cells was similar to that in the mock-infected control cells (unpublished data), additional rounds of DNA replication appeared to have occurred in the separase-defective MEF cells. Interestingly, the actual DNA content of each MOI 200 cell correlated roughly with the size of the nuclei (), suggesting that enlargement of the nuclei in MEFs infected with AxCre may follow the additional rounds of DNA replication. 4 d after AxCre infection, the numbers of centrosomes per cell also increased in MOI 200 MEFs (6.4 ± 4.3) from the numbers observed in mock-infected MEFs (2.1 ± 1.2). Therefore, we concluded that separase deficiency suppressed nuclear division and centrosome separation accompanied by cytokinesis in MEFs, resulting in the accumulation of cells with a high DNA content. We also generated a mutant mouse line deficient in Pttg, a mammalian homologue of securin. We used homologous recombination followed by Cre-loxP–mediated recombination in ES cells to generate a securin/Pttg mutant () with the deletion of exon 2 (). As reported, homozygous mutants for were viable and fertile (; ). We crossed heterozygotes with heterozygotes. Double heterozygotes ( ) were obtained at Mendelian ratios (unpublished data). These mice were subsequently crossed with homozygotes. mutants, suggesting that this genotype is also embryonically lethal (). mutant embryos died by E11.5. These embryos were easily distinguishable from their littermates as early as E9.5 by severely retarded growth (). These results strongly suggest that heterozygosity for separase function in the absence of function exhibits haploinsufficiency. A more detailed analysis of mutant embryos will be required to elucidate the molecular mechanisms responsible for this phenotype. embryos by E11.5 suggested a genetic interaction between these genes during embryonic development. To elucidate the molecular nature of this interaction, we analyzed separase and securin functions in MEF cells. mice were crossed to homozygotes; MEFs prepared from the resulting and embryos at E14.5 were infected with AxCre to inactivate separase. MEFs, which did not carry any floxed alleles, after AxCre infection. This growth suppression was increasingly evident in cells infected at higher MOIs (). As AxCre infection did not alter the growth profile of wild-type or MEF cells ( and not depicted), this result suggests that Cre expression alone by recombinant adenovirus may suppress the growth of MEFs lacking a functional securin gene. MEFs exhibited severely restrained increases in cell number after infection with AxCre (), suggesting that a single functional allele is not sufficient to support normal growth in securin-deficient MEF cells. To examine this defect in detail, we analyzed the ploidy profiles of these cells by FACS analysis (). MEF cells possessed 2C or 4C DNA, even 4 d after infection with AxCre, indicating that the additional DNA replication resulting in polyploidy that was observed in MEF cells did not occur in MEF cells. We also did not observe any changes in the incidence of mitotic cells, suggesting that mitosis proceeded normally in the presence of one intact separase gene, despite the absence of securin. We also could not detect any increased incidence of cell death in these cells. In contrast, the proportion of 4C cells significantly increased; 27.4% of and 35.6% of MEF cells exhibited 4C DNA at 4 d after infection with AxCre. MEF cells. To examine the chromosomal structure of MEF cells with abnormally high ploidy, we performed karyotype analysis. 3 d after infection with AxCre, MEF cells were mitotically arrested using colcemid; prepared chromosomal spreads were then observed. Wild-type cells arrested at prometaphase each contained 40 chromosomes, including a pair of acrocentric chromatids (). In separase-deficient cells, the majority of the spreads contained chromosomal clusters (, B and C, inset). The total number of chromosomal clusters per cell was ∼40; each chromosomal cluster contained two or four pairs of chromatids. These pairs were attached at their centromeric regions to form diploid or quadruple chromosomes, respectively (). To determine if these abnormal chromosome clusters contained multiple copies of the same chromatid, we applied the spectral karyotyping (SKY) method to these chromosome spreads. All chromosomes within clusters stained with a single color, demonstrating that they were copies of the same sister chromatid (, E and F; D was control). In separase-deficient cells, these abnormal connections of chromosomal pairs and tetramers at their centromeric regions likely resulted from defects in chromosome segregation during previous rounds of mitosis. It is also possible that in the absence of separase cohesin is abnormally abundant at the centromeric regions of chromosomes, resulting in aberrant DNA replication that creates abnormal chromosome connections. We next analyzed the localization of genomic DNA in the nuclei of interphase cells by FISH (). In control MEF cells expressing normal separase, all of the probes detecting specific chromosomal regions (chromosome 5–specific centromeric and telomeric probes are shown in green and red, respectively) detected two spots within each nucleus (). This result likely reflects normal diploidy; sister chromatids were located in such close vicinity that resolution of the two was not possible, even after DNA replication. In separase-deficient cells, telomeric probes detected several scattered spots reflective of the high ploidy. The centromeric probes, however, detected either two spots or clusters (, and N). These results indicated that abnormal connections of chromatids at their centromeric regions existed in separase-deficient cells, even during interphase. Separase thus plays an essential role in the separation of the centromeric regions of sister chromatids in mouse cells; separase deficiency resulted in the formation of aberrant centromeric connections between chromosomal pairs or tetramers. MEF cells after infection with AxCre (), we also analyzed chromosomal structures of these cells by karyotype analysis. MEF cells infected with AxCre, 14 spreads, each of them containing 80 chromosomes (8C), were picked up and analyzed. In these spreads, the centromeres of most chromosomes (78.4%) were apart from each other and localized free from other chromosomes. MEF cells infected with AxCre (79.6%). MEF cells suffering from the arrest or severe delay of cell cycle. These results suggested that the cell cycle progresses in separase-deficient cells, despite the persistence of connections between multiple chromosomes at their centromeric regions. To analyze the defect in cell division observed in the absence of separase, we infected and MEF with the recombinant adenoviruses AxCre and AxH-GFP, which encode GFP-tagged histone H2B. Using this technique, we monitored chromosomal dynamics during the cell cycle by time-lapse microscopy (). In cell cultures infected with AxCre, the majority of cells that had initiated mitosis during the observation period underwent normal mitotic division into two daughter cells (). In contrast, a normal pattern of chromosomal segregation was rarely observed in separase-deficient cells. The proportion of cells undergoing chromosomal condensation, however, was similar in both separase-deficient and control cell populations, indicating that the defect in segregation occurred after condensation. In the majority of separase-deficient cells (62.8%), the condensed chromosomes aligned to form the metaphase plate, but never segregated. As the cells could not enter anaphase (), the nonsegregated chromosomes then decondensed, reforming a single nucleus (). Nuclear reformation was accompanied by a cytokinesis-like cytoplasmic division, resulting in the production of a subset of anuclear cells (, arrowhead). In 9.3% of all mitoses in separase-deficient cells (4/43), cytokinesis divided the decondensing chromosomes (, asterisk), separating the nuclear chromatin into two blocks (arrows). This type of abnormal mitosis (cytokinesis in the absence of sister chromatid separation) is reminiscent of the “cut” (cell untimely torn) phenotype, which is observed in cut1/separase-deficient fission yeast (; ). MEF cells infected with both AxCre and AxH-GFP. In these cultures, cells undergoing chromosome condensation were identified and the mitotic process was then analyzed by time-lapse microscopy. The majority of these cells (87.0% or 20/23) segregated their chromosomes normally. No mitotic defects could be detected. In this study, we reported the construction and characterization of separase mutant mice. Using a conditional inactivation system, we also characterized separase-deficient MEF cells, which showed severely restrained increases in cell numbers. Our analysis demonstrated that separase is essential for chromosome segregation in mammalian cells. This result correlates well with those of previous studies in lower eukaryotes, including yeast, flies, and worms (; ; ). Although chromosome segregation was significantly impaired, time-lapse analysis of separase-deficient MEF could not identify any apparent defect or delay in either the condensation or metaphase alignment of chromosomes. Chromosome decondensation and cytokinesis also proceeded in the absence of chromosome segregation in separase-deficient cells. This abnormal form of cell division typically produced one cell containing a single large nucleus and a second anuclear cell-like structure, explaining the accumulation of cells with an abnormally large nucleus of high DNA content containing amplified centrosomes. We also observed this phenotype in separase-deficient blastocysts in vivo. Therefore, the presence of nonsegregated chromosomes in separase-deficient cells does not appear to cause additional defects in cell cycle progression events, such as DNA replication or centrosome duplication. In fission yeast, the /separase mutation results in cell death after the appearance of unsegregated chromosomes torn apart by cytokinesis (the cut phenotype; ; ). Inhibition of cytokinesis in cut1/separase mutant cells, however, prevents cell death and allows cells to enter the cell cycle (; ), suggesting that failure of chromosomal segregation is not the cause of cell death in separase-deficient cells. From yeast to mammals, accumulating evidence suggests that although separase inactivation may prevent chromosome segregation, it does not interfere with other events in cell cycle progression, such as chromosome condensation and decondensation, metaphase chromosome alignment, cytokinesis, DNA replication, and centrosome duplication. All of the chromosomes in our chromosome spreads from separase-deficient MEFs exhibited abnormal chromosomes, which were connected only at the centromeres. This chromosome structure suggested that separase was required for the separation of sister centromeres, but not of the arm regions. Indeed, SKY analysis demonstrated that these abnormal chromosomes were multiples of identical sister chromatids. Retention of these centromeric connections was confirmed by FISH analysis of interphase cells. As cell cycle progression occurred normally in separase-deficient MEF, these abnormal chromosomes may be generated by extra rounds of DNA replication of nonsegregated chromosomes, suggesting that selective centromeric linkages are maintained throughout the cell cycle in the absence of mammalian separase. In this model, the sister chromatid pairs that had failed to segregate in separase-deficient cells would replicate again during the next S phase, then condense normally at the next mitosis, retaining their abnormal centromeric connections. Although this type of aberrant chromosome was not observed in separase-deficient HeLa cells generated with knockdown technology (), inactivation of separase expression may not have been complete. Our data strongly suggest that mammalian separase is essential for centromere separation, but not for chromosome arm separation. There does not appear to be a checkpoint system capable of detecting and/or repairing the abnormal centromeric connections anywhere throughout the cell cycle in mammals. Although homozygous mutant mice deficient in separase underwent embryonic lethality, heterozygous mutants were viable and apparently normal. On a securin-deficient background, however, heterozygous separase mutants also exhibited embryonic lethality. Heterozygous separase-deficient MEF on a securin-deficient background also exhibited severely restrained increases in cell numbers, as seen in homozygous separase-deficient MEF. These results suggest that securin may play a positive role in promoting separase function. Securin has been reported to function as a chaperone to stabilize separase in human cells and fission yeast (; ). Therefore, separase heterozygosity would be insufficient to support increases in cell number on a securin-deficient background. In contrast to the phenotype of separase-deficient MEF, we could not identify any apparent mitotic abnormalities in heterozygous cells on securin-deficient background by time-lapse analysis of living cells. Karyotype analysis also failed to detect abnormal chromosomes, such as diploid or quadruple chromosomes, indicating that separase heterozygosity is sufficient for sister chromatid separation, even on a securin-deficient background. Instead, we observed an accumulation of 4C cells, suggesting a possible defect in interphase. Although we are not able to exclude the possibility that additional mitotic defects might be concealed by the limited number of mitosis observed on a securin-deficient background, these results suggest that the cell cycle was significantly delayed in G2 phase. A function for separase in interphase has recently been reported. In fission yeast, separase-mediated cleavage of cohesin during interphase was essential for DNA repair (). Autocleavage of human separase also plays a role at the G2/M transition (). Our results also suggest an interphase function for separase in cell cycle progression that is independent of its role in mitosis. DNA damage that occurs spontaneously in these cells may not be efficiently repaired, causing the cell cycle to be delayed in G2 by activation of the damage checkpoint. Our analyses of mutant mice established that mammalian separase is essential for the separation of sister chromatid centromeres, probably through the separase-mediated proteolytic cleavage of cohesin in the centromeric regions. Cohesin complexes in mammalian cells are released from the chromosome arm regions without the requirement of separase-mediated cleavage (). If, however, a small amount of separase was present in the heterozygous cells, mitosis progressed normally; no polyploid chromosomes could then be observed, even in the absence of securin, whereas the cell number increase was severely retarded. These results suggest that a small amount of separase may be sufficient for the removal of centromeric cohesin during mitosis, even in the absence of securin, but more separase is required in the absence of securin for progression through interphase, a phase in which separase performs a function that remains to be identified. We obtained an ∼20-kb mouse genomic DNA fragment containing the NH-terminal separase sequence by screening a 129SVJ mouse genomic DNA phage library with an NH-terminal human separase cDNA fragment as a probe. To construct a targeting vector, we inserted an 11.5-kb mouse genomic fragment between the ApaLI site in separase intron 5 and the NotI site within the cloning site of a phage clone. A pGK-neo-polyA fragment flanked by a pair of sequences was inserted into the EcoRI site of intron 5. An additional sequence was inserted into the SmaI site of intron 6. A DT-A fragment was ligated to the 5′ end of the targeting vector to facilitate negative selection. After linearization by digestion with SacII, the targeting vector was electroporated into J1 ES cells as previously described (). After selection in G418, homologous recombinants were identified by Southern blot analysis using a 375-bp HindIII–EcoRI fragment containing separase exon 5 as a probe. Positive clones were electroporated with pCre-PAC (), which transiently expresses the Cre recombinase. Clones containing the or loci were identified by Southern blot hybridization. After HindIII digestion, hybridizing fragments of 11.3, 5.4, 3.7, or 2.4 kb should correspond to the wild-type, , , or alleles, respectively. We injected the mutant clones into C57BL/6J blastocysts to create chimeric mice. These animals were crossed to C57BL/6J mice, and germline transmission was confirmed by either genomic Southern blotting or PCR analysis of mouse tail DNA. For PCR, the combination of primer #2 (5′-CAGATCCTTGCCCTAGATCTCAGGC-3′) and primer #3 (5′-CTACCCAGGCTAGTGCCCTCTACTG-3′) detected a 272-bp fragment derived from the wild-type allele and a 414-bp fragment derived from the allele. The combination of primer #1 (5′-TCCTGGCACTTGGGAACCAGAGGTG-3′) and primer #3 detected a 356-bp fragment derived from the allele. The use of animals in this research study complied with all relevant guidelines for the ethical treatment of animals of the Japanese government and the Japanese Foundation for Cancer Research Cancer Institute. Total mRNA of ES cell clones was isolated using Micro-Fast Track (Invitrogen) and reverse transcribed using random primers. A cDNA fragment, including the region surrounding exon 6, was amplified using primers #12 (5′-TGTTGGAGGCCTTAGAGGGCCTGTC 3′) and #13 (5′-CTCTCCACATGCAGCCTGAAGCACC-3′), which correspond to sequences within exons 5 and 7 of the separase gene, respectively. The amplified fragments were separated by electrophoresis and analyzed by sequencing after subcloning into a plasmid. Blastocysts, obtained at E3.5 from female mice crossed with male mice, were cultured on Terasaki plates at 37°C. Bright field images were acquired with an inverted microscope (DM IRE2; Leica). For immunofluorescence microscopy, blastocysts were fixed in 4% PFA and permeabilized with Triton X-100. Cytox green (Invitrogen) and an anti-pericentrin polyclonal antibody (Covance) were used to stain DNA and centrosomes, respectively. Immunofluorescence images were taken through a microscope (DM RE; Leica) with a confocal microscopy system (TCS SP2; Leica). Each blastocyst was carefully recovered and genotyped by PCR, as described in the previous section. MEFs were obtained from embryos at E14.5, as previously described (), and were maintained in DME containing 10% fetal bovine serum at 37°C. Cells were used for analyses within three passages. Exponentially growing cells were plated at 5 × 10 cells per well in 6-well dishes. After a 12-h incubation, cells were infected with AxCre (3.3 × 10 plaque-forming units) at an MOI of 20 for chromosome analysis or 200 for FISH analysis. Cre-mediated recombination was confirmed by both genomic Southern blot and PCR analysis. To count centrosome numbers, cells were fixed with cold methanol and stained with anti–γ-tubulin antibody (Sigma-Aldrich). For flow cytometric analysis, cells were fixed in 70% ethanol and stained with 100 μg/ml propidium iodide solution after treatment with 2.5 mg/ml RNase A for 30 min. Cellular DNA content was also analyzed by laser scanning cytometry (LSC2 system; Olympus). For time-lapse imaging, cells were plated in 35-mm dishes before coinfection with AxH-GFP (MOI of 2), which encodes GFP-tagged histone-H2B, and AxCre (MOI of 50). Time-lapse images were taken at 1-min intervals though an inverted microscope (Leica) with a time-lapse system (AS MDW; Leica) at 37°C. To obtain chromosome spreads, MEF were exposed to 0.1 μg/ml colcemid for 2 h, treated with a hypotonic 0.075 M KCl solution for 15 min, and fixed in ice-cold Carnoy's fixative. For SKY analysis, chromosome spreads were treated with a 0.003% pepsin solution (0.01 M HCl) for 15 min and stained with a SkyPaint kit (Applied Spectral Imaging). Chromosomes were also counterstained with DAPI. SKY images were acquired through a fluorescence light microscope (BX50; Olympus) with a spectral imaging system (SpectraView SD-300; Applied Spectral Imaging). FISH analysis was performed as previously described (; ) using bacterial artificial chromosomes as probes. Centromeric (RP23-315O5) and telomeric (RP23-159N17) probes specific for chromosome 5 were labeled with biotin-16-dUTP and digoxigenin-11-dUTP (Roche), respectively, by nick-translation. These labels were detected with FITC-avidin and anti–digoxigenin-rhodamine, respectively. FISH images were acquired through a microscope (Axioplan2; Carl Zeiss MicroImaging, Inc.) with a confocal microscopy system (LSM 510; Carl Zeiss MicroImaging, Inc.).

Cohesion between sister chromatids is essential for the biorientation of chromosomes on mitotic spindles (). By resisting the tendency of microtubules to pull sister chromatids apart, cohesion creates the tension needed to stabilize the attachment of microtubules to kinetochores (). Chromosome segregation cannot, however, actually take place until the links holding bioriented sister chromatids together are broken, a process that occurs simultaneously on all chromosomes a few minutes after the last chromosome has bioriented (). Thus, loss of sister chromatid cohesion triggers what is possibly one of the most dramatic events in the life of any eukaryotic cell—the sudden migration of sister chromatids to opposite poles, an event known as the metaphase–anaphase transition. Sister chromatid cohesion is mediated by a complex called cohesin () whose two structural maintenance of chromosomes proteins (Smc1 and -3) and a single α kleisin (Scc1/Rad21) subunit join together to create a tripartite ring within which, it has been proposed, sister DNAs are topologically entrapped (). Crucially, sister chromatid cohesion is suddenly destroyed at the onset of anaphase by the cleavage of cohesin's α kleisin subunit by a protease called Separase (), which opens the cohesin ring and causes it to dissociate from chromosomes. Because loss of sister chromatid cohesion before chromosome biorientation is disastrous for chromosome segregation, cleavage of cohesin by Separase is tightly controlled. For most of the cell cycle, Separase is bound by a chaperone called Securin, which inhibits its proteolytic activity (; ; ; ). Once all chromosomes have been bioriented, Securin is targeted for proteasomal destruction by a ubiquitin ligase called the anaphase promoting complex or cyclosome (APC/C; ; ; ), resulting in Separase activation. In vertebrate cells, Separase is inhibited not only by Securin but also by phosphorylation at the hands of Cdk1 (). In these cells, therefore, APC/C triggers Separase activation through the simultaneous destruction of Securin and of Cdk1's activating subunit cyclin B. In most, if not all, organisms, Securins have both positive and negative effects on Separase activity. Thus, in and , inactivation of the Securins () and (), respectively, is lethal and causes phenotypes very similar to inactivating Separase. Though not lethal, deletion of the genes in mice (), human tissue culture cells (), or () also has adverse effects on sister chromatid separation. In the yeasts and , either inactivation of Separase or expression of noncleavable α kleisin subunits prevents sister chromatid separation (; ), and in , α kleisin cleavage is even sufficient for triggering anaphase (). Given the importance of the metaphase–anaphase transition and the degree of control that is exerted over this process, it is essential to know whether the chemistry of sister chromatid separation unearthed in yeast is shared by all eukaryotes, including humans. However, little is known about the functions of both α kleisin cleavage and Separase in organisms other than yeast. Thus, in () and (), Separase is known to be required for sister chromatid separation, but whether it triggers anaphase by cleaving α kleisins is not known. In mammals, α kleisin can be cleaved by Separase purified from tissue culture cells, a small fraction is indeed cleaved at the metaphase–anaphase transition (), and expression of a noncleavable version interferes with chromatid segregation at anaphase (). Investigation of Separase's in vivo function has hitherto been confined to the use of RNA interference to deplete it from tissue culture cells, which interferes with chromosome segregation and causes the production of highly abnormal (polyploid) nuclei (; ). However, it has so far not been possible to directly observe the entry into and passage through mitosis of cells known to lack Separase. It is therefore not yet known for certain whether Separase is essential for sister chromatid separation in mammalian cells. It is in fact not a forgone conclusion that Separase is essential for sister chromatid separation in mammals because most cohesin dissociates from chromosome arms (but not centromeres) during prophase and prometaphase (; ; ). This process is called the prophase pathway and is at least partly dependent on phosphorylation of cohesin's Scc3-SA2 subunit () but not, apparently, by cleavage of its α kleisin subunit. A similar process could conceivably also contribute to sister chromatid separation at anaphase, when cohesin persisting at centromeres disappears from chromosomes. To address as rigorously as possible the role of Separase during the chromosome cycle of mammalian cells, especially when they are growing in the context of real tissues within animals, we used homologous recombination in embryonic stem (ES) cells to replace the wild-type gene by a version in which the eight COOH-terminal exons encoding part of its conserved protease domain are flanked by sites and can therefore be deleted from the genome by Cre-recombinase expression. We chose this approach because manipulation of the genome has proven to be a more reliable method of altering gene function than methods that merely alter the abundance or activity of gene products. We find that deletion of Separase specifically blocks sister chromatid separation but not other aspects of mitosis, mitotic exit, cytokinesis, or even chromosome rereplication. To generate a conditional allele, we created a targeting vector in which the most COOH-terminal eights exons of the locus, which encode part of the conserved COOH terminus, including its catalytic dyad, were flanked by sites. An identical strategy was used to create a targeting vector for the gene in which its three COOH-terminal exons were flanked by sites (). HM1 ES cells were transfected separately with the and targeting vectors, and G418-resistant HM1 ES cell clones in which a single or locus (allele) had been replaced by the targeting construct were identified by Southern blotting. Transient transfection of these clones with a plasmid expressing Cre recombinase created or deletion alleles. Three independent ES cell clones carrying or deletion alleles of or were injected into C57BL/6 blastocytes. Chimeras were crossed with C57BL/6 mice to obtain germ-line transmission (). mice were intercrossed. , 42 were , and none were (). We detected no obvious difference between the development, health, or behavior of +/+ and /+ mice. These data imply that a single copy of the gene is both necessary and sufficient for embryonic development. intercrosses at 6.5, 7.5, and 8.5 days post coitus (dpc). 14 were +/+, 40 were /+, and none were /, which implies that Separase is required for early embryogenesis (). We have no explanation for the slight but significant excess of /+ embryos. Other crosses yielded at roughly expected frequencies mice homozygous for alleles () as well as mice with one deleted and one allele ( ). The normal appearance of such mice implies that the allele of is possibly as functional as wild type. mice at roughly the expected frequencies. These mice were fertile, both as males and females, albeit less so than wild type (unpublished data). This confirms previous reports that the mouse gene is not essential for either mitosis or meiosis (; ). and mice. , 22 were , 25 were , and none were (). genotype should have been as frequent as the other three classes, and their absence suggests that embryonic development in the absence of Securin requires both copies of the gene. embryos are smaller than embryos at 10.5 dpc, they have irregular somites and abnormal neurale tubes (), and they die at 11.5 dpc. Hoechst staining of longitudinal paraffin sections from 9.5 dpc embryos revealed extensive cell death and larger lobed nuclei in several organs (somites, heart, and brain) of but not embryos (). To determine whether such nuclei arise from mitotic defects, we cultured mouse embryonic fibroblasts (MEFs) from 10.5 dpc embryos. embryonic cells after 10 d in culture. MEFs possessed abnormally large lobed nuclei (). Out of 24 mitotic cells, 16 had multipolar spindles and 8 were undergoing anaphase with lagging chromosomes. embryos, only one was undergoing anaphase with lagging chromosomes and none contained multipolar spindles. MEFs did not reveal the diplochromosomes characteristic of MEFs completely lacking Separase (see Loss of Separase causes polyploidy). embryos that a further twofold reduction in its activity leads to lethal chromosome missegregation. Separase is mainly expressed in proliferating cells. However, modest amounts can be detected by Western blotting in liver extracts from adult mice (; unpublished data). To determine whether Separase has a function in resting hepatocytes, we used an transgene to delete both copies of from liver cells of mice. Despite efficient deletion of the gene (), which leads to a reduction in the level of Separase in their livers (), 10 out of 10 mice survived >6 mo without any overt pathology (not depicted). To address the function of Separase in proliferating hepatocytes, we induced entry into the cell cycle of resting hepatocytes by surgical removal (hepatectomy) of two thirds of the liver. Mice can survive with the reduced liver mass for several days, but regrowth is required for long-term survival. 10 mice, 10 , and 10 mice were first injected with 400 μl poly(I)poly(C) (pI/C), and hepatectomy was performed 3 d later. Remarkably, all 30 mice, including all those with a genotype, survived for several months after the hepatectomy. Livers from all three sets of mice reached their original size ∼3 wk after hepatectomy. Southern and Western blotting confirmed that Separase had been efficiently deleted in the livers, both before and 17 d after the two thirds hepatectomy (). This result was surprising because it implies that liver regeneration does not require Separase. To further investigate the process of liver regeneration in the absence of Separase, 10 10 , and 10 mice were analyzed at different time points after pI/C injection and two thirds hepatectomy. Histological analysis at 3, 5, and 17 d after hepatectomy revealed that the size of cells and their nuclei was greatly increased after regeneration in livers lacking Separase (), whereas that of control livers was unaltered (not depicted). Feulgen-staining revealed that in the livers, some hepatocytes had a 2- or 8C DNA content but most had a 4C DNA content both before and 17 d after the two thirds hepatectomy (unpublished data). The DNA contents of hepatocytes from mice ( = 3) resembled that of their controls before hepatectomy, but their DNA contents had increased to 8C, 16C, 32C, or even higher 17 d after hepatectomy (). These data indicate that liver regeneration without Separase is accompanied by several rounds of genome rereplication in the absence of cell proliferation, which leads to the production of highly polyploid, albeit apparently functional, hepatocytes. mice (after collagenase perfusion) 3 d after injection with pI/C and analyzed them by live cell video microscopy (Fig. S1, available at ). Under these conditions, hepatocytes divide just once, and a maximum mitotic index of 5–10% is reached 48 h after cultivation. hepatocytes invariably produced two identically sized nuclei, but it was only sometimes accompanied by cell division. This means that mitosis usually creates a binucleate hepatocyte (). hepatocytes entered mitosis and aligned their chromosomes on metaphase plates, but in 80% of anaphase cells, sister chromatids failed to disjoin properly at the onset of anaphase. Cells nevertheless exited from mitosis and produced progeny containing micronuclei as well as abnormally large nuclei ( and Fig. S1; = 100 mitotic cells per experiment). These data suggest that the abnormal polyploidy of Separase-deficient hepatocytes arises not because of their failure to enter mitosis but because of their failure to undergo anaphase. pI/C causes efficient expression of Mx-Cre in bone marrow cells as well as in hepatocytes (). To investigate the consequences of Separase depletion in this compartment, we isolated hematopoietic cells from bone marrow at 2 and 3 d after pI/C injection of 10 , 10 , and 10 mice. and mice. By day 3, bone marrow from mice contained only erythrocytes (Fig. S2 A, available at ). mice survived despite their severe bone marrow aplasia and fully reconstituted their bone marrow by day 31 with or hematopoietic cells (Fig. S2 A) whose nuclear divisions and cell sizes appeared normal. Unlike hepatocytes, diploid hematopoietic cells appear to undergo rapid cell death in the absence of Separase. embryos and used the 3T3 protocol to create an immortalized MEF (iMEF). To inactivate the allele, we used adenovirus expressing Cre recombinase (AdCre). Infection with AdCre but not with adenovirus expressing GFP (AdGFP) within 4 d caused deletion of most alleles (), which was accompanied by a reduction in the level of Separase as measured by Western blotting () and the accumulation of cells with large and multilobed nuclei (). iMEFs infected with AdGFP () and must therefore be caused by deletion of the locus and not by infection with adenovirus or expression of Cre by itself. Remarkably, cells with even larger multilobed nuclei accumulated 3 wk after infection, and such cells entered mitosis with huge numbers of chromosomes (). These observations suggest that, like hepatocytes, iMEFs lacking Separase undergo multiple rounds of DNA replication despite failing to segregate their chromosomes at mitosis. At early stages, rereplication of chromosomes was accompanied by centrosome reduplication. iMEFs infected with AdCre, but 5% or fewer of those infected with AdGFP, contained multipolar spindles (). iMEFs were first grown to 100% confluency, which led to contact inhibition, and then infected with AdCre or -GFP, and 48 h later the cultures were split to stimulate their entry into the cell cycle (). FACS sorting revealed an increase in 4- and 8C cells relative to 2C cells 48 h after splitting of iMEFs infected with AdCre () and an appreciable number of cells with DNA contents of 16C or more after 72 and 96 h, whereas cultures infected with AdGFP showed no increase in DNA content. AdCre but not -GFP also caused a large increase in the number of cells with <2C DNA contents, and more floating cells were observed in these tissue culture plates, which indicates the accumulation of apoptotic cells. To analyze the state of chromosomes, cells at each time point after splitting were incubated for 5 h in the presence of nocodazole to enrich mitotic cells, which were then collected by shakeoff, spread on glass slides, and stained with Giemsa. iMEFs infected with AdCre (but none infected with AdGFP) contained diplochromosomes in which two sets of sister chromatids were either closely aligned in parallel or remained attached at their centromeres (). Samples collected at 72 h frequently contained quadrupled chromosomes, that is, four sets of sister chromatids associated with each other in the region of their centromeres. In cells that had undergone yet another round of DNA replication, we sometimes observed, albeit rarely, karyotypes in which eight sets of sister chromatids remained associated (). Spreads from mitotic cells sampled 96 h after splitting had even higher numbers of chromosomes, but most were single chromosomes containing a single pair of sister chromatids. These observations imply that cells lacking Separase fail to separate sister chromatids when they enter mitosis but nevertheless subsequently reduplicate their chromosomes. Remarkably, reduplication gives rise to chromosomes in which the two sets of sister chromatids produced by reduplication frequently remain associated with each other. We do not know whether the association between pairs of sister chromatids after reduplication in the absence of Separase is, like that between sister chromatids themselves, mediated by cohesin. iMEFs infected with AdCre or -GFP or with no virus. Cells were cultured as in and either collected by mitotic shakeoff and stained with Giemsa (Fig. S3, available at ) or processed for immunofluorescence microscopy (). Infection with AdCre caused a reduction in the frequency of anaphase and telophase cells but not that of prometaphase or metaphase cells (Fig. S3). AdCre also caused the appearance of abnormal metaphase-like cells that contained partly decondensed chromosomes (Fig. S3 B, a–c) and three types of highly abnormal telophases (Fig. S3 B), namely, cells whose chromosomes were untimely torn by cell cleavage (28% from abnormal telophases), cells with chromatin bridges connecting highly asymmetric chromosome masses (31%), and cells containing a single nucleus on one side of their cleavage furrow (40%). A strong reduction in normal anaphases and telophases was also observed by immunofluorescence microscopy when cells infected with AdCre were stained with DAPI and antibodies to the spindle midzone protein MkLp1 and the mitotic kinase Aurora B (). Aurora B normally relocates from centromeres to the midspindle in anaphase and accumulates at the midbody in telophase. In yeast the association of Aurora B/Ipl1 with the spindle depends on Separase (). Our experiments revealed that Aurora B is located at centromeres in prometaphase cells that are lacking Separase, but in many metaphase-like cells, Aurora B was distributed throughout the cytoplasm (). These observations imply that Aurora B can dissociate from chromosomes in the absence of Separase but then fails to associate with microtubules. It is unclear whether cells lacking Separase can form a proper midspindle. The inability of Aurora B to associate with microtubules could thus be either a direct or an indirect consequence of Separase depletion. These cells were nevertheless able to undergo cytokinesis, and during this process Aurora B became enriched in the cortical region of the ingressing cleavage furrow and later in the bridge that connects daughter cells (). Separase activity is therefore not essential for cytokinesis in mouse fibroblasts. To determine whether the lack of anaphase or telophase cells could be caused by a failure to activate the APC/C, we measured cyclin B levels by immunofluorescence. In cells infected with AdGFP, most prometaphase and metaphase cells were cyclin B positive, whereas anaphase and telophase cells were negative. In cells infected with AdCre, 66% of metaphase cells were cyclin B negative (). These cells have presumably activated the APC/C but despite this failed to separate their sister chromatids. cells that stably express an mRFP-tagged version of histone H2B and filmed cells as they formed metaphase plates. All eight cells infected with AdGFP underwent anaphase and cytokinesis, yielding daughter cells whose nuclei had equal amounts of mRFP fluorescence (). In contrast, none of the 16 cells infected with AdCre managed to segregate their chromosomes, despite forming apparently normal metaphase plates. All 16 cells nevertheless formed cleavage furrows even though their chromosomes remained at the spindle equator. In two cells, the furrows attempted to bisect the chromosomes as they decondensed, causing constriction of the chromosomes. These furrows later regressed, leading to the formation of binucleate cells. In two other cells, the cleavage furrow constricted the chromosomes in an asymmetric fashion. These cells completed cytokinesis, producing one interphase cell and another that did not attach to the plate. In the remaining 12 cells, cleavage produced one cell that contained all or most chromosomes and another that contained few if any and did not attach to the plate (). Irrespective of their precise fate, all 16 cells decondensed their chromosomes, indicating that they exited from a mitotic state. We conclude that Separase is necessary for segregating chromosomes at anaphase but not for mitotic exit or cytokinesis. cell line expressing a myc-tagged version of the cohesin subunit Scc1 (). This coimmunoprecipitated with Smc1 and -3 () and was concentrated between pairs of CREST (calcinosis, Raynaud's phenomenon, esophageal dysmotility, sclerodactyly, and telangiectasia) dots at centromeres of mitotic cells (treated with nocodazole) after extraction of bulk soluble cohesin (). On the diplochromosomes produced by infection with AdCre, Scc1-myc staining was enriched at both centromeres but not between them (). Faint staining was also observed on chromosome arms, exclusively between the sister chromatids. In cells infected with AdGFP, Scc1-myc staining on chromosomes was detectable in two thirds of mitotic cells positive for phosphorylated histone H3 (P-H3; , D [top] and E). Thus, cells that have initiated anaphase remain P-H3 positive for a considerable period after Scc1-myc has disappeared from chromosomes. In contrast, in cells infected with AdCre, Scc1-myc staining was detectable on chromosomes in nearly 99% of P-H3–positive cells (, D [bottom] and E). We also observed a small portion (∼5%) of cells positive for Scc1-myc on chromosomes but negative for P-H3 that we never observed in cells infected with AdGFP. These data suggest that cohesin persists at mitotic centromeres longer than in wild-type cells. The failure of cells lacking Separase to segregate their chromosomes may therefore be caused by their failure to remove cohesin. In the yeast , separation of sister chromatids is triggered by cleavage of cohesin's Scc1 subunit by a site-specific protease called Separase. Previous work using RNA interference is consistent with the notion that Separase has a similar function in mammalian tissue culture cells (). However, this method has two important limitations. First, it can produce off-target or nonspecific phenotypes; second, gene product depletion is rarely complete, often yielding hypomorphic phenotypes. We therefore introduced into the mouse germ-line a allele that permitted us to induce deletion of Separase's conserved protease domain upon induction of the Cre recombinase. We also created a allele of , Separase's inhibitory chaperone. By these means, we have shown that Separase is essential for mammalian embryonic development, that its activity is severely compromised in mice lacking Securin, and that Separase is essential for chromosome segregation at the onset of anaphase, both in hepatocytes in vivo and in iMEFs in vitro. It has been recently suggested that Separase is required for timely entry into mitosis (). Our experiments would not have detected a modest delay in the G2–M phase transition. However, our finding that cells lacking Separase repeatedly enter mitosis after reduplicating their chromosomes is inconsistent with the notion that Separase has a major role in promoting M phase entry. It is important to point out that our Separase deletion could in principle lead to the continued synthesis of a truncated polypeptide that, though lacking any protease activity, might nevertheless have other cell cycle functions. iMEFs lacking Separase enter mitosis and align chromosomes on metaphase plates but fail to completely segregate their chromosomes. This catastrophic failure is not detected by any surveillance mechanism (checkpoint) capable of arresting cell cycle progression. Thus, cells lacking Separase form cleavage furrows. These furrows sometimes bisect the chromosomes but more often leave them on one side of the dividing cell, producing one daughter cell with most chromosomes and another with few, if any. The former invariably exit from mitosis and with high frequency reenter the cell cycle, reduplicate their chromosomes, and fail again to undergo anaphase after entering mitosis. The polyploidization caused by inactivation of Separase in mammalian cells is broadly similar to that observed in fungi (; ; ; ). It also resembles the phenotype caused by inactivating Separase in (; ; ). Remarkably, both iMEFs and larval brain cells in can undergo many rounds of chromosome reduplication in the absence of Separase, creating huge cells containing hundreds of chromosomes. A very similar sequence of events appears to occur in hepatocytes in vivo when they are stimulated by hepatectomy to undergo cell division in the absence of Separase. In this case, the polyploid cells produced in the absence of Separase appear fully capable of sustaining liver function. Several lines of evidence suggest that the lack of chromosome segregation in cells lacking Separase might be caused by a failure to destroy sister chromatid cohesion. First, cells lacking Separase clearly biorient their chromosomes on metaphase plates. If we assume that the spindle forces during biorientation are similar to those that segregate chromosomes during anaphase, then the largely successful biorientation of chromosomes in Separase-deficient cells indicates that their spindles should also be capable of pulling chromatids to opposite poles of the cell during anaphase were it not for some other defect, for instance, in sister chromatid separation. The lack of chromosome segregation cannot be caused by a failure to activate the APC/C because cells destroy cyclin B, exit from mitosis, and usually undergo cytokinesis. Second, centromeric cohesin fails to disappear in mitotic cells lacking Separase, which indicates that sister chromatid cohesion may never be destroyed. Third, the finding of diplochromosomes after one round of rereplication in the absence of Separase and quadruplochromosomes after two rounds implies that sister chromatids that should have been separated at anaphase remain in sufficient proximity to each other during and after the next round of DNA replication that the two pairs of sister chromatids so produced remain closely associated. Fourth, the spectrum of phenotypes caused by inactivation of Separase in iMEFs resembles in many regards the spectrum of phenotypes caused by expression of noncleavable versions of cohesin's Scc1 subunit (). This includes the production of polyploid cells, cells that attempt to cleave through a central mass of decondensing chromosomes, and the production of diplochromosomes. It should be noted that most of these phenotypes are much more penetrant in Separase-deficient iMEFs than they are in HeLa cells expressing noncleavable Scc1. There is little or no chromosome segregation in iMEFs lacking Separase, and diplochromosomes are generated in virtually all cells that reduplicate their chromosomes. In contrast, many HeLa cells in which noncleavable Scc1 expression has been induced still undergo chromosome segregation, and it is difficult to know whether this is due to insufficient expression of noncleavable Scc1. Furthermore, diplochromosomes accumulated in only 5.4% of cells. There are two possible explanations for this discrepancy. Either the supposedly noncleavable Scc1 alleles are still partly cleavable in vivo or Separase has functions besides Scc1 cleavage that are necessary for efficient chromosome segregation at anaphase. To distinguish between these possibilities, it may be necessary to test whether artificial cleavage of Scc1 is sufficient to trigger anaphase, as has been performed in yeast (). Our filming of iMEFs suggests that they exit from mitosis and undergo cytokinesis with high efficiency in the absence of Separase or sister chromatid separation. Cells in embryos behave likewise in the absence of the Securin-like protein Pimples, which appears to be essential for Separase activity (). There is therefore no evidence so far that Separase has an essential role in promoting mitotic exit in somatic animal cells. The formation of diplo-, quadruplo-, and even octuplochromosomes in cells lacking Separase has also been observed in mutants (). In iMEFs, cohesin is clearly associated with the sister centromere pairs of diplochromosomes, but it is unclear whether appreciable amounts also connect the two sets of sister chromatids. This raises an interesting question as to the fate of sister chromatid cohesion inherited from the previous cycle in Separase mutants. Do cohesin-mediated bridges survive the next round of DNA replication and thereby somehow hold sister chromatid pairs together as well as sister chromatids, or do these bridges merely survive until the next round of DNA replication, and does the resulting close proximity of sister chromatids during replication cause the de novo production of connections, albeit abnormal ones, between the two sets of sister chromatids? Such abnormal connections could be mediated either by cohesin or by DNA catenation. In summary, our data show that Separase is essential for sister chromatid separation in mammalian cells as well as in yeast (), flies (), and worms (). The so-called prophase pathway that involves phosphorylation of cohesin's Scc3-SA2 subunit is capable of removing cohesin from chromosome arms in mammalian cells () but is prevented from removing cohesin from centromeres by the presence of shugoshins at this location of the chromosome (). Separase alone can resolve centromeric cohesion in mammalian cells. This dependence on Separase is a crucial aspect of mitosis because Separase (via control of the APC/C) and not the prophase pathway is subject to the highly sophisticated regulation needed to ensure that sister chromatid separation does not commence until all chromosomes have bioriented on mitotic spindles. Mouse and genomic DNA were isolated from a 129/Sv bacterial artificial chromosome (BAC) library (Research Genetics) by using a cDNA probe derived from dbEST AA165880 for and from AA790273 for . BAC clone 297K4 was used for construction of the targeting vector and BAC clone 435D15 for the targeting vector. The construction of the targeting vector and the gene targeting in HM1 ES cells was performed as described in . Correct integration of the targeting construct at the genomic locus was analyzed on EcoRV-digested ES cell DNA by using an external probe. The presence of the COOH-terminal site was confirmed by using an digest and an internal probe. or alleles were obtained by electroporation with 25 μg and confirmed by Southern blot analysis of an EcoRV digest with an internal probe. Correct integration of the targeting vector at the genomic locus was analyzed by an digest using an external probe. With an internal probe and an digest, the presence of the third site was confirmed. and alleles were identified on Southern blot using a BamHI and digest and an internal probe. Chimeric mice were created as described in . Cre-recombinase expression was induced as described in . In hepatocytes, pI/C was injected 2×, and in hematopoietic cells 1×, within an interval of 72 h. BAC library high-density filter hybridization and Southern blot analysis were done as described previously (). Two thirds hepatectomy and single-cell DNA measurement were done as described previously (; ). Hepatocyte culture, live cell video microscopy, and immunofluorescence microscopy of hepatocytes were performed as described previously (). For paraffin sections, embryos were fixed overnight in 4% PFA/PBS, dehydrated to 100% EtOH, embedded in paraffin, and sectioned at 5 μm. After the deparaffinization procedure, sections were treated with Proteinase K, stained with Hoechst, and mounted. The following primers were used: 5′-ACATGACTCTGGGTGTGTCTTCTC-3′ and 5′-TTCATCACCCAAGCTCCAGCAG-3′ for the deletion allele; 5′-ACTGACCGTGACATTGACCGTTAC-3′ and 5′-TTCATCACCCAAGCTCCAGCAG-3′ for the allele; and 5′-ATGAGGAACTTCAAAGGAGTCAACTTC-3′ and 5′-GCGCAAGCCTTTAATCCCAG-3′ for the loading control. To detect Separase on Western blot, we used mouse monoclonal (7A6) antibody (). Other antibodies used in this study were as follows: mouse anti–Aurora B antibody (anti–AIM-1; BD Biosciences), CREST serum (a gift from A. Kromminga, Univiersity of Mainz, Mainz, Germany), mouse anti-myc antibody (clone 4A6; Upstate), antibody to phosphohistone H3 (Ser10[P-H3], a mouse monoclonal antibody that detects histone H3 when phosphorylated at serine 10; Cell Signaling Technology), anti-Topoisomerase IIα (Chemicon), mouse anti–cyclin B1 antibody (Santa Cruz Biotechnology, Inc.), and anti–γ-tubulin (Sigma-Aldrich). iMEFs were cultured in DME supplemented with 10% FCS, 0.2 mM -glutamine, 100 U/ml penicillin, 100 μm/ml streptomycin, 1 mM sodium pyruvate, 0.1 mM 2-mercaptoethanol, and nonessential amino acids. For infection, cells were grown to 100% confluency, washed with PBS, and infected with the virus (7,000 particles/cell) in DME supplemented with 2% FCS. After 24 h, cells were transferred to fresh medium. Cells were split after 24 h, and samples were harvested at different time points. Nocodazole was used at a final concentration of 100 ng/ml. AdCre and adenovirus expressing EGFP (Ad5 CMV Cre and EGFP) were purchased from The University of Iowa (Iowa City, IA). For generation of the H2B-mRFP stably expressing line, cells were infected with plasmid (a gift from J. Ellenberg, European Molecular Biology Laboratory, Heidelberg, Germany) using lipofectamine reagent (Invitrogen). Stable expressants were selected in a complete medium containing 800 μg/ml G418 and were screened by fluorescence microscopy for expression of H2B-mRFP. iMEF cell line stably expressing Scc1, murine Scc1 was COOH-terminally tagged with nine myc epitopes and inserted into vector (CLONTECH Laboratories, Inc.). The resulting plasmid was infected into φNX-Eco cells (Stanford University Medical Center, Stanford, CA) using lipofectamine reagent. cells containing transactivator. Selection with 100 μg/ml hygromycin B was started 48 h later. After 2 wk, cell lines arising from single cells were picked and tested for Scc1-myc expression by immunofluorescence microscopy. For flow cytometric analysis, cells fixed in 70% methanol were washed with PBS and subsequently stained in PI buffer (10 μm/ml propidium iodide, 10 mM Tris-Hcl, pH 7.5, 5 mM MgCl, and 200 μm/ml RNase A) for 20 min at 37°C. Preparation of hepatocytes and iMEF cell extracts for Western blotting was done as described previously (). Immunofluorescence microscopy and chromosome spreads were done as described previously (). Fig. S1 shows that hepatocytes undergo abnormal mitosis. Fig. S2 shows aplastic bone marrow in Separase-deficient mice. Fig. S3 shows that there is no anaphase but abnormal telophases in iMEFs lacking Separase. Online supplemental material is available at .

Kinetochores are multiprotein complexes that assemble on centromeric () DNA and attach chromosomes to spindle microtubules (MTs; ). Kinetochore–MT attachments generate the forces required for sister chromatid biorientation during metaphase and poleward movement during anaphase (). Evidence from a variety of organisms suggests that regulation of MT dynamics by kinetochores is critical to both of these processes and that multiple motor and nonmotor MT-associated proteins (MAPs) are involved (). The comparative simplicity of budding yeast s makes an attractive organism in which to undertake a thorough study of this aspect of kinetochore biology (). has six kinesins and a single dynein heavy chain (for review see ), but only the four nuclear kinesins—Cin8p, Kip1p, Kip3p, and Kar3p—are potential kinetochore subunits. Yeast nuclear kinesins belong to different subfamilies with distinct directionalities, structures, and functions. Cin8p and Kip1p are members of the kinesin-5 family of plus end–directed motors (BimC motors; ) that form homotetramers that are active in cross-linking parallel and antiparallel MTs (; ). Cin8p and Kip1p function in spindle assembly and in other MT-based processes (). mutants are viable at 25°C, but have high rates of chromosome loss and undergo frequent spindle collapse (); at 37°C, cells are dead. and are synthetically lethal, and KIP1 overexpression suppresses the spindle collapse phenotype of , though does not cause elevated chromosome loss. , but not , is synthetically lethal with (), presumably because checkpoint-mediated cell cycle delay is required for cells to complete mitosis successfully. Overall, these data show that Kip1p and Cin8p are functionally redundant (; ), but that Cin8p plays the larger role under normal circumstances. Kip3p belongs to either the kinesin-8 or -13 families (formerly Kip3 and KinI kinesins; Table S1, available at ). These families include the kinetochore motors MCAK in mammals; XKCM1 in ; KLP10A, KLP59C, and KLP59D in ; and Klp5 and Klp6 in (Table S1). Kinesin-13 motors destabilize MT protofilaments, causing MT depolymerization primarily at plus ends (). KLP10A and KLP59C mediate the disassembly of MTs from the plus and minus ends, respectively (). cells are resistant to the MT-depolymerizing drug benomyl, which is consistent with a role for Kip3p in MT destabilization in yeast (). Kinesin-8 and -13 motors are also thought to function during metaphase to correct improper kinetochore–MT attachment and to align chromatid pairs at the metaphase plate (for review see ). Thus, functions for kinesin-8 and -13 motors in vivo include kinetochore–MT attachment during metaphase and kinetochore MT (kMT) depolymerization during anaphase. Kar3p, the fourth nuclear motor in budding yeast, is a minus end–directed kinesin-14 family member that localizes to spindle pole bodies (SPBs) and the tips of cortical MTs (). Kar3p destabilizes MT minus ends in vitro and cytoplasmic MTs in vivo (; ) and has been found at low levels in biochemical preparations of the CBF3 -binding complex (; ). Like Cin8p (), Kar3p associates with DNA when assayed by chromatin immunoprecipitation (ChIP; ). Kar3p is involved in the sliding of minichromosomes laterally along MTs when newly formed kinetochores are captured by MTs. Endogenous chromosomes are bound to MTs throughout the cell cycle, however, making it unclear whether Kar3p functions at kinetochores during normal cell division. Functional analysis of nuclear kinesins in budding yeast is complicated by their involvement in multiple mitotic processes either individually or in combination. This multiplicity of function creates complex loss-of-function phenotypes. To begin to understand kinesin functions specifically at kinetochores, we have applied a series of fixed and live-cell assays that focus on kinetochore biology. We find that all four nuclear kinesins localize to kinetochores and perform the following three distinct functions: Cin8p and Kip1p are required for correct alignment and clustering of kinetochores on the metaphase spindle; Kip3p is required for coordinated movement of sister chromatids to spindle poles at anaphase; and Kar3p appears to function specifically at a subset of kinetochores on which MT attachments are slow to form. Thus, although nuclear kinesins in budding yeast are best known as essential players in spindle assembly, they also have important roles in ensuring the accurate attachment of kinetochores to MTs. To determine whether Cin8p, Kip1p, Kip3p, and Kar3p localize to kinetochores, we applied three criteria previously used in the analysis of other kinetochore proteins (). First, GFP-tagged kinesins were examined in fixed cells and localization patterns were compared with patterns for known kinetochore proteins; second, tagged kinesins were tested for association by ChIP; and third, the role of CBF3 in binding of kinesins was examined by using a temperature-sensitive mutation () in a subunit of CBF3. CBF3 is an essential four-protein complex that is required for initiating kinetochore assembly and for the recruitment of all known kinetochore proteins to DNA (; ; ). Ndc10p-dependent localization to kinetochores and association with DNA are diagnostic of kinetochore proteins. To image motor proteins, kinesins were fused at their COOH termini to GFP and integrated into endogenous loci in a strain containing Spc42p-CFP–labeled SPBs. Genetic tests established that the tagged motors were biologically active (see Materials and methods). In early mitosis, kinetochore proteins localize to a single focus spanning the short (<1-μm-long) distance between the spindle poles. Subsequently, at spindle lengths of 1.0–1.2 μm, kinetochores resolve into two distinct foci lying between the SPBs. This bilobed pattern is analogous to the metaphase plate in metazoans and is maintained until anaphase, at which time chromatids move poleward and become tightly associated with SPBs (; ). When metaphase cells were examined by three-dimensional (3D) deconvolution microscopy, Cin8p-GFP, Kip1p-GFP, and Kip3p-GFP were found to have bilobed localization patterns similar to that of Ndc80p-GFP, a well characterized kinetochore protein (, top). In addition, these kinesins also decorated interpolar MTs during metaphase (unpublished data). In anaphase, Cin8p-GFP and Kip3p-GFP were found at the spindle midzone and Kip1p-GFP localized to faint puncta along pole-to-pole MTs (pMTs), whereas Ndc80p-GFP was visible only near kinetochores (, bottom; ). Biochemical experiments have shown that the majority of Kip1p is degraded at the metaphase–anaphase transition (), implying that the Kip1p-GFP visible in anaphase cells represents a small fraction of undegraded MT-bound protein. Overall, imaging data suggest that Cin8p, Kip1p, and Kip3p localize to kinetochores, as well as to other MT-based structures. To further demonstrate this point, we established that the bilobed localization of Cin8p, Kip1p, and Kip3p was lost in cells, but that GFP fluorescence along spindle MTs was maintained (). Cin8p-GFP localized close to the poles in , whereas Kip1p-GFP was found along the spindle () and Kip3p-GFP was concentrated at the spindle midzone. We interpret these data to mean that in the absence of CBF3 the association of Cin8p, Kip1p, and Kip3p with kinetochores was disrupted, whereas localization to other MT-based structures was retained. In contrast to Cin8p-GFP, Kip1p-GFP, and Kip3p-GFP, Kar3p-GFP was found primarily along the nuclear face of SPBs and did not appreciably colocalize with Ndc80p-CFP ( and ). Localization of Kar3p-GFP to SPBs in living cells confirms previous immuno-EM data (). However, faint Kar3p-GFP foci were also visible along the spindle in ∼30% of early (<1.5-μm-long spindles) and 20% of late (1.5–2.5-μm-long spindles) metaphase cells (). These Kar3p-GFP foci were rarely, if ever, seen during anaphase () and did not adopt the bilobed pattern typical of core kinetochore proteins. To test the idea that Kar3p might associate specifically with detached or partially attached kinetochores (which are more abundant early in mitosis), cells were arrested in α-factor and released into the MT poison nocodazole. Most kinetochores in nocodazole-treated cells migrate to SPBs, apparently by following shrinking MTs (), but a subset becomes detached, moves farther from the SPBs, and recruits high levels of Bub1p, Mad1p, and Mad2p checkpoint proteins (). In nocodazole-treated cells, we observed that the majority of the Kar3p-GFP signal remained associated with SPBs (which were visualized with Spc42p-CFP), but in 75% of cells one or more faint foci were visible distal to the SPBs (). These fainter, more distant Kar3p-GFP foci colocalized with Ndc80p-CFP (, bottom). Based on our previous analysis (), Kar3p-GFP signals distant from SPBs almost certainly represent kinetochores that have detached from MTs (). SPB distal Kar3p-GFP represented 8 ± 2.5% of the total punctate GFP signal; implying efficient recruitment of Kar3p to detached kinetochores. We conclude that in vegetatively growing cells Kar3p becomes bound to detached or improperly attached kinetochores, but that most Kar3p is associated with SPBs. In this paper, we determine which kinesin motor proteins in localize to kinetochores and analyze the functions of kinetochore kinesins in metaphase and anaphase. Because has a closed mitosis, only the four nuclear kinesins Cin8p and Kip1p (kinesin-5 family members), Kip3p (a kinesin-8,-13/KinI motor), and Kar3p (a minus-end directed kinesin-14) have the potential to bind to kinetochores. ChIP has previously established that Cin8p and Kar3p associate with , and we show that this is also true of Kip1p and Kip3p, the two remaining nuclear motors. Live- and fixed cell imaging shows that kinetochores are one of the primary structures to which Kip1p-, Cin8p-, and Kip3p-GFP are localized during mitosis in normally growing cells. However, association with other structures is observed in cells lacking active kinetochores (as a consequence of disrupting the -binding complex CBF3), which is consistent with previous data showing that kinesins play important roles in spindle assembly. How are kinesins recruited to kinetochores? In the case of Kip3p, it appears that the motor binds directly to core kinetochore components; Kip3p remains -bound in and mutants, despite the dissociation of chromosomes from MTs. In this respect, Kip3p is similar to the human kinesin-13/KinI motor MCAK, which is a component of the inner kinetochore (). The finding that binding by Cin8p and Kip1p is partially, but not entirely, dependent on and is ambiguous in respect to the role of MT attachment, but it seems likely that both motors require MTs to associate with kinetochores. Other yeast kinetochore proteins, including members of the MT-binding Dam1–DASH complex, require MTs for kinetochore association (; ; ), as do plus end MAPs, such as CLIP-170, in higher eukaryotes (). Therefore, it seems that MT-binding kinetochore proteins in fall into two classes: those that are recruited directly by core kinetochore proteins and those that bind to, or are transported to, MT plus ends and then associate with kinetochores. These two classes of kinetochore MAPs must then interact to form a fully functional kinetochore–MT attachment site. In contrast to Kip1p, Kip3p, and Cin8p, which mainly localize to kinetochores, Kar3p-GFP is found primarily on the nuclear face of SPBs. It has been suggested that Kar3p might be a kinetochore motor, based on its copurification with CBF3 (; ) and genetic interaction with other motors (). However, previous immuno-EM data () are consistent with our live-cell imaging in showing Kar3p to be primarily SPB bound. Low levels of Kar3p can be detected at a subset of kinetochores early in mitosis, and higher levels can be detected on kinetochores that are detached from MTs by nocodazole-treatment. Kar3p has recently been implicated in lateral MT sliding of newly captured ectopic kinetochores (). However, kinetochores normally remain MT-bound throughout the cell cycle (), and MT capture is probably important only during a brief period in S phase. This may explain both the low levels of Kar3p on kinetochores under normal conditions and the absence of elevated chromosome loss in cells. Moreover, whereas Kar3p may function in de novo kinetochore–MT attachment, we have not been able to detect a kinetochore function for Kar3p in cells under normal growth conditions. Our data show that Cin8p and, to a lesser extent, Kip1p are involved in the generation or stabilization of the distinctive bilobed kinetochore clusters found in budding yeast from mid- to late metaphase. The disruption of bilobed clustering in mutants does not appear to reflect gross disorganization of the spindle, dramatic increases in the number of detached chromosomes, or changes in the fraction of transiently separated sister s. Instead, we speculate that Cin8p and Kip1p, like other kinesin-5 motors that can cross-link parallel and antiparallel MTs (), are involved in cross-linking kMTs. Because associate with a single MT, kinetochore-bound Cin8p and Kip1p must cross-link MTs from different kinetochores. Metaphase sister kinetochores can transiently separate by 0.5 μm or more; therefore, it seems unlikely that Cin8p and Kip1p are able to cross-link sisters; instead, we propose that cross-linking involves kMTs from different chromatids, though not necessarily kMTs emanating from the same pole (). If Cin8p and Kip1p, like human Eg5, can translocate to MT plus ends and remain attached (), the motors may actively bundle and link kinetochores together. Kinetochore–MT cross-linking could couple the polymerization of multiple MTs, perhaps explaining the requirement for Cin8p/Kip1p in forming bilobed kinetochore clusters. The significance of clustering is suggested by the appearance of detached kinetochores in mutants and an elevated rate of chromosome loss. Higher eukaryotes contain kinetochore fibers made up of 20 or more MTs. Bundles of yeast kMTs created by Cin8p and Kip1p may therefore resemble metazoan multistranded kinetochore fibers, except that multiple chromatids would be involved in the yeast MT bundles. Further analysis of Cin8p and Kip1p function during metaphase will require a deeper understanding of the forces that generate the bilobed configuration of yeast kinetochores, an effort that is currently underway in several laboratories. Live- and fixed cell imaging of cells reveals abnormally asynchronous sister chromatid separation during anaphase. A subset of chromatids in cells lags behind the majority and is found arrayed along spindle MTs at a point when the bulk of disjoined sisters have already arrived at the spindle poles. Surprisingly, a second subset of chromatids exhibits the opposite behavior—prolonged hyperstretching. Transient sister separation and chromosome stretching are observed in wild-type cells, but coordinated dissolution of sister cohesion and poleward movement generate only a brief period of hyperstretching at anaphase A onset. In cells, stretching is greater in magnitude and duration. Hyperstretching presumably reflects the initiation of poleward movement before the complete degradation of cohesin. Despite these problems early in anaphase, chromatids in cells are disjoined correctly by the end of anaphase B, consistent with a normal rate of chromosome loss in mutants (). The simultaneous generation of lagging and hyperstretched chromatids in cells implies a role for Kip3p in ensuring the synchronicity of poleward movement, presumably by coupling plus end MT depolymerization to the release of tension on sisters after cohesion degradation. In , a similar function has been proposed for kinesin-13 motors in . Kip3p function in yeast does not appear to be restricted to anaphase, however, because kinetochore dynamics during G1 and MT length in α-factor are altered in cells. double mutants. Overall, we conclude that budding yeast Kip3p, like kinesin-13/KinI motors in higher eukaryotes, plays an important role in the timely and efficient depolymerization of kMTs during anaphase and probably also during other phases of the cell cycle. We have established that all four nuclear kinesins localize to mitotic kinetochores in , implying considerable complexity in kinetochore–MT interaction. During normal cell division, Cin8p, Kip1p, and Kip3p are found at high levels on most, if not all, kinetochores, whereas Kar3p is found transiently on only a subset of maloriented or unattached kinetochores. The absence of Kar3p from the majority of metaphase chromatids suggests that kinetochores do not normally move poleward along the sides of MTs, though such motion may by observed during MT capture by newly assembled ectopic kinetochores (). Instead, it appears that in yeast, as in other organisms, the primary way that kinetochores move is by binding to MT plus ends and then altering their dynamics. Our data suggest that Kip3p is one protein involved in this regulation. Among our most striking observations is that Cin8p and Kip1p are important in organizing the bilobed metaphase configuration of yeast kinetochores. No precedent exists for this in higher cells, but we speculate that kinetochores with a single bound MT, such as those in , present mechanical problems not found in complex kinetochores that bind multiple MTs. Perhaps by bundling ∼16 kMTs (the number bound to one pole in a haploid) in cells creates a structure similar to a kinetochore fiber in higher cells, thereby strengthening MT attachment. Strains were derived from W303 or S288C. GFP-tagged proteins were constructed as previously described () and integrated into the genome to replace the endogenous wild-type copy. Because loss-of-function phenotypes for individual motor deletions are subtle, GFP-tagged kinesins were tested in strains carrying deletions exhibiting synthetic lethality, and the resulting compound mutants were tested for viability and growth. Kip1p-GFP was examined in cells; Cin8p-GFP in cells; Kar3p-GFP in cells; and Kip3p-GFP in cells. In all cases, compound mutants were viable, with growth rates that were indistinguishable from wild type. Kar3p-GFP has previously been shown to localize to the plus ends of cortical MTs in cells treated with α-factor (). In α-factor–arrested cells, Kar3p-GFP bound MT plus ends, furthering the belief that the GFP fusion was active. KanMX deletion strains were constructed by amplifying the deleted gene of interest from American Type Culture Collection deletion strains via PCR, using primers that were located 500 bp upstream and downstream of the deleted gene. PCR products were transformed into fresh cells, and correct integrants were confirmed by PCR. pFA6a-HisMX6 deletion strains were made as described in , using primers with at least 50 bp of homologous sequence. Image acquisition and processing were performed as previously described (), using a microscope with 100×, 1.4 NA, optics (Deltavision RT; Applied Precision) and a CoolSnap camera (Photometrics). Proteins were localized in fixed cells. and show consecutive frames from live-cell videos. All images are two-dimensional (2D) projections of 3D data. Fixed cells were prepared by treatment with 2% formaldehyde for 2–5 min, followed by 0.1 M phosphate buffer, pH 6.6, for at least 10 min before microscopy analysis. Live cells were grown in SD media for several hours and then resuspended in fresh media before imaging at either room temperature or at 30°C. Temperature-sensitive strains and wild-type controls were grown at 37°C for 3 h before cross-linking (), with minor modifications, as described in . Untagged strains served as a control. To establish the linearity of the ChIP assay, serial dilutions of immunoprecipitated or total DNA were used as substrates for PCR amplification of 200-bp or flanking fragments. ChIP signals were determined as a ratio of DNA in the immunoprecipitation to in the total DNA preparation. ChIP data is presented as a ratio of signals for mutant versus wild-type strains or versus flanking DNA. Table S1 lists the known kinetochore function of kinesin family members in yeast and other organisms. We have also included two figures further describing our live-cell imaging data. Fig. S1 shows graphs depicting live-cell movements of sister chromatids in metaphase in relation to a reference SPB in wild type, as compared with kinesin mutant cells. Fig. S2 shows a comparison of the probabilities of G1 dynamics data from wild-type and cells that are depicted graphically in . Online supplemental material is available at .

The aim of this study was to identify the homologue of the vertebrate Cajal body (CB). CBs were discovered >100 yr ago by the Spanish neurobiologist Ramón y Cajal (), but their major molecular components have been described in only the past 15 yr, and few specific biochemical functions have been assigned to them. There is a general consensus that steps in the assembly and modification of the RNA processing machinery of the nucleus take place in vertebrate CBs, including the machinery for splicing, preribosomal RNA processing, and histone pre-mRNA processing (for reviews see , ; ; ; ). Cajal's original studies involved mammalian neurons, and even today the majority of studies on CBs make use of cultured mammalian cells. Nevertheless, CBs occur in a wide variety of other organisms, including amphibians, insects, plants, and probably budding yeast (for review see ). The identification of CBs relies heavily on specific biochemical markers, of which the protein p80-coilin is the most widely used (; ). Orthologues of human coilin are known from several other vertebrates, including the mouse (), (), and (), as well as the plant (Shaw, P.J., personal communication). However, the overall sequence of coilin is not highly conserved, and attempts to identify coilin have so far been unsuccessful. Fortunately, four potentially specific CB markers, two proteins and two RNAs, have recently been described. Three of these—dLsm10, dLsm11 (, ; ), and dU7 small nuclear RNA (snRNA; )—are components of the U7 snRNP, which is required for histone pre-mRNA maturation. In the amphibian oocyte nucleus () and in HeLa cells (), U7 snRNA is localized almost exclusively in CBs. The fourth marker is dU85 (), which functions in the CB as a guide RNA for modifications on U5 snRNA (). U85 and related RNAs have been called small CB-specific RNAs (scaRNAs) because of their high concentration in vertebrate CBs. Significantly, in situ hybridization of dU85 revealed a single small focus of label in the nuclei of S2 cells, strongly suggesting that dU85 recognizes the CB (). We began our study of by examining the U7 snRNP on the assumption that U7 would be specific for CBs, as it is in and human cells. We identified a nuclear body that contains the U7 snRNP and showed that this body is physically associated with the histone gene locus. However, when we probed for four other CB components—dU85, dU2 snRNA, the survival of motor neurons (SMN) protein (dSMN), and fibrillarin—we found them colocalized in a nuclear body separate from the body that contains the U7 snRNP. These findings pose both substantive and terminological questions. Based on its apparently greater complexity, we designate the dU85/dU2/dSMN/fibrillarin body as the CB and the second nuclear body as the histone locus body (HLB). These two bodies are frequently close to one another or actually touching, although they may lie far apart in the nucleus. Our findings suggest that the CB, like the CB in other organisms, is a composite structure whose subunits in some cases fuse together or reside next to each other but sometimes lie in separate parts of the nucleus. dU85 scaRNA is a 316-nt nuclear RNA that contains sequences characteristic of both classes of small nucleolar RNAs (snoRNAs), the box C/D motif and the box H/ACA motif (). It is a guide RNA that simultaneously specifies the modification of two bases in dU5 snRNA, methylation at C46 and pseudouridylation at U47. Human U85 scaRNA was shown by in situ hybridization and biochemical fractionation to be localized exclusively in the CB, hence the name small CB-specific RNA (; ). U85 scaRNA has a sequence similar to that of human U85 and was localized by in situ hybridization to a discrete focus within cultured S2 cells (). In our experiments, we used a full-length antisense RNA probe to detect dU85 scaRNA. In situ hybridization was performed on a variety of tissues, including brains and salivary glands of third instar larvae and nurse and follicle cells from adult ovaries. dU85 was detected in sharply defined foci with only background levels of label elsewhere in the nucleus or cytoplasm. Most nuclei in the brain, salivary glands, and follicle cells displayed a single focus of label (). This was also true of most nurse cell nuclei up to about stage 4, but nurse cell nuclei from larger egg chambers often contained two or three foci (see ). Rarely did a large nurse cell nucleus contain as many as 10 foci. In situ hybridization was performed for dU2 snRNA in larval brains, salivary glands, and adult ovaries. In keeping with findings from vertebrate cells (; ; ), dU2 snRNA exhibited a speckled pattern superimposed on a more diffuse nuclear distribution. When double in situ hybridization was performed with dU2 and dU85, the most prominent “speckle” overlapped the dU85 signal (). However, the speckle was usually less discrete than the dU85 focus, and in some nuclei there were other equally prominent speckles that were not associated with dU85 (). In nurse cell nuclei, the strongest dU2 signals always coincided with dU85 signals (unpublished data). published an image of a S2 cell after double in situ hybridization for dU85 and dU2. In this case, the dU85 signal was more discrete than the dU2 signal and there was a moderate level of dU2 signal throughout the nucleus. In vertebrate cells, SMN is a ubiquitous cytoplasmic protein involved in the assembly of the Sm snRNPs (; ; ). A small amount of SMN is found in the nucleus, where it is concentrated in CBs or in bodies closely associated with CBs, called gems (). The distribution of dSMN has been examined with antibodies in cultured cells (), but its distribution in tissues of the fly has not been reported. Our observations were made on transgenic flies that express dSMN-YFP under the control of GAL4. Although we have not performed extensive genetic studies with these flies, we know that the transgene rescues hypomorphic and null mutations (Matera, A.G., personal communication). In keeping with the known distribution of the SMN protein in vertebrate () and cells (; ), dSMN-YFP is strongly expressed in the cytoplasm of all cells examined, including larval salivary glands and brains and adult ovaries (). Although there is a low level of dSMN expression in most somatic nuclei, we have seen discrete foci of protein only in follicle cell nuclei of the ovary and in nurse cell nuclei and the germinal vesicle (GV). The GV always has a single bright focus up to about stage 9; the nurse cell nuclei usually have a single focus, but two or even three foci are common (). Immunostaining with an antibody against dSMN reveals a similar pattern in wild-type flies (unpublished data). To determine whether the dSMN foci in the germline are related to the similar dU85 foci seen in nurse cell nuclei, we performed in situ hybridization for dU85 in ovaries from flies that expressed dSMN-YFP. There was precise colocalization of the dU85 and dSMN signal in the nurse cell nuclei (). Fibrillarin is the methyl transferase that carries out 2′--methylation of ribose moieties using box C/D snoRNAs as guides (). Although its name derives from its localization in the fibrillar part of the nucleolus (), fibrillarin is also found in CBs and in fact was one of the first proteins demonstrated in CBs by immunofluorescent staining (). The sequence of fibrillarin is highly conserved evolutionarily, and antibodies exist that react with many species. We used mAb 72B9 to detect fibrillarin in various tissues of , including brain and salivary gland cells of third instar larvae and follicle and nurse cells of the ovary. In cells double stained with mAb 72B9 for fibrillarin and DAPI for DNA, one usually sees only a single contiguous area of stain within the nucleolus (). We made this observation early in our investigation and were puzzled that we did not see a separate focus of fibrillarin corresponding to a CB. Only when cells were simultaneously labeled by in situ hybridization for dU85 did it become clear that the CB (U85 body) is always colocalized with fibrillarin stain. In most cases, the CB touches or lies within the nucleolus so that its fibrillarin stain is contiguous with that in the fibrillar zone of the nucleolus. Sometimes the CB is completely independent of the nucleolus, and in these cases there is a separate focus of fibrillarin stain (). The situation in the giant nurse cell nuclei can be equally confusing. In these nuclei, the nucleolus expands enormously and takes on an irregular lobulated form. The CB lies somewhere within the giant nucleus, generally surrounded by nucleolar material. The CB often stains more intensely than the nucleolus with mAb 72B9 and can be recognized without an additional probe. However, reliable identification of the CB requires in situ hybridization for dU85, in which case one can see that the CB always stains with the antibody against fibrillarin (). In the amphibian oocyte () and in HeLa cells (), the U7 snRNP colocalizes with coilin in the CB. Some of these CBs are associated with the histone genes. In , the situation is somewhat different. Here the U7 snRNP resides in a separate body that is invariably located next to the histone genes. The U7 snRNP consists of U7 snRNA associated with a ring of seven core Sm proteins. Five of the proteins are identical to the B/B', D3, E, F, and G proteins found in the U1, U2, U4, and U5 splicing snRNPs (for review see ). The other two, Lsm10 and Lsm11, take the place of D1 and D2 in the ring and, so far as is known, are unique to the U7 snRNP (, ; ). Our first observations of the U7 snRNP were based on the expression of YFP-dLsm11 in transgenic flies. To obtain expression, we crossed flies homozygous for a YFP-dLsm11 insert with flies homozygous for , which carry the yeast gene on the third chromosome under control of the () promoter. The promoter supports the constitutive expression of the GAL4 protein in many cells of most tissues. Four different inserts gave similar expression. Two of these were on the X chromosome and had YFP fused to the carboxy terminus of dLsm11 and two were on the second chromosome with YFP at the amino terminus. Many interphase nuclei of maintain the so-called Rabl orientation, in which the centromeres and associated heterochromatin are on one side of the nucleus and the free chromosome ends are on the opposite side. This orientation is particularly evident in blastoderm nuclei after cellularization (; ). Early in our observations it became clear that the HLB often lies at the periphery of the nucleus near the heterochromatin, suggesting a specific chromosomal association (; and ). Because of the well-known association of CBs with the histone genes in both amphibian (; ) and human nuclei (), we postulated that the HLB is also associated with these genes, which are located in polytene chromosome bands 39D2-3 to 39E1-2 near the centromeric heterochromatin of chromosome 2L (). This hypothesis was verified by in situ hybridization with a probe against histone H4. Hybridization was performed on squashes and whole mounts of various tissues from flies that expressed YFP-dLsm11. In larval brain cells, there was essentially complete colocalization of histone genes with the YFP-dLsm11 signal (). The same was true in the giant salivary gland nuclei and ovarian nurse cells. In all cells of all tissues, the HLB was invariably associated with histone genes. This situation is in striking contrast to that of amphibian oocytes and mammalian tissue culture cells, where a minority of CBs is associated with histone genes. Various tissues were examined that had been probed simultaneously for a CB component and an HLB component. In most cases, the CB and HLB lay close to each other or were in contact, but they never completely overlapped (). We made quantitative measurements on follicle cell nuclei that had been simultaneously hybridized with probes against dU7 and dU85. Multiple confocal images through individual nuclei were projected onto a single plane for analysis. The dU7 and dU85 bodies were said to be touching if their center-to-center distance was less than their diameter (0.5–1.0 μm for both bodies). They were said to be close if the center-to-center distance was greater than one but less than three diameters, and they were considered far away if the center-to-center distance was greater than three diameters. In a sample of 163 follicle nuclei, the two bodies were touching in 58% of the cases (94/163), close in 28% (45/163), and far away in 15% (24/163). These numbers overestimate the degree of association because they do not take into account the three-dimensional nature of the nucleus. In the larger nurse cell nuclei, there are multiple HLBs but only 1–3 CBs. Nevertheless, in these cells as well, most CBs lie close to or touch an HLB (). CBs in vertebrate cells are usually identified by immunostaining with antibodies against the marker protein p80-coilin (; ). Because an orthologue of coilin has not yet been recognized in , we used other markers to search for the CB. CBs in vertebrate cells contain many components involved in RNA processing. These include snRNAs and related factors that carry out pre-mRNA splicing; snoRNAs, fibrillarin, and other factors involved in preribosomal RNA processing; and the U7 snRNP and the stem-loop binding protein that are required for 3′-end cleavage of histone pre-mRNA. We examined the intranuclear distribution of proteins and RNAs from each of these categories: pre-mRNA processing (dU2, dU85, and dSMN), preribosomal RNA processing (fibrillarin), and histone pre-mRNA 3′-end cleavage (dU7, dLsm10, and dLsm11). We find that components in the first two categories reside in the same body, but components of the U7 snRNP are found in a separate nuclear body. In many nuclei, the two bodies appear to touch at the resolution of the light microscope. Furthermore, the body that contains the U7 snRNP is invariably associated with the histone genes. These relationships are shown diagrammatically in . The CB was identified in three earlier publications. The first was an electron microscope study by , who described a nearly spherical structure in the GV of the early oocyte, which they called the endobody. The CB, identified by dSMN-YFP expression (, inset), has the same size and relationship to the condensed chromatin as the endobody and occurs in the same early stages of the oocyte. Endobody is the English translation of the German Binnenkörper, a term originally applied to spherical, nonnucleolar structures in various insect oocytes, including flies, by . We previously showed by immunostaining that the endobody of the cricket is the oocyte equivalent of the somatic CB (). The second study, by , concerned the localization of a neuron-specific protein encoded by the gene (). The ELAV protein was detected by immunostaining in a small structure inside the nuclei of larval brain cells. This structure was also stained by a polyclonal serum (R288) raised against the carboxy terminus of human coilin (), from which the authors concluded that they had identified the CB (then referred to as the coiled body). Further experiments are needed to determine whether the ELAV-positive body corresponds to either of the nuclear bodies we describe here and what epitope is recognized by the antibody against human coilin. The third report, by , concerned the localization of dU85 scaRNA in cultured cells (S2 cells). These authors reported a sharply defined focus in the nucleus after in situ hybridization with a probe against dU85. Simultaneous hybridization with a dU2 snRNA probe showed a major concentration of U2 in the U85 focus, with additional diffuse labeling throughout the nucleus. called this focus the CB, based on the fact that human U85 is limited to the CB in HeLa nuclei. There is little question that the CB we describe in somatic and germline tissues of the fly corresponds to the CB in S2 cells, as our identification also relies on U85 and U2 probes. The HLB was first described in a study that dealt with chorion gene amplification (). “A subnuclear sphere of unknown identity” was seen in follicle cell nuclei after staining with mAb MPM-2, a relatively nonspecific antibody against phosphorylated proteins (). Staining of MPM-2 spheres depended strongly on cyclin E/Cdk2 levels in the cell. Subsequently, cyclin E and Cdk2 were shown to associate with vertebrate CBs in a cell cycle–dependent manner (; ), suggesting that MPM-2 bodies might be related to vertebrate CBs. Double-label experiments now show precise colocalization of MPM-2 and dLsm11 staining in blastoderm cells, thereby demonstrating identity of the MPM-2 body and the HLB (White, A., B. Calvi, W.F. Marzluff, and R.J. Duronio, personal communication). The existence of two separate nuclear bodies in that each contain canonical CB components emphasizes the composite nature of the CB. A similar composite nature is well documented in earlier studies on other organisms. One of the best-known examples concerns SMN and its associated gemin proteins, which can reside in separate nuclear bodies, the gems or Gemini of the CB (). In most mammalian cultured cells and adult tissues, SMN colocalizes precisely with coilin in the CB, but in fetal tissues and in the line of HeLa cells (PV) in which gems were first described, gems exist either as completely separate bodies or as close partners with CBs (; , ). The degree of association depends in part on the extent to which coilin is methylated (, ). In , we find that SMN is colocalized with U2 snRNA and U85 scaRNA in CBs in the female germline and in follicle cell nuclei. A diffuse nuclear distribution of endogenous SMN was previously demonstrated in S2 cells after staining with an antibody against dSMN. Cells transfected with dSMN also gave a diffuse distribution, whereas some cells transfected with human SMN had dotlike structures in their nuclei (). Colocalization of these dots with other CB markers was not tested. A second partner of CBs is the cleavage body, so called because it contains pre-mRNA cleavage factors, such as the 64-kD cleavage stimulation factor (CstF64) and the 100-kD cleavage and polyadenylation specificity factor (CPSF100). Cleavage bodies reside next to or overlap with CBs. When transcription was inhibited with α-amanitin or 5,6-dichloro-1-β--ribofuranosylbenzimidazole, cleavage factors redistributed and colocalized with coilin in the CB (, , ). Cleavage factors have not yet been looked for specifically in . Coilin knockout mice provide an especially informative case of CB heterogeneity. In the original description of fibroblasts derived from coilin −/− embryos (), the authors described “residual CBs,” which contained fibrillarin and Nopp 140 but no splicing snRNPs. Some of these cells also displayed separate bodies that contained SMN. In a subsequent study of coilin −/− fibroblasts (), a third nuclear body was found, which contained scaRNAs U85 and U93, along with their cognate substrates snRNAs U5 and U2. In mouse cells, coilin clearly plays an important role in bringing CB components together into a single nuclear body. The same conclusion was reached in a study of CBs in pronuclei (). CBs in pronuclei contain coilin, fibrillarin, and splicing snRNAs, but when coilin was depleted from the extract in which pronuclei were formed, snRNAs no longer accumulated in the residual CBs. The existence of the HLB provides a fortuitous opportunity to study the histone mRNA processing machinery separate from other CB components. For instance, the HLB should make it easier to examine molecules that are shared with other complexes, such as the Sm proteins that also occur in the splicing machinery and the recently discovered shared components of the cleavage and polyadenylation machinery (; ). A separate HLB also permits insight into the relationship of the body to the histone genes themselves. In amphibian oocytes (; ) and HeLa cells (; ), a minor fraction of CBs is associated with the histone genes, whereas in , HLBs are invariably present at the histone locus. One obvious result of this association is that the processing machinery is brought into proximity to the site of transcription of histone pre-mRNA. However, histone transcripts are synthesized and processed only during the relatively short S phase. Why, then, are HLBs present throughout the interphase period? Part of the answer may be that the processing machinery is simply stored where it will be used. Of greater interest, however, is the possibility that steps in the assembly or modification of the histone processing machinery occur in the HLB and that these events take place throughout the interphase. It is known that assembly and modification of the splicing machinery occur in CBs (; ; ). It may well be that the HLB plays multiple roles in assembly, modification, storage, and delivery of the histone mRNA processing machinery. Efforts to find a orthologue of coilin have so far proved unsuccessful. A coilin gene could reside in an unsequenced heterochromatic region, or its sequence could be so divergent that it is not recognizable by the usual similarity comparisons. However, if coilin is indeed absent, cells may lack an important component of the “glue” that holds CBs together in vertebrate cells. It will be particularly instructive to examine transgenic lines that express a vertebrate coilin gene. Such lines could provide important clues concerning the molecular interactions of coilin as well as its role in the physical organization of CBs. Based on sequences published by , we constructed plasmids that contained and genes tagged with the HA epitope or with a modified YFP called Venus (). Transcripts from these constructs were injected into the cytoplasm of oocytes, and the newly translated proteins were identified in the oocyte nucleus by Western blotting and immunostaining. YFP-labeled dLsm11 was properly translated and efficiently targeted to CBs in the oocyte nucleus, whether the YFP tag was at the amino or carboxy terminus of the protein. YFP-labeled dLsm10 was properly translated but not well targeted to CBs in these preliminary experiments. Therefore, we concentrated on YFP-labeled dLsm11 in subsequent studies on . We made two element constructs of , with YFP at either the amino or carboxy terminus of the protein. The element was pUASp, in which the cloned protein is under control of the yeast upstream activating sequence (UAS; ). pUASp is a modified version of pUASt () and was used because it shows enhanced expression in the ovary. element transformation was performed by standard procedures. Four different homozygous viable lines were obtained, two with YFP at the amino terminus and two at the carboxy terminus. The precise positions of the inserts were determined by sequencing () and in situ hybridization as follows: for VW-1 and VW-5, single elements on chromosome 2 at 5,981,114 and 7,576,630, respectively; for WV-1, two elements on the X chromosome at 11,567,054 and 19,472,038; and for WV-2, a single element on the X chromosome at 11,567,054. We also constructed a pUASp plasmid that contained the gene upstream of the YFP tag. Two different homozygous viable lines were isolated, one with on chromosome 2 and one on chromosome 3. Brains and salivary glands from third instar larvae and ovaries from adult flies were examined as whole mounts or as squashes. Fresh tissues were isolated in OR2 medium () or Grace's insect medium (). Whole mount samples were fixed in 2% paraformaldehyde and 0.1% Triton X-100 in OR2 or Grace's medium for 5–30 min, rinsed in PBS (135 mM NaCl, 2.5 mM KCl, 4.3 mM NaHPO, and 1.5 mM KHPO, pH 7.2) and stained. Squashes were prepared essentially as described previously (; ). Small pieces of fresh tissue were transferred to an 8-μl drop of medium in the middle of an 18-mm coverslip (in some cases siliconized). The coverslip was inverted over a 3- × 1-inch glass slide, and gentle pressure was applied. The slide was submerged in liquid nitrogen until bubbling ceased, the coverslip was flipped off with a razor blade, and the still-frozen preparation was placed immediately in the fixative. Tissues were fixed 5–30 min in 95% ethanol, 2% paraformaldehyde in 86% ethanol, or 2% paraformaldehyde in PBS. Slides were then washed in PBS and stained. Eggs and embryos were fixed after removal of the chorion and vitelline membrane as previously described () with the following modifications. Washes were done with PBS + 0.1% Tween 20, the chorion was removed in 50% commercial bleach for 5 min, and fixation was performed with 4% paraformaldehyde in OR2 buffer. Rabbits were injected with GST-tagged fragments of dLsm10 and dLsm11 that had been expressed in . The fragments consisted of amino acids 61–142 for Lsm10 and 1–123 for Lsm11. Crude sera from the second or third bleed were diluted 1:1,000 for immunostaining. Other primary antibodies were as follows: mAb Y12 against the Sm epitope (; provided by J. Steitz, Yale University, New Haven, CT), mAb 72B9 against mouse fibrillarin (; provided by K.M. Pollard, the Scripps Research Institute, La Jolla, CA), affinity-purified rabbit polyclonal against dSMN (; provided by J. Zhou, University of Massachusetts Medical School, Worcester, MA), and rabbit polyclonal anti-GFP (Torrey Pines BioLabs). Secondary antibodies were goat anti–mouse IgG or goat anti–rabbit IgG labeled with Alexa 488, 546, or 594 (Invitrogen). Tissue squashes were blocked with 10% horse serum for 10 min, stained with a primary antibody for 1–2 h, rinsed in PBS, and stained with a secondary antibody for 1–2 h. Whole ovaries or other tissues were blocked for 10 min or longer, stained with primary antibody overnight, washed extensively with PBS, and stained with secondary antibody for 2 h or longer. To facilitate penetration of reagents into whole tissues, 0.1% Triton X-100 was included in all solutions. Most specimens were also stained for a few minutes with 0.1–0.5 μg/ml of the DNA-specific dye DAPI for easy recognition of nuclei. Tissues for in situ hybridization were prepared essentially as described for immunostaining, with the following modifications. After a tissue sample was squashed and frozen and the coverslip was removed, it was fixed in 4% paraformaldehyde in PBS for 15 min. The slide was washed in PBS, transferred to 95% ethanol and then acetone, and dried in air. About 7 μl of probe was applied to the specimen, an 18-mm coverslip was added, and the edges were sealed with rubber cement. The preparation was incubated at 42–52°C for several hours or overnight, depending on the probe. The coverslip was removed under 2× SSC (1× SSC is 150 mM NaCl and 15 mM Na citrate, pH 7.0), and the specimen was stained with DAPI before being mounted in 50% glycerol + 1 mg/ml p-phenylenediamine. For greater permanence, some preparations were sealed with nail polish. Tissues for whole mounts were fixed in 4% paraformaldehyde in OR2 or Grace's medium for 15 min. They were then washed in excess medium and placed directly in hybridization mix for several hours or overnight at 42–52°C. Subsequent washing, staining with DAPI, and mounting were performed essentially as for squashes. Hybridization probes were diluted in the following hybridization mix: 50% formamide, 5× SSC, 10 mM citric acid, 50 μg/ml heparin, 500 μg/ml yeast tRNA, and 0.1% Tween 20. Sense and antisense RNA probes were made by in vitro transcription from DNA clones or PCR products as described previously (). Part of the U2 gene was cloned in pUC19 as an EcoRI–XbaI fragment that consisted of the T3 phage promoter, the first 53 bases of U2, and the T7 phage promoter. The fragment was generated by PCR from genomic DNA, based on the published sequence of the dU2 gene (). The dU85 clone was constructed in the same way from genomic DNA, based on the published sequence of dU85 scaRNA (). The clone contains the entire dU85 sequence. The histone H4 clone consisted of the entire 312-nt coding region, generated by PCR from clone GH10208 (). The U7 clone was prepared by H. Gao (Carnegie Institute of Washington, Baltimore, MD), based on the published sequence of dU7 (). It was constructed by annealing two partially overlapping deoxyoligonucleotides, one of which included a PvuII site, the T7 promoter, and part of the dU7 sequence, whereas the other contained an HpaI site, the T3 promoter, and the rest of the dU7 sequence. The single-stranded ends were filled in by Klenow enzyme, and the product was cloned into pUC19 at the restriction sites. The inserts of all clones were fully sequenced. Confocal images were taken with a 40× (NA 1.25) or a 63× (NA 1.40) Plan Apo objective on laser-scanning confocal microscopes (NT or SP2; Leica). Images were taken with the laser intensity and photomultiplier gain was adjusted so that pixels in the region of interest were not saturated (“glow-over” display). In most cases, contrast and relative intensities of the green (Alexa 488), red (Alexa 546 or 594), and blue (DAPI) images were adjusted with Photoshop (Adobe).

Extracellular signal-regulated kinases (ERKs) are members of the MAPK family of signaling proteins, which play a crucial role in the intracellular transmission of extracellular signals (; ). Induction of this signaling cascade leads to phosphorylation of several target proteins that eventually regulate proliferation and other cellular processes (). The ERK-induced proliferation is regulated by a multistep mechanism that involves several cell-cycle stages (), including the regulation of G0, G1, S, and M (; ; ; ; ). Thus, aside from its role in the acute transmission of extracellular signals, the ERK cascade plays a role in the regulation of other cellular processes, which are mediated via a large set of effectors (). One role of the ERK cascade is the regulation of G2/M and mitosis progression. Indeed, all components of the cascade were shown to undergo activation during the late G2 and M phases of the cell cycle (; ; ). In addition, inhibition of MEKs' activities by dominant-negative constructs or with pharmacological inhibitors delayed the progression of cells through the same stages (; ). Several molecular mechanisms have been implicated in the regulation of G2/M by the ERK cascade, including the phosphorylation of centromere protein E (), SWI–SNF (), and polo-like kinase 3 (Plk3) (), as well as the indirect activation of Plk1, Cdc2 (), and Myt1 (). However, one of the best studied mechanisms by which the cascade can influence mitosis is the regulation of Golgi fragmentation, which is the focus of this study. During mitosis, a mammalian cell needs to split its Golgi apparatus between two daughter cells. The mechanism that allows the proper division to occur is a massive fragmentation of the Golgi into thousands of vesicles that are later shared by the splitting cells (; ). This process occurs during the prophase/anaphase stages of mitosis, and is essential for the proper progression of cell division (). One of the kinases that participates in the regulation of this process is MEK1 (), which normally acts as an activator of ERK1 and ERK2 (ERK1/2; ). Interestingly, these ERKs were not found to be associated with the fragmented Golgi. In addition, it was later shown that MEK action in the Golgi can proceed even in the absence of their NH-terminal D domain, which is essential for their activity toward ERK1/2 (), indicating the presence of a different MEK substrate in the Golgi (). In recent years, several MEK1-induced, ERK1/2-independent proteins were proposed to play a role in mitotic Golgi fragmentation, including an ERK-like protein () that appears to be mono-Tyr phosphorylated before fragmentation (). The mono-Tyr phosphorylation of the ERK-like protein suggests that its mode of regulation is distinct from that of ERK1/2, which are usually found either nonphosphorylated, under resting conditions, or double Thr and Tyr phosphorylated, upon activation (). The mechanism by which the ERK-like kinase executes MEKs' signals in the Golgi is not fully understood, but may involve phosphorylation of the Golgi reassembly stacking protein of 55 kD, which serves as an ERK2 substrate in the Golgi (), or activation of Plk3, which was recently proposed to mediate MEK1 signals in the Golgi (). It should be noted, however, that the exact role of MEKs and their substrates in Golgi fragmentation is controversial, as it was shown that the phosphorylation of GM130, which is an important component of this process, mainly requires the activity of CDC2 without the involvement of MEKs (; ; ). Therefore, this study may provide new information on this controversial issue. Recently, we cloned an alternatively spliced isoform of ERK1 in human cells and named it ERK1c (). This ERK isoform is regulated differently from ERK1/2, mainly because of its altered cytosolic retention sequence/common docking motif. In addition, it was shown that it is localized in the Golgi of confluent cells and that its localization is regulated mainly by monoubiquitination (). We extended our studies on the Golgi function of ERK1c, and found that its expression, phosphorylation, activity, and Golgi localization are increased in mitosis. Knockdown experiments revealed that ERK1c attenuates Golgi fragmentation during mitosis, and as a consequence it slows down cell cycle progression. Similar to the situation in high density cell culture, the ERK1c effect could not be substituted by ERK1 or ERK2, making it a unique MEK effector in mitosis. These results suggest that ERK1c mediates the MEK-regulated Golgi fragmentation during mitosis. In addition, this unique role of ERK1c provides a molecular mechanism by which ERK cascade may execute its multiple distinct, and even opposing, effects. Golgi fragmentation in mitosis is regulated by several protein kinases, including MEKs (). Because ERK1/2, which are the known downstream targets of MEKs, are probably not involved in this Golgi fragmentation (), the mechanisms by which MEKs regulate this process are not clear. Recently, we identified ERK1c, which is an ERK1 isoform, and showed that it is localized in the Golgi of cells from high density cultures (). Therefore, we undertook an examination of whether ERK1c plays a role in mitosis-related, MEK-dependent Golgi fragmentation. To do so, we synchronized HeLa cells using the double thymidine block procedure (), which arrests cells in early S phase and allows a synchronized cell cycle progression upon block release. Indeed, most of the cells were in S phase shortly after the release, shifted to G2 within 6–9 h, and cycled back to the next G1 within 12 h (). These results were complemented by mitotic index determination, in which we found that the mitosis of the synchronized cells peaked 11 h after release ( and Fig. S1, available at ). We then examined the behavior of ERK1c, as compared with ERK1 in the synchronously cycling cells, 6 h (G2), 9 h (late G2), and 11 h (peak mitosis) after release. The expression levels of ERK1c, which were obtained by blotting with specific ERK1c antibody, were significantly increased during mitosis (). The apparent lower levels of ERK1c in S and G2 cells, as compared with the nonsynchronized cells, were probably attributable to the observation that in the latter cells the expression of ERK1c was the mean between the low amount in S phase and the much higher amount in mitosis. On the other hand, the expression levels of ERK1/2 were unchanged during the times examined, although the regulatory phosphorylation of the ERKs, which was determined using diphospho-ERK (pERK) antibody (), was elevated in late G2 and during mitosis, supporting previous works that demonstrated activation of MEKs and ERK1/2 during these stages of the cell cycle (; ). Because ERK1c migrates together with ERK2 on an SDS-PAGE (), it was not clear from the results whether the 42-kD band recognized by the pERK antibody contained any phosphorylated ERK1c or consisted only of phosphorylated ERK2, which is much more abundant. Therefore, to follow the specific activation of ERK1c and compare it to ERK1, these two proteins were immunoprecipitated with either specific ERK1c or ERK1 antibodies and subjected to an in vitro kinase assay using myelin basic protein (MBP) as a substrate. Using this protocol, nonsynchronous cells were found to contain relatively low ERK1c activity and even lower ERK1 activity (). The activity of ERK1c was lower in the S and G2 phases and significantly increased during late G2 and mitosis (sevenfold at 11 h). This rate of activation was different from that of ERK1, which constantly increased throughout the experiment, up to 12-fold higher than its activity in S phase. Because the increased activity of ERK1c could be caused by its elevated expression, we equilibrated the amount of ERK1c protein from nonsynchronized and mitotic cells to measure changes in their specific activity. Elevated levels of pERK1c were detected in the mitotic cells as compared with the nonsynchronous cells (), indicating that the elevation of ERK1c activity in mitosis is achieved by elevation in its expression, as well as in its activatory Thr and Tyr phosphorylation. To eliminate the possibility that ERK1c activation was induced by the thymidine treatments and not by cell cycle progression, we used nocodazole, which arrests cells in mitosis via a distinct mechanism (). Indeed, application of nocodazole to HeLa cells resulted in their mitotic arrest within 24 h, as judged by FACS analysis () and mitotic index determination (not depicted). The treated cells were then used to examine the expression and activation of ERK1c and ERK1 in a manner similar to that described for the double thymidine–based synchronization. Thus, similar to the double thymidine results, ERK1c expression, as well as its phosphorylation and activity, were elevated in the nocodazole- arrested cells (). On the other hand, in agreement with a previous study (), the expression of ERK1 was not affected, and the activity of this protein was only slightly elevated by the nocodazole treatment, suggesting a specific mode of ERK1c activation in these stages of mitosis. Collectively, our results clearly show that both the expression and activity of ERK1c are elevated by cell cycle progression and not by individual chemical treatments. In a previous study, we showed that ERK1c resides in the Golgi of confluent cells (); therefore, we undertook to examine whether ERK1c can be localized in the Golgi apparatus during mitosis as well. To do so, the double thymidine–synchronized mitotic HeLa cells were costained with ERK1c, together with either the general Golgi marker GM130 or the cis-Golgi marker p58 antibodies. The exact stage of mitosis was determined by DAPI staining, which showed a typical chromosomal arrangement for each phase and confirmed that the sample depicted 11 h after the release from the thymidine block contained mainly cells ranging from late G2 to telophase. The staining of these cells with the Golgi markers () revealed the expected () two-step fragmentation, first into bulbs (prophase and early prometaphase) and later into microvesicles (late prometaphase and metaphase), which were spread throughout the cell. We then followed ERK1c localization in the various stages of mitosis (). In interphase cells, ERK1c was diffusely spread throughout the cells, without any Golgi preferences. In prophase, which is when the DNA started to condense, much of the ERK1c staining was found colocalized with the partially fragmented Golgi in the perinuclear region, although some ERK1c remained spread throughout the cells. In prometaphase, when the DNA was already condensed and on its way to the equator, ERK1c was mostly colocalized with the Golgi markers in some bulbs and microvesicles that covered a significant portion of the cytosol. In metaphase, when the condensed DNA was localized in the cell equator, both ERK1c and the Golgi markers were homogenously spread throughout the cells; similar distribution was also observed with the mitotic nocodazole- arrested cells (not depicted). Interestingly, during telophase there was still interaction of ERK1c with the recomposing Golgi, as well as with the kinetochore (), but this colocalization was rapidly changed, and in new G1 cells ERK1c was again diffusely spread. The observation that ERK1c is concentrated in the Golgi in the various stages of mitosis was also confirmed by costaining with lamin, which is disassembled together with the entire nuclear envelope after prophase (Fig. S2, available at ). It should also be noted that the Golgi localization was specific to ERK1c, as the more abundant ERK1 isoform was not detected in the Golgi during prophase () or any other stage of the cell cycle. The colocalization of ERK1c with p58 or GM130 was confirmed by an electronic merge of the two stainings, showing again that the major portion of ERK1c is colocalized with the Golgi markers that are already in prophase (Fig. S3). Quantification of mitotic cells (prophase/telophase) revealed costaining of ERK1c and the Golgi markers in ∼75% of the cells as compared with only ∼23% in interphase cells (). However, it is possible that the percentage of mitotic cells with Golgi-localized ERK1c is even higher because cells in which the Golgi or the ERK1c were not properly stained were not counted as ERK1c-Golgi–colocalized cells. These results indicate that ERK1c specifically interacts with the Golgi in mitosis and not only upon an elevation in cell density. We have previously shown that, similar to ERK1/2, ERK1c can be activated by the upstream MEK kinases (). However, the different kinetics of ERK1c and ERK1 activation during the cell cycle () challenged the involvement of MEKs in ERK1c activation; therefore, it was important to examine whether ERK1c is directly activated by MEKs in the mitotic Golgi. For this purpose we first overexpressed a constitutively active (CA) form of MEK1 () or added the MEK inhibitor PD98059 to synchronized G2/M HeLa cells. Immunoprecipitation of ERK1c from the treated cells, followed by an in vitro kinase assay, revealed that ERK1c was highly activated in the mitotic cells containing CA-MEK1 and that its activation was reduced with PD98059 (). These activations correlated with the double-Thr and -Tyr phosphorylation of ERK1c, again indicating that ERK1c is a substrate for MEK1 during mitosis. In addition, ERK1c appeared to be heavily mono-Tyr phosphorylated in the mitotic cells, and this monophosphorylation was increased in the mitotic cells expressing the CA-MEK and prevented by PD98059 (). This phosphorylation was distinct from that of ERK1/2, which were mainly diphosphorylated (not depicted). Therefore, these results indicate that although both ERK1c and ERK1/2 are activated by MEKs (; ) they are subjected to different modes of regulation in mitotic cells. It was previously reported that monophosphorylated ERK–like proteins are present in the Golgi just before fragmentation (). Indeed, staining of cells in prophase or interphase confirmed the reported appearance of monophosphorylated ERK in mitosis, without an accompanied accumulation of diphosphorylated ERK in the Golgi under any conditions examined (). This staining experiment also revealed that the monophosphorylated ERK colocalized with ERK1c but not with ERK1/2, supporting the possibility that the monophosphorylated ERK in the mitotic Golgi is ERK1c. Because our results indicate that ERK1c is activated by MEKs, we examined whether, unlike the diphosphorylation of ERK1/2 (), MEK1 phosphorylates ERK1c preferentially on its regulatory Tyr residue. Therefore, we examined the possibility that ERK1c, which is not recognized by most general ERK (gERK) antibodies but can be identified by the monophosphorylated ERK antibody, is the monophosphorylated ERK in the mitotic Golgi. To do so, we immunoprecipitated MEK1-GFP from EGF-treated cells and, in parallel, we purified recombinant GST-ERK1c or GST-ERK1 from bacteria. Incubation of the active MEK1 with ERK1c in the presence of ATP and magnesium resulted in a fast accumulation of monophosphorylated ERK1c () that was much faster than the appearance of pERK1c that correlated with the low activity of the kinase. Under these conditions, the diphosphorylation of ERK1/2 was much faster than the mono-Tyr phosphorylation, which is in agreement with previously published studies (; ). These results indicate that MEKs phosphorylate ERK1c preferentially on its Tyr204 and suggest that the monophosphorylated ERK in the Golgi can be ERK1c, which is directly mono-Tyr phosphorylated by MEKs. Therefore, the presence of monophosphorylated ERK1c in the Golgi does not seem to require the activity of the Ser/Thr phosphatase PP2A, which had previously been suggested as a potential cause for the presence of monophosphorylated ERK because of its activity on the phosphorylated Thr of diphosphorylated ERK-like protein in the Golgi (). The observation that monophosphorylated ERK1c was identified only in the Golgi, although it was also localized to some extent in other parts of the cytosol, suggests that the regulation of this enzyme in the Golgi is distinct from its regulation in other cellular locations. Our results clearly indicated that ERK1c is localized and active in the Golgi during mitosis, but its exact role there was not resolved. Therefore, we undertook to study whether ERK1c is the enzyme that participates in the regulation of mitotic Golgi fragmentation downstream of MEKs. To this end, we altered ERK1c (and ERK1 as control) expression in the mitotic cells by either overexpressing the GFP-conjugated wild-type constructs (68 and 70 kD of GFP-ERK1c and GFP-ERK1, respectively; ) or by knocking down the expression of the endogenous proteins using specific small interfering RNA (siRNA). To confirm a specific reduction of either ERK1c or ERK1, the interfering sequences were directed toward the unique COOH-terminal sequences of the ERKs. Indeed, expression of the ERK1c-interfering sequence (siRNA of ERK1c [si-ERK1c]) in HeLa cells, resulted in a reduction of 75 ± 7% in the expression of endogenous ERK1c, with no effect on the expression of ERK1/2 (). Similarly, expression of the ERK-interfering sequence (si-ERK1) specifically reduced the expression of ERK1 (80 ± 10%) without any effect on ERK1c or ERK2. The reduction in ERK1c expression by its interfering sequence was also demonstrated by immunostaining with ERK1c antibody (), which showed a marked reduction in the amount of ERK1c with no significant change of its localization in the si-ERK1c–expressing cells ( and Fig. S4, available at ). It should also be noted that this correlation between the Western blot and the immunostaining confirmed the specificity of ERK1c staining by the antibody used in this study. As expected, activity assays showed that overall ERK1c activity in mitotic cells overexpressing GFP-ERK1c was significantly higher than the activity derived from vector control cells, whereas the activity derived from the cells transfected with si-ERK1c was markedly lower (). Therefore, this system was judged suitable for the study of downstream effects that are mediated either by phosphorylation or by changes in expression and localization of ERK1c. To determine whether ERK1c is indeed involved in the regulation of Golgi fragmentation, we used both GM130 and p58 markers to follow Golgi architecture in cells expressing GFP, GFP-ERK1c, and si-ERK1c. In these experiments, the ERK1c constructs clearly affected Golgi fragmentation during interphase and the different stages of mitosis (). Thus, in interphase, overexpression of ERK1c increased Golgi fragmentation as previously reported (), whereas the si-ERK1c had no significant affect (). Acceleration of Golgi fragmentation by cells overexpressing ERK1c was also observed in prophase and to some extent in prometaphase, when the Golgi appeared much more broken than in cells transfected with GFP control. In these stages of mitosis, si-ERK1c expression significantly attenuated fragmentation, as only few Golgi fragments could be detected in most of the cells (quantified in , right bottom). We next followed the Golgi architecture in metaphase, when the Golgi markers appeared to be spread throughout the GFP control cells (), indicating that the Golgi was completely fragmented, as expected (). Interestingly, the overexpression of GFP-ERK1c did not change the distribution of the Golgi markers. However, in 55% of the siRNA-expressing cells the Golgi was not completely spread, but was rather broken into bulbs that were still concentrated in the cytoplasm, similar to its appearance in control prophase cells. Thus, si-ERK1c significantly inhibited Golgi fragmentation in the prophase/metaphase stages of the cell cycle. In telophase (), the Golgi was already rebuilt and equally divided between the daughter cells in the GFP- and ERK1c-expressing cells, whereas in the si-ERK1c cells the division of Golgi between the daughter cells was not equal. In these cultures, 60% of cell pairs had one cell with a bigger Golgi, whereas the Golgi of the mate cell was much smaller. These results could be a consequence of the incomplete fragmentation of the Golgi at metaphase that allowed a distribution of some unprocessed Golgi bulbs. These results may also suggest that the pace of mitotic progression is not fully synchronized with or controlled by Golgi fragmentation, as previously suggested (). As expected from previous studies involving overexpression of ERK1 () and the absence of ERK1 from mitotic Golgi (), ERK1 had no significant effect on Golgi architecture under any of the conditions used (Fig. S5, available at ; and not depicted), again supporting the specific role of ERK1c in the regulation of Golgi fragmentation. Golgi fragmentation was found to play a critical role in the progression of mitosis (). Therefore, it was important to examine whether ERK1c regulates G2/M progression via its involvement in Golgi breakdown. To this end, HeLa cells overexpressing GFP only, GFP-ERK1c, GFP-ERK1, si-ERK1c, or si-ERK1 were synchronized using the double thymidine block, and examined by FACS for their cell cycle stage. Thus, 9 h after block release the various transfected cells were found in the G2 phase of the cell cycle, indicating that neither ERK1c nor ERK1 play a role in S/G2 progression of HeLa cells (). On the other hand, 11 h after release the GFP-ERK1c, but not GFP-ERK1, had a marked influence on the mitosis progression (). As expected, the control GFP-expressing cells were reproducibly found equally in their 2N or 4N forms, indicating that most cells are in the midst of mitosis at this stage. Overexpression of ERK1c shifted this distribution toward G1 (28% in G2 and 66% in G1; quantification in ), whereas the si-ERK1c inhibited the transition (72% in G2 and 22% in G1). These effects were restricted to 10–12 h after release from the double thymidine block because when these experiments were performed 13 h after the release the cells were all found in the next G1 phase with no significant effect of any of the ERK constructs. These results indicate that the effect of ERK1c is restricted to the initial stages of mitosis and is not mimicked by ERK1. Interestingly, the effects of ERK1c during mitosis were similar to those obtained by either inhibiting or activating the upstream kinase MEK1 (, 11 h). However, unlike ERK1c, the modulation of MEKs activity still maintained a minor, but significant, effect 13 h after release. Thus, PD98059 retained 31% of the cells in G2 as compared with 12% observed in GFP control, and the CA-MEK1 reduced the percentage of cells in G2 to 7%. The longer effect of the MEKs inhibitor, as compared with si-ERK1c, may suggest that MEKs operate in part via Golgi-independent mechanisms to regulate the progression of mitosis (). To further establish the mitotic role of ERK1c and MEKs, we monitored the percentage of mitotic cells 11 h after the double thymidine release, as well as the percentage of cells in telophase, out of the total mitotic cells. The percentage of mitotic cells was not affected by si-ERK1c expression (, left), but the percentage of cells in telophase was significantly inhibited (, right), indicating again that ERK1c affects cell cycle at early stages of mitosis, without a significant effect on the G2 phase and the entrance to mitosis. The overexpression of ERK1c or CA-MEKs reduced the percentage of cells in mitosis but increased the portion of cells in telophase. This could be explained by the accelerated rate of mitotic progression in these cells, which pushed most of the cells to either telophase or to the G1 phase of the following cycle. Addition of PD98059 9 h after release from the double thymidine arrest inhibited the entrance and, even more so, the progression of cells in mitosis, indicating that unlike the si-ERK1c, it had some role at the G2/M phase, probably by its additional effect on ERK1/2. This conclusion is supported by the stronger effect of PD98059, when it was combined with the release from the double thymidine arrest (not depicted). Finally, the ERK1 constructs had no effects at these stages. These results further corroborate the specific role of ERK1c in the progression of mitosis, which is probably attributable to its effects on Golgi fragmentation at this stage. In this study, we show that ERK1c is an important regulator of Golgi fragmentation during mitosis and thereby plays a role in controlling mitotic progression in human cells. Mitotic Golgi fragmentation is a two-step process in which the pericentriolar Golgi stacks are converted into bulbs and then either undergo further vesicular spreading or fuse with the ER (). The first part of this process seems to be regulated by phosphorylation, which is mediated by several protein kinases, including MEK1 (). The role of MEK1 in Golgi fragmentation was clearly demonstrated in various cell types. However, the mechanisms by which MEK1 functions during this process was not clear because the known MEK1 substrates (ERK1/2) were not found to participate in the process (). Rather, it was suggested that mitotic MEKs may function via an ERK-like protein (), which may be mono-Tyr phosphorylated (), or via other protein kinases such as Plk3 (). In this study, we present data suggesting that ERK1c, but not ERK1/2, might be the mediator of the MEK-induced Golgi fragmentation during mitosis. This is strongly supported by the following observations: (a) the expression levels of ERK1c, but not ERK1/2, are elevated in late G2 and mitosis (); (b) the activity of ERK1c is elevated in late G2 and during mitosis (); (c) ERK1c, and not the other ERKs, is localized in the Golgi during the early stages of mitosis (); (d) si-ERK1c, but not of ERK1, reduces Golgi fragmentation during mitosis (), which attenuates mitotic progression (); and (e) ERK1c is preferentially phosphorylated on its Tyr204, which correlates with the appearance of pY-ERK in the mitotic Golgi (). This is unlike ERK1/2, which under most conditions are phosphorylated on both their activatory Thr and Tyr residues at a comparable rate (; ; ). This observation, together with the mitotic Golgi distribution of ERK1c, strongly suggests that ERK1c is the monophosphorylated ERK that was suggested to play a role in Golgi fragmentation during mitosis (). It has previously been demonstrated that Golgi fragmentation is a crucial step in the progression of mitosis, and it was suggested that this fragmentation, in fact, provides a sensor for controlling entry into mitosis in mammalian cells (). Our results demonstrate that, indeed, inhibition of Golgi fragmentation achieved with the si-ERK1c caused a delay in the onset of mitosis. However, the block was never full and could not be detected 13 h after the double thymidine release. The incomplete arrest of the HeLa cells in mitosis could be caused by an incomplete inhibition of ERK1c activity because of the partial effect of the siRNA. However, it is also possible that Golgi fragmentation is not the only sensor for mitotic progression. This suggestion is supported by the observation that in cells expressing si-ERK1c, the distribution of Golgi fragments to daughter cells during telophase was not equal (). Therefore, it is possible that the progression through mitosis is governed by a multistep mechanism, which requires Golgi fragmentation at only one point during mitosis. Thus, partial Golgi breakdown (such as the one observed in si-ERK1c prometaphase in ) is probably a sufficient signal to allow the mitotic progression. Because some of the Golgi membranes were still in bulb form and not in spread vesicle form, and the cells proceeded through telophase without waiting for the Golgi to complete its breakdown, an unequal distribution of particles resulted, and thus, an unequal Golgi size in the daughter cells. Interestingly, the cells overexpressing GFP-ERK1c contained similar amount of Golgi fragments in the daughter cells, although the progression through early mitosis was accelerated in these cells. This indicates that once the Golgi is fully degraded it can be homogenously divided into the daughter cells at the right stage of mitosis. Therefore, our results may indicate that Golgi fragmentation is an important step in the initiation of mitosis, but that once the set of events that are required for this progression have started, the synchronization of the process is probably regulated by other mechanisms. ERK1c is an alternatively spliced form of ERK1, which is altered in the cytosolic retention sequence/common docking regulatory region; therefore, many of its regulatory aspects are distinct from those of ERK1/2. Indeed, in a previous study we showed that the kinetics of activation of this protein is distinct from its relative ERKs upon EGF or NaCl stimulation (). Furthermore, the subcellular localization of ERK1c seems to be distinct from that of ERK1/2, as it was found mainly in the Golgi under various conditions (; ). In our previous work, we suggested that the Golgi translocation of ERK1c is regulated by its monoubiquitination. In mitosis, this is probably not the situation, as we failed to obtain ERK1c at any other molecular weight than the nonmodified endogenous 42-kD protein (). Therefore, the mechanism that allows ERK1c to specifically associate with the Golgi has yet to be clearly identified. Additional information on ERK1c and ERK1 during mitosis was observed using nocodazole, which arrests the HeLa cells in a prometaphase-like stage. Similar to cells synchronized by double thymidine block, nocodazole induced higher expression () and Golgi localization (not depicted) of ERK1c, indicating that these effects are indeed mediated by cell cycle arrest and not by the individual effect of the drugs. However, despite the pronounced activation of ERK1c after nocodazole treatment (), which was higher than that observed upon double thymidine treatment (), the activation of ERK1 was much smaller (). This result is in agreement with a previous work, which showed that nocodazole does not induce ERK1 activation (). The reason for the difference is not clear and could be the outcome of some nonspecific effects of the cell cycle–arresting drugs. However, this difference could also be attributable to the homogenous arrest of cells at the prometaphase-like stage upon nocodazole treatment, unlike the wider cell cycle distribution in cells 11 h after release from double thymidine block (G2/anaphase). This may indicate that the activation of ERK1/2 occurs mainly during the G2–prophase stage and that this activity is reduced at prometaphase, whereas ERK1c is activated somewhat later than ERK1/2; its activity is maximal at prometaphase and may proceed even later in mitosis. Importantly, the results of this study strongly suggest that ERK1c has unique functions that are not shared by ERK1/2 in regulating Golgi fragmentation. The functional difference between the isoforms was also manifested in the stronger block in cell cycle progression by the MEK inhibitor PD98059, as compared with that of the si-ERK1c (). Because both si-ERK1c and PD98059 reduced ERK1c activity to a similar level ( and ), it is likely that MEKs have some ERK1c-unrelated effects on the progression of mitosis that are probably mediated by ERK1/2. This is supported by the fact that ERK1/2 are activated in the onset of mitosis as well (). In addition, ERK1/2 are not localized in the Golgi at any stage of the cell cycle, suggesting that they participate in other, Golgi-independent signaling events. The observation that reductions in ERK1 levels and activity did not exert any significant effect on the progression through mitosis () may be explained by the ability of its close homologue ERK2 to compensate for its shortage. On the other hand, our results clearly indicate that ERK1c cannot overcome the effects of ERK1/2 and vice versa, which can be explained by their distinct localization and differential regulation (). Thus, we propose a pathway in which activated MEKs colocalize with ERK1c in the Golgi, activate it, and thereby induces the process of Golgi remodeling during mitosis. At the same time, other MEK molecules phosphorylate ERK1/2, and those molecules participate in the regulation of distinct mitotic processes. Therefore, MEKs seem to orchestrate a large part of the mitotic events by activating three ERK isoforms that may each function in a different location and regulate distinct processes. In summary, we characterized the function of ERK1c in the Golgi and found that it plays an important role in the regulation of Golgi fragmentation during mitosis and thereby in the regulation of cell cycle progression. The ERK1c effects could not be complemented by ERK1/2, indicating that this isoform is the unique MEK effector in the mitotic Golgi. These results shed light on the MEK-dependent regulation of Golgi fragmentation and division into the daughter cells during mitosis. In addition, this unique role of ERK1c extends the substrate specificity of the signaling by the ERK pathway and may suggest a molecular mechanism for the ability of the ERK cascade to regulate distinct, and even opposing, cellular processes. Thymidine, nocodazole, MBP, propidium iodide, EGF, and ATP were purchased from Sigma-Aldrich. GM130, lamin A/C antibodies, and protein A/G PLUS–agarose beads were obtained from Santa Cruz Biotechnology, Inc. Sepharose-immobilized protein A, an ECL kit, glutathione beads, and γ[P]ATP were purchased from GE Healthcare. DAPI was purchased from Invitrogen, and PD98059 was obtained from Calbiochem. p58, pERK, monophosphorylated ERK, gERK, and ERK1 antibodies were obtained from Sigma-Aldrich. The ERK1c antibody was prepared by the Antibody Unit of the Weizmann Institute of Science, as previously described (). GFP antibody was purchased from Roche. The developing substrate NBT/BCIP was obtained from Promega. FITC-, rhodamine-, alkaline phosphatase–, and horseradish peroxidase–conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories. Buffer A consists of 50 mM β-glycerophosphate, pH 7.3, 1.5 mM EGTA, 1 mM EDTA, 1 mM DTT, and 0.1 mM sodium vanadate. Buffer H consists of Buffer A supplemented with 1 mM benzamidine, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 2 μg/ml pepstatin A. Buffer RM (reaction mixture at threefold concentration) consists of 30 mM MgCl, 4.5 mM DTT, 75 mM β-glycerophosphate, pH 7.3, 0.15 mM sodium vanadate, 3.75 mM EGTA, 30 μM calmidazolium, and 2.5 mg/ml bovine serum albumin. Radioimmunoprotein assay buffer consists of 137 mM NaCl, 20 mM Tris, pH 7.4, 10% (vol/vol) glycerol, 1% Triton X-100, 0.5% (wt/vol) deoxycholate, 0.1% (wt/vol) SDS, 2.0 mM EDTA, 1.0 mM PMSF, and 20 μM leupeptin. Buffer HNTG consists of 50 mM Hepes, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, and 10% glycerol. GFP-ERK and GFP-ERK1c were prepared in plasmid of EGFP (pEGFP) vector (). CA-MEK1 (DN-EE-MEK1) was prepared in phosphorylated cDNA1 (). GFP in pEGFP was purchased from CLONTECH Laboratories, Inc. GST-ERK1c was prepared by inserting ERK1c into the BamHI and EcoRI sites of pGEX-2T. The GST protein was purified according to the manufacturer's instructions and eluted from the glutathione beads using 10 mM of reduced glutathione. To generate the pSUPER-ERK1c and pSUPER-ERK1 we used the pSUPER plasmid (). The sequence that was used (CGACGGATGAGGTGGGCCA; 1,007–1,025 bp) was derived from the unique sequence of human ERK1c. For human ERK1, the sequence that was used (AGCTGGATGACCTACCTAA; 1,052–1,070 bp) was derived from the COOH terminus of human ERK1 that is not present in ERK1c. HeLa cells were grown in DME supplemented with 10% FCS (Invitrogen). Transfection into the cells was performed using the polyethylenimine method, as previously described (). In brief, the cells were grown to 50–70% confluence in 12-well plates. The 1.5-μg plasmid was suspended in 125 μl PBS and mixed with polyethylenimine solution (5 μl of 3 mM polyethylenimine in 125 mM NaCl). The mixture was left at 23°C for 15 min and then incubated with the cells for 90 min, after which the cells were washed and placed in DME + 10% FCS. Cells were fixed (20 min in 3% paraformaldehyde in PBS or 10 min in methanol at −20°C), followed by a 20 min permeabilization/blocking solution (0.1% Triton X-100/2% BSA). Antibodies of choice were added for 1 h, washed, and developed with fluorescence-tagged secondary antibodies (Cy2, FITC, or rhodamine; Jackson ImmunoResearch Laboratories) for 1 h at 23°C. Nuclei were stained with 0.1 mg/ml DAPI (Invitrogen) in PBS. The slides were visualized using a fluorescence microscope (Optiphot; Nikon) with 100×, 1.3 NA, or 40×, 0.7 NA, immersion oil objectives (Nikon). Digital images of cells were captured using a camera (DVC) and C-view V2.1 Imaging software (DVC). Images were processed using Photoshop 7.0 (Adobe). After treatment, the cells were rinsed twice with ice-cold PBS and once with ice-cold buffer A. Cells were scraped into buffer H (0.5 ml per plate) and disrupted by sonication (two 50-W pulses for 7 s). The extracts were centrifuged (20,000 for 15 min at 4°C) and the supernatants were kept at 4°C. The supernatants were then separated by a 10% SDS-PAGE, transferred onto a nitrocellulose membrane, and probed with the appropriate antibodies, as previously described (). Cells extracts prepared as described in Cell culture and transfection were incubated (2 h at 4°C) with ERK1c (affinity purified), ERK1, or GFP antibodies coupled to protein A/G–Sepharose. For determination of ERK1c activity, the beads were washed once with HNTG buffer, twice with 0.5 M LiCl in 0.1 M Tris, pH 8.0, and once with 1 ml of buffer A. Immunoprecipitates were subjected either to Western blotting or subjected to an in vitro kinase assay. Immunoprecipitated ERKs attached to 15-μl beads were used as kinases by mixing them with MBP (8 μg/reaction). Similarly immunoprecipitated active MEKs attached to 15-μl beads were mixed with either recombinant GST-ERK1c or GST-ERK1 (0.5 μg per reaction). Buffer RM containing 100 μM γ[P]ATP (4,000 cpm/pmol) was added to the reaction at a final volume of 30 μl and incubated for 20 min at 30°C. The reaction was terminated by adding 10 μl of 4× sample buffer, and the phosphorylated proteins were resolved on SDS-PAGE and subjected to autoradiography and Western blot analysis with the proper antibodies. Cells were synchronized at the G1/S boundary by the double thymidine block (). In short, HeLa cells were treated with 2 mM thymidine in DMSO, washed twice with PBS, grown for 8 h in regular medium, and then treated again with 2 mM thymidine for 16 h and washed with PBS. This marks time 0, after which the cells were grown under the regular conditions for the indicated times (). HeLa cells were also synchronized at the M phase using 100 ng/ml nocodazole for 24 h (). For FACS analysis, HeLa cells were trypsinized, washed with PBS, and fixed in 70% ice-cold methanol for at least 1 h. The samples were then centrifuged (500 for 2 min) and resuspended in 0.5 ml of staining solution (0.001% Triton X-100, 0.1 mM EDTA, 100 μg/ml RNase, and 50 μg/ml propidium iodide in PBS). The cells were analyzed by FACsort (Becton Dickinson), and the percentage of cells at different stages was calculated using the CellQuest software (BD Biosciences). Fig. S1 shows the mitotic index of the synchronized cells. Fig. S2 shows that ERK1c translocation to the Golgi correlates with nuclear envelope breakdown. Fig. S3 shows the subcellular localization of ERK1c during mitosis (color images). Fig. S4 shows si-ERK1c in transfected cells. Fig. S5 shows that ERK1 does not affect the Golgi architecture. Online supplemental material is available at .

Asexual stages of the malaria parasite invade and replicate in human RBCs. During the 48 h of its asexual life cycle, the parasite dramatically remodels the host RBC (for reviews see , ). This includes the generation of unique, flattened membranous structures in the cytoplasm of the infected RBC (IRBC) called Maurer's clefts and protrusions on the IRBC surface termed knobs. also exports several proteins into the host RBC by a mechanism that depends on the presence of an NH-terminal motif, termed either the export element (PEXEL) or vacuolar transport signal (VTS), which is conserved across the genus (; ). One of the most important of these exported proteins is PfEMP1 ( erythrocyte membrane protein 1), a protein of parasite origin that is exposed on the surface of IRBCs and mediates adhesion to several receptors expressed on the surface of vascular endothelial cells (). Ultimately, this results in the adhesion of mature parasite-infected RBCs to the vascular endothelium, which is a process that underpins the development of severe and often fatal complications that accompany malaria infection in humans (; ). Therefore, understanding the pathways and molecular mechanisms by which parasites export proteins to the RBC membrane is critical for a complete understanding of the pathogenesis of malaria and, in the longer term, could lead to the development of novel therapeutic approaches to prevent cytoadherence and severe pathological sequelae. For most parasite proteins that are destined for the RBC membrane skeleton and the surface, trafficking seems to involve Maurer's clefts (; ; ). Clefts appear at the early trophozoite stage of parasite development and persist throughout the remainder of the intra-erythrocytic cycle. The final destination of Maurer's clefts in IRBCs does appear to be in juxtaposition with the RBC membrane skeleton, and there is some evidence to suggest that they may interact with actin, ankyrin, or other proteins associated with the RBC membrane skeleton such as LANCL1 (, ). It is possible that Maurer's clefts dock with the RBC membrane skeleton and provide the exit site for proteins, such as PfEMP1, that are exposed on the surface of the IRBC; however, such a mechanism has not yet been demonstrated. Several proteins have been described that are resident within Maurer's clefts, but there is no information about their functional roles and whether they may be involved in protein trafficking (; ). One of the first Maurer's cleft proteins to be described was SBP1 (skeleton-binding protein 1), a 48-kD integral membrane protein that spans the Maurer's cleft membrane (). The NH-terminal domain is found within the cleft, whereas the COOH-terminal domain is exposed within the IRBC cytoplasm and interacts with a RBC membrane skeleton protein, possibly participating in anchoring the clefts to the RBC membrane skeleton (). To determine the function of SBP1 in IRBCs, we have generated clonal transgenic parasite lines in which SBP1 is not expressed and have extensively examined the biological properties of these mutant IRBCs. Analysis of the SBP1-deleted parasite line revealed that the major virulence factor PfEMP1 is not expressed on the surface of the IRBC and that the wild-type phenotype can be restored when the gene deletion is complemented. Importantly, the SBP1-deleted parasite line (SBP1 knockout [KO]) represents the first parasite line showing a knock-down in PfEMP1 surface expression. (PlasmoDB accession no. PFE0065w) is a 1.2-kb gene comprising two exons separated by a 170-bp intron and is located in the subtelomeric region of chromosome 5 (). To determine the function of SBP1, we disrupted the gene in 3D7 parasites (). Integration of the drug resistance cassette into 1 by homologous recombination would disrupt the coding region at residue 192. Eight clonal lines were selected from two independent transfection events. Of the eight selected clones, three were selected for further analysis (1G8, 2D8, and 1G5). Analysis of intact chromosomes derived from these transgenic clonal lines, which were separated by pulsed field gel electrophoresis as previously described (), confirmed that integration into chromosome 5 had occurred (not depicted). Analysis by Southern blotting using genomic DNA digested with ClaI and EcoRI and probed with either h (human dihydrofolate reductase) or F2 confirmed that the gene had been disrupted in all three parasite clones (). No episomal plasmid could be detected in any of the clones by PCR analysis. Restriction fragments of the predicted sizes were also obtained on similar Southern blots using DNA digested with alternative endonucleases EcoRI–XhoI or NsiI (not depicted). As expected, the analysis of three selected parasite clones by Western blotting using polyclonal antibodies raised against either NH- or COOH-terminal regions of SBP1 () revealed that SBP1 expression had been ablated in all three transgenic clones (1G5, 1G8, and 2D8; ). There was no reactivity of any of the bands with antibodies against the NH-terminal region of SBP1 with any of the transgenic clones, which is indicative of the loss of translation of full-length SBP1. One of the three clones (1G8) was selected for more detailed analysis. The examination of IRBCs infected with 1G8 parasites by immunofluorescence using mouse polyclonal antisera specific for SBP1 failed to show any reactivity, suggesting that there were no truncated SBP1 products and that SBP1 expression had been completely ablated (). Because SBP1 is known to be a major structural component of Maurer's clefts, we wanted to assess Maurer's cleft structure and function in the SBP1 KO line compared with the 3D7 wild-type parasites. We first analyzed the localization of several intrinsic Maurer's cleft markers (; ; ) in the two parasite lines. As shown in , the deletion of SBP1 from 3D7 parasites had no apparent affect on the expression or internal cellular localization of two resident Maurer's cleft proteins, Pf332 and REX (ring-stage exported protein). In addition, we analyzed proteins that are transiently associated with Maurer's clefts. Both PfEMP1 and KAHRP (knob-associated histidine-rich protein [HRP]), the major structural component of knobs, are still present in the Maurer's clefts of the SBP1 KO line, and, within the limits of discrimination of immunofluorescence, there appears to be no gross change in the amount of either protein within Maurer's clefts (). Similarly, PfEMP3, a protein transported to the IRBC membrane skeleton, is still able to reach its customary location (not depicted). Treating IRBCs with the pore-forming agent streptolysin O, which increases antibody accessibility to Maurer's clefts, revealed similar levels of PfEMP1 colocalization with another resident Maurer's cleft marker, MAHRP (Maurer's cleft–associated HRP; ) in both parental and SBP1 KO lines (). Examination of transmission electron micrographs of RBCs infected with either 3D7 or 1G8 parasites () revealed the presence of identifiable Maurer's clefts in both types of IRBCs. Precise quantitation revealed no significant difference in the number of Maurer's clefts in 1G8 parasites when compared with 3D7 (1.28 ± 0.16 vs. 1.30 ± 0.16 clefts/μm; mean ± SEM for 1G8 and 3D7 parasites, respectively). In both parasite lines, IRBCs with more than one parasite tended to have proportionately more clefts. In a total of 64 random IRBCs examined, only one 3D7 IRBC and one 1G8 IRBC had no detectable clefts at all. When examined at higher magnification, however, there were subtle detectable differences in the fine morphological structure of the clefts, which were typically narrower in RBCs infected with 1G8 parasites when compared with RBCs infected with 3D7 (). We also examined Maurer's cleft location in the parental 3D7 and KO 1G8 lines, measuring the minimum distances of the cleft from the nearest IRBC surface. In both sets of samples (number of clefts examined = 37 for each), most clefts were approximately parallel to the surface, and, in both, about one fifth of clefts were angled so as to approach the surface closely. The minimum distance from the cell surface to the Maurer's cleft in 3D7 IRBCs was 139 ± 12 nm (mean ± SEM), whereas in the SBP1 KO line, Maurer's clefts were on average slightly further from the IRBC surface (160 ± 15 nm). By scanning electron microscopy, the external appearances of 3D7 or 1G8 IRBCs were similar, with irregular surfaces bearing numerous knobs (). The knobs on both 3D7 and 1G8 IRBCs also looked identical by transmission electron microscopy (; black arrows). Precise quantitation revealed that there were no significant differences between the number of knobs on 1G8 IRBCs when compared with 3D7 IRBCs (3.43 ± 0.31 vs. 3.45 ± 0.31 knobs/μm; mean ± SEM for 1G8 and 3D7, respectively). In both parasite lines, IRBCs with more than one parasite tended to have proportionately more knobs. In the total of 64 random IRBCs examined, only one 3D7 IRBC and one 1G8 IRBC had no detectable knobs at all. To examine the fate of PfEMP1 in detail, we performed various immunochemical and proteolytic assays on surface-exposed PfEMP1 in both parental and mutant parasite lines. It has previously been shown that PfEMP1 changes its solubility during transport to the IRBC surface. Specifically, the protein is soluble in Triton X-100 during transport in the IRBC but becomes detergent insoluble/SDS soluble when incorporated into the knob structure on the IRBC surface (). In addition, the large NH-terminal PfEMP1 exodomain is sensitive to trypsin when the protein is surface exposed (). We analyzed the distribution of PfEMP1 in both wild-type and SBP1 KO parasites by Western blot analysis of the Triton X-100–insoluble/SDS-soluble fraction of membrane proteins with and without trypsin treatment of intact IRBCs (). Thus, in the parental 3D7 line, PfEMP1 was present in the Triton X-100–insoluble/SDS-soluble fraction and was cleaved after exposure to trypsin. In contrast, PfEMP1 was not present in the Triton X-100–insoluble/SDS-soluble fraction in the SBP1 KO line, and the protein was not sensitive to trypsin treatment (). To control for trypsin digestion, we examined the fate of an integral membrane and surface-exposed protein of human RBCs, glycophorin C, in both parental and mutant parasite lines. In both cases, glycophorin C was cleaved by the exogenous trypsin, demonstrating that PfEMP1 in the SBP1 KO line was not present on the IRBC surface but remained internal within the Maurer's clefts as shown by immunofluorescence assay (IFA; ). The relative differences in the amount of PfEMP1 in this Triton X-100–insoluble fraction can be more readily visualized after the extensive preabsorption of anti-PfEMP1 antiserum against inside-out RBCs to reduce cross-reactivity of the antiserum with RBC spectrin (). There is almost no detectable product in the KO parasite line, suggesting that there is no intermediate compartment between residence in the Maurer's cleft and exposure on the surface. It also suggests that PfEMP1 has not exited the cleft but becomes trapped in association with the membrane skeleton en route to the IRBC surface. Together with our IFA experiments (), these data are consistent with a block in the surface expression of PfEMP1 at the level of the Maurer's clefts in the SBP1 KO parasite line. Cytoadherence of IRBCs is mediated by the differential binding properties of variant PfEMP1 exodomains to several host cell receptors. The absence of PfEMP1 expression on the surface of IRBCs should, therefore, result in reduced cytoadherence in the 1G8 parasite line compared with wild-type 3D7 parasites. Because 3D7 parasites bind only to the endothelial cell surface–expressed receptor CD36, we quantified the adhesion of IRBCs to this receptor using both static and flow-based adhesion assays. In static assays, the ability of RBCs infected with 1G8 parasites to adhere was significantly reduced by >96% when compared with the level of adhesion for 3D7 IRBCs (mean reduction in adhesion = 96.7 ± 1.3%; mean ± SEM for five experiments; P < 0.01 by Mann-Whitney U test). Similarly, when tested under flow conditions (at 0.1 Pa), all three transgenic -deleted parasite clones showed significantly reduced adhesion to CD36 (). The low level of binding was similar for all three transgenic clones, averaging only 12% of the level for 3D7 (P < 0.01 by Kruskall Wallace test). We reintroduced a full-length gene from 3D7 parasites under control of the constitutive 86 (heat shock protein 86) promotor into 1G8 parasites by transfection of the modified Gateway expression vector pHrB1-1/2 containing the 3D7 gene that had been amplified from cDNA and appended to the reporter protein EYFP. Examination of drug-resistant parasites by Western analysis ∼80 d after transfection revealed that the resulting parasites, 1G8/, were clearly producing SBP1, although at a somewhat lower level than seen in wild-type parasites (). When live IRBCs were viewed under 513-nm blue light, Maurer's clefts were visualized as fluorescing structures throughout the cytoplasm of the IRBC, with a proportion of clefts in close association with the RBC membrane skeleton (). Analysis of these 1G8/ parasites by immunofluorescence and confocal microscopy revealed that the SBP1-EYFP chimeric protein was trafficked correctly to Maurer's clefts that were located at the RBC membrane skeleton ( and ). The SBP1-EYFP protein also colocalized with REX () and other Maurer's cleft markers (Pf332 and KAHRP; not depicted), demonstrating that the SBP1-EYFP chimeric protein was being trafficked to its native cellular location. The effect of complemented SBP1 on PfEMP1 was examined in two independent cloned lines derived from 1G8, 1G8/, and 1G8/. Reintroduction of SBP1 led to the reappearance of PfEMP1 on the surface of the IRBC and the restoration of cytoadherence to CD36. Surface expression could be demonstrated by the reappearance of PfEMP1 in the Triton X-100–insoluble/SDS-soluble fraction, demonstrating an association with the membrane skeleton (). Importantly, PfEMP1 was once again sensitive to trypsin digestion, confirming an exposed location on the IRBC surface. Of interest is that the two complemented lines derived from 1G8 differ in the variant of PfEMP1 they express on the IRBC surface. 1G8/ expresses a variant of PfEMP1 that is unchanged in size from 3D7, whereas 1G8/ has undergone a gene switch to express a different variant of PfEMP1. Importantly, this demonstrates that SBP1 has a general role in PfEMP1 trafficking and is not specific to a particular PfEMP1 variant. Reintroduction of the full-length gene into 1G8 parasites also dramatically restored the ability of both complemented lines to adhere to CD36 under both static and flow conditions. For example, in static assays, the reintroduction of into 1G8 parasites restored adhesion such that 1G8/ IRBCs bound at levels ∼60-fold higher than 1G8 IRBCs (76.1 ± 16.5% vs. 1.3 ± 1.3%; mean ± SEM for 1G8/ and 1G8 parasite clones, respectively). Similarly, when tested under flow, adhesion of 1G8/ IRBCs showed a level of binding similar to that measured for 3D7 IRBCs (). Interestingly, the 1G8/ complemented line showed a consistently higher level of adhesion in both static and flow assays when compared with the wild-type 3D7 (not depicted). The amount of surface-exposed PfEMP1 that exerts this adhesive effect would appear to be quite small, as there is no detectable difference in appearance by IFA of either 1G8 or the complemented parasite lines (). Trafficking of PfE MP1 in –IRBCs has been examined by several groups recently and has been shown convincingly to occur in association with Maurer's clefts (; ; ; ; ; ). Exit from the parasite into the RBC cytoplasm is under the direction of a protein export motif known as the PEXEL or VTS (; ), and initial transport within the IRBC cytoplasm appears to be as soluble protein complexes rather than via vesicles (; ; ). PfEMP1 then appears to be inserted into the Maurer's cleft membrane, although is not clear whether the protein is transported within the cleft itself or attached to the cytoplasmic face of the cleft, as the resolution of light microscopic studies is insufficient to answer this question. Once Maurer's clefts reach the RBC membrane skeleton, PfEMP1 becomes anchored at the knob structures, and its extracellular binding domains become exposed on the RBC surface. Although it is clear that the export of PfEMP1 from the parasite to the IRBC cytoplasm is dependent on a functional PEXEL/VTS motif, it is still unknown precisely how PfEMP1 accomplishes the final translocation across the RBC membrane and what, if any, parasite or host molecules are involved in this process. It is likely that such molecules would either be present at the RBC surface, probably within the knob structure, or would be found in association with Maurer's clefts. SBP1 has previously been shown to be a membrane protein resident in Maurer's clefts that is oriented so that its COOH-terminal domain is present in the RBC cytoplasm (). Its function was previously unknown, although it has been suggested that it might be involved in binding the cleft to the RBC membrane skeleton through a specific interaction with proteins of the RBC membrane skeleton such as actin () or RBC membrane–associated proteins such as LANCL1 (). SBP1 is clearly not an essential structural protein of the cleft, as its absence does not affect cleft morphology or number markedly. The subtle changes in cleft morphology related to the deletion of SBP1 noted in our studies here may merely represent the loss of SBP1 itself or the loss of a group of associated proteins that are anchored to the cleft by interaction with SBP1. We also conclude that it is not a required part of the general trafficking mechanism for parasite proteins exported to the RBC as Maurer's cleft–associated proteins because PfEMP3, Pf332, KAHRP, REX, and MAHRP are trafficked normally to the membrane skeleton of RBCs infected with -deleted parasites. Furthermore, the cytoadherence-related knob complexes on the surface of the IRBC form normally in -deleted parasite clones. In fact, SBP1 appears to have a precise function, which is to mediate the final step in the translocation of PfEMP1 on to the surface of the IRBC. We conclude this from our observations that PfEMP1 is present in Maurer's clefts in an -deleted parasite line but does not appear on the surface of RBCs infected with mature stage parasites. The failure of PfEMP1 to reach the RBC surface leads to a loss of binding of IRBCs to CD36 both in static and flow-based adhesion assays. Reintroduction of the SBP1 protein leads to the reappearance of PfEMP1 on the IRBC surface and restoration of the capacity of IRBCs to adhere to CD36. What is the mechanism by which SBP1 has this effect on PfEMP1 trafficking? Our evidence would suggest that this is not caused by a direct physical interaction between PfEMP1 and SBP1. First, analysis of the SBP1 protein sequence does not reveal any domains with homology to known protein chaperones or to proteins known to be involved in protein trafficking or vesicular transport of proteins. Second, immunoprecipitation of SBP1 from IRBC lysates does not lead to the coprecipitation of PfEMP1 (Fig. S1, available at ). Third, acceptor photobleaching fluorescent resonance energy transfer (FRET) experiments in which PfEMP1 and SBP1 were labeled with FRET-compatible fluorophores did not lead to the detectable transfer of energy from one molecule to the other (Fig. S1). This suggests that they are not found in close contact as would be expected in directly interacting proteins. Finally, a recent study devoted to the interactome reported the interactions of PfEMP1 and SBP1 with other proteins in the parasite but did not find a direct interaction between the two (). Collectively, these data suggest that the effect of SBP1 on PfEMP1 transport is indirect. At present, the data are most compatible with the hypothesis that the alterations in Maurer's cleft morphology and distance to the IRBC surface leads to changes that make the final translocation step of PfEMP1 markedly inefficient. These changes are subtle, and this is likely to be the reason why trafficking of only one protein, PfEMP1, is interfered with and not others such as KAHRP, which presumably have more margin for error in the relatively straightforward process of binding to nearby spectrin. In contrast, PfEMP1 must successfully perform translocation across another membrane to reach its final destination, presumably using the transmembrane and COOH-terminal domains of PfEMP1, as these have been demonstrated to be sufficient for translocation of the protein across the RBC membrane (). Whatever the ultimate mechanism, it is clear that SBP1 has a general role in PfEMP1 trafficking as surface localization of PfEMP1 is restored in complemented lines expressing variant forms of the protein. The specific function of SBP1 in PfEMP1 translocation is consistent with its phylogenetic distribution within malaria species. We have examined several . genomes and are unable to identify a homologue of either by direct homology searching using the Basic Local Alignment Search Tool (BLAST) or Position-Specific Iterated BLAST. The gene is located between the genes for the exported parasite proteins and at the subtelomeric end of chromosome 5 in Orthology searches in the genome database () revealed that synteny between chromosome 5 and a contig from both and breaks down at the locus. This suggests that the more subtelomeric part of chromosome 5, including , , and , is specific only to (). Other species of lack a PfEMP1-like molecule, so it is reasonable that a protein specific for translocation should also be absent from the same species. In this context, it is interesting to note that deletion of the subtelomeric end of chromosome 2 of (which frequently occurs within the locus) results in the loss of knobs and cytoadherence of IRBCs, although PfEMP1 is still expressed on the IRBC surface (). In addition, the locus is also the breakpoint of synteny between chromosome 2 and a large contig (). This supports the idea of a few –specific subtelomeric loci that are critical for the transport and function of PfEMP1 and, consequently, in cytoadherence of IRBC to host vascular endothelium. The absence of in other malaria species that express proteins on the surface of the IRBC and that are involved in antigenic variation such as the , , and gene families (collectively termed the interspersed repeat superfamily []; ) also suggests that these molecules reach the surface of the IRBC by some other mechanism. As RIFINs are the representatives of the family in that species, it seems very likely that their trafficking to the IRBC surface in occurs by an SBP1-independent mechanism. We cannot examine this directly in our KO line because it is generated on a 3D7 parasite background, a line that does not express detectable RIFINs on the RBC surface (). In conclusion, we show here, for the first time, that the Maurer's cleft–associated protein SBP1 plays a major role in translocation of the cytoadherence ligand PfEMP1 onto the surface of –infected RBCs. We have demonstrated this via complementation of a gene KO parasite line, which is the first study to our knowledge of the complementation of a blood stage gene in . Future studies will be aimed at identifying the precise mechanisms and interactions involved in this process and whether other molecular players are involved. Such knowledge might be used to interfere with the translocation process, which in vivo would be likely to be associated with a marked decrease in parasite virulence and may constitute a new therapeutic approach to the control of malaria infection in humans in the future. The availability of specific PfEMP1-null lines also makes possible new studies examining such questions as the importance of PfEMP1 in the pathogenesis of malaria by, for example, the disruption of SBP1 in monkey-adapted parasite lines and subsequent infection of primate hosts. If such parasite lines are avirulent in the vertebrate host, they could become the basis of a genetically defined attenuated vaccine for use in humans. parasites (3D7) were maintained in continuous in vitro culture in human RBCs suspended in Hepes-buffered RPMI 1640 supplemented with 0.5% AlbumaxII using standard procedures (). Cultures were kept synchronous, and the knob-positive phenotype was maintained by gelatin floatation (). Clonal parasite lines were derived by the method of limiting dilution. To disrupt the gene in 3D7 parasites, 5′ and 3′ sequences of ∼1 kb flanking the gene were cloned into the transfection plasmid pHHT-TK (gift from A. Cowman and B. Crabb, The Walter and Eliza Hall Institute, Melbourne, Australia; ) to derive pHTKΔ (). Specifically, the 5′ segment of (including 458 bp of the 5′ untranslated sequence) was amplified from genomic DNA from 3D7 parasites using forward and reverse primers 5′TCcgatacaaccctccttttatg3′ (SacII site underlined) and 5′GGgacatagattcggctgga3′ (SpeI site underlined), respectively, and was subcloned into the SacII and SpeI sites of pHHT-TK upstream of the h resistance cassette. The 3′ segment of (including 437 bp of the 3′ untranslated sequence) was amplified using the forward and reverse primers 5′CGgcagattttgcaaaacaagc3′ (EcoRI site underlined) and 5′CATGcatatacataaacgatcaaaag3′ (NcoI site underlined), respectively, and was introduced into the EcoRI and NcoI sites downstream of the h cassette. Plasmid DNA was amplified in and purified using MegaPrep (QIAGEN). For complementation, the full-length gene (excluding the stop codon) was PCR amplified from 3D7 cDNA using the forward and reverse primers 5′CACCTATATACAatgtgtagcgcagctcgagca3′ and ggtttctctagcaactgtttttg, respectively, and was cloned using topoisomerase into the multisite Gateway entry vector pENTR/D-TOPO (Invitrogen). This was recombined together with pDONR P4-P1R vector containing the promoter and pDONR P2R-P3 vector containing the EYFP reporter gene into a destination vector, pHrB1-1/2, that had been previously modified for use in () to produce the expression plasmid pHrB1-1/2-. Ring-stage parasites were transfected by electroporation with 150 μg of purified supercoiled plasmid DNA (pHTkΔ) diluted in cytomix according to standard procedures () but using modified electroporation conditions to enhance DNA delivery (). Transfected parasites were cultured in the presence of 2.5 nM of the antifolate drug WR99210 () for ∼30 d until viable parasites were observed in Giemsa-stained smears. 4 μM ganciclovir (Roche Diagnostics) was then added to select for parasites having only double crossover homologous recombination (). For complementation, positive selection for parasites transformed with pHrB1-1/2- was performed using 8 μM blasticidin S hydrochloride (Calbiochem) as previously described (). Genomic DNA was extracted from parasite culture using the Nucleon BACC2 kit (GE Healthcare), digested with ClaI and EcoRI, separated on 1% agarose gels, and transferred to nylon membranes. Southern blot hybridization was performed using standard procedures. Cultured IRBCs were harvested on Percoll and solubilized in 2× reducing SDS sample buffer containing protease inhibitor cocktail (Roche Diagnostics). These total parasite extracts were separated on 10% SDS-PAGE gels and transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were probed with rabbit polyclonal anti-HSP70 (1:10,000), mouse polyclonal anti-SBP1 (1:400), or mouse monoclonal antiglycophorin C (1:500; Sigma-Aldrich) antibodies. Detection by enhanced chemiluminescence (Lumi-light Western blotting substrate; Roche Diagnostics) was performed after secondary detection with either sheep anti–rabbit or sheep anti–mouse Ig-HRP conjugate (1:2,000; Silenus). The trypsin cleavage assay was performed as previously described () to visualize PfEMP1 expressed on the IRBC surface. In brief, mature IRBCs were enriched on Percoll as previously described () and were incubated with 100 μg/ml TPCK-treated trypsin (Sigma-Aldrich) in the presence or absence of 1 mg/ml soybean trypsin inhibitor (STI; Sigma-Aldrich) for 15 min at 37°C. Reactions were stopped by the addition of STI to a final concentration of 1 mg/ml followed by a further incubation of 15 min at room temperature. Membrane proteins (including PfEMP1) were extracted using Triton X-100 and SDS solubilization as described previously () and diluted in reducing Laemmli sample buffer. Samples were separated on 6% SDS-PAGE gels and transferred for 4 h at 4°C onto PVDF membranes. The cytoplasmic tail of PfEMP1 (VARC) was detected using the mouse monoclonal antibody 1B/98-6H1-1 (1:100; gift from The Walter and Eliza Hall Institute). For indirect immunofluorescence, cultured RBCs were fixed in suspension with 4% PFA containing 0.0075% glutaraldehyde in PBS, permeabilized with 0.1% Triton X-100, and blocked with 3% BSA in PBS as previously described (). Cells were then incubated for 1 h with either mouse polyclonal anti-SBP1 (1:500), rabbit polyclonal anti-KAHRP (1:500), rabbit polyclonal anti-Pf332 (1:200), rabbit polyclonal anti–REX-1 (1:2,000; gift from D. Gardiner, Queensland Institute of Medical Research, Brisbane, Australia), mouse polyclonal anti–MAHRP-1 (1:200; gift from C. Spycher and H.-P. Beck, Swiss Tropical Institute, Basel, Switzerland), or rabbit polyclonal anti-GFP (1:1.000; Invitrogen). For PfEMP1 localization using rabbit polyclonal anti-VARC (1:100), detection was performed using thin culture smears that had been air dried and fixed with cold acetone/methanol (9:1) because this antibody showed poor reactivity when used on RBCs that had been fixed with PFA/glutaraldehyde in suspension using the method of . Primary antibodies were then detected using either anti–mouse or anti–rabbit IgG conjugated to AlexaFluor488 or -568 (Invitrogen) and visualized using a laser-scanning confocal microscope (model TCS NT; Leica) equipped with a krypton/argon laser (488/568 nm). The confocal scan head was mounted on an inverted microscope (model DM RBE; Leica) equipped with a 100× NA 1.4 oil plan-Apo objective. All images of individual RBCs shown in figures are representative of numerous similar RBCs in multiple fields of view. Green fluorescence of EYFP-expressing transformants was observed using live unfixed RBCs directly from culture under 513-nm light using a fluorescence microscope (BX51; Olympus). RBCs from synchronous cultures containing predominantly mature stage parasites (late trophozoites/young schizonts) were fixed by the dilution of packed RBCs into isotonic PBS containing 2.5% (vol/vol) electron microscopy grade glutaraldehyde (Sigma-Aldrich). After routine preparative procedures (), ultrathin sections were prepared for transmission electron microscopy and pelleted for SEM. Random images of IRBCs were captured digitally at a magnification of 20,000× in an electron microscope (H7600; Hitachi) and analyzed using Scion Image software (v4.0.2). Electron-dense knobs and Maurer's clefts were counted manually for each IRBC. A total of 32 IRBCs were examined for each of the parasite lines 3D7 and 1G8. Parameters measured were the perimeter of each IRBC, the total sectional area of each IRBC, and the sectional area of the parasite (if more than one parasitewas present, their areas were added together). From the second and third parameters, the area of IRBC external to the parasite was obtained by subtraction. Knobs and Maurer's clefts were counted manually to determine their frequencies. The adhesive properties of IRBCs were quantified using both static and flow-based assays. Parasite cultures were tested when the majority of parasites were pigmented trophozoites as assessed by Giemsa-stained smears. Cultured RBCs were resuspended in adhesion buffer (Hepes-buffered RPMI 1640 supplemented with 1% BSA and pH adjusted to 7.0) to a concentration of ∼3 × 10 RBCs/ml for static adhesion assays or 1.5 × 10 RBCs/ml for flow-based assays. For all adhesion assays, the parasitemia averaged ∼4.4% trophozoites (range of 2.2–7.8%). Static assays were performed in 36-mm petri dishes as previously described () except that purified recombinant CD36 (R&D Systems) was immobilized as the target receptor (100 μg/ml). Adhesion to CD36 under flow conditions that mimic those in postcapillary venules was visualized and quantified in vitro by direct microscopic observation on a microscope (IMT-2; Olympus) with a 40× water immersion objective (Olympus) using flat, rectangular glass microcapillary tubes (Microslides; VitroCom, Inc.) connected to a flow-control system as previously described (,). Fig. S1 shows the results from immunoprecipitation and FRET experiments to demonstrate that there is no direct molecular interaction between SBP1 and PfEMP1 in IRBCs infected with 3D7 parasites. Supplemental text provides details of these FRET and immunoprecipitation experiments. Online supplemental material is available at .

The p53 transcription factor is activated by inappropriate cell growth stimulation or by certain types of DNA damage and regulates the expression of other genes involved in cell growth arrest, DNA repair, and apoptosis (). These various p53-mediated effects suppress tumorigenesis, and mutation of the p53 gene or of the p53 signaling pathway is commonly found in most human cancers (). Although the ability of p53 to regulate cell growth after exposure to stress has been well established, the role of p53 in regulating normal (nondamaged) cell growth and in tissue homeostasis is uncertain. Mice deleted for p53 will form tumors with 100% penetrance yet undergo normal development (), albeit with a reduction from the expected numbers of female births and a small percentage of embryos presenting at midgestation with exencephaly (; ). In addition, transgenic mice bearing a reporter gene under transcriptional control of a p53 response element reveal little or no postnatal p53 activity in the absence of DNA damage (). These findings suggest that p53 is important in suppressing tumorigenesis but is largely dispensable for normal cell growth, cell differentiation, and development. In contrast, a recently generated mouse model bearing a mutated p53 allele () that increased p53 activity in vitro displayed early aging-like phenotypes, including reduced mass of various internal organs, thinning of the dermis, hair loss, and osteoporosis (). Interestingly, the p53 m/+ mice also had reduced incidence of cancer relative to p53 heterozygous (−/+) mice, suggesting that the mutant allele increased p53 activity encoded by the wild-type p53 gene in the m/+ mice. Although this increase in basal levels of p53 activity offered further protection from neoplasia, the authors hypothesized that the slight increase in p53 activity also reduced stem cell proliferation in affected tissues leading to reduced tissue cellularity. These results suggest that negative regulation of p53-induced apoptosis or inhibition of cell growth might be important to maintain proper tissue homeostasis in adult mice. Mdm2 is a key negative regulator of p53 activity in the cell. Mdm2 complexes with p53 and negatively regulates p53-induced transcription of target genes, including the gene (for review see ). During times of cellular insult, p53 activates gene expression by binding to a p53 response element within the first intron of the gene (). Induction of Mdm2 protein levels leads to an increase in Mdm2–p53 complex formation that interferes with the ability of p53 to transactivate Mdm2. Thus, Mdm2 expression is autoregulated because of the ability of Mdm2 to negatively regulate p53 (). Mdm2 has been shown to interfere with the ability of p53 to transactivate target genes by binding and stearically hindering the NH-terminal activation domain of the p53 protein (; ) or by altering p53 protein modifications that regulate p53 transcriptional activation (). In addition, Mdm2 can function as an E3 ligase to coordinate the ubiquitination of p53 () and can induce the degradation of p53 by the 26S proteasome (; ; ). Mdm2 can also assist in shuttling p53 from the nucleus into the cytoplasm (; ). The importance of Mdm2 in negatively regulating p53 activity is perhaps best illustrated by the finding that the early (embryonic day [E] 4–5) lethal phenotype of Mdm2-null mice can be fully rescued by the concomitant deletion of p53 (; ). Although the requirement for Mdm2-mediated inhibition of p53 activity during early development has been well established, the role of Mdm2 in regulating p53 functions in later stages of embryogenesis or in adult tissues is unclear. However, several lines of evidence suggest that Mdm2 does function to regulate p53 activity in postnatal tissue. EuMyc transgenic mice display a delayed onset of B cell lymphoma when haploinsufficient for Mdm2, suggesting that a reduction in Mdm2-mediated suppression of p53 can reduce tumorigenesis (). In addition, mice bearing a hypomorphic allele of Mdm2 that have ∼30% of the normal endogenous levels of Mdm2 are smaller in size, have reduced numbers of hematopoietic cells, and display excess apoptosis in the lymphoid compartment (). Crossing the Mdm2 hypomorphic allele onto a p53-deficient background reversed the various phenotypes observed in these mice, demonstrating that the phenotypic effects caused by Mdm2 reduction in this model were induced by p53. These data suggest that Mdm2 is capable of negatively regulating p53 activity in hematopoietic tissues. To determine the absolute requirement for Mdm2 during development and in adult tissues, we have recently used Cre-loxP technology to generate Mdm2-conditional mice. Gene targeting experiments in embryonic stem cells flanked the last two exons of the gene encoding the zinc RING (really interesting new gene) finger domains and polyadenylation signals with loxP sites. Cre-mediated recombination of the loxP sites in the conditional allele destabilizes Mdm2 transcripts and results in loss of Mdm2 message. (). Because studies of p53 mutant mice suggest that excess p53 activity might have a deleterious effect on normal bone homeostasis, we sought to determine whether Mdm2 regulates p53 activity during osteogenesis. To this end, Mdm2- or p53-conditional mice were bred with transgenic mice in which the Cre-recombinase gene has been placed under transcriptional control of a 3.6-kb fragment of the promoter. These Col3.6-Cre–transgenic mice have been previously reported to express Cre in cells of the osteoblast lineage (). Mdm2-conditional mice bearing the Col3.6 transgene have multiple skeletal defects, including fused or otherwise altered lumbar vertebrae, reduced mineralized bone, and reduced bone length. Osteoblasts deleted for Mdm2 do not undergo apoptosis but do have elevated p53 activity, increased transactivation of p53 target genes, reduced cell proliferation, and reduced levels of the osteoblast transcriptional regulator Runx2, which is essential for osteoblast differentiation (; ; ). In contrast, osteoblasts deleted for p53 display elevated Runx2 levels, enhanced cell proliferation, and increased maturation and mineralization. Furthermore, mice specifically deleted for p53 in osteoblast progenitor cells develop osteosarcomas. These results demonstrate that p53 is an important negative regulator of osteogenesis and that Mdm2-mediated inhibition of p53 function is a critical requirement for Runx2 activation and proper osteoblast differentiation and skeletal formation. To understand the contribution of Mdm2 and p53 signaling to bone development, we used Mdm2-conditional mice bearing Mdm2 genes (Mdm2) with loxP recombination sites flanking exons 11 and 12 (). Deletion of these exons will result in loss of Mdm2 in cells expressing the P1 bacteriophage Cre-recombinase transgene under transcriptional control of a rat 3.6-kb type I collagen promoter fragment (). This transgene has been reported to induce Cre-mediated gene excision in postnatal calvarial bone and long bone osteoblasts, as well as in skin and tendons. We initially verified the expression of Cre in skeletal elements of adult mice. RT-PCR reactions using RNA isolated from the rib and femur of adult Col3.6-Cre mice demonstrated Cre expression in these tissues, whereas Cre transcripts were not detected in other tested organs (). To identify regions in Col3.6-Cre developing embryos that express Cre and induce Cre-mediated excision, we mated the Col3.6-Cre–transgenic mice to ROSA26 (R26R) reporter mice that express β-galactosidase upon Cre-induced deletion of a floxed cassette that inhibits the reporter gene expression (). Our findings demonstrate that this promoter becomes activated at two distinct points during development. Expression of the Col3.6-Cre transgene as determined by β-galactosidase activity is first detected in whole-mount stainings of E8–19 embryos and is localized initially to the caudal portion of the embryo. Reporter gene expression is strongest in this region at E10 (). This region of the developing embryo contains both neural and mesenchymal progenitor cells that give rise to the caudal axial skeleton and spinal cord. Later in development (∼E14), the Col3.6-Cre transgene undergoes robust activation in relation to the formation of connective tissue. Staining of sagital sections of R26R/Col3.6-Cre embryos for β-galactosidase activity indicates that Cre-mediated excision occurs in both the skin and developing skeletal elements (). These findings indicate that Cre transgene expression will induce Mdm2 deletion in Col3.6-Cre × Mdm2-conditional mice in the surface ectoderm and developing tail bud of E9 mice as well as in developing bone and connective tissue of these mice during midgestation. We crossed Mdm2-conditional mice (Mdm2 or Mdm2) to Col3.6-Cre–transgenic mice containing one conditional Mdm2 allele (Mdm2,Col3.6-Cre). The resulting litters contained Mdm2,Col3.6-Cre mice (designated as mutant), as well as Mdm2,Col3.6-Cre mice and Mdm2,Col3.6-Cre mice. These last two groups of mice were indistinguishable from wild-type mice, which was expected, given that Mdm2 heterozygous mice were previously found to display no morphologic developmental defects (). To examine the effects of Mdm2 loss during development, timed matings of Mdm2,Col3.6-Cre–transgenic mice were performed and embryos harvested and genotyped at various times after coitum. Expected numbers of mutant Mdm2 embryos were recovered between E8.5 and -12.5 (). However, reduced numbers of mutant embryos were found at E13.5 and throughout later stages of development, and no viable Mdm2,Col3.6-Cre mutant mice were recovered at weaning or at birth, demonstrating that deletion of Mdm2 in Col3.6-Cre–transgenic mice induces lethality during later stages of embryogenesis. Defects in the developing tail bud region of mutant mice were apparent as early as E10.5, and caudal runting was seen in all mutant mice throughout later stages of development (). Col3.6-Cre–induced deletion of Mdm2 expression in the caudal region of mutant mice resulted in a complete absence of tissue surrounding the somites at E10.5, as well as exposure of somitic mesenchyme at the surface of the embryo, absence of a tail, and a severe invagination in the posterior dorsal region encompassing the lumbar vertebrae (). Histologic analysis of mutant and wild-type embryos harvested at E10 was performed. Wild-type embryos had clearly segmented somites that will give rise to axial skeletal components, including vertebrae and ribs (, arrows). These somites are surrounded by primitive neural tissue and by developing dorsal root ganglia (, arrowheads). In contrast, mutant embryos lack developing neural tissue and posterior dorsal root ganglia, leaving caudal somites externalized (). To understand the mechanism underlying this tissue loss, we performed TUNEL staining on serial sections. Our results show that there is a dramatic increase in TUNEL-positive apoptotic cells in the caudal somites and surrounding tissue of mutant E10 embryos in comparison to wild-type littermates (). Skeletal preparations of wild-type and Mdm2,Col3.6-Cre embryos revealed that the apoptosis observed in the E10 caudal somatic cells resulted in highly dysplastic axial skeletal elements in the mutant embryo with fused cartilaginous lumbar vertebrae (). However, the appendicular skeleton and bones of the skull were unaffected by the initial Cre activation. Activation of Col3.6-Cre transgene expression in the limbs and skull did not occur until the latter half of gestation (beginning at E14.5). These skeletal elements were more porous in mutant animals (), and measurements of the long bones (tibia, radius, ulna, humerus, and femur) revealed that the total long bone length was 12.1 ± 3.9% shorter in the mutant embryos and the length of mineralized portion of long bones was 12.3 ± 4.4% shorter in the mutant embryos than in wild-type littermate embryos at E17.5 (). In contrast, skeletal preparations of embryos deleted for both Mdm2 and p53 () revealed no vertebral dysplasia or reduction in bone length (). These results reveal that the deleterious effects of Mdm2 loss on skeletal formation during development are p53 dependent. Appendicular and skull bones from E18.5 Mdm2, Col3.6-Cre mice showed significantly less mineral deposition than wild-type littermates, as assessed by microcomputed tomography (micro-CT) scan () and by silver nitrate (Von Kossa) staining (). Decreased bone mineralization was observed in the skull (), femur, and vertebra () of Mdm2 mutant mice. Histomorphometric analysis of micro-CT scans performed on litter-matched, E18.5 embryos revealed reduced bone density in the calvaria, femur, and vertebra of Mdm2 mutant mice (). However, these defects were not observed in embryos lacking both Mdm2 and p53. These results demonstrate that loss of Mdm2 in developing skeletal tissues negatively affects multiple parameters of bone quality in a p53-dependent manner. Furthermore, Von Kossa and toluidine blue staining of femur and vertebra sections () revealed that the vascularized marrow cavity and other normal compartments of bone are present in the mutant mice and that Mdm2 mutant bone has a morphologically normal growth plate. These data suggest that Mdm2 mutant mice do not have a chondrogenic defect but rather a defect in ossification. However, unlike the proapoptotic effects of Mdm2 deletion in undifferentiated progenitor cells in E10 mutant mice, TUNEL assays performed on E14.5 skeletal rib elements from mutant embryos revealed no increase in the number of apoptotic cells (unpublished data). To better understand the underlying cause of the decrease in bone quality in Mdm2-conditional Col3.6-Cre mice, calvarial osteoprogenitor cells were isolated from E19 wild-type and mutant embryos as well as from R26R/Col3.6-Cre embryos to visualize the pattern of Cre-mediated excision in these cultures. Calvarial osteoprogenitors were induced to undergo osteogenic differentiation ex vivo by allowing cultures to proliferate for several days followed by the addition of ascorbic acid and inorganic phosphate to the media postconfluence. The addition of ascorbic acid to the cultured osteoprogenitors stimulates postconfluent proliferation, resulting in the formation multilayered nodules that later become mineralized (). Examination R26R/Col3.6-Cre osteoprogenitor cells reveals that few osteoprogenitor cells had activated the transgene before isolation from the embryo and that Cre expression is induced in the osteoblast cultures during differentiation (). Because Cre-mediated recombination occurs in these cultures in maturing osteogenic nodules after reaching confluence, Mdm2 expression in Mdm2-conditional Col3.6-Cre cultures should be lost in multilayering nodules of maturing osteoblasts. Toluidine blue staining of calvarial osteoprogenitor cells cultured from wild-type and Mdm2-conditional Col3.6-Cre mice revealed that both wild-type and Mdm2 mutant–derived osteoprogenitor cultures achieved confluence simultaneously (, B and C, left), and BrdU incorporation indicated no difference in the rates of cell proliferation in these cultures before confluence (not depicted). However, upon reaching confluence and upon robust induction of Cre expression, wild-type osteoprogenitor cell cultures underwent robust nodule formation (), whereas Mdm2,Col3.6-Cre cultures were unable to form a significant number of multilayered nodules. Subsequently, only a small fraction of the mutant cells were able to undergo osteoblast differentiation as reflected by alkaline phosphatase activity (). Consistent with our in vivo findings of reduced mineralization in Mdm2 mutant bone, wild-type osteoblasts ultimately formed heavily mineralized nodules in culture, whereas mutant cells failed to deposit significant mineral in the extracellular matrix () as determined by silver nitrate staining (right). Quantitative analysis by real-time PCR of osteogenic gene expression revealed that both wild-type and mutant cultures began to activate early osteogenic genes type I collagen and alkaline phosphatase upon reaching confluence. However, osteogenic gene expression was abrogated in Mdm2 mutant cultures shortly after achieving confluence (, dashed lines). Osteocalcin, a marker of late osteoblast differentiation, was not activated in mutant cultures at any time during the culture period (). Because osteoblast differentiation and expression of these osteoblast phenotypic genes is dependent on the activity of the Runt-related transcription factor Runx2 (; ; ), we next examined the impact of Mdm2 loss on the activity of this gene. Real-time PCR analysis revealed that Mdm2 and Runx2 are expressed at low levels in wild-type cultures during the period of proliferation but are strongly activated in postconfluent cultures during multilayering and differentiation. However, as expected, Mdm2 up-regulation is abrogated in osteoprogenitor cultures derived from Mdm2-conditional Col3.6-Cre mice during the course of differentiation (). Interestingly, both Runx2 protein and message levels were lost in mutant cultures concomitantly with the loss of Mdm2 expression (). Furthermore, although the G1/S ratio of preconfluent osteoprogenitor cells was unchanged in Mdm2-conditional Col3.6-Cre cells (before induction of Cre expression), postconfluent cycling of mutant cells deleted for Mdm2 was strongly inhibited as determined by BrdU uptake assays, with a G1/S ratio of 7.5:4.2 for nonmutant cells. In addition, no differences in apoptotic cell numbers were detected by TUNEL assays during differentiation of cultured osteoblast progenitor cells after Cre-mediated deletion of Mdm2 (unpublished data). Collectively, our results indicate that Mdm2 activity is required for postconfluent cell proliferation and nodule formation in osteoblast cultures and the subsequent activation of the master osteoblast transcriptional regulator Runx2. Failure of cultures lacking Mdm2 to activate the Runx2 gene ultimately results in inhibition of osteoblast differentiation and inactivity of osteoblast phenotypic genes. To confirm that reduced Runx2 expression in the Mdm2 mutant cells is the underlying cause of the maturation defect, recombinant adenovirus vectors were generated to transduce either lacZ (control) or Runx2 cDNA into the osteoblast progenitor cultures. Addition of exogenous Runx2 into Mdm2 mutant cells induced maturation of these progenitor cells and partially or fully restored the expression of mature osteogenic genes such as collagen type 1, alkaline phosphatase, and osteocalcin (). Examination of p53 levels in ex vivo osteoblast cultures revealed no change in total p53 protein levels in cultures undergoing deletion of Mdm2 (). However, a difference in the amount of activated p53 transcription factor present in the cultures was detected using a phospho-Ser15 specific antibody, with a marked induction in P-Ser15 p53 levels observed in mutant cultures relative to levels in wild-type cells (), suggesting that Mdm2 negatively regulates p53 activity but not overall p53 protein levels in differentiating osteoblasts. Furthermore, real-time PCR analysis revealed up-regulation of Ptprv and p21 gene expression () in postconfluent osteoblast cultures derived from Mdm2-conditional Col3.6-Cre mice (dashed lines) relative to expression levels observed in Col3.6-Cre cells containing wild-type Mdm2 alleles (solid lines). These genes are targets of p53 transactivation known to be involved in regulating the progression of primary cells from G1 into S phase of the cell cycle (; ). This increase in Ptprv and p21 expression is consistent with the inhibition of postconfluent osteoprogenitor cell growth observed in the Mdm2-conditional Col3.6-Cre cultures. Interestingly, the expression levels of and , two proapoptotic p53 response genes, were not significantly altered in osteoblast progenitor cells after deletion of Mdm2 (unpublished data), consistent with the unaltered level of apoptosis in Mdm2 mutant and wild-type cells. Our findings indicate that loss of Mdm2 results in a block in osteoblast differentiation that is due to excess p53 activity. To confirm that p53 negatively regulates bone maturation and mineralization, we examined the differentiation of osteoprogenitor cells harvested from the calvaria of p53-null mice. Calvarial osteoprogenitor cells were isolated from E19 wild-type or p53-null mice and induced to undergo osteogenic differentiation. The addition of ascorbic acid to the cultured osteoprogenitors greatly increased postconfluent proliferation in the p53-null cultures relative to wild-type cultures, resulting in increased formation of multilayered nodules and excess mineralized bone formation (). BrdU staining and FACS analysis of postconfluent osteoprogenitor cells revealed fewer p53-null cells in G1 and more in S phase (). Furthermore, real-time PCR analysis of Runx2 transcripts in the ex vivo cultures revealed a dramatic increase in the Runx2 expression levels during maturation and mineralization when p53 was absent (). These data confirm that p53 negatively regulates osteoblast proliferation and differentiation. Notably, there was a difference in the ex vivo maturation of p53-null and Mdm2/p53 double-null osteoblast progenitor cells (), underscoring our previous in vivo findings ( and ) of a p53-dependent role for Mdm2 in regulating skeletal development. Increased proliferation of osteoprogenitor cells in the calvarial cultures suggests that negative regulation of osteoblast progenitor cell proliferation and differentiation by p53 may be an important component of p53-mediated suppression of bone tumorigenesis. To explore a link between negative regulation of osteoblast growth and neoplasia, we bred the Col3.6-Cre mice with p53-conditional mice (). Cohorts of transgenic Col3.6-Cre mice that were either homozygous or heterozygous for the p53-conditional allele were used to perform a tumor assay. Col3.6-Cre mice heterozygous for the p53-conditional allele developed mostly bone masses starting at 20 wk of age, and all mice in the colony presented with cancer by 82 wk, whereas Col3.6-Cre mice homozygous for the p53-conditional allele all presented with tumors by 42 wk (). Histopathologic analysis of p53-conditional Col3.6-Cre mouse tumors identified 60% of these cancers as osteosarcomas, with a 20% incidence of lymphoma or fibrosarcoma. Most of the osteosarcomas were classified as high grade and were of intermediate differentiation, producing osteoid but not mature lamellar bone (, samples PT2, -50, and -38). However, a few of the osteosarcoma tumors were especially aggressive and very poorly differentiated (, sample PT58). Western blot analysis of representative primary tumor samples harvested from these mice indicated that Runx2 expression was greatly elevated specifically in osteosarcomas, regardless of their differentiation (, lanes 1–3). These results confirm that p53 not only down-regulates osteoblast cell growth and differentiation during development but also plays a critical role in suppressing osteosarcoma formation in adult bone tissue. The rescue of Mdm2-null mice from peri-implantation lethality by deletion of p53 demonstrates that Mdm2 plays an important role in early development by negatively regulating p53 activity (; ), but little is known about the requirement for Mdm2 throughout embryogenesis and in postembryonic tissues. We previously documented that the growth characteristics of p53-null primary fibroblasts and the tumorigenic potential of p53-null mice are indistinguishable in the presence or absence of Mdm2 (). Thus, if Mdm2 was important in the latter stages of development, in organogenesis, or in postnatal cell growth control, it is likely due to the ability of Mdm2 to down-regulate p53 activity and not to p53-independent effects of Mdm2. In support of a role for Mdm2 in regulating p53 during the latter stages of development, recent studies of mice that contain reduced amounts of Mdm2 relative to wild-type levels indicate that Mdm2 regulates p53 activity in hematopoietic development and in B cell tumorigenesis (; ). However, the effect of complete ablation of Mdm2 activity on cellular differentiation and in organogenesis is unknown. Coordinated proliferation and differentiation of bone-forming osteoblast progenitor cells and the deposition of extracellular matrix proteins by osteoblasts and subsequent deposition of crystalline salts for mineralization of the skeleton are two critical steps in bone modeling both during development and in the adult skeleton (). Although little is known regarding the role of p53 in this process, previous studies of mice bearing a hypermorphic p53 mutation revealed that mice with increased amounts of p53 activity exhibit symptoms of rapid aging, including osteoporosis (). Furthermore, a subset of mice haploinsufficient for functional p53 in all tissues develop osteosarcomas (). These data suggest that p53 may regulate normal bone growth and that alterations in the levels of p53 activity can contribute to abnormal bone phenotypes. To further explore a role for p53 in osteogenesis and to determine whether Mdm2–p53 signaling is important in bone growth and development, we bred Mdm2-conditional mice with Col3.6-Cre–transgenic mice. Deletion of Mdm2 upon expression of the Col3.6-Cre transgene resulted in midgestational caudal defects, including loss of tissue surrounding the somites, exposure of somitic mesenchyme at the surface of the embryo, absence of a tail, and a severe caudal invagination. Furthermore, TUNEL staining of serial sections of mutant embryos revealed increased apoptosis in the caudal somites and surrounding tissue of mutant embryos, suggesting that these defects arose through unregulated p53 apoptosis. Harvests of embryos from timed matings of Mdm2-conditional Col3.6-Cre–transgenic mice revealed a marked decrease in the recovery of Mdm2,Col3.6-Cre embryos at E13.5, coincident with the robust activation of the Cre transgene in developing skeletal elements. Skeletal preparations of wild-type and of Mdm2,Col3.6-Cre embryos documented numerous skeletal defects during the latter stages of development, including a reduction in mineralized bone and in length of appendicular bone, abnormal bone architecture, and an increase in bone porosity. However, analysis of skeletal preparations of embryos deleted for both Mdm2 and p53 revealed no skeletal defects, and no difference was observed in the growth and maturation of cultured calvarial cells deficient for p53 or for both Mdm2 and p53, indicating that the effects of Mdm2 loss on skeletal formation and osteoblast maturation are p53 dependent. In contrast to what was observed in the caudal mesoderm of E10 mutant embryos, TUNEL assays performed on E14.5 bone isolated from Mdm2-conditional Col3.6-Cre embryos and on osteoprogenitor cells cultured from E19 Mdm2-conditional Col3.6-Cre embryos revealed no increase in the number of apoptotic cells. This finding indicates that deletion of Mdm2 does not induce p53-mediated apoptosis in these cells but rather induces p53-mediated effects that block osteoblast proliferation or differentiation. To confirm that p53 plays a role in regulating osteoblast differentiation, we harvested osteoblast progenitor cells from the calvarial of p53-null mice just before birth (E19). Analysis of the growth and development of these cultured cells revealed that p53-null osteoprogenitor cells proliferated far faster than wild-type progenitor cells and underwent more robust differentiation, confirming that p53 functions to negatively regulate osteoblast maturation and mineralization. Surprisingly, the overall level of p53 protein did not change in osteoblast cells during differentiation in the presence or absence of Mdm2; however, deletion of Mdm2 did result in an increase in the level of activated p53 as judged by the increased levels of phosphorylated p53. Furthermore, the message levels of and the cyclin-dependent kinase inhibitor , two p53 target genes involved in regulating cell cycle progression from G1 to S phase, were increased in osteoblast cells after Mdm2 deletion. These results indicate that Mdm2 regulates p53 activity during osteoblast differentiation not by altering p53 stability but by inhibiting p53-mediated transactivation of genes involved in regulating osteoblast growth and differentiation. Runx2 is a critical inducer of osteoblast differentiation in vitro and in vivo (). Interestingly, levels of Runx2 message and protein were reduced in cells deleted for Mdm2, as were the message levels of Runx2 target genes type I collagen and alkaline phosphatase. In addition, expression of osteocalcin, a marker of late osteoblast differentiation, was not activated in cultures deleted for Mdm2 during differentiation, providing further molecular evidence for a block in osteoblast development upon deletion of Mdm2. In contrast to the reduction in Runx2 levels observed in osteoblasts deleted for Mdm2, Runx2 message levels were found to be greatly elevated in maturing osteoblasts deleted for p53. As Runx2 is a well-established master regulator of osteoblast differentiation, it is possible that p53 directly controls osteoblast maturation by negatively regulating Runx2 expression. However, analysis of the Runx2 promoter sequences failed to identify any p53 canonical binding sites, and there is no evidence present in the literature to suggest that Runx2 expression is directly regulated by p53. Therefore, we hypothesize that proper osteoblast differentiation and bone development require Mdm2 to inhibit a p53-mediated block on osteoprogenitor cell division. By permitting the postconfluent proliferation of osteoblasts through the down-regulation of p53 activity, Mdm2 indirectly facilitates Runx2 induction and osteoblast maturation. In support of this hypothesis, expression levels of mature osteogenic genes were found to be elevated in Mdm2 mutant osteoblasts after restoration of Runx2 expression. Our results indicate that up-regulation of p53 activity due to Mdm2 deletion induces a block in bone differentiation and mineralization and causes profound skeletal defects in the developing embryo. Furthermore, deletion of p53 in osteoblasts induces hyperproliferation, greatly elevated levels of Runx2 expression, and increased bone maturation in vitro. These findings indicate that p53 is an important negative regulator of bone growth and development. Interestingly, loss of cell differentiation and reduced expression of mature osteogenic genes such as are prognostic indicators in human osteosarcomas, with poorly differentiated or dedifferentiated tumors usually associated with the high-grade category (). In addition, Runx2 is down-regulated in various human osteosarcoma cell lines, suggesting a link between loss of Runx2 expression, dedifferentiation, and cancer (). However, we observed increased osteoblast differentiation and elevated Runx2 expression in osteoblast progenitor cells derived from p53-null mice. Therefore, we examined the ability of p53 to suppress tumorigenesis in osteoprogenitor-derived cells by crossing Col3.6-Cre–transgenic mice with p53-conditional mice (). Our results indicate that loss of p53 in osteoblasts induces a fairly rapid tumorigenesis in mice. Interestingly, Col3.6-Cre–transgenic mice heterozygous or homozygous for the p53-conditional allele display kinetics of tumor onset similar to those that have been previously documented for p53 knockout heterozygous or homozygous mice, though the tissue specificity of tumorigenesis was greatly altered. Mice deleted for p53 in all tissues die predominantly from lymphomas, chiefly of the thymus, and only occasionally will present with bone tumors (). Depending on the genetic background of the mice, between 3 and 8% of p53-null mice develop osteosarcomas (; ). However, this tumor spectrum may reflect the critical importance of p53 in suppressing thymic lymphomas in relatively young mice and not a reduced role for p53 in suppressing bone cancer, as osteosarcomas do constitute approximately one third of all tumor types observed in the longer lived p53 heterozygous mice (). A majority of the Col3.6-Cre–transgenic, p53-conditional heterozygous mice or p53-conditional homozygous mice in our study developed osteosarcomas. Furthermore, Runx2 expression was elevated in primary osteosarcoma samples harvested from these mice, in agreement with our finding of increased Runx2 expression in the p53-null calvarial cell cultures. The results of our p53-conditional Col3.6-Cre–transgenic mouse cross confirm that p53 is a critical tumor suppressor in bone tissue and indicate that osteosarcoma formation does not require loss of Runx2 expression. Instead, we propose that p53 inhibition of osteoblast cell proliferation is the mechanistic basis for suppression of bone osteosarcomas in this model. As we have demonstrated that Mdm2 is a key regulator of p53 activity in osteoblasts, disrupting the ability of Mdm2 to down-regulate p53 activity in these cells may prove to be a useful therapeutic strategy in treating osteosarcomas. Mdm2-conditional mice were developed in our laboratory using standard gene targeting techniques in embryonic stem cells. The resulting allele contains two sites flanking exons 11 and 12. Upon -mediated recombination, exons 11 and 12 are excised, rendering the allele inactive (described in ). Mdm2 excision was mediated by crossing Mdm2-conditional (Mdm2) mice to transgenic mice in which the 3.6-kb type I collagen promoter governs the expression of the -recombinase enzyme (). Visualization of tissues in which the -recombinase activity has recombined target alleles was facilitated by mating –transgenic mice to R26R reporter mice (The Jackson Laboratory) in which expression results in the removal of a -flanked DNA segment that prevents expression of a gene. The p53-conditional mouse model () was obtained from A. Berns (Netherlands Cancer Institute, Amsterdam, Netherlands). All animals were maintained and used in accordance with the University of Massachusetts Animal Care and Use Committee. Embryos were harvested from timed pregnant mothers at various time points during gestation followed by fixation in 4% paraformaldehyde. Tissues destined for histological sectioning were dehydrated in a graded series of ethanol and xylene, followed by infiltration with paraffin wax. Tissues from R26R/ crosses were fixed in 4% paraformaldehyde, equilibrated overnight in 30% sucrose, and embedded in optimal cutting temperature for cryosectioning. Some embryos were fixed in 4% paraformaldehyde followed by whole-mount staining for β-galactosidase activity. Parrafin sections were cut at 7 μm and counterstained with eosin. Frozen sections were cut at 10–12 μm, stained for β-galactosidase activity, and counterstained with eosin. β-Galactosidase activity was visualized by staining whole embryos, cryosections, or calvarial cultures in a solution containing 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl, 0.2% NP-40, 0.01% sodium deoxycholate, and 1 mg/ml X-Gal in PBS, pH 7.4, at 37°C for 1–6 h. Calvarial cultures were stained for mineral content using the method of Von Kossa. In brief, sections were exposed to a solution of 3% silver nitrite under direct sunlight for 15 min, after which mineral deposits were visualized as black precipitate under brightfield microscopy. Alkaline phosphatase activity in calvarial cultures was visualized by colorimetric enzymatic reaction to a solution containing 0.5 mg/ml napthol as MX phosphate disodium salt (Sigma-Aldrich), 2.8% N,N dimethylformamide (Sigma-Aldrich), 0.1 M Tris-maleate buffer, pH 8.4, and 1 mg/ml fast red salt (Sigma-Aldrich). The reaction was performed at 37°C for 10 min. Decalcified femora and vertebrae were embedded in a mixture of methyl methacrylate: glycol methacrylate, sectioned at 2 μm, stained with Von Kossa, and counterstained with toluidine blue. TUNEL staining was performed on dewaxed paraffin sections using a fluorescein-conjugated in situ cell death detection kit (Roche) according to the manufacturer's protocol. Skeletons were prepared for visualization of cartilage and bone using alcian blue and alizarin red stains, respectively. Embryos were eviscerated followed by overnight fixation in 100% ethanol. The next day, cartilaginous elements were stained overnight in a solution containing 4 parts ethanol, 1 part glacial acetic acid, and 0.3 mg/ml alcian blue 8GX (Sigma-Aldrich). The next day, soft tissues were dissolved for 6 h in a 2% KOH solution followed by an overnight staining in a 1% KOH solution containing 75 μg/ml alizarin red S (Sigma-Aldrich). Skeletons were then destained in 20% glycerol and 1% KOH for several days and stored in 50% glycerol and 50% ethanol. Embryos were fixed overnight in 4% paraformaldehyde, rinsed three times with PBS, and scanned on a Skyscan 1072 instrument (Skyscan). Image acquisition of the head was performed using 25× magnification (50× for the limb) at 45 kV and 222 μA, with a 0.45° rotation between frames to obtain two-dimensional images. Three-dimensional reconstruction and quantitative analyses were performed on a computer (Dell) using the NRecon, ANT, and CTAn software supplied with the Skyscan instrument. Calvarial osteoblasts were isolated from E19 embryos by enzymatic digestion of calvarial bones. In brief, calvaria were minced and subjected to three sequential digestions (8, 10, and 26 min) with collagenase P (Roche) at 37°C. Osteoblasts in the second and third digest were collected and resuspended in α-MEM supplemented with 10% FBS (HyClone). Cells were plated at a density of 10 cells/6-well plate (). Differentiation was initiated after confluence by the addition of ascorbic acid and β-glycerol phosphate. Cultures were harvested at various time points and stained for β-galactosidase activity, mineral content, alkaline phosphatase activity, or total cellularity using toluidine blue. All osteoblast differentiation experiments were performed a total of three times, and each experiment used embryos of different genotypes harvested on the same day (E19) from the same litter. The proliferation, confluence, maturation, and mineralization stages of differentiation are defined as days 5, 10, 14, and 20 in culture, respectively, except for , where confluence, maturation, and mineralization stages of differentiation were reached on days 8, 12, and 17 of culture, respectively. Whole-mount photographic images of embryos ( and ), histologic sections of embryos ( and ), skeletal preps (), and stained osteoblast cultures (; and ) were obtained using a stereoscope (MZ8; Leica) with either a 1× or 0.63× reduction lens and a digital camera (3008 Prog/Res; JenOptik) coupled to a computer (G4; Macintosh), using Photoshop 4 software (Adobe). Bone histology images () were captured using an Axioskop 40 (Carl Zeiss MicroImaging, Inc.) equipped with a camera (AxioCamMRc; Carl Zeiss MicroImaging, Inc.), a Dell computer, and MRGrab software. Magnifications at source are 2.5× (low power) and 20× (high power). TUNEL photographic images () were obtained using an inverted microscope (model 405; Carl Zeiss MicroImaging, Inc.) with a Fluor 10 (10×) plan (Nikon) and a digital camera (SPOT; Diagnostic Instruments) connected to a Dell computer using Photoshop 4 Imaging. Tumor hematoxylin and eosin–stained sections () were imaged using a microscope (Eclipse E400; Nikon) with a 20× plan (Nikon) coupled to a SPOT digital camera and a computer (IBM) with Spot acquisition software (4.0.1) and Photoshop 7 software. Adenoviral infection of primary mouse osteoblasts was performed at day 10 of culture (confluence) with either a vector expressing Xpress-tagged mouse under transcriptional control of the cytomegalovirus CMV5 promoter (pAd/CMV5/Xpress-Runx2/IRES/GFP) or a control vector expressing LacZ (pAd/CMV5/LacZ/IRES/GFP). Infections were performed at a multiplicity of infection of ∼100 in α-MEM containing 5% FBS (HyClone). 12 h after infection, the media was replaced with α-MEM containing 10% FBS and ascorbic acid to initiate osteogenic differentiation. Cultures were harvested 7 d after infection, and quantitative RT-PCR was performed on total RNA isolated from each sample. RNA was isolated from tissue or cell cultures using Trizol reagent (Invitrogen) according to the manufacturer's protocol. After purification, 5 μg of total RNA was DNase treated using a DNA-free RNA column purification kit (Zymo Research). 1 μg RNA was then reverse transcribed using Oligo-dT primers and a first-strand synthesis kit (SuperScript; Invitrogen) according the manufacturer's protocol. Gene expression was assessed by semiquantitative (Cre, Mdm2, and glyseraldehyde-3-phosphate dehydrogenase [GAPDH]; 25 cycles) and quantitative real-time PCR (Runx2, alkaline phosphatase, osteocalcin, collagen type I, and GAPDH). Quantitative PCR was performed using SYBR green 2× master mix (Eurogentec) and a two-step cycling protocol (anneal and elongate at 60°C and denature at 94°C). Specificity of primers was verified by dissociation temperature of amplicons. Results are representative of two or more independent experiments. Total protein was isolated from calvarial cultures or from primary tumor samples in the presence of direct lysis buffer (0.1 M Tris-HCl, pH 7.5, 10% glycerol, 0.01 M DTT, 12% urea, and 2% SDS) followed by heating for 5 min at 100°C. 40 μg of total protein was then electrophoresed through a 10% acrylamide gel followed by transfer onto a polyvinylidene fluoride (Immobilon) membrane. Membranes were blocked in PBS Tween 20 (PBST) containing 2% nonfat dry milk (Bio-Rad Laboratories) before incubation with antibodies. Antibodies were incubated with membranes in the presence of PBST containing 2% nonfat dry milk for 1 h at room temperature. Excess primary antibody was removed with three 10-min washes of PBST. Secondary antibodies were incubated with membranes for 1 h at room temperature followed by three 10-min washes with PBST to remove excess antibody. Proteins were visualized on the membrane by exposure to Western lightning chemiluminescent reagent. The Runx2 antibody was a gift from Y. Ito (National University of Singapore, Singapore). Total p53 was detected using a 50:50 mix of ab-1 and -3 (Oncogene Research Products). Activated p53 was detected using an anti-pSer15 p53 antibody (Cell Signaling Technology). All antibodies were used at a concentration of 50 ng/ml.

The classic model of retrovirology describes virion budding and virion–cell fusion as virus-mediated processes that do not follow any normal cell biological pathway (; ). However, animal cells are capable of secreting proteins and lipids in exosomes. Exosomes are single membrane organelles that are secreted from the cell and have a similar size and the same topology as retroviral particles (, , ; ; ; ; ). Exosomes share many additional properties with retroviral particles, including similar lipid and protein compositions. These similarities have raised the possibility that retroviruses are “Trojan exosomes” that use the exosome biogenesis pathway for virion biogenesis and can also use an exosome uptake pathway as an alternative, Env-independent pathway for infecting neighboring cells (). One approach to studying the possible relationship between exosomes and retrovirions is to directly compare the composition of exosomes with retrovirions released by a given cell type. The only study that directly compares exosomes with retrovirions was performed with human macrophages, and it identified several similarities between the host cell proteins that were included in or excluded from both exosomes and HIV particles (). Another approach is to determine whether cells bud exosomes and retrovirions from similar or distinct locations of the cell. Several studies in macrophages demonstrated that exosome-like vesicles and HIV particles are formed in a two-step process. First, outward vesicle budding (outward = away from the cytoplasm) at the limiting membrane of endosomes generates discrete vesicles, and second, fusion of vesicle-laden endosomes with the plasma membrane secretes vesicles into the extracellular milieu (; ; ; ). These macrophage studies fit the common conception that exosome biogenesis occurs only by outward budding at endosomal membranes followed by fusion of vesicle-laden endosomes with the plasma membrane (; ; ; ). In fact, the dearth of evidence for exosome budding at the plasma membrane and the budding of HIV from the plasma membrane of T cells (; ) have been cited as reasons to doubt that retroviruses are a variant form of exosomes (). In this study, we examined the cellular distribution of exosomal proteins and lipids in the human CD4+ T cell line Jurkat and found that exosomal markers colocalized primarily at discrete domains of the plasma membrane. We also observed presumptive outward budding intermediates at the plasma membrane of Jurkat T cells and found that these structures were associated with exosomal markers. Moreover, Jurkat T cells appear to treat HIV Gag as an exosomal cargo molecule, sorting it to domains of plasma membrane that are enriched in exosomal markers and secreting it from the cell in exosomes. Human CD4 T cells and certain T cell lines, such as Jurkat T cells, are permissive for HIV infection and bud HIV and HIV Gag from their plasma membrane (). To determine whether human T cells also bud exosomes from their plasma membrane, we first tested for the presence of known exosomal marker proteins on exosomes that are secreted from Jurkat T cells, which were grown under our laboratory conditions. Jurkat T cells were adapted to growth in serum-free medium, and exosomes were collected from the medium by filtration and differential centrifugation. The exosome-containing pellet was resuspended, subjected to sucrose density gradient centrifugation, and fractions across the gradient were assayed for density by refractometry and for specific proteins by immunoblotting. CD81 has previously been detected on the limiting membrane of endosomes (), is particularly enriched in exosomal membranes (; ), and was enriched in exosomes secreted by Jurkat T cells grown in serum-free medium (, top). CD63 has also been detected on the limiting membrane of endosomes () and exosomal membranes (; ; ; ; ; ; ). CD63 was abundant in the peak fractions of Jurkat-derived exosomes, although perhaps not as enriched as CD81 (). In contrast to CD81 and CD63, the plasma membrane protein CD45 is excluded from T cell–derived exosomes () and was also excluded from exosomes secreted by Jurkat T cells (, bottom). Interestingly, CD45 is also excluded from T cell–derived HIV virions (; ). K562 cells also secreted exosomes that were enriched in CD81 and CD63 but lacked CD45 (). We next examined the subcellular distribution of CD81 and CD63. Immunofluorescence analysis of fixed and permeabilized Jurkat cells revealed that CD81 was located predominantly at the plasma membrane where it was enriched at discrete domains (), a distribution that was consistent with what we observed for CD81 at the surface of fixed, unpermeabilized Jurkat cells (). The subcellular distribution of CD63 in permeabilized cells was a bit different, with much of the protein residing in internal organelles, although some CD63 appeared to be at the cell periphery (). Immunofluorescence analysis of fixed and unpermeabilized Jurkat cells confirmed that the CD63 present at the cell periphery was exposed to the extracellular milieu (). Moreover, double-label immunofluorescence analysis of fixed, unpermeabilized Jurkat cells revealed that CD81 and CD63 colocalized at discrete domains of the T cell surface (). In contrast, CD45 and 10 other plasma membrane proteins were distributed evenly over the surface of these cells (supplemental material and Fig. S1, available at ). Similar results were observed in K562 cells (not depicted). and previously established that N-Rh-PE (1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine--[lissamine rhodamine B sulfonyl]), a fluorescent phosphatidyl ethanolamine (PE) analogue, is taken up at the plasma membrane of cells, sorted to endosomes, and secreted in exosomes. When added to human T cells, N-Rh-PE was taken up at the plasma membrane (), and, by 1 h, much of this lipid had been endocytosed (). However, by 3–4 h, a significant portion of the cell-associated N-Rh-PE was present at discrete domains of the plasma membrane. This distribution could be detected up to a day () or more after labeling. In addition, N-Rh-PE was secreted from the cells on exosomes, which could be captured, bound to coverglasses, and visualized by fluorescence microscopy (). Double-label fluorescence microscopy experiments revealed that the plasma membrane domains enriched for N-Rh-PE were the same as those enriched for CD81 () and CD63 (). Jurkat T cells sorted a fluorescein-labeled PE derivative, N-F-PE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine--[carboxyfluorescein]), in much the same way, as shown by its colocalization with surface CD63 (). When Jurkat T cells were pulse labeled with N-Rh-PE, incubated for 1 d, pulse labeled with N-F-PE, and examined 16 h later, one could detect the colocalization of these exosomal lipids at discrete domains of the plasma membrane () and their secretion on overlapping populations of exosomes (). Fixed, unpermeabilized, N-Rh-PE–labeled Jurkat T cells were also decorated with antibodies specific for the extracellular domain of Lamp1. This protein also colocalized with N-Rh-PE, although not completely, as Lamp1 was also detected at additional points along the plasma membrane (). Similar results were obtained in K562 cells (not depicted). Approximately one fourth of Jurkat T cells possess a single, large domain enriched for these exosomal markers. An even greater proportion of the cell population, approximately two thirds, possess multiple plasma membrane domains enriched for exosomal proteins. The presence of multiple domains was evident in cells stained for CD81 and CD63 (such as the middle cell in and the upper cell in ) and is shown here for N-Rh-PE–labeled cells stained for surface CD63 (). We failed to detect such domains in ∼10% of the cells (for quantitation, see supplemental material). Similar results were observed in primary human T cells (). To determine whether these domains remain for just a few seconds or persist for longer periods of time, we examined the distribution of N-Rh-PE in live cells by time-lapse microscopy. The plasma membrane domains marked by this lipid appeared to be semistable, with many of them persisting for periods of 15 min or more (). Similar results were observed in K562 cells (). However, these domains were not static entities. Some domains moved laterally in the plasma membrane, whereas others formed or dispersed over time, typically by the movement of smaller domains toward or away from one another but also by movement of intracellular organelles to the cell surface. The presence of discrete domains in the T cell plasma membrane raises the question of how these domains are organized and how the cell retains lipids such as N-Rh-PE and N-F-PE at these domains. To determine whether membrane lipid composition plays a role in retaining N-Rh-PE at these domains, we pulse labeled Jurkat T cells with N-Rh-PE for 1 h, incubated them for 16 h, exposed them to methyl-β-cyclodextrin for 30 min to deplete sterols and glycosphingolipids from the plasma membrane, removed the methyl-β-cyclodextrin, and followed N-Rh-PE distribution over time. Before the addition of methyl-β-cyclodextrin and after mock treatment, N-Rh-PE is concentrated at discrete domains of the plasma membrane (). Incubation with methyl-β-cyclodextrin for 30 min released a significant amount of N-Rh-PE from these domains (). This phenotype was reversible, as removal of the methyl-β-cyclodextrin allowed Jurkat T cells to redirect N-Rh-PE to discrete domains of the plasma membrane (). Thus, it appears that lipid composition is important for the retention of N-Rh-PE at discrete domains of the plasma membrane. Although most Jurkat T cells possess one larger or several small discrete domains, there is no reason to presume they are present in the plasma membrane of all cells. Monocytic cell types, including macrophages and dendritic cells, are one group of cells that might be expected to lack such domains. These cells appear to use the delayed mode of exosome biogenesis, in which vesicles bud into endosomes, accumulate, and are secreted as exosomes when the endosomes fuse with the plasma membrane. To test for the presence of discrete domains on the plasma membrane of monocytic cells, we pulse labeled the human monocytic cell line U937 with N-Rh-PE and examined N-Rh-PE distribution 1 d later, counterstaining the cells with antibodies specific for the plasma membrane protein CD43. Unlike Jurkat T cells, which direct N-Rh-PE to discrete domains of the plasma membrane, U937 monocytes retain N-Rh-PE in discrete endosomes (). To determine whether these domains of the T cell plasma membrane share any functional similarities with bona fide endosomal membranes, we examined the subcellular distribution of a class E vacuolar protein sorting (VPS) protein, AIP1/VPS31. Class E VPS proteins are thought to mediate outward vesicle budding from the limiting membrane of endosomes, and AIP1/VPS31 has been localized to endosomes and implicated in a late stage of protein sorting to outward budding vesicles in yeasts and in animal cells (; ). We observed that a human AIP1/VPS31-YFP fusion protein was present at both internal organelles and at discrete domains of the T cell plasma membrane (). A similar distribution was observed for an AIP1/VPS31-DsRed fusion protein (). Another class E VPS protein, VPS4, is an ATPase that is also required at a late stage in protein sorting to outward budding vesicles. Human VPS4B has previously been described as largely cytoplasmic (), and we found that a VPS4B-DsRed fusion protein was localized to the cytoplasm of Jurkat T cells (not depicted). However, ATPase-defective forms of VPS4 have been reported to concentrate on endosomal membranes (; ; ; ), and we detected human VPS4B/K180Q-DsRed at the same domains of the plasma membrane as CD63 () as well as at internal organelles, presumably endosomes (). Similar results were observed in K562 cells (). An antibody to yet a third class E VPS protein, TSG101/VPS23, also decorated these domains of plasma membrane in some N-Rh-PE–labeled Jurkat T cells (). However, we were unable to detect TSG101-GFP fusion proteins at such domains (not depicted). The colocalization of outward vesicle budding factors and exosomal proteins at discrete domains of the plasma membrane indicates that these domains might be sites of outward vesicle budding (i.e., sites of exosome biogenesis). To explore this possibility, we examined Jurkat T cells and K562 cells by electron microscopy. Most thin sections of these cells lacked presumptive outward vesicle budding intermediates. However, we did occasionally detect such structures (). Using HRP-conjugated secondary antibodies and 3,3′-DAB–based cytochemistry for HRP activity (which deposits an electron-dense reaction product in the vicinity of the antigen), we observed CD63 enrichment at discrete domains of the plasma membrane (), one of which contained a presumptive outward budding intermediate (). Such domains were more common than presumptive outward vesicle budding intermediates but were still fairly rare. Similar results were obtained using anti-CD81 antibodies and 6-nm gold-conjugated secondary antibodies. Most regions of the plasma membrane were devoid of CD81 label (), but clusters of CD81 spanning 1–3 μm of the plasma membrane () could be detected in ∼5% of sections (for quantitation of immunogold labeling, see supplemental material). For those rare cell sections that possessed presumptive outward vesicle budding intermediates (∼1%), CD81 labeling was detected on or near ∼50% or more of the buds (). N-Rh-PE–labeled Jurkat T cells were also examined, this time using rabbit polyclonal antibodies specific for rhodamine and mouse monoclonal antibodies specific for CD81 followed by 18-nm gold-conjugated anti–mouse IgG secondary antibodies and 6-nm gold-conjugated anti–rabbit IgG secondary antibodies. Both markers colocalized on presumptive outward budding intermediates (). Like CD81, N-Rh-PE was detected at discrete domains of the plasma membrane (). In contrast to what we observed for the exosomal markers CD81 and N-Rh-PE, labeling Jurkat T cells for the nonexosomal protein CD45 decorated the entirety of the plasma membrane (not depicted). We next compared the sorting of HIV Gag in Jurkat T cells to that of exosomal proteins and lipids. HIV Gag mimics the budding of the intact retrovirus (; ; ; ; ; ), and GFP-tagged forms of HIV Gag can be used to follow its biogenesis (, ). When expressed in human T cells, HIV Gag-GFP, like HIV Gag, is detected at discrete domains of the plasma membrane (; ; ). This curiously focal distribution appears to reflect the trafficking of HIV Gag to endosome-like domains of the plasma membrane, as demonstrated by its colocalization with surface CD81 (), surface CD63 (), N-Rh-PE (), and AIP1/VPS31 (). Similar results were obtained in K562 cells, which also colocalized VPS4B/K180Q-DsRed with HIV Gag-GFP (). It should be noted that HIV Gag was also present at flanking regions of the plasma membrane, raising the possibility that HIV Gag might reach endosome-like domains by lateral sorting within the plasma membrane. Moreover, those Jurkat T cells that contained multiple endosome-like domains of plasma membrane sorted HIV Gag to each of these domains (). The trafficking of HIV Gag-GFP to endosome-like domains of the T cell plasma membrane indicates that it buds from these domains just as CD81, CD63, and N-Rh-PE bud from these domains. To test this directly, we examined Gag-expressing T cells and K562 cells by immunoelectron microscopy. During budding, HIV Gag oligomerizes to form an electron-dense protein core, which marks sites of Gag budding. Exosomal markers were detected using mouse monoclonal primary antibodies and gold-conjugated secondary antibodies. Although most of the Jurkat T cell plasma membrane was devoid of budding intermediates, we could occasionally detect nascent HIV Gag buds at the surface of Jurkat T cells () in the vicinity of the exosomal marker CD81 (). Surface labeling of K562 cells revealed that the exosomal marker CD63 was also present on or near nascent HIV Gag budding intermediates (). Similar results were observed in 293T cells expressing a proviral clone of HIV (). Another way to probe the relationship between exosomes and retroviruses is to collect exosomes secreted by Gag-expressing cells and test them for the presence of Gag. HIV Gag-GFP was expressed in Jurkat T cells that had been pulse labeled with N-Rh-PE, and exosomes were collected from the medium of these cells, bound to poly--lysine–coated coverglasses, and examined by fluorescencemicroscopy. Exosomes were detected on the basis of N-Rh-PE fluorescence () and HIV Gag-GFP fluorescence colocalized with this exosomal marker (). A parallel exosome preparation from HIV Gag–expressing cells contained discrete, single membrane organelles of ∼100-nm diameter, which is typical of exosomes, and a subset of these exosomes contained prominent electron-dense cores of oligomerized HIV Gag (). Further purification of a similarexosome preparation by sucrose density gradient centrifugation revealed that Jurkat T cells secreted HIV Gag in vesicles that cofractionated with the exosomal markers CD81 and CD63 (). Immunoblot analysis for CD45 demonstrated that these exosome fractions were significantly depleted of contaminating plasma membrane microvesicles, which can potentially contaminate exosome preparations (; ; ). HIV Gag also copurified with exosomes that were secreted by K562 cells (), providing additional evidence that human cells secrete HIV Gag from the cell in exosomes. #text Jurkat T cells were maintained in a serum-free medium (AIM-V; Invitrogen). K562 and primary human T cells were grown/maintained in RPMI 1640 supplemented with 10% FCS. N-Rh-PE and N-F-PE were obtained from Avanti Polar Lipids, Inc. The plasmid that expresses VPS4B(K180Q)-DsRed was provided by W. Sundquist (University of Utah, Salt Lake City, UT). The HIV Gag expression vector p96ZM651gag-opt was obtained from Y. Li, F. Gao, and B.H. Hahn through the National Institutes of Health (NIH) AIDS Research and Reference Reagent Program. The Gag-GFP expression vector was generated by amplifying the optimized HIV Gag ORF from p96ZM651gag-opt using the primers 5′-GAAAGGTACCATGGGCGCCCGCGCCAGCATC-3′ and 5′-CCCGGATCCCTGGCTCAGGGGGTCGCTGCC-3′. The resulting PCR products were digested with Asp718 and BamH1 and inserted between the Asp718 and BamH1 sites of pcDNA3-GFP. A sequence-confirmed clone (pcDNA3-HIVGag-GFP) with no errors was used for the experiments described in this paper. The AIP1/VPS31-YFP and -DsRed expression vectors were generated by amplifying a fragment of the AIP1 cDNA (synthesized by PCR from a human muscle cDNA library using the primers 5′-CAAGGTACCATGGCGACATTCATCTCGGTG-3′ and 5′-CAAGGATCCCTGCTGTGGATAGTAAGACTG-3′), cleaving the fragment with Asp718 and BamH1, and inserting it between the Asp718 and BamH1 sites of pcDNA3-YFP and pcDNA3-DsRed, respectively. Only sequence-confirmed clones lacking undesired mutations were used for experiments. FITC-conjugated anti–human CD63, FITC-conjugated anti–human CD81, phycoerythrin-conjugated anti–human CD81, and unlabeled monoclonal antibody directed against human CD30 (BerH8) and CD81 (JS-81) were obtained from BD Biosciences. Unlabeled mouse monoclonal antibodies directed against human Lamp1 (H4A3), CD7 (3A1E-12HT), CD43 (DF-T1), CD45 (2D-1), and CD63 (H5C6) and polyclonal rabbit antibodies specific for CD81 were obtained from Santa Cruz Biotechnology, Inc. Rabbit anti-rhodamine antibodies were obtained from Invitrogen. Polyclonal antibodies directed against HIV Gag were generated against purified p24. Secondary antibodies were obtained from Jackson ImmunoResearch Laboratories. Directly labeled anti-CD63 and anti-CD81 antibodies were used in , , , and . Unlabeled primary antibodies were used in all other experiments. 5 × 10 cells were electroporated with 2 μg of plasmid using a Nucleofactor apparatus (Amaxa) and Cell Line Nucleofector Kit V (Amaxa) or by using an electroporator (ECM600; BTX) and standard protocols. Pulse labeling of cells with N-Rh-PE, N-F-PE, or N-biotin-PE involved solubilizing the phospholipid in absolute ethanol, injecting it into cold AIM-V media with a Hamilton syringe to a final concentration of 5 μM, incubating cells in this solution for 1 h at 4°C, washing the cell extensively with cold AIM-V media, and resuspending the cells in AIM-V medium at 37°C. Labeling of transfected cells with N-Rh-PE was performed ∼14 h after transfection. Cells labeled with N-Rh-PE or N-F-PE were fixed ∼20 h after labeling unless otherwise stated. Treatment with 15 mM β-cyclodextrin was for 30 min at 37°C. For fluorescence microscopy, cells were fixed with 3% PFA in 1× PBS, pH 7.2, for 15 min and washed twice in PBS, all at 20–25°C. For the experiments with permeabilized cells (, and H; and ), fixed cells were incubated in 1% Triton X-100 in 1× PBS, pH 7.2, for 5 min followed by two washes in 1× PBS, pH 7.2. After fixation or fixation and permeabilization, cells were incubated with either labeled primary antibodies (experiments in ) or sequentially with unlabeled primary and labeled secondary antibodies, each diluted in 1× PBS, pH 7.2. After the antibody incubations, samples were washed extensively in PBS. Before viewing, samples were placed in mounting medium (90% glycerol, 100 mM Tris, pH 8.5, and 0.1% paraphenylenediamine). Unless otherwise noted, cells were examined at 20–25°C, and images were obtained using a fluorescence microscope (BH-2; Olympus) equipped with a 60× NA 1.4 lens and captured with a CCD camera (Sensicam QE; Cooke) linked to a MacG4 computer using IPlab software (Scanalytics). Exceptions are listed as follows. Images for (A–D) and 4 (Q–S) were obtained using a confocal fluorescence microscope (LSM510 Meta; Carl Zeiss MicroImaging, Inc.) with a 63× NA 1.4 lens and dedicated imaging software (LSM image browser; Carl Zeiss MicroImaging, Inc.). Images for (E–P), 4 (I–L), 6 (A–L), and 9 (A and B) were obtained using an UltraView Live Cell Imaging system (PerkinElmer) with a 63× NA 1.4 lens (Nikon), a spinning Nipkow disk with microlenses, and UltraView software (PerkinElmer). For live cell imaging (), cells were incubated in growth medium supplemented with 100 mM Hepes on a peltier modulated stage warmer (Instec) maintained at 37°C. Digital images were false colored, converted to Adobe Photoshop TIFF files, adjusted for brightness and contrast, and assembled in Adobe Illustrator. Fluorochromes for labeled primary antibodies are described above. For all experiments using secondary antibodies, FITC and Texas red conjugates were used, and staining patterns were dependent on prior incubation with primary antibodies. False color images correspond to idealized hues for these fluorochromes. For experiments using fusion proteins to fluorescent proteins, images correspond to direct fluorescence of the fusion protein. Images shown are representative of results from a minimum of four replicates of each experiment. Immunogold surface labeling of cells was performed essentially as described previously (). In brief, cells were fixed for 1 h at RT in 2% PFA in 100 mM PO, pH 7.4. Cells were washed, blocked in 10% FCS in 100 mM PO, pH 7.4, for 20 min, and incubated overnight with anti-CD63 antibodies in 100 mM PO, pH 7.4. After washing, cells were incubated for 2 h at RT with donkey anti–mouse 5-nm Au conjugate, washed, and again fixed for 1 h in 2% glutaraldehyde, 2.5% sucrose (wt/vol), and 100 mM cacodylate, pH 7.4. The cells were spun into a tight pellet, fixed in Palade's osmium for 1 h, en bloc stained overnight in Kellenberger's uranyl acetate, dehydrated through a graded series of ethanol, and embedded in epon. 80-nm ultrathin sections were cut on an ultramicrotome (model UCT; Leica) collected onto 400 mesh high transmission nickel grids, post-stained in uranyl acetate and lead citrate, and observed on a transmission electron microscope (EM 410; Philips). Immunoperoxidase labeling was performed essentially as described previously (). In brief, cells were fixed in 4% PFA and contained in 0.1 M phosphate buffer, pH 7.4, for 1 h at RT. Fixed cells were sequentially incubated with primary antibodies, washed, and incubated with HRP-conjugated secondary antibodies, all diluted in 1× PBS. Cells were washed in 0.1 M cacodylate-HCl, pH 7.4, and 7.5% sucrose followed by postfixation with 4% glutaraldehyde in 0.1 M cacodylate- HCl, pH 7.4, for 1 h. Cells were then incubated in DAB reaction buffer (3 mg/ml DAB, 0.01% HO, and 0.1 M Tris-HCl, pH 4) and postfixed in 0.1 M cacodylate-HCl, pH 7.4, 1% reduced OsO, and 1% KFeCN. Cells were dehydrated through a graded series of ethanol and embedded in epon. 80-nm sections were prepared on an ultramicrotome (model UCT; Reichert), collected onto nickel grids, and stained with lead citrate. For cryoelectron microscopy, cells were fixed at 4°C in 100 mM PO containing 2.5% sucrose, pH 7.4, and 4% PFA overnight, washed, and harvested. Cell pellets were cryoprotected in 2.3 M sucrose containing 30% polyvinyl pyrollidone, mounted on aluminum cryopins, and frozen in liquid nitrogen. Ultrathin cryosections were then cut on an ultramicrotome (Reichert) equipped with an FCS cryostage, and sections were collected onto 300 mesh formvar/carbon-coated nickel grids. Grids were washed, blocked in 10% FCS, and incubated overnight with 10 μg/ml primary antibodies; in the case of double indirect labeling, antibodies were mixed. After washing, grids were incubated with gold-conjugated secondary antibodies for 2 h, washed, and embedded in a mixture containing 3.2% polyvinyl alcohol (10 molecular mass), 0.2% methyl cellulose (400 centiposes), and 0.2% uranyl acetate. Sections were analyzed on a transmission electron microscope (Philips), and images were collected with a digital camera (Megaview III; Soft Imaging System). Digital images were adjusted for brightness and contrast and assembled into figures using Adobe Illustrator or Canvas. Control experiments in which primary antibodies were omitted showed no labeling, and primary antibodies used for immunoelectron microscopy were also tested for labeling specificity by immunofluorescence microscopy. Cells were removed from the culture supernatant by centrifugation at 2,000 for 5 min followed by passage of the supernatant through a 0.22-μm filter. The supernatant was then subjected to two to three successive spins at 10,000 for 30 min to remove cell debris and at 70,000 for 1 h to pellet exosomes. Further purification of exosomes was accomplished by resuspending the 70,000 pellet fraction and centrifuging the exosome sample for 16 h at 270,000 through a continuous sucrose gradient. 11 fractions were collected, assayed for density by refractometry, and particulate material was collected by spinning each sample for 1 h at 350,000 . Pellets were resuspended in 1× SDS-PAGE loading buffer, separated by SDS- PAGE, and processed for immunoblotting using specific antibodies (mouse monoclonal antibodies to CD45 and CD63 and rabbit polyclonal antibodies for CD81 and HIV Gag together with HRP-conjugated secondary antibodies to mouse or rabbit, respectively) followed by chemiluminescent detection of HRP-conjugated secondary antibodies. For fluorescence microscopy of exosomes, 70,000 exosome pellet fractions were resuspended in 1× PBS, pH 7.2, applied to poly--lysine–coated coverslips, and imaged on a fluorescence microscope (BH-2; Olympus) at 20–25°C equipped with a 60× NA 1.4 lens, and images were captured using a CCD camera (Cooke) linked to a MacG4 computer using IPlab software. Fig. S1 shows the distribution of other plasma membrane proteins on Jurkat T cells. Supplemental material provides details about the quantitation of immunofluorescence data and immunogold labeling. Online supplemental material is available at .

The myelin sheath is a multilamellar, spirally wrapping extension of the plasma membrane of oligodendrocytes that is essential for rapid impulse conduction in the central nervous system. This specialized membrane exhibits a unique composition with >70% of the dry weight consisting of lipids and the remainder being comprised of a restricted set of proteins, of which most are exclusively found in myelin (; ). The major central nervous system myelin proteins, the myelin basic protein, and the proteolipid proteins (PLPs/DM20) are low molecular weight proteins found in compact myelin that constitute 80% of the total myelin proteins. PLP is a highly hydrophobic protein with four transmembrane domains that interact with cholesterol and galactosylceramide-enriched membranes during its biosynthetic transport in oligodendrocytes (; ; ). To form the myelin sheath, oligodendrocytes must deliver large amounts of myelin membrane to the axons at the appropriate developmental stage of the oligodendroglial and neuronal cell lineage (; ). On the other hand, axons produce signals that regulate the differentiation of oligodendrocytes (; ). This led us to postulate that neuronal signals could be involved in the coordination of the trafficking of myelin membrane in oligodendrocytes. In this study, we show that the transport of PLP in oligodendrocytes is under neuronal control. PLP is initially targeted to late endosomes/lysosomes (LEs/Ls) by using a cholesterol-dependent and clathrin-independent endocytosis pathway. PLP is then redistributed from LEs/Ls to the plasma membrane upon activation by neuronal cells. We provide evidence that this development-dependent regulation of PLP localization occurs by the down-regulation of endocytosis and by the stimulation of exocytosis from LE/L storage sites. To analyze the localization of PLP in immature oligodendrocytes, oligodendroglial precursor cells growing on top of a layer of astrocytes were shaken off and cultured for 3 d to induce the expression of PLP. By confocal immunofluorescence microscopy, extensive colocalization of PLP and Lamp-1, a marker for LEs/Ls, was observed () as reported previously (; ). The same striking colocalization of PLP and Lamp-1 was observed in an immortalized oligodendroglial precursor cell line, Oli-neu. Fusion of either a myc tag or EGFP to PLP did not affect the LE/L targeting of PLP (). To obtain further support for the localization of PLP to LEs/Ls in immature cells, we used a spontaneously transformed oligodendroglial precursor cell line, OLN-93. When these cells were incubated for 5 h with rhodamine–dextran followed by a 2-h chase or were treated with LysoTracker red DND-99 to stain for LEs/Ls, colocalization with PLP was observed ( and not depicted). Staining with filipin revealed a partial colocalization of PLP with cholesterol in LEs/Ls (). To resolve the ultrastructure of the PLP-containing organelles, we performed immunoelectron microscopy (). Both endogenous PLP and PLP-EGFP colocalized with Lamp-1 in vacuolar structures that contained abundant lumenal multilamellar and/or multivesicular membrane arrays. To determine whether the subcellular localization of PLP is influenced by the presence of neuronal cells, oligodendroglial progenitor cells were added to a neuronal cell culture. Similar to the cultures without neurons, oligodendrocytes started to express PLP during the first 2 d in culture, and an accumulation of PLP in LEs/Ls was observed in ∼90% of the cells after 3 d in culture (). However, in marked contrast to cultures without neurons, we observed that PLP disappeared from LEs/Ls 2 d later (only ∼20% of the cells showed an accumulation of PLP in LEs/Ls) in the presence of neurons (). Also, the staining of Lamp-1–containing structures decreased in intensity after PLP removal. To follow the developmental regulation of PLP trafficking in Oli-neu cells, we produced PLP-EGFP–stably expressing cell lines. Fusion of EGFP to PLP did not interfere with transport to the cell surface, as indicated by the positive staining of transfected Oli-neu cells with O10 mAb. This anti-body recognizes a conformation-dependent epitope of PLP on the surface of living cells (). In addition, transfection of primary myelinating oligodendrocytes confirmed that PLP-EGFP is transported to myelin (Fig. S1, available at ). When PLP-EGFP–expressing Oli-neu cells were added on top of a neuronal culture, a dramatic change in the localization of PLP was observed (). Quantitative analysis showed that 71.5% of PLP colocalized with Lamp-1 in Oli-neu cells alone, whereas only 11.5% of PLP colocalized with Lamp-1 in cells that had been in coculture with neurons for 2 d (). Moreover, surface staining of living cells at 4°C with O10 mAb showed that the majority of PLP-EGFP was located at the plasma membrane in cells cultured with but not without neurons (). To test whether the localization of PLP shows the same developmental regulation in vivo, we performed immunohistochemistry on brain sections of young (postnatal day [P] 7) and adult mice (P60). Significant colocalization of PLP and Lamp-1 was only observed in cells of P7 mice but not in sections prepared from adult mice (). Analysis of the sections indicated that the colocalization of PLP and Lamp-1 was increased >20-fold in P7 mice as compared with adult mice. Together, our data demonstrate that PLP disappears from LEs/Ls upon oligodendroglial maturation and emerges at the surface of the oligodendrocyte in a process that is dependent on the presence of neuronal cells. There are several possibilities to explain our results. One possibility is that less PLP is transported into and/or more PLP is transported out of LEs/Ls in the presence of neurons. An alternative explanation is that the degradation of PLP in lysosomes increases during development. To test the latter hypothesis, we performed pulse-chase experiments. Primary oligodendrocytes (for 2 d in culture) grown with or without neurons were metabolically labeled with [S]methionine/cysteine for 18 h and chased for 2 d. Hence, the metabolic labeling was performed at the stage of development (day 2–3) when PLP accumulates in LEs/Ls. The cells were chased up to the developmental stage (day 5) when PLP is almost completely removed from LEs/Ls in cells cultured with neurons but not in cells cultured without. Nevertheless, we did not observe any significant differences in the amount of labeled PLP and its alternatively spliced isoform DM20 (). Thus, differential proteolysis of PLP does not seem to be the underlying reason for the removal of PLP from LEs/Ls. Therefore, it is more likely that neuronal signals influence the transport of PLP into and/or out of LEs/Ls in oligodendrocytes. Next, we determined whether endocytosis accounts for the transport of PLP to LEs/Ls and, if so, how its endocytic trafficking is regulated. To block endocytosis, we transiently transfected Oli-neu cells with a mutant form of dynamin-II that is defective in GTP binding (K44A; ) and PLP-myc. We found that dynamin-II (K44A) reduced the colocalization of PLP with Lamp-1 and, at the same time, increased the fraction of PLP at the cell surface (), suggesting that endocytosis is required for the transport of PLP to LEs/Ls. A reduction of the intracellular accumulation of PLP was also observed when dynamin K44A was expressed in OLN-93 cells (not depicted). To gain more insight into the endocytosis process of PLP, the endocytic uptake of PLP was determined by additional antibody uptake experiments using the O10 mAb. To determine whether clathrin function is involved, we either depleted cells of potassium to disrupt the formation of clathrin-coated pits or used the dominant-negative mutant of Eps15 (EΔ95/295; ). Despite that EΔ95/295 and K depletion had no significant effect on the uptake of O10, it led to the reduction of clathrin-mediated transferrin–rhodamine internalization (). In addition, intracellular accumulation of PLP was not reduced when EΔ95/295 was coexpressed in OLN-93 (not depicted). Furthermore, we did not detect a significant colocalization of PLP and clathrin on the surface of OLN-93 (colocalization was ∼5%; ) or primary oligodendroglial cells (Fig. S2; available at ). These results strongly suggest that the internalization of PLP occurs by a clathrin-independent endocytosis pathway. Most clathrin-independent endocytosis pathways are sensitive to cholesterol depletion or actin depolymerization (). Therefore, we used methyl-β-cyclodextrin (mβCD) to selectively extract cholesterol from the cell surface and latrunculin A to prevent actin polymerization. Treatments with either mβCD or latrunculin A led to an almost complete inhibition of O10 uptake (). The conditions of the cholesterol depletion experiments were established so that clathrin-dependent endocytosis was not affected as evaluated by the uptake of transferrin–rhodamine. Some clathrin-independent endocytosis pathways require dynamin, whereas others are independent of dynamin function (; ; ; ; ). The uptake of O10 was clearly reduced by interfering with dynamin function (), which is consistent with the redistribution of PLP to the cell surface by dynamin-II (K44A; ). Because both cholesterol and dynamin are essential for caveolar-dependent uptake, we compared the localization of PLP to caveolin-1 and GFP-caveolin. Caveolin-1 and GFP-caveolin were detected in punctate arrays on the plasma membrane and on intracellular compartments, but no colocalization with PLP was seen (Fig. S2 and not depicted), suggesting that caveolae are not involved in the endocytosis of PLP. The Rho family of small GTPases differentially regulate nonclathrin and noncaveolar endocytosis pathways. Although cdc42 is involved in the endocytosis of glycosyl-phosphatidylinositol–anchored proteins by a pinocytic pathway to recycling endosomes (), rhoA has been implicated in the dynamin-dependent uptake of interleukin 2 receptor to LEs/Ls (). When O10 uptake experiments were performed with dominant-negative mutants of either cdc42 or rhoA, we observed a significant reduction of internalization when the function of rhoA but not cdc42 was inhibited (). In summary, our results show that OLN-93 cells use a clathrin-independent but cholesterol-dependent endocytosis pathway that requires a functional actin cytoskeleton and the rhoA GTPase. To test whether the capacity for endocytosis of PLP changes after contact with neurons, we added OLN-93 cells to neuronal cultures and compared the uptake of O10 into cells that were cultured without neurons. We observed a significant reduction in the internalization of PLP in cells in coculture as compared with cells cultured without neurons (). We also analyzed the reduction of cell surface PLP upon various times in culture with or without neurons using the O10 mAb internalization assay. The amount of O10 mAb remaining at the cell surface was analyzed by incubating cells with [I]-labeled secondary antibody at 4°C. We found that PLP was cleared more efficiently over time from the surface of OLN-93 cells that were cultured without neurons as compared with cells in coculture. Together, these results indicate that neurons reduce the endocytosis of PLP (). Reduction of endocytosis appears to be one reason why PLP disappears from LEs/Ls after contact with neurons. Another event that could simultaneously contribute is the increased resorting of PLP to the plasma membrane by retrograde transport from LEs/Ls. There are many examples (e.g., wound healing, cytotoxic lymphocyte killing, major histocompatibility complex [MHC]–II processing, and melanin secretion) that show that lysosomes are not merely degradative dead ends but are able to store and release proteins in a regulated fashion (). To analyze the putative exocytic trafficking of PLP from LEs/Ls, we performed live cell imaging experiments with LysoTracker in PLP-EGFP–expressing Oli-neu cells. There was an almost complete colocalization of PLP-EGFP and LysoTracker () similar to the observation in OLN-93 cells. Analysis of the dynamics of PLP-EGFP/LysoTracker vesicles in Oli-neu cells revealed that most vesicles were clustered perinuclearly and did not show any significant movement ( and Video 1, available at ). Next, we investigated whether the movement of the vesicles changes as a result of the presence of neurons. When live cell imaging experiments were performed shortly (6–12 h) after the addition of Oli-neu cells to neuronal cultures, an extensive colocalization of PLP-EGFP and LysoTracker () was still observed. However, the movement of these vesicles was markedly increased. Two pools of vesicles could be distinguished: a perinuclear, immobile pool and a peripheral pool of highly mobile vesicles ( and Video 2). Both pools of vesicles colocalized with Lamp-1. Quantitative analysis revealed that ∼29% of the vesicles were mobile and exhibited a mean speed of ∼0.56 μm/s (). The pool of perinuclear vesicles not only decreased in size (), but the individual vesicles also became smaller with increasing time in coculture (∼33% decrease after 16 h of coculture; reduction from 0.96 ± 0.2 μm to 0.63 ± 0.19 μm; = 96). In several cases, PLP-EGFP and LysoTracker-filled vesicles emanated from larger perinuclear vesicles and moved radially toward the plasma membrane at the cell periphery (). To analyze the behavior of the peripheral vesicle pool, we used total internal reflection fluorescence microscopy (TIRFM). TIRFM allows the selective illumination of a region within a 70–120-nm distance of the plasma membrane as the excitatory evanescent field decays exponentially from the interface of the cell membrane with the coverslip. We observed PLP-EGFP and LysoTracker-containing vesicles within the 100-nm vicinity of the plasma membrane in living Oli-neu cells that were cocultured with neurons (). In contrast, no PLP-EGFP and LysoTracker-containing vesicles were observed in the proximity of the plasma membrane when cells were cultured without neurons. To determine whether the acidic vesicles fuse with the plasma membrane, we used time-lapse TIRFM imaging. The cells we studied contained 6 ± 2.4 vesicles per 100-μm area of the plasma membrane. The vesicles that lost their fluorescence during the time of observation fell into two groups. We found vesicles moving in and out of the evanescent field without fusing and vesicles fusing with the plasma membrane (, top). Fusion was defined by the loss of vesicular fluorescence and the concurrent lateral spread of the released dye into the medium (). We detected one to two fusion events/minute per cell at the plasma membrane. When Oli-neu cells were cultured without neurons, we did not observe any fusions in agreement with the absence of PLP-EGFP and LysoTracker-containing vesicles at the plasma membrane. Because the redistribution of PLP from LEs/Ls to the surface of the plasma membrane was only observed in Oli-neu cells grown in the presence of neurons, neuronal signals are likely to activate this pathway. We wanted to determine whether this neuronal signaling is transferred as a soluble factor or is a consequence of direct cell-to-cell contact. Oli-neu cells were either directly added on top of a neuronal culture or placed on a separate coverslip to prevent cell contact, allowing diffusible factors to reach the cells. We found that diffusible factors were sufficient to redistribute PLP from LEs/Ls to the surface of the cell (). To analyze the signals involved, we treated Oli-neu cells cultured in the presence of neurons with various pharmacological kinase inhibitors. Colocalization of PLP-EGFP and Lamp-1 increased markedly when cocultures were treated for 1 d with Rp-cAMPs to inhibit protein kinase A (). In contrast, treatment of cocultures with db-cAMP, a protein kinase A agonist, promoted the localization of PLP to the plasma membrane (). Quantitative analysis of time-course experiments showed that db-cAMP accelerated the redistribution of PLP to the surface of the cell (). To test whether similar effects were observed in Oli-neu cells cultured without neurons, we treated Oli-neu cells for 1 d with db-cAMP and quantified the amount of PLP-EGFP in LEs/Ls and on the surface of the cell. In cells treated with db-cAMP, more PLP was found on the surface of the cell, whereas, at the same time, the fraction within LEs/Ls decreased (). Furthermore, a peripheral pool of highly mobile PLP-EGFP/LysoTracker-containing vesicles was observed by live cell imaging (). These data suggest that cAMP-dependent signaling is part of the developmental switch that is triggered by neurons, leading to the redistribution of PLP from LEs/Ls to the surface of the plasma membrane. Our data demonstrate that the trafficking of PLP in oligodendrocytes is under neuronal control. PLP is initially targeted to LEs/Ls by using a cholesterol-dependent and clathrin-independent endocytosis pathway. The situation changes dramatically upon receiving maturation signals from neurons. PLP is then redistributed from LEs/Ls to the plasma membrane. We provide evidence that the developmental regulation of PLP localization occurs by the down-regulation of endocytosis and by the transport from LEs/Ls to the cell surface. The regulation of the transport of PLP is strikingly similar to the trafficking of MHC-II in dendritic cells (). Immature dendritic cells have a high rate of endocytosis and target MHC-II to lysosomes (). After exposure to inflammatory mediators, the endocytosis of MHC-II is reduced, and the transport of MHC-II from lysosomes to the cell surface is triggered (; ; ). The transport pathway of PLP from LEs/Ls can also be related to the release of secretory lysosomes from hematopoietic cells. However, unlike the classic secretory lysosomes that are specialized to release luminal content, oligodendrocytes mainly transport membrane, and this may occur without significant extracellular release of lysosomal content, as is the case for dendritic cells (; ). For oligodendrocytes, LE/L compartments may be particularly useful as storage compartments, as they are able to harbor large amounts of membrane in a multilamellar and multivesicular fashion for myelin biogenesis. In most cells, however, the majority of molecules that localizes to internal vesicles of the endosomal system are destined for lysosomal degradation. This raises the question of how PLP survives in an environment where protein degradation usually occurs. One possibility is that immature oligodendrocytes have specialized LEs/Ls with low proteolytic capacity. Our unpublished observation that the vesicular stomatitis virus glycoprotein accumulates undegraded in LEs/Ls of Oli-neu cells supports this notion. In dendritic cells, for example, lysosomal proteolysis is regulated in a developmentally linked fashion (; ). Another possibility is that PLP is poorly degradable and, therefore, accumulates within LEs/Ls. A second issue is how PLP escapes from a compartment associated with the limited capacity for membrane recycling. Previous work has provided evidence that not all intralumenal membranes of LEs/Ls are destined for lysosomal degradation. It has been suggested that some vesicles may undergo back fusion with the limiting membrane, and, in some instances, this membrane is sorted via tubulovesicles to the plasma membrane (; ; ; ; ). Whether PLP is sorted by back fusion and tubules to the surface of oligodendrocytes are issues that have to be addressed in future studies. It is important to note that the accumulation of PLP at the surface of the plasma membrane after the receipt of maturation signals most likely reflects the contribution of multiple factors. Our finding that the endocytosis of PLP is reduced after receiving signals from neuronal cells suggests that the regulation of endocytosis may play an essential role in this process. It will be interesting to elucidate the molecular mechanisms of how neurons control the rate of endocytosis in oligodendrocytes. One attractive possibility is that the endocytic activity is controlled by the RhoA GTPase. Importantly, not all myelin components were found to be internalized into LEs/Ls. Although PLP and cholesterol resided in LEs/Ls, myelin basic protein and galactosylceramide were mainly found in or at the plasma membrane (unpublished data). The differential compartmentalization of myelin components before the onset of myelination might be a mechanism to prevent premature and inappropriate assembly. Our results suggest that an external soluble factor regulates myelin membrane assembly by controlling the trafficking of PLP to and from the surface of the cell. Among the many potential candidates are soluble mediators such as neurotrophins, neuregulin, or adenosine that can now be tested with the described experimental system (; ). Together, our findings reveal an unexpected and novel role of LEs/Ls in oligodendrocytes. It provides a striking example of how cell-to-cell communication regulates trafficking to and from a cellular compartment to guide the development of a multicellular tissue. The proposed role of LEs/Ls in myelin biogenesis may help to explain the cellular mechanisms of dysmyelination that is observed in many lysosomal storage diseases. The mutant and wild-type cDNAs of GFP-Eps15 and GFP–dynamin-II were provided by A. Benmerah (Institut Pasteur, Paris, France) and S. Schmid (Scripps Research Institute, La Jolla, CA), respectively. The following primary antibodies were used: myelin basic protein (monoclonal IgG; Sternberger, Inc.), PLP (polyclonal, P6; ), O10 (monoclonal mouse IgM), βIII-tubulin (Promega), neurofilament (monoclonal IgM; Qbiogene), τ (polyclonal; DakoCytomation), myc tag (monoclonal IgG; Cell Signaling), clathrin heavy chain (monoclonal IgG; BD Transduction Laboratories), GFP (Synaptic Systems GmbH), caveolin-1 (monoclonal IgM; BD Biosciences), and Lamp-1 (CD107a, rat monoclonal; BD Biosciences). Secondary antibodies were obtained from Dianova and GE Healthcare. Primary cultures of mouse oligodendrocytes were prepared as described previously (). After shaking, cells were plated in MEM containing B27 supplement, 1% horse serum, L-thyroxine, tri-iodo-thyronine, glucose, glutamine, gentamycine, pyruvate, and bicarbonate on poly--lysine–coated dishes or glass coverslips. Cocultures of neurons and oligodendrocytes were produced by preparing mixed brain cultures from 16-d-old fetal mice that were cultivated for 2 wk, to which the primary oligodendrocytes or Oli-neu cells were added. The mixed brain cultures were prepared at a density of ∼50,000 cells/cm. Cocultures without direct neuron–glia contact were prepared by growing neuronal cultures on glass coverslips, which were placed upside down on a metal ring positioned in a culture dish. Oligodendrocytes were added on an additional coverslip facing upwards. The oligodendroglial precursor cell line, Oli-neu (provided by J. Trotter, University of Mainz, Mainz, Germany), and OLN-93 cells (provided by C. Richter-Landsberg, University of Oldenburg, Oldenburg, Germany) were cultured as described previously (; Richter-Landsberg and Heinrich, 1996). Transient transfections were performed using FuGENE transfection reagent (Roche) according to the manufacturer's protocol. PLP-EGFP was generated by fusing EGFP to the COOH terminus of PLP by gene fusion PCR. The fusion product was cloned into pEGFPN1 vector using the EcoRI–NotI site. Stable cell lines were obtained by the cotransfection of PLP-EGFPNI and pMSCV-hygro (CLONTECH Laboratories, Inc.) followed by the selection of clones by incubation with hygromycin. Immunofluorescence and immunohistochemistry were performed as described previously (). For assaying endocytosis, living cells were incubated with O10 antibody in medium for 30 min at 4°C, washed, and incubated in medium at 37°C for 60 min. The antibody remaining on the surface was removed under low pH conditions in 0.2 M glycine and 0.5 M NaCl, pH 4.5, for 30 min at 4°C. Cells were washed three times in PBS, fixed, and stained by immunofluorescence. Additionally, OLN-93 cells were transiently transfected with PLP and added on a neuronal culture or left alone. Cells were incubated at 4°C for 30 min with O10 antibody in binding medium consisting of HBSS and 10 mM Hepes supplemented with 0.2% BSA 16 h after transfection. After washing, O10 internalization was allowed to continue for 0, 45, and 90 min. O10 antibody remaining at the cell surface was detected with 5–20 μCi/μg [I]-labeled mouse secondary antibody in binding medium for 30 min at 4°C. Next, the cells were washed five times, lysed in 0.2 M NaOH, and the amount of radioactivity was determined by γ counting. Image processing and analysis were performed using Meta Imaging Series 6.1 software (Universal Imaging Corp.). Quantification of colocalization was performed with the colocalization module of the software. Vesicle movement was analyzed by subtracting from each image in a time stack preceding its image. The different image stack thus generated was used to identify vesicle motility events. The velocity of individual vesicles was determined using the Manual Tracking plug-in for ImageJ software (National Institutes of Health). Statistical differences were determined with a test. TIRFM was performed on a custom-built prism-based evanescent field microscope using an HCX Apo L 63× water immersion objective (NA 0.90; Leica; ). Evanescent field excitation was obtained by focusing 488- and 568-nm laser light onto a hemicylindrical prism at 68 and 71° incidence angles, respectively, leading to a field depth of ∼80–100 nm. Images were acquired with a back-illuminated 16-bit CCD camera (Cascade 512B; Roper Scientific) with on-chip charge multiplication. Each pixel corresponded to 0.25 μm in the specimen plane. For analysis of individual fusion events, a small circle was positioned on the vesicular fluorescence, a concentric ring was placed around the circle, and fluorescence intensity was plotted against time. Fusion events were identified by the increase of fluorescence in the central region that spread into the surrounding annulus followed by a sudden decline (; ; ; ). Immuno-EM was performed as described previously (). For metabolic labeling, cells were pulsed with 265 μCi [S]methionine (1 Ci = 37 GBq; GE Healthcare) in methionine and cysteine-free DME for 18 h, and chase was performed for 0 or 48 h. Immunoprecipitation was performed as described previously (). Autoradiographs were scanned and quantified with ScionImage software (Scion Corp.). Values are shown as means ± SD. Statistical differences were determined with a test. Fig. S1 shows that EGFP-tagged PLP is sorted to myelin. Fig. S2 shows the absence of PLP colocalization with clathrin heavy chain or caveolin-1. Video 1 shows the movement of LEs/Ls in Oli-neu cells, whereas Video 2 shows this in Oli-neu cells cultured with neurons. Video 3 shows the fusion of vesicles with the plasma membrane. Online supplemental material is available at .

The transport of GPI-APs in polarized cells was first investigated in the late 1980s in epithelial MDCK cells, a popular model system for polarized membrane trafficking, as they form a well-defined epithelial monolayer with apical and basolateral domains that are separated by tight junctions (). Two seminal papers showed that the GPI anchor serves as a signal for transport to the apical membrane (; ). Soon afterward it was established that newly synthesized GPI-APs are delivered to the apical membrane of MDCK cells directly and do not make a detour to the basolateral membrane, as later observed in hepatocytes (; ). This result indicated that MDCK cells sort GPI-APs at some intracellular site, but the sorting mechanism posed a puzzle. In transmembrane proteins, sorting signals typically reside in the part exposed to the cytoplasm, which enables recognition by the machinery for the generation and transport of vesicles (). GPI-APs, however, have no cytoplasmic part. A landmark study then found that a GPI-AP became associated with glycolipid-enriched detergent-resistant membranes while moving through the Golgi (). This discovery provided the first experimental support for the idea that glycolipids generate apical transport platforms in the Golgi (), a notion that later gave rise to the raft hypothesis (). This initial work produced an appealingly simple model: correct targeting of GPI-APs is ensured by association with lipid rafts, which are destined for the apical membrane. In MDCK cells, sorting takes place intracellularly, probably at the Golgi, and apical delivery occurs along a direct transport route. New results soon disagreed with this model. Epithelial Fisher rat thyroid cells, for instance, send most GPI-APs to the basolateral membrane (). It also became clear that some apical proteins in MDCK cells do not associate with rafts, whereas some basolateral proteins do. Thus, raft association alone is insufficient to dictate apical targeting, and additional mechanisms must be at work. Glycosylation was proposed to govern apical versus basolateral targeting of GPI-APs (). Furthermore, oligomerization seems to be important, as apical but not basolateral GPI-APs form oligomers in the Golgi (). These findings have led to a refinement of the original model. It is now thought that the oligomerization or lectin-mediated cross-linking GPI-APs drive their inclusion into, and perhaps also the generation of, clustered rafts, which then facilitate apical transport of their constituents (; ). The original model was questioned even more fundamentally by a recent high-profile publication from the group of Jennifer Lippincott-Schwartz (). Using live-cell imaging of nonpolarized cells, the authors first provided evidence that GPI-APs and basolateral proteins leave the Golgi in the same transport carriers. They then treated polarized MDCK cells with tannic acid, a fixative used for leather production but here applied for the first time to study polarized membrane trafficking. Striking images showed that GPI-anchored YFP failed to reach the apical membrane after tannic acid had inactivated transport from the basolateral domain. This startling result indicated that GPI-APs need to traverse the basolateral membrane. Finally, the authors demonstrated that GPI-anchored GFP undergoes transcytosis, i.e., that it can be endocytosed from the basolateral membrane and then travel to the apical side. Although it was not shown that this occurs for newly synthesized GFP-GPI, concluded that GPI-APs reach the apical membrane via the basolateral domain. They proposed that GPI-APs are sorted at the basolateral membrane rather than at the Golgi. These results contradicted many previous studies. Earlier live-cell imaging of nonpolarized cells had shown that apical and basolateral proteins leave the Golgi in separate transport carriers (). Biochemical experiments had never detected the bulk of newly synthesized GPI-APs passing through the basolateral domain in polarized MDCK cells. suggested that endocytosis was rapid, so that only few GPI-AP molecules would be present at the basolateral membrane at any given time and might have been missed. This explanation was hard to reconcile with the slow basolateral endocytosis of GPI-APs in MDCK cells (). showed that the transcytosis of GPI-APs has a half-time of >60 min, consistent with their slow exit from recycling endosomes and their slow transcytosis in hepatocytes (; ). However, GPI-APs typically appear at the apical membrane within 15 min of leaving the Golgi, making it difficult to see how they could complete the circuitous journey via the basolateral membrane quickly enough. These discrepancies emphasized the need for more incisive assays and more quantitative data. The contradicting imaging results in particular highlighted that live-cell imaging had to be extended to fully polarized cells, a difficult task given that MDCK monolayers are ∼10 μm thick. The groups of Chiara Zurzolo and Ira Mellman have now independently revisited the issue of the routes taken by newly synthesized GPI-APs (Paladino et al., p. 1023, and Hua et al., p. 1035, this issue). first improved the biochemical analysis of GPI-AP transport in polarized MDCK cells using the same GFP-GPI construct as . To ensure that rapid passage through the basolateral domain would be detected, they treated the basolateral side with low concentrations of trypsin during transport. In this way, GFP-GPI appearing at the basolateral membrane would be cleaved. Cleavage did indeed occur, but only for the small fraction of GFP-GPI missorted to the basolateral side. The vast majority of newly synthesized GFP-GPI reached the apical membrane uncleaved, indicating that it travels along a direct pathway. Interestingly, the authors noticed that the accuracy of targeting increases during polarization. At early stages, the delivery of GPI-APs showed little preference for either surface domain but was restricted mostly to the apical membrane in fully polarized cells. Next, the authors reexamined the results obtained with tannic acid. When transport from the basolateral membrane was inhibited by a brief tannic acid treatment, GFP-GPI still reached the apical membrane. Importantly, when the conditions of were replicated by prolonging the exposure to tannic acid, tight junction integrity was compromised and the segregation of apical and basolateral membranes was abolished. These observations stress that tannic acid should be used with utmost caution. Finally, the authors used spinning-disc confocal microscopy to achieve live-cell imaging of polarized MDCK cells. This allowed them to show that GFP-GPI accumulates only at the apical side after exit from the Golgi, reinforcing the conclusion that GPI-APs take a direct route. approached the problem of visualizing surface transport in polarized MDCK cells more generally and focused on imaging Golgi to plasma membrane trafficking quantitatively. Using laser-scanning confocal microscopy, they were able to follow the transport of fluorescent apical and basolateral marker proteins for ∼30 min. They then derived rate constants for Golgi exit and for passage through the cytosol by quantifying the amount of marker protein present at the Golgi, at the apical or basolateral membrane, and in the intervening cytosol at 1-min intervals. As one way of demonstrating the utility of their system, the authors used GPI-anchored YFP as an apical marker, a construct also analyzed by . The rate-limiting step for YFP-GPI transport was exit from the Golgi, the marker accumulated at the apical but never at the basolateral membrane, and surface arrival was largely complete within 20 min. These data support direct apical transport. Interestingly, the authors noted that YFP-GPI seemed to pass through a subapical kinetic intermediate en route to the apical membrane. The live-cell imaging results were confirmed by antibody uptake experiments. An antibody against YFP was added to the basolateral side during transport so that YFP-GPI passing through the basolateral domain would be labeled. Only a small fraction of the YFP-GPI that appeared on the apical side was antibody bound. A known transcytotic protein, on the other hand, efficiently picked up the antibody during transport to the apical membrane. Also, these observations argue against transcytosis being a major pathway during the biosynthetic delivery of GPI-APs. These reports leave little room for biosynthetic apical delivery of GPI-APs by a transcytotic route in fully polarized MDCK cells. It is not entirely clear why the data of differ so much from all other studies, but the evidence against their conclusions seems overwhelming. So, was this a fairly unremarkable affair after all? We do not think so. Rather, we believe that the debate sparked by has been quite productive in both a technical and a conceptual sense. First, the controversy has stimulated the development of new methodology. Live-cell imaging of polarized MDCK cells holds great promise for resolving other open issues. For example, the question of exactly when and where apical and basolateral proteins separate may now be answered by tracking individual transport carriers. The tracking of early endosomes was recently achieved by fast live-cell imaging of nonpolarized cells (). If imaging were restricted to the supranuclear Golgi region of polarized MDCK cells, multicolor tracking of Golgi-derived vesicles might now be feasible. Second, the new results herald yet more refinements of the original model for GPI-AP trafficking. The observation that polarizing MDCK cells increasingly restrict the delivery of GPI-APs to the apical membrane is reminiscent of the situation in hippocampal neurons, in which the GPI-anchored prion protein is present in all neurites early during differentiation but localizes only to the axon later on (). In the developing embryo, GPI-APs are basolateral in the surface ectoderm but apical in ectoderm-derived internal epithelia (). Such shifts could be achieved by a reorganization of intracellular trafficking. For instance, Fisher rat thyroid cells use a transcytotic mode of apical delivery at early stages of polarization but then switch to direct targeting (). Perhaps epithelial cells more generally abandon the transcytotic mode during differentiation (). n y q u e s t i o n s r e m a i n . T h e r e i s g r e a t v a r i e t y i n t h e s t r u c t u r e o f t h e G P I a n c h o r i t s e l f , b u t w e h a v e l i t t l e c l u e w h y . M a y b e d i f f e r e n t a n c h o r s d e t e r m i n e t o w h a t d e g r e e a G P I - A P c a n b e s o r t e d b y i n c l u s i o n i n t o c l u s t e r e d r a f t s . T h e p o s s i b l e s w i t c h i n g b e t w e e n t r a n s p o r t r o u t e s d u r i n g d i f f e r e n t i a t i o n i s v e r y i n t r i g u i n g , b u t w h a t c o u l d i t b e g o o d f o r ? T o e x t e n d t h e l i s t , n e i t h e r d o w e u n d e r s t a n d h o w G P I - A P s a r e e n d o c y t o s e d b y d i f f e r e n t p a t h w a y s . W e a r e c o n f i d e n t t h a t t h e t r a f f i c k i n g o f G P I - A P s w i l l c o n t i n u e t o p r o v i d e a m p l e o p p o r t u n i t y f o r d e b a t e .

The idea that nuclear actin plays a role in gene transcription is gaining strong support. Two milestone studies published in the 1980s presented circumstantial evidence that actin is implicated in the transcription of protein-coding genes (; ). However, biochemical observations that supported the idea encountered criticism, and the findings were immediately dismissed as resulting from contamination or being artifacts, primarily because of the notorious abundance of actin. Moreover, “traditional” filamentous actin structures that are commonly observed in the cytoplasm, made visible by certain drugs, such as phalloidin, were not revealed in the cell nucleus even by advanced light microscopy methods. Skepticism in the field remained, and the question of whether actin is present in the nucleus was left to drift as an “untouchable” topic (for review see ). 20 yr later, we now agree that β-actin plays a role in gene transcription associated with three different entities: components of ATP-dependent chromatin remodeling complexes (for reviews see ; ), RNP particles (, ), and the three RNA polymerases in the eukaryotic cell nucleus (; ; ; ; ). Mass spectrometry () and immunoreactivity criteria () have suggested that the actin in the nucleus is β-actin. However, the studies that have investigated the function of actin in gene transcription (see the following paragraphs) have not revealed whether the actin molecules involved are in monomeric or polymeric form (for review see ). F-actin was detected in the nuclei of neuronal cells by heavy meromyosin labeling (). Nuclear actin filaments have also been observed attached to the nuclear pore complexes in amphibian oocytes (). Recent immunochemical studies have detected several types of actin structures in the nucleus (). Furthermore, a pool of nuclear polymeric actin has been detected by FRAP in cell lines that express GFP-tagged actin and in cells microinjected with fluorescent actin binding proteins (). However, we do not know whether the actin that is engaged in transcription is monomeric, oligomeric, or filamentous. Much work is currently being performed to reveal the molecular mechanisms that underlie the role of nuclear actin in transcription. Recent investigations have examined different systems and reached somewhat different conclusions. We summarize and discuss these recent studies and describe a general model for the function of nuclear actin in transcription. NM1 was found in the cell nucleus (), which suggested that nuclear actin and myosin may function in a concerted manner. In fact, both proteins have been found in mammalian nucleoli (; ; ). Chromatin immunoprecipitation has been used to study the association of actin and NM1 with rDNA promoters. In spite of technical limitations due to epitope accessibility, the chromatin immunoprecipitation data strongly supports the view that both actin and NM1 are present on actively transcribing ribosomal genes (). Actin is associated with both active and inactive pol-I, whereas NM1 binds to the transcription machinery through the pol-I–specific transcription initiation factor IA (). Based on these data, one could speculate that recruitment of the actin–pol-I complex to the rDNA promoter brings actin and NM1 close to each other. Actin and NM1 would then be able to interact and presumably activate rDNA transcription (). Indeed, posttranscriptional NM1 gene silencing down-regulates preribosomal RNA synthesis, and both actin and NM1 are necessary for in vitro rDNA transcription (). These results suggest that the actin–NM1 complex plays a role in the assembly of a transcription-competent pol-I. An intriguing question is whether NM1 plays a role also in transcription by pol-II. Pol-II transcription is inhibited by a peptide-specific anti-NM1 antibody (), but there is no evidence that NM1 sits on the promoter of protein-coding genes. Recent immunoelectron microscopy studies have shown that NM1 is located at nucleoplasmic transcription sites (), but there is evidence that NM1 is physically connected to transcribed loci via RNA, not via TFII factors, as RNase treatment of living HeLa cells specifically depletes cellular NM1 (). #text Important clues concerning the role of nuclear actin came when nuclear proteins associated with actin were identified. Studies of chromatin regulation induced by antigen-receptor signaling in lymphocytes led to the identification of β-actin and BAF53, an actin-related protein (Arp), as components of the BAF (BRG-associated factor) chromatin remodeling complex (). After this initial observation, many other Swi/Snf-like complexes and histone-modifying factors in different organisms have been found to be associated with actin and/or Arps, and numerous investigations have reinforced the idea that there is a functional link between actin and the regulation of chromatin structure (for reviews see ; ). The molecular mechanisms by which actin and Arps contribute to chromatin remodeling are not fully understood. The ubiquity of actin and Arps in a large variety of chromatin remodeling complexes suggests that these proteins may be involved, directly or indirectly, in many different nuclear processes, including transcription (for review see ). In the case of the mammalian BAF complex, which participates in transcription activation in response to external stimuli, actin is directly associated with the ATPase subunit BRG1. In this context, actin is necessary for the association of BRG1 with chromatin and is required for the full activation of the BRG1's ATPase activity (). As summarized in , actin can potentially be recruited to transcription complexes through different types of interactions. The presence of actin in some chromatin remodeling complexes suggests a possible mechanism for the recruitment of such complexes to active genes (). Studies in the dipteran suggest that actin can also influence chromatin structure through the recruitment of histone-modifying enzymes. Actin binds directly to the nuclear protein HRP65-2 in A synthetic peptide that can disrupt the interaction between actin and HRP65-2 inhibits pol-II transcription in living cells, which suggests that the actin–HRP65-2 interaction is required for transcription in vivo (). A recent study has shown that the inhibitory effect of the peptide can be counteracted by trichostatin A, which is a general inhibitor of histone deacetylases. This suggests that the actin–HRP65-2 interaction is involved in acetylation/deacetylation events. This suggestion is supported by the observation that HRP65-2 binds directly to a histone acetyltransferase (HAT) called p2D10, and the interaction between actin and HRP65-2 is required for p2D10 to associate with the transcribed chromatin (). Moreover, the association of p2D10, actin, and HRP65-2 with chromatin is sensitive to ribonuclease digestion (; ; ), which indicates that these proteins are tethered to the transcribed genes by binding to the nascent transcript. In summary, these findings suggest that actin, HRP65-2, and p2D10 become assembled into nascent pre-mRNPs during transcription. Based on these observations, it has been proposed that the actin–HRP65-2–p2D10 complex is part of the nascent pre-mRNP, and it can travel along the transcribed gene, allowing p2D10 to acetylate histone H3 (). According to this proposal, the actin–HRP65-2–p2D10 complex maintains the chromatin in a transcription-competent conformation (). This model is supported by the observations that H3 acetylation is reduced () and transcription is inhibited () when the interaction between actin and HRP65-2 is disrupted. A recent study by suggests that the cotranscriptional binding of actin to mRNA binding proteins with the concomitant recruitment of chromatin modifiers has been conserved throughout evolution. DNase I affinity chromatography has shown that in human cells there is a specific association between actin and heterogeneous nuclear RNP (hnRNP) U. In vitro reconstitution experiments with purified proteins have shown that the interaction between actin and hnRNP U is direct and that actin binds a short amino acid sequence that is similar to the actin binding motif of HRP65-2. Moreover, actin must interact with hnRNP U for pol-II transcription to take place (). Interestingly, hnRNP U associates with the transcription activator p300/CBP, a potent HAT (). One hypothesis is that the actin–hnRNP U complex recruits HAT activity. The HRP65-2–p2D10 case in is thus similar to the hnRNP U–p300/CBP case in human cells, and we suggest that actin–hnRNP complexes serve as molecular platforms for the recruitment and tethering of chromatin-modifying enzymes in both insect and mammalian cells. The investigations described in this section support the view that actin modulates the structure of the chromatin template during the transcription of class II genes. In this context, actin functions as a platform for protein–protein interactions through a mechanism that requires ongoing RNA synthesis. This mechanism is coupled to transcription elongation and does not seem to require myosin or motor activity. The role of actin in the modulation of chromatin structure does not, however, exclude more direct actions of actin on the basal transcription apparatus. For both pol-I and -II, anti-actin antibodies inhibit the transcription of naked DNA templates in transcription systems containing partially purified polymerases and the necessary transcription factors (; ). Further, β-actin is needed for the full activation of partially purified pol-III preparations from HeLa cells, again when a pure DNA template is used (). These results demonstrate that actin plays a role in transcription even when chromatin is not present. This conclusion is supported by the findings that actin is associated with pol-I (; ), pol-II (; ), and pol-III (). The identities of the actin binding partners in each transcription machinery are not known, but coimmuno precipitation experiments in HeLa cells have identified three pol-III subunits associated with actin (). Two of them, RPABC2 and -3, are present in all three RNA polymerases, and the solution of the crystal structure of pol-II shows that these two subunits are located close to each other at the surface of the polymerase (). suggested that RPABC2 and -3 form an actin binding patch that is common to all three RNA polymerases. On the other hand, coimmunoprecipitation experiments have also shown that actin binds to the COOH-terminal domain of the largest subunit of pol-II (CTD; ), a domain that is absent in pol-I and -III. These findings lead us to propose that actin can interact with the eukaryotic RNA polymerases in two different ways. The first way is general and is probably mediated by RPABC2 and -3, whereas the second way is specific to pol-II. The latter interaction takes place via the CTD, presumably in its phosphorylated state (). We do not yet know how actin acts at a molecular level when it associates with the RNA polymerases. Actin binds to the three RNA polymerases, as described in the previous paragraph, and anti-actin antibodies reduce pol-I, -II, and -III activities to a similar extent in transcription assays in vitro (). The roles of actin may thus be similar in all three systems. Abortive transcription-initiation assays in a pol-I system have demonstrated that anti-actin antibodies do not affect the synthesis of a 3-nt-long abortive transcript, whereas they do inhibit the subsequent elongation of the 3-nt product. This suggests that actin is not involved in initiation per se but in a process that occurs shortly after transcription initiation. This process may be promoter clearance or elongation (). Similar experiments performed in a pol-II system at a lower resolution could not determine the exact stage of transcription at which actin acts, but these experiments confirmed that actin plays an early role in the transcription of class II genes (). Copurification and colocalization studies have led researchers to propose that actin plays a role in the assembly of pol-II preinitiation complexes (). Many questions about the mechanisms by which actin functions in transcription remain unanswered. The simplest model that is compatible with our current knowledge of nuclear actin places actin in three different scenarios. First, actin is a component of some chromatin remodeling factors. Second, actin appears to be necessary in an early step of the transcription process, a step in which actin interacts directly with the transcription apparatus. This step is independent of chromatin structure and may be common for all three RNA polymerases. Third, actin becomes incorporated into the nascent mRNPs, where it is involved in the recruitment of factors that regulate chromatin structure. The molecular function of actin may be the same in all the scenarios. The structure of actin is complex (for review see ), and actin can adopt several molecular conformations in a reversible and regulated manner (). It has been proposed that actin works as a conformational switch to control the assembly or the activity of chromatin remodeling machines (). In agreement with this idea, it is tempting to suggest that the fundamental function of actin in transcription is to mediate dynamic protein–protein interactions and to act as an allosteric factor in the remodeling of large multimolecular complexes, such as the transcriptional apparatus or the nascent mRNP. The transcription apparatus undergoes major conformational changes after assembly of the preinitiation complex, and these changes lead to transition into the elongation phase and to promoter clearance (for review see ). We speculate that actin acts in association with the eukaryotic RNA polymerases as a molecular switch in these structural transitions. The involvement of actin in the recruitment of chromatin-modifying factors while it is part of the nascent mRNP is a more specialized function of actin in pol-II transcription. We suggest that actin has the potential to trigger the release of chromatin modifiers from the mRNP concomitantly with the termination of transcription. The newly synthesized mRNP must undergo remodeling while the pre-mRNA is processed, surveyed, and exported, and it will be interesting to determine whether actin plays a role in these events. Finally, we note that actin is involved in signal transduction pathways in the cytoplasm, and it is therefore interesting to consider that actin-based mechanisms of transcription regulation may sense extracellular signals via cytoplasmic changes in the actin pools. The idea that actin plays a central role in the coordination of signal transduction is not new (; ). What is new is the suggestion that actin can modulate the overall transcriptional activity of the cell in response to extracellular signals.

Historically, epithelial and mesenchymal cells have been identified on the basis of their unique visual appearance and the morphology of the multicellular structures they create (). A typical epithelium is a sheet of cells, often one cell thick, with individual epithelial cells abutting each other in a uniform array. Regularly spaced cell–cell junctions and adhesions between neighboring epithelial cells hold them tightly together and inhibit the movement of individual cells away from the epithelial monolayer. Internal adhesiveness allows an epithelial sheet to enclose a three-dimensional space and provide it with structural definition and mechanical rigidity. The epithelial sheet itself is polarized, meaning that the apical and basal surfaces are likely to be visually different, adhere to different substrates, or have different functions. Mesenchymal cells, on the other hand, generally exhibit neither regimented structure nor tight intracellular adhesion. Mesenchymal cells form structures that are irregular in shape and not uniform in composition or density. Adhesions between mesenchymal cells are less strong than in their epithelial counterparts, allowing for increased migratory capacity. Mesenchymal cells also have a more extended and elongated shape, relative to epithelial cells, and they possess front-to-back leading edge polarity. Unlike epithelia, the irregular structure of mesenchyme does not allow for rigid topological specialization. Moreover, mesenchymal migration is mechanistically different from epithelial movement. Epithelial cells move as a sheet en block, whereas mesenchymal migration is considerably more dynamic. Mesenchymal cells move individually and can leave part of the trailing region behind. Elizabeth Hay (Harvard University, Boston, MA), who first described the EMT (), illustrated the fundamental differences of such movement in embryogenesis (subtle/controlled) and tumorigenesis (aggressive/uncontrolled) to define the distinct EMT mechanisms at the EMT conference. Turning an epithelial cell into a mesenchymal cell requires alterations in morphology, cellular architecture, adhesion, and migration capacity. Commonly used molecular markers for EMT include increased expression of N-cadherin and vimentin, nuclear localization of β-catenin, and increased production of the transcription factors such as Snail1 (Snail), Snail2 (Slug), Twist, EF1/ZEB1, SIP1/ZEB2, and/or E47 that inhibit E-cadherin production. Phenotypic markers for an EMT include an increased capacity for migration and three-dimensional invasion, as well as resistance to anoikis/apoptosis. A summary of common EMT markers is listed in . Importantly, these developmental regulators can induce EMT in a nondevelopmental context and thereby have an important role in cancer and fibrosis. Much of the meeting highlighted signaling pathways that regulate or mediate the EMT, focusing both on refinement and extension of known pathways, but also on the discovery of new regulators and novel pathways (). One of the first cell surface receptors identified that was able to stimulate scattering of epithelial cells was the Met receptor tyrosine kinase. Activation of Met by its ligand, hepatocyte growth factor, enhances the migration of multiple cell lines in vitro, and scattering of cultured multicystic dysplastic kidney cells is a classical EMT assay. Morag Park (McGill University, Montreal, Quebec, Canada) reported that transgenic mice expressing wild-type or active variants of Met under the control of the mouse mammary tumor virus promoter develop nodal and ductal hyperplasia and spontaneous mammary tumors, albeit with a long latency period (∼1.5 yr). Park suggested that Met cooperates with the Her2/neu oncogene in activating EMT, and that the Crk family of SH2 and SH3 adaptor proteins are critical in Met-mediated EMT. Crk proteins are highly expressed in human breast tumors, and Park reported that small interfering RNA (siRNA) ablation of Crk inhibits Met-dependent cell migration and EMT. Although the Met receptor-mediated signaling results in cell scattering, it has not been made clear whether Met signaling also has a more permanent effect on the expression or localization of some of the effectors of EMT, such as E-cadherin and β-catenin. Recent work by Walter Birchmeier (Max Delbruck Center, Berlin, Germany) suggests that Met also regulates intracellular localization of β-catenin. β-Catenin has a dual role in the EMT; it enhances cell–cell adhesion when bound to cadherin complexes in adherens junctions and also functions as a transcriptional coactivator upon entry into the nucleus (). The ability of β-catenin to enhance cadherin-dependent adhesion depends on β-catenin binding to α-catenin and on α-catenin binding to the cadherin (). Phosphorylation of β-catenin residue Y142 prevents α-catenin interaction and enhances the binding of β-catenin to BCL9-2, which is the vertebrate homologue of the legless gene (). Interaction of β-catenin with BCL9-2 enhances nuclear accumulation of both proteins, simultaneously decreasing cadherin-mediated adhesion and activating catenin target gene transcription. Ectopic BCL9-2 expression is sufficient to induce EMT in cultured cells, and siRNA-mediated BCL9-2 inactivation drives the reverse mesenchymal–epithelial transition (MET). Birchmeier reported that Y142 can be phosphorylated by the Met tyrosine kinase, indicating the existence of an EMT activation pathway where Met induces β-catenin nuclear translocation by enhancing BCL9-2 interaction. This pathway satisfactorily links these two well known EMT regulators. Interestingly, Pez/PTPN14, which is a tyrosine phosphatase that is frequently mutated in colorectal tumors (), induces Snail1 expression and can also activate cell migration (Yeesim Khew-Goodall, Hanson Institute, Adelaide, Australia). Pez can dephosphorylate β-catenin on tyrosine residues that regulate its interaction with the adherens junction complex, suggesting that Pez mutations contribute to EMT by preventing cytoplasmic β-catenin–cadherin interaction and enhancing its nuclear translocation. However, Pez overexpression in MDCK and MDA-MB468 cells was shown to be sufficient to cause EMT, and knockdown in zebrafish causes multiple developmental abnormalities, including aberrant pigmentation and craniofacial deformation. These defects are broadly consistent with dysfunctional neural crest EMT in the absence of Pez. Cancer-relevant insights into EGF signaling were provided by Erik Thompson (University of Melbourne, Melbourne, Australia), who has identified EGF as a novel EMT inducer in human breast cancer, as measured by EGF's ability to decrease E-cadherin and increase vimentin production in PMC42 cells. Interestingly, EMT may influence the response of certain cancers to EGF receptor (EGFR)–targeted therapeutics. John Haley (OSI Pharmaceuticals, Melville, NY) presented data showing that the sensitivity of nonsmall cell lung cancer cell lines to erlotinib, which is an EGFR-targeted monoclonal antibody, did not correlate with EGFR levels, but rather depended on their EMT status, with those having undergone EMT showing resistance (). An interesting and novel aspect of EGFR signaling was presented by Mien-Chie Hung (The University of Texas MD Anderson Cancer Center, Houston, TX), who reported that EGFR, which is a transmembrane receptor tyrosine kinase, complexes with the STAT3 transcription factor in the nucleus and can be immunoprecipitated from the EGF-responsive iNos promoter (). The role that promoter-complexed EGFR has in EMT is uncertain, but high nuclear EGFR is associated with a poor prognosis in breast carcinoma (). The observation that a transmembrane receptor is found in functional promoter complexes in the nucleus was one of the meeting's most surprising observations, and it will be of great interest to characterize the topological and structural mechanisms through which a membrane receptor enters the nucleus and activates transcription (). TGF-β is a major regulator of EMT and has been implicated in skin cancer development (). Jiri Zavadil (New York University School of Medicine, New York, NY) reported that TGF-β activates EMT through Smad-3–dependent activation of the HEY1 gene, a member of the Hairy/Enhancer-of-split family of transcriptional repressors. Zavadil used extensive gene expression profiling to identify HEY1 targets that are important in EMT induction (). He reported on the profiling of EMT in the following three different contexts: HaCaT human keratinocyte EMT in response to TGF-β, mouse model of aristolochic acid nephropathy, and human kidney-proximal tubule cells. Satisfyingly, one of these targets is (, which is a gene that regulates EMT by repressing the production of Notch, GSK3β, and β-catenin. Another HEY1 target seen in all three systems was the polycomb family histone methyltransferases EZH1/EZH2, suggesting that TGF-β–activated EMT could be controlled through structural histone modification. Other TGF-β targets include integrins β4 and α6. Richard Bates (University of Massachusetts, Worcester, MA) reported that the integrin αvβ6 is up-regulated during colon cancer development and highly expressed in metastatic samples (). Christopher Gebeshuber showed that TGF-β induced Smad-2 tyrosine phosphorylation and that TGF-β–induced EMT was blocked upon expression of nonphosphorylatable Smad-2 mutant, the expression of which inhibited metastases formation. Gebeshuber also reported that this mutant had a reduced ability to interact with the Tcf–Lef1 transcription factor. This suggests that tyrosine phosphorylation of Smad-2 may potentiate Tcf–Lef1 interaction and stimulate both EMT and metastatic induction. Ali Nawshad (University of Nebraska, Lincoln, NE) and Elizabeth Hay reported a similar noncanonical role for TGF-β in the EMT of mouse palatal epithelial seam and kidney-proximal tubule cells. They reported that Smad-2/4 repressed E-cadherin transcription through Tcf–Lef1 (; ). One of the functions of TGF-β is to stimulate expression of ECM proteins. Do ECM proteins initiate EMT? Andre Menke (University of Ulm, Ulm, Germany) showed that extracellular collagen that is deposited during a fibrotic disease can be an initiator of EMT. Menke reported that pancreatic cancer cell lines cultured on collagen I have a reduced capacity to cluster E-cadherin at points of cell–cell contact and have a more mesenchyme-like morphology. Menke postulated an EMT pathway where collagen induces both the recruitment of FAK to cadherin adhesion complexes and the phosphorylation of β-catenin. Phosphorylated β-catenin then translocates to the nucleus, activating EMT target genes. Conceptually, this may be similar to work by Mina Bissell describing the capacity of mechanical forces or the shape of the cell to initiate EMT. The Snail1 transcriptional repressor is a key EMT regulator (). There was much interest in signaling pathways converging on Snail1 production, stability, and intracellular localization. Derek Radisky (Mayo Clinic, Jacksonville, FL) reported that matrix metalloproteinase-3 (MMP-3) activates Snail1 production in mammary cells. MMP-3 is expressed in many primary breast tumors, induces mammary carcinogenesis in transgenic mice, and causes an in vitro EMT in mouse mammary cells (; ). Radisky reported that MMP-3 activates EMT by inducing the production of an alternatively spliced variant of Rac1, which is a small GTPase that regulates cell migration through control of actin polymerization (). This splice variant, termed Rac1b, activates the mitochondrial production of reactive oxygen species (ROS), which subsequently activates Snail1 production (). However, the mechanism by which MMP-3 stimulates alternative splicing, or how the Rac1 variant activates ROS, is unclear. Snail genes can be considered regulators of cell survival, adhesion, and migration, and the triggering of the EMT is just one of the mechanisms they use to promote cell movement (). Pierre Savagner (Batiment de Recherche en Cancerologie, Montpellier, France) reported that Snail2-deficient mice show delayed mammary gland tubule growth, and precocious branching morphogenesis similar to that seen in the mammary gland lacking P-cadherin, which is a cadherin that is selectively expressed in myoepithelial cells (). Snail2-deficient mammary gland retained normal smooth muscle actin-staining myoepithelial cells. These cells lack P-cadherin, suggesting that Snail2 controls a progenitor-like phenotype in the mammary gland through P-cadherin. Several investigators reported new insights into the control of Snail1 expression. Shoukat Dedhar (University of British Columbia, Vancouver, British Columbia, Canada) reported that integrin-linked kinase (ILK) activates Snail1 expression. Using proteomic approaches, Dedhar and coworkers made the surprising finding that ILK-mediated induction of Snail1 transcription maps to a portion of the Snail1 promoter that is bound by poly-ADP-ribose polymerase 1 (PARP-1). PARP-1 regulates transcription by modifying chromatin structure and through interaction with other transcription factors (). ILK activation promotes PARP-1 binding to the Snail1 promoter, whereas siRNA ILK knockdown and drug inhibition of ILK activity prevents PARP-1 from binding to the promoter. siRNA knockdown of PARP-1 in mesenchymally transformed PC-3 cells inhibited Snail1 expression and stimulated E-cadherin expression, suggesting the novel idea that PARP-1 itself is an important factor in EMT control. It is unclear whether direct phosphorylation of PARP-1 by ILK controls its ability to interact with the Snail1 promoter. Inhibiting ILK activity with the small molecular inhibitor QLT0267 inhibited production of urokinase type plasminogen activator and the invasion of MDA-MB231 breast cancer cells (Nancy Dos Santos, University of British Columbia, Vancouver, British Columbia, Canada). Anna Bagnato (Regina Elena Cancer Institute, Rome, Italy) also reported that endothelin 1 induced EMT in ovarian carcinomas in in vitro and in vivo cells through a phosphoinositide 3 kinase– and ILK-mediated signaling pathway, leading to glycogen synthase kinase-3β (GSK-3β) inhibition, Snail and β-catenin stabilization, and transcriptional programs that control repression of E-cadherin. Inhibition of the endothelin A receptor reversed the EMT, suppressed ILK and Snail1 expression, and restored E-cadherin expression. Snail1 represses E-cadherin expression by binding to three independent E-boxes in the cadherin promoter. Snail1 prevents E-cadherin expression through at least two pathways, one dependent on class I histone deacetylases and the other independent of it (Antonio Garcia de Herreros, Universitat Pompeu Fabra, Barcelona, Spain). Snail transcription is regulated by the estrogen receptor (ER; Paul Wade, National Institute of Environmental Health Sciences, Research Triangle Park, NC). ER is an EMT inhibitor and is critical in maintaining the epithelial status of normal breast cells. Wade reported that MTA3, which is a component of the Mi-2–NuRD transcriptional repressor complex, is an ER-responsive gene, and its expression correlates well with ER expression in primary breast tissue samples. Wade reported that MTA3 binds to the Snail1 promoter and inhibits Snail1 transcription (). Because expression of the ER is a marker for good breast cancer prognosis, the observation that ER is an EMT inhibitor provides further evidence in support of a role for EMT in oncogenesis. Snail1 levels can also be controlled posttranslationally, and Garcia de Herreros and Hung both reported that Snail1 is a phosphoprotein. Garcia de Herreros reported that Snail1 phosphorylation prevents its nuclear accumulation and inhibits its ability to activate EMT (). Hung reported that Snail1 is phosphorylated by GSK-3β on two distinct motifs. Phosphorylation of two serines in the first motif directs Snail1 ubiquitination and proteolytic destruction. Phosphorylation of four serines on the second motif directs nuclear export. Mutation of all six GSK-3β phosphorylation sites increased the half-life of the Snail1 protein and ensured that it was constitutively nuclear. Consistent with a role for Snail1 phosphorylation in EMT, expression of Snail1 that could not be phosphorylated caused a loss of E-cadherin production and an EMT-like morphological change in human tumor lines (). Jim Woodgett (Samuel Lunenfeld Research Institute, Toronto, Ontario, Canada) described an important role for GSK-3β in controlling embryonic stem cell differentiation and the maintenance of pluripotency. During embryogenesis, the neural crest develops from a small portion of the dorsal neural tube (; ). After an EMT, neural crest cells migrate away from the neural tube and differentiate into bone, smooth muscle, peripheral neurons and glia, and melanocytes. Don Newgreen (Murdoch Children's Research Institute, Melbourne, Australia) reported that the Sox transcription factors control this EMT and subsequent migration. Using an electroporation system that delivers Sox genes to cells on one side of the neural tube in living chicken embryos, Newgreen reported that ectopic expression of Sox-8, -9, or -10 was sufficient to induce EMT and activate migration away from the neural tube while suppressing terminal differentiation. This migratory capacity was conferred to all cells of the neural tube, indicating that Sox expression was overriding inhibitory signals that normally restrict neural tube EMT to cells of the neural crest. Nelly Auersperg (University of British Columbia, Vancouver, British Columbia, Canada) provided evidence that EMT occurs in the ovaries of adult women. The mature mammalian ovary is enveloped by the ovarian surface epithelium (OSE), and the bulk of ovarian carcinomas arise from these cells. As a result of wound repair after egg extrusion, OSE cells are trapped in the ovarian follicle or stroma of postovulatory ovaries. Dr. Auersperg presented evidence showing that normal human OSE cells have a strong propensity to undergo EMT in vitro and in vivo in response to growth factor stimulation and alteration in their extracellular matrix. Auersperg suggested that normal OSE trapped within the ovary may undergo EMT as a means of maintaining ovarian homeostasis. A defining feature of EMT is a reduction in E-cadherin levels and a concomitant production of N-cadherin. Cadherins are transmembrane proteins whose homotypic interaction between neighboring cells creates adherens junctions (). Alteration of cadherin-based adhesion has a key role in modulating development and organogenesis. At the cell membrane, cadherin proteins are found as homodimers tethered to the actin cytoskeleton by a multiprotein complex that includes α-, β-, and p120-catenin. To characterize the physical forces underlying cadherin-based adhesion, Jean-Paul Thiery (Institut Pasteur, Paris, France) reported on an elegant system designed to measure the force necessary to separate two cells that are adhered solely to each other (, ). Thiery reported that the development of intercellular adhesion by N- or E-cadherin is a two-step process. The first step relies on interactions between the cadherins on the surface of adjacent cells. This interaction takes 30 s to develop and requires a force of ∼10 nanoNewtons to break apart. The second step, which takes up to 30 min to maximize, strengthens the initial interaction and requires ∼200 nanoNewtons to separate it. This strengthening depends on Rac- and Cdc42-mediated induction of actin polymerization, presumably to anchor the cell surface cadherins to the cytosol. Thiery also reported that four times more force is required to separate adhesions between E-cadherin molecules compared with N-cadherin ones. In addition, there is no detectable interaction strength between E- and N-cadherin. This supports the current EMT paradigm, where the presence of E-cadherin in epithelial cells allows for greater cell–cell adhesive strength compared with that of the N-cadherin–expressing mesenchyme. Moreover, the minimal adhesive interaction between E- and N-cadherin would be predicted to allow an N-cadherin–expressing cell to migrate through a layer of E-cadherin–expressing cells. Alpha Yap (University of Queensland, Brisbane, Australia) reported evidence that E-cadherin clustering at cell–cell junction sites requires dynamic microtubules. Yap reported visual evidence that the plus ends of microtubules terminate in E-cadherin puncta and that agents that block dynamic plus ends inhibit the ability of cells to concentrate cadherin at cell–cell contacts. This suggests that the actin and microtubule cytoskeletons both serve to anchor E-cadherin adhesions. This would contrast cadherin adhesions to integrin-containing focal adhesions because microtubule association with focal adhesions triggers their disassembly (). Mina Bissell (Lawrence Berkeley National Laboratory, Berkeley, CA) described data suggesting that cell shape changes brought about by the destruction of the basement membrane cause EMT. She then described a model of branching morphogenesis of the mammary gland and showed data to support a transient EMT at the tip of the branching structures. This was demonstrated by the activation of the vimentin promoter (visualized by a GFP reporter) at the branch tip. Bissell went on to describe studies that provided an understanding of how branching structures are created. She used engineered matrices and biomaterials to show that the architecture of the created vessel in collagen gels can determine where and how branches are created. Although the role of cell geometry in growth (; ), apoptosis (), and metabolic regulation () has been known for decades, the molecular pathways that link cell shape to these events, and also to EMT, are only now beginning to be elucidated (; ; ). The orientation of a cell to its growth substrate may also regulate EMT. Marcia McCoy and Calvin Roskelley (University of British Columbia, Vancouver, British Columbia, Canada) reported that overexpression or mislocalization of the apical marker podocalyxin destabilized cell polarity in vitro, which may explain why podocalyxin overexpression is an independent marker of in vivo breast carcinoma progression (). The occurrence of EMT during tumor progression allows benign tumor cells (i.e., ones that are noninvasive and nonmetastatic) to acquire the capacity to infiltrate surrounding tissue and to ultimately metastasize to distant sites. The pathological staging of tumors supports this paradigm. The most compelling evidence for the involvement of EMT in oncogenesis is the ability of multiple EMT regulators to enhance tumor formation and/or metastasis (). For example, expression of Snail1 increases the aggressiveness of experimentally induced breast tumors, and high Snail1 expression correlates with an increased risk of tumor relapse and poor survival rates in human breast cancer (). Loss of E-cadherin is a hallmark of metastatic carcinoma (), and proteomic analysis of breast cancer reveals that circulating mammary tumor cells, or those found as micrometastases, show evidence of mesenchymal conversion (). The EMT meeting added to the growing list of EMT regulators that control some aspect of oncogenesis, which includes MMP-3, BCL9–2, EGFR, Met, Goosecoid, Kaiso, TGF-β, FOXC2, GSK-3β, Smad-3, Pez, Snail1, Snail2, and ILK (). However, there remains some controversy in the cancer community, particularly among pathologists, as to whether the transformation of a normal cell into a cancerous cell or a noninvasive tumor into a metastatic tumor is truly an EMT (). Skepticism about the role of EMT in cancer stems from the apparent rarity of the EMT–like morphological changes that are observed in primary tumor sections, and also from the observation that metastases appear histologically similar to the primary tumor from which they are derived. Of central importance, therefore, is the direct visualization of EMT during tumor progression. Garcia de Herreros used a new Snail1 antibody that is suitable for mouse and human immunohistochemistry (EC3) to show that Snail1 protein is expressed specifically at the invading front of colorectal tumors. Snail antibodies have been difficult to use in immunohistochemistry, and Karl-Friedrich Becker (Technical University of Munich, Munich, Germany) used another new Snail1 antibody (Sn9H2; ) to demonstrate nuclear Snail1 in gastric, mammary, and endometrial tumors. Richard Bates reported that integrin αvβ6 is specifically expressed at the invading edge of colorectal cancer xenografts. Thomas Brabletz (University of Erlangen, Erlangen, Germany) reported that tumor cells at the invading edge of colorectal carcinomas have nuclear β-catenin and loss of E-cadherin. Nuclear localization of β-catenin is frequently used as an EMT marker, and nuclear β-catenin is a marker for a poor prognosis in colorectal cancer. The ability of EMT markers to identify a subset of tumor cells raises the possibility that EMT could be associated with the maintenance of cancer stem cells. Brabletz reported that invading cells with nuclear β-catenin also express the stem cell markers hTert and survivin, possibly implicating EMT in cancer stem cell maintenance (). The presence of EMT markers at the tumor–host interface, but not in the bulk tumor, is strong evidence that EMT occurs during tumor development and that it regulates invasiveness and tumor aggressiveness. The histological similarity of secondary, metastasis-derived tumors to the primary tumor indicates that EMT-mediated metastatic development must be followed by a reverse MET to allow colonization of secondary sites. Brabletz reported that metastases derived from tumors originally expressing nuclear β-catenin were found to reexpress E-cadherin, and their β-catenin became cytoplasmic, which is suggestive of a MET (). Similarly, Christine Chaffer (Bernard O'Brien, Institute of Microsurgery, Melbourne, Australia) reported that variants of the metastatic T24/TSU-Pr1 bladder carcinoma line that were selected for enhanced metastatic potential have more epithelial markers (E-cadherin and keratins) than their less metastatic counterparts, but continue to express some mesenchymal markers (vimentin and MMPs). This ability of cells to express attributes of both epithelial and mesenchymal phenotypes was referred to by Savagner as a “metastable phenotype” (). Consistent with this idea, Savanger reported that Rac distribution can be found with both epithelial-like (adherens junctions) and mesenchyme-like (lamellopodia) patterns during the migration of cohesive epithelial cells, and probably during tumor invasion as well. Metastability is consistent with the expression of stem cell markers in colorectal cells undergoing EMT and suggests that such plasticity may be found in progenitor cells in various organs. This plasticity could also be an explanation for the difficulty in observing EMT in cancer development; acquisition of mesenchymal characteristics may be transitory and undergo a reversal during later tumorigenesis. Robert Weinberg (Whitehead Institute, Cambridge, MA) reported that three transcription factors regulating developmental EMT—Twist, Goosecoid, and FOXC2—have important roles in metastasis. Each of these gene products enhances metastasis in experimental mouse models and is highly expressed in primary human tumors and metastases. Twist is a basic helix-loop-helix transcription factor that was originally identified as a EMT activator (). Weinberg reported that Twist expression is sufficient to induce an in vitro EMT in breast cells and that Twist inactivation inhibits metastasis development in vivo (). Goosecoid is a homeobox transcriptional repressor that marks the Spemann organizer in vertebrate gastrulation and is one of the first identified regulators of embryological patterning (). Both Twist and Goosecoid regulate FOXC2, which is a transcription factor of the FOX family of forkhead helix-turn-helix DNA-binding proteins that regulates EMT and organ development in multiple tissues (). Twist, Goosecoid, and Snail1 all repress E-cadherin and induce FOXC2; they also enhance cell migration in vitro and metastatic potential in vivo. It is not yet known whether these three genes regulate individual or overlapping pathways of EMT and metastases. Importantly, FOXC2 also directly up-regulated mesenchymal gene transcription, rather than causing an EMT through E-cadherin repression. Frans van Roy and Geert Berx (Ghent University, Ghent, Belgium) reported on the identification of a series of novel target genes of the E-cadherin repressors Snail1 and SIP1/ZEB2 that control the establishment of junctional complexes, intermediate filament networks, and the actin cytoskeleton (). They also showed some direct effects on mesenchymal factor transcription via these pathways. Christine Gilles (University of Liege, Liege, Belgium) reported that vimentin transcription was activated by SIP1/ZEB2, as well as a Tcf– β-catenin complex. The accumulation of fibroblasts, excess collagen, and other matrix components at sites of chronic inflammation lead to scar tissue formation and progressive tissue injury. These fibroblasts derive from the bone marrow, but also arise from an EMT of cells at injury sites (; ). EMT is likely involved in the progressive fibrotic diseases of the heart, lung, liver, and kidney. Eric Neilson (Vanderbilt University, Nashville, TN) presented work using fibroblast-specific protein 1 (FSP1) as a marker for EMT that occurs during fibrosis (). FSP1-positive cells appear during kidney fibrosis and in IgA nephropathy; increased expression of FSP1 correlates with the prognosis and extent of fibrosis (). The ablation of FSP1 cells attenuates fibrosis and collagen deposition, indicating a causal role for these cells in fibrotic disease (). Kidney FSP1-positive cells derive from two sources; from the bone marrow and from an EMT at sites of renal fibrosis (). Inactivation of FSP1 with a LacZ “knock in” mouse produced fibroblasts that were less motile in wound healing assays and had impaired angiogenesis in an aortic ring outgrowth model. Neilson also introduced studies on the FSP1 promoter and reported the identification a new zinc finger protein, fibroblast transcription factor 1, which binds in the FSP1 promoter. Fibroblast transcription factor 1 also up-regulates Twist and Snail1 and suppresses β-catenin, E-cadherin, and ZO-1 during EMT, indicating that it may be a key regulator of the EMT transcriptome. Raghu Kalluri (Harvard University, Boston, MA) introduced the novel concept of endothelial–mesenchymal transition, which is probably an important process in TGF-β1–mediated cardiac fibrosis. Kalluri also reported that an inhibitor of TGF-β signaling, bone-morphogenic protein 7 (BMP7), could inhibit cardiac fibrosis in two mouse models of this disease. BMP7 belongs to the BMP family of TGF-β growth factors, and has a specific role as a morphogen during liver development. Kalluri also discussed the functional interconnection between EMT and angiogenesis, suggesting that angiogenesis inhibition could be therapeutic for fibrosis as well as cancer. Michael Zeisberg (Harvard University, Boston, MA) reported that BMP7 can inhibit fibroblast migration and prevent fibrotic disease in mouse models of liver fibrosis. The detection of EMT in vivo during disease progression in adult organisms remains one of the central challenges of EMT physiology. Pioneering work by established that fibrosis involves EMT, and this approach has been extended to include the formation of metastatic tumor cells (). Evidence of EMT markers at the leading edge of invading tumors was provided by Bates (integrin αvβ6), Garcia de Herreros (using a new Snail1 antibody), and Brabletz (nuclear β-catenin), and these new findings were some of the highlights of the meeting, strongly suggesting an important role for EMT in driving tumor invasion and metastasis. Because it is now possible to visualize the movement and morphology of individual tumor cells in real-time in a living animal (), the examination of EMT in real-time is a possibility for the future. The detailed molecular studies of many investigators at the EMT meetings will hopefully provide additional markers for this task (). These markers may allow further investigation into the role of metastability in cancer. Metastability indicates the existence of cells with features of both epithelial and mesenchymal cells. This concept is consistent with the sequential steps of junctional dissolution that were described by Thiery () and is gaining momentum through the accumulation of evidence in favor of such hybrid states. The predominantly epithelial, yet somewhat mesenchymal, phenotype of highly aggressive and metastatic bladder cancer cells presented by Chaffer reinforces the potential of many cancer cells for plastic differentiation. In addition, Savagner showed evidence of both epithelial and mesenchymal patterning of Rac in epithelial cells that were induced to migrate. The importance of MET or other partial loss of mesenchymal markers in the successful growth of metastases could add further opportunities for therapies that block metastases. The possibility that softer boundaries exist between epithelial and mesenchymal tumor cells and the possibility of hybrid cells may help explain the current lack of robust clinical evidence for EMT as a metastasis mediator (). Most importantly, the meeting witnessed the emergence of EMT as a target for drug development in cancer and fibrosis. For example, BMP7 mimetics antagonize TGF-β–driven EMT in fibrotic kidney and heart and inhibit disease development. In addition, small molecule ILK inhibitors inhibit Snail1 production, induce E-cadherin expression, and inhibit invasion. Also discussed at the meeting was the possibility that angiogenesis, EMT, fibrosis, and cancer have common regulatory pathways and that the angiogenesis inhibition may be useful in both fibrosis and cancer. The involvement of ILK in angiogenesis, EMT, fibrosis, and cancer suggest that ILK inhibition may be one useful therapy. In addition, EMT could be used as a functional screen for novel anticancer agents, a strategy that led to the identification of motuporamine (Calvin Roskelley). Motuporamine was derived from a library of marine invertebrate compounds and inhibits in vitro invasion and migration by activating the Rho GTPase and stimulating actin stress fiber formation. Continued identification of new EMT inhibitors holds the promise of novel cancer and fibrosis treatment options. We anticipate considerable progress in this field in the year leading up to the 2007 EMT meeting, which is planned to take place in Montpellier, France (), building on the current exponential trend of EMT observations in numerous cellular systems of physiological and pathophysiological importance.

Squamous cell carcinoma antigen (SCCA) was first discovered as a marker of squamous cell carcinomas in the cervix (). Cloning of the SCCA gene demonstrated that SCCA belongs to the serpin superfamily of serine proteinase inhibitors (). However, it soon became apparent that SCCA is a cross-class inhibitor and that its target molecules include cysteine proteinases such as cathepsin L and papain (; ). Recent work has revealed the presence of tandemly aligned homologous genes. The telomeric DNA segment contains a gene that was cloned by and has since been designated SCCA1 (). The centromeric gene, which is 92% identical at the nucleic acid level, was named SCCA2. Interestingly, SCCA2 inhibits chymotrypsin and its relatives, as is expected for a serpin (). SCCAs are also expressed in psoriatic epidermis (), which is where abnormal proliferation and aberrant differentiation are characteristic features. The epidermis is the outermost tissue, whose primary role is to form a barrier against hostile environmental factors, including UV, and it consists of four kinds of cells (i.e., cornified, granular, spinous, and basal cells). Although the epidermis has effective countermeasures against UV irradiation, its protective mechanisms, other than melanin, are still unknown. We show that SCCA1 is a specific endogenous inhibitor of c-Jun-NH-terminal kinase-1 (JNK1) and acts to protect UV-exposed keratinocytes from apoptotic cell death. During studies to analyze the localization of SCCAs in normal and diseased skin, we observed strong up-regulation of SCCAs in sun-exposed epidermis. In normal, sun-protected skin, SCCAs were only weakly stained in the upper epidermis (). When the cheeks and eyelids of 22–84-yr-old subjects were examined, all the skin tissues showed marked elevation of SCCAs, although SCCA1 was predominant. Interestingly, some of the nuclei in sun-exposed epidermis were heavily stained, as was the cytoplasm (, inset). A study involving UV irradiation of the buttocks of healthy volunteers confirmed strong induction of SCCA1 in the spinous to granular layers of the irradiated skin. An in situ hybridization study showed that SCCA1 mRNA is weakly detectable in normal epidermis (). UV irradiation caused strong induction of SCCA1 mRNA in the top layers of the epidermis. In cultured neonatal human keratinocytes (NHK), quantitative PCR analysis showed induction of both SCCAs, but a larger amount of SCCA1 mRNA is synthesized in the later stage, after UV irradiation (). This is presumably the reason why the overall production of SCCA1 is much higher than that of SCCA2 in sun-exposed skin. In mouse, serpin b3a is considered the mouse orthologue of human SCCA because of its strong inhibition of chymotrypsin (). Semiquantitative PCR analysis showed that mouse epidermis expressed serpin b3a (mouse SCCA), although 3T3/J2, which is a mouse fibroblast cell line, did not produce mouse SCCA (Fig. S1 A, available at ). To investigate the photobiological role of SCCA, we stably transfected SCCA cDNAs into 3T3/J2; clones were established under G418 selection. Each clone expressed only one transfected gene, SCCA1 or SCCA2 (Fig. S1 B). FACS analyses demonstrated that 3T3/J2 cells undergo apoptosis when exposed to UV at doses >30 mJ/cm (). At 30 mJ/cm, 37.2 ± 1.3% of the mock-transfected cells (control) died within 48 h (). When each transfected pool was used, SCCA1- or SCCA2-transfected cells showed resistance to UV-induced apoptosis (P < 0.001) and only 23.8 ± 3.3% or 19.9 ± 4.4% of transfected cells died after UV irradiation, respectively. There was no significant difference in antiapoptotic activity between SCCA1 and SCCA2. We established 12 individual clones for SCCA1 and examined the effect of UV irradiation. Only 20.7% of the minimally expressing cells survived at 50 mJ/cm, whereas 74.7% of the maximally expressing cells (2,772-fold expression) survived in the same condition (). Clearly, SCCA1-expressing clones showed a strong correlation between the expression levels and the suppression of UV-induced apoptosis ( = 0.734). We then examined the effect of the suppression of SCCAs using small interfering RNA (siRNA). An H1 promoter-driven siRNA sequence directed against homologous sites of SCCA1 and SCCA2 was introduced into HaCaT cells. We used the HaCaT cell line because it was derived from spontaneous transformation of human adult keratinocytes () and synthesizes relatively high levels of both SCCAs under proliferative conditions. A stably transfected cell line was established (small interfering SCCA [siSCCA]/HaCaT), which showed >90% suppression of both SCCA1 and SCCA2 mRNA levels, compared with the control siRNA (targeted to GFP mRNA) as judged by quantitative PCR (Fig. S1 C). Western blot analysis showed that SCCA protein levels were also decreased to approximately one-tenth of the control in siSCCA/HaCaT (Fig. S1 D). The siSCCA- and control siRNA-expressing cells did not show any changes by FACS analyses (). However, UV irradiation caused a significant decrease of cell survival in siSCCA/HaCaT cells (). Under conditions where 42.8 ± 5.0% of control cells survived, only 11.8 ± 6.3% of siSCCA/HaCaT cells survived (P < 0.001). Overall, the results suggest that SCCA1 and SCCA2 can similarly suppress UV-induced apoptosis and that the expression levels of SCCA regulate the antiapoptotic activity under UV irradiation. The antiapoptotic activity may be independent of proteinase-inhibitory activity because these molecules possess highly specific inhibition profiles for cysteine and serine proteinases, respectively. Furthermore, SCCA1 may play a primary role because of its abundance after the UV induction. To determine the intracellular localization of SCCAs in NHK, we used laser confocal microscopy. As shown in , SCCA1 was present exclusively in the cytoplasm of the proliferating keratinocytes. After UV irradiation at 50 mJ/cm, SCCA1 was translocated into the nucleus and cytoplasmic localization was greatly reduced at 24 h. Because the SCCAs do not contain known nuclear translocation signals, it is likely that SCCAs are transported to the nucleus via binding with a target molecule. To look for SCCA-binding molecules, we used a signal transduction antibody array. This screening resulted in the identification of JNK as a protein that binds to both SCCA1 and SCCA2 (). Interestingly, the binding of SCCAs was observed only with active JNK (p-JNK), but not with inactive JNK. Anti-p38, ERK1, and ERK2 antibodies were also present on the membrane, but showed a negative result (unpublished data). Interaction between SCCA1 and JNK1 was confirmed with immunoprecipitation. Interestingly, SCCA1 was detected in the precipitates prepared with anti-JNK1 antibody (), indicating the binding of SCCA1 to JNK1 within the cell. SCCA1 precipitated with anti-JNK1 antibody was markedly reduced in siSCCA/HaCaT. We examined the UV responsiveness of JNK1 in cultured NHK cells because growing evidence suggests that JNK1 is responsible for UV-induced apoptotic cell death in other cell types (; ; ). JNK1 was found in the cytoplasm, as well as in the nucleus, of proliferating NHK cells (Fig. S2 A, available at ). After UV irradiation, JNK1 accumulated into the nucleus (77.9 ± 9.4%; Fig. S2, A and B). To investigate whether translocation of SCCAs is caused by binding with JNK1, we used a peptide inhibitor of JNKs, TAT-containing JNK-interacting protein-1 (JIP1) peptide (; ). First, we confirmed that in the presence of 5 μM of JIP1 peptide nuclear translocation of JNK1 was no longer observed, even after 50 mJ/cm UV irradiation. JNK1 remained in the cytoplasm of most NHK cells (exclusively cytoplasmic, 65.4 ± 11.5%; cytoplasmic and nuclear, 31.3 ± 8.9%) after UV irradiation (Fig. S2 B). JIP1 control peptide had no effect on the localization of JNK1 (Fig. S2 C). We then analyzed the effect of JIP1 peptide on SCCA1 localization. SCCA1, which was present mostly in the cytoplasm of control keratinocytes (92.5 ± 1.5%), promptly accumulated in the nucleus after UV irradiation (confined nuclear, 63.6 ± 9.4%; peripheral nuclear, 32.0 ± 7.0%; ). JIP1-negative control peptide had no effect on JNK or SCCA1 localization (Fig. S2 C). JIP1 peptide alone did not affect the localization of SCCA1. Interestingly, in the presence of 5 μM of JIP1 peptide, SCCA1 remained in the cytoplasm of UV-irradiated cells (81.1 ± 8.5%; ). The fact that JIP1 peptide inhibited translocation of SCCA1 into the nucleus after UV irradiation strongly suggests that SCCA1 requires direct or indirect association with JNK1 for the translocation. Our results indicate that the interaction of SCCA1 with the complex may be involved in the UV-induced JNK activation. Gene disruption studies have revealed that JNK1, but not JNK2, plays a key role in UV-induced apoptosis (). Therefore, we investigated the effect of SCCA1 on JNK1 kinase activity. Because c-Jun is a well-known substrate of JNK, we used in vitro kinase assays to examine whether c-Jun phosphorylation by JNK1 is affected by SCCA1. In the presence of SCCA1, c-Jun phosphorylation was significantly reduced (). Inhibition was nearly parallel with the active JNK1 concentration and was observed at a uniform rate with increasing amounts of active JNK1 (), suggesting stoichiometric interaction between SCCA1 and p-JNK1. We also confirmed that SCCA1 did not show any suppression on the kinase activities of other MAPKs, p38α (), ERK1, or ERK2 (). Interestingly, SCCA1 altered the phosphorylation of c-Jun in siSCCA/HaCaT keratinocytes (). Down-regulation of SCCA1 resulted in marked increase of phospho-c-Jun, and UV irradiation further enhanced the levels of c-Jun phosphorylation. On the other hand, down-regulation of SCCA1 did not show any detectable changes in either the protein or phosphorylation levels of ERK1/2, p38α, and its primary substrate, MAPK-activated protein kinase-2 (MAPKAPK2; ; ). p38α is also known to respond to UV stress (). After UV irradiation, phosphorylated p38α increased its intensity at 5–15 min in control HaCaT and siSCCA/HaCaT cell extracts, whereas the p38α protein remained at the same level. Similar changes were observed for MAPKAPK2. These results suggest that the kinase activity of JNK1 is specifically regulated by SCCA1 that is also in the cell. It is well known that death enzyme caspases play essential roles in apoptotic cell death. It might be possible that suppression of UV-induced apoptosis by SCCA1 could include inhibition of certain caspases because SCCA1 is a strong inhibitor of cysteine proteinases. We tested this possibility and confirmed that even a 50-fold excess of SCCA1 had no effect on the activities of caspase-1–10 (Fig. S3 B, available at ), although an equimolar amount of SCCA1 strongly inhibited a typical cysteine proteinase, known as papain (Fig. S3 A). Thus, in the UV-signaling pathway, it is suggested that the antiapoptotic property of SCCA1 depends on the inhibition of JNK1. To examine the physiological importance of SCCA up-regulation in the epidermis, we generated transgenic mice with a hairless phenotype (HR-1) that overexpress SCCA1 driven from the involucrin promoter (). We chose involucrin promoter because UV irradiation induced strong up-regulation of SCCA1 mRNA in the upper epidermis (). Immunostaining of skin sections revealed strong expression of SCCA1 in the upper epidermis of the transgenic mice, but not in the skin of wild-type HR-1 mice (). When 200 mJ/cm UV was irradiated twice on the backs of mice, the epidermis of wild-type HR-1 mice was seriously damaged, as seen in loss of the granular layer and spongiosis in the suprabasal layer (). Surprisingly, the same dose of UV irradiation did not cause any degeneration in the epidermis of SCCA1-overexpressing mice, although the epidermis of the transgenic mice showed hypertrophic and parakeratotic changes. These results clearly reveal that SCCA1 plays a critical role in the UV protection mechanism in the skin. UV irradiation profoundly damages living organisms, which have had to evolve effective countermeasures. It is well known that UVB (290–320 nm) has the most toxic effect on chromosomal DNA. Basal cells, which are replicating cells, are physically protected with melanin, which forms a supranuclear melanin cap to shield the nucleus against UV irradiation. Furthermore, keratinocytes as well as melanocytes produce the potent antiapoptotic factors bcl-2 and bcl-xL (; ). Deficiency of these survival factors sensitizes keratinocytes to apoptotic stimuli, including UV irradiation. However, UV irradiation induces down-regulation of bcl-2 expression (). On the contrary, SCCA1 is up-regulated in the upper epidermis with UV irradiation. Considering the importance of reproductive basal cells and barrier-forming cornified cells, the middle to top layers of spinous and granular cells are thought to represent transient states that bridge the early and terminal differentiation of keratinocytes. We have identified a novel UV protection mechanism that will function to protect the middle to top layers of epidermis. Observations from bullous diseases such as pemphigus vulgaris clearly demonstrate the importance of maintaining these layers. Our study suggests that under UVB exposure SCCAs are highly synthesized in spinous and granular layers and suppress UV-induced cell death via the inhibition of JNK1. This is supported by our current study on the SCCA mutant mice, showing that up-regulation of SCCA1 in top layers of epidermis was enough to block death-promoting UV stress. It would be beneficial for humans because this mechanism is not functioning in replicating basal cells, but is working for the differentiated spinous and granular cells, which are destined to be shed after cornification. An SCCA1-specific sequence was PCR amplified from exon 8 (642–1,001 nt) and cloned into pCRII vector. Complementary RNA probes were prepared using DIG RNA labeling mix (Roche). HRP-conjugated antidigoxigenin antibody (DakoCytomation) was used for visualization. SCCA1 and SCCA2 cDNAs cloned from a psoriatic cDNA library () were subcloned into pTarget vector and transfected into 3T3/J2 cells using Lipofectamine Plus (Invitrogen). After 4 wk of culture in 500 μg/ml of G418, 12 colonies were isolated for SCCA1 and established as SCCA-expressing cell lines. Levels of SCCA expression were determined with Taqman PCR (Fig. S1). Expression levels among 12 colonies varied from 1–2,772-fold. A double-stranded oligonucleotide was designed corresponding to a common sequence of human SCCAs (5′-AAGCCAACACCAAGTTCATGT-3′) to allow formation of the hairpin structure in the expressed oligo-RNA, cloned into the pSilencer vector (Ambion), and transfected into human keratinocyte cell line HaCaT cells. siRNA against GFP mRNA (Ambion) was used as a control. Transfection was performed with Lipofectamine 2000 (Invitrogen) and stable cell lines were established in hygromycin-B media for 4–6 wk. HaCaT, HSC-4, and 3T3/J2 cells were cultured in DME (Invitrogen) supplemented with 10% FBS. NHK was cultured with EpiLife (Kurabo). Cells were irradiated with UVB at 60–70% confluency using a transilluminator (model TOREX FL205-E-30/DMR; Toshiba Medical Supply) emitting UVB at 0–100 mJ/cm. The cells were trypsinized and stained using the Annexin V-FITC and propidium iodide double-staining method (Immunotech). The cells were analyzed on a FACS Coulter EPICS XL-MCL (Beckman Coulter) with acquisition of a total of 10,000 events/samples to ensure adequate data. SCCA1 cDNA was subcloned into pQE-30 (QIAGEN) with a poly-histidine tag (6× His). His-tagged SCCA1 was expressed in stain BL21 (DE3) cells and purified on Ni-nitrilotriacetate resin (QIAGEN) and a Mono-Q column (GE Healthcare). To screen protein–protein interactions, a Signal Transduction AntibodyArray (Hypomatorix, Inc.) was used. The extract from a squamous cell carcinoma cell line, HSC-4, showed the highest expression of both SCCAs among NHK, HaCaT, and other SCCA cell lines. The cells were irradiated with 50 mJ/cm UVB were lysed and extracted with 1% Triton X-100 extraction buffer containing 15 mM Tris-HCl, pH 7.5, 120 mM NaCl, 25 mM KCl, 2 mM EGTA, 2 mM EDTA, 0.1 mM DTT, 80 μM bestatin, and 10 μM pepstatin. After incubation with the extract, the antibody array membrane was blotted with HRP-conjugated anti-SCCA1 or -SCCA2 mAb and detected with ECL Plus kit (GE Healthcare). Kinase activity of JNK was determined using a stress-activated protein kinase (SAPK)/JNK assay kit (Cell Signaling Technology). Inhibition of JNK kinase activity by SCCA1 was assessed using phosphorylated-JNK1a1/SAPK1c (Upstate Biotechnology). Phosphorylated JNK, c-Jun fusion beads, and SCCA1 were incubated with gentle rocking overnight at 4°C. The phosphorylated c-Jun was separated by SDS-PAGE and probed with anti–phospho-c-Jun antibody (1:1,000 dilution). ERK1 and ERK2 kinase activities were measured using MAPK (ERK1/2) activity assay kit (Chemicon International, Inc.). Kinase activity of p38α was assessed using active p38α and GST-tagged ATF2 (19–96; Upstate Biotechnology) as a substrate. Antibodies to ERK1/2 (Chemicon International, inc.), p-ERK1/2 (Biosource International), p38α (Upstate Biotechnology), p-p38α, p-ATF2, MAPKAPK2, and phosphorylated MAPKAPK2 (Cell Signaling Technology) were used for immunoblots. SCCA1 cDNA was fused to the involucrin promoter (L.B. Taichman, State University of New York, Stony Brook, NY; ) and transgenic mice were generated using BDF1 mice. To obtain the hairless phenotype, the involucrin-SCCA1 transgenic mice were crossed with HR-1 mice, and SCCA1+/+ mice were obtained after three passages. UVB irradiation was done with a transilluminator at 200 mJ/cm/day for 2 d. Skin specimens were taken the following day and stained with hematoxylin and eosin or immunostained with anti-SCCA1 or -SCCA2 mAb using MoMap kit (Ventana Medical Systems, Inc.). Anti-SCCA1 and -SCCA2 mAbs and rabbit anti-JNK1 antibody (Santa Cruz Biotechnology, Inc.) were used. Alexa Fluor 555–, 488–, and 546–conjugated secondary antibodies (Invitrogen) were used for immunofluorescence detection. Nuclei were stained with DAPI (Invitrogen). Actin filaments were visualized with Alexa Fluor 488–conjugated phalloidin (Invitrogen). Fluorescence images were collected using a microscope (model BX51WI; Olympus) equipped with a confocal system (Radiance 2100; BioRad Laboratories) at 21°C. UPlanApo 20×/0.70 ∞/0.17, UPlanApo 40×/0.85 ∞/0.11–0.23, and PlanApo 60×/1.00 WLSM ∞/0.17 were used. Fig. S1 shows establishment of SCCA-overexpressing (A) and knockdown (B) cell lines. Strong suppression of the SCCA mRNA and protein in the knockdown cells was shown in B. Fig. S2 shows localization of JNK1 in cultured NHK cells before and after UV irradiation (A). Comparison of cell numbers in cytoplasmic or nuclear localization of JNK1 before and after UV irradiation was summarized in B. Effects of the JIP1 control peptide (C) after UV irradiation were shown. Fig. S3 shows effect of SCCA1 on papain and various caspase activities. Online supplemental material is available at .

In plants, rigid cell walls restrict changes in cell shape and size. As a result, polarized secretion of cell wall components takes on particular importance during growth and development. Polar expansion in root hairs, a polarized plant cell type, is accompanied by accumulation of secretory compartments behind the growing tips of these cells (for reviews see ; ). The Rab GTPase, RabA4b, specifically labels TGN-like compartments displaying polarized localization in expanding root hair cells (). Although RabA4b-labeled compartments are thought to deliver new cell wall components to expanding root hair tips, little is known about mechanisms for sorting and targeting secretory vesicles. Rab GTPases regulate membrane trafficking steps by recruiting cytosolic effector proteins to their specific subcellular compartment (for review see ; ). Therefore, to better understand the role RabA4b GTPases play in trafficking secretory cargo, we characterized proteins that selectively interact with RabA4b in its active (GTP bound) conformation. It is becoming increasingly clear that phosphoinositides play key roles in membrane trafficking steps along the secretory pathway. Specific phosphoinositide isoforms, and proteins that specifically bind these lipids, preferentially mark different subcellular membranes (; for reviews see ; ). Despite their importance in membrane trafficking, little is known about how their generation and turnover is regulated upon specific elements of the secretory system. We show that the RabA4b GTPase specifically interacts with the phosphatidylinositol 4-OH kinase, PI-4Kβ1, and both colocalize to tip-localized membranes in growing root hairs. In transfer DNA (T-DNA) insertional mutants, where both PI-4Kβ1 and its close relative PI-4Kβ2 are disrupted, root hairs have aberrant morphology. The novel homology (NH) domain, specific to this class of PI-4Ks, is sufficient for interaction with RabA4b, and the NH-terminal domain of PI-4Kβ1 specifically interacts with calcineurin B–like protein (AtCBL1), a Ca-sensor protein. Finally, tip localization of RabA4b membranes is disrupted by collapsing the tip-focused Ca gradient in root hair cells. Based on these observations, we propose a model for RabA4b and PI-4Kβ1 action during polarized root hair expansion. Rab GTPases perform their regulatory activities through specific recruitment of cytosolic proteins when the Rab GTPase is in its active (GTP bound) state (for reviews see ; ). Therefore, we screened a yeast two-hybrid expression library for interaction with a constitutively active (GTP bound) form of RabA4b. This resulted in identification of a clone containing the COOH-terminal portion of PI-4Kβ1 (PI-4Kβ1Δ1-421), which interacted with the constitutively active (GTP bound) form of RabA4b but not the dominant-negative (GDP bound) form (). Further, interaction of PI-4Kβ1Δ1-421 with RabA4b was selective, and no interaction with vacuole-localized RabG3c was detected (). contains 12 PI-4Ks in three separate families: PI-4Kα, -β, and -γ (; ). In yeast and animals, these PI-4K families localize to distinct subcellular compartments and have nonredundant functions (; ; ). Consistent with this, we detected no interaction of RabA4b with either PI-4Kα1 or -4Kγ6 (). Endosomal Rab GTPases from yeast (Ypt51) and mammals (Rab5) recruit phosphoinositide 3-OH kinases (PI-3Ks), which are necessary for PI-3P accumulation on endosomes (; ; for review see ). AtVPS34, the plant PI-3K, also failed to interact with either active or inactive RabA4b (). Collectively, these results suggested that recruitment of PI-4Ks by RabA4b was selective for PI-4Kβ1. We next determined which PI-4Kβ1 domains were responsible for RabA4b interaction. PI-4Kβ1 contains several domains (; ), including the catalytic domain at the COOH terminus and a lipid kinase unique (LKU) domain that is conserved in type III PI-4Ks of both the α and β families (; ). The NH domain is specific to β subfamily members in yeast, animals, and plants (), and a repetitive motif is unique to PI-4Kβ1 and -4Kβ2 in (; ). Testing different combinations of these domains indicated that the NH domain interacted with RabA4b (). Surprisingly, in the yeast two-hybrid system, full-length PI-4Kβ1 was not able to interact with RabA4b. This occurred even though the full-length PI-4Kβ1 was expressed at levels similar to the PI-4Kβ1Δ1-421 construct that did interact with RabA4b (unpublished data). Biochemical methods were used to confirm the RabA4b–PI-4Kβ1 interaction (). Affinity columns were generated using –expressed GST-RabA4b, loaded with either GTPγS (active form) or GDP (inactive form). [S] Met-labeled, in vitro–translated PI-4Kβ1 was passed over the column, and unlike the yeast two-hybrid assay, full-length PI-4Kβ1 was recruited to GST-RabA4b–GTPγS (). This indicated that the presence of the NH terminus did not abrogate PI-4Kβ1 interaction with RabA4b. The minimal piece necessary for interaction in the yeast two-hybrid system, the NH domain, also associated with the GST-RabA4b–GTPγS at levels similar to the full-length construct. Specificity of the PI-4Kβ1–RabA4b interaction was again demonstrated as PI-4Kγ6 and AtVPS34 were not recruited. Unlike yeast and mammals, has two type IIIβ PI-4Ks. At the protein level, PI-4Kβ2 is 83% identical to PI-4Kβ1. Like PI-4Kβ1, the PI-4Kβ2 NH domain also interacted with the constitutively active form of RabA4b (). Therefore, we concluded that both PI-4Kβ1 and -4Kβ2 proteins are effector proteins that are selectively recruited to the RabA4b GTPase in its active (GTP bound) form. We next examined the intracellular localization of PI-4Kβ1. EYFP-RabA4b–labeled membranes localize to the tips of growing root hairs (). Therefore, if PI-4Kβ1 and RabA4b interact in vivo, they should colocalize at the tips of these cells. We generated anti–PI-4Kβ1 antibodies that recognized an endogenous plant protein band of ∼125 kD, the predicted size of PI-4Kβ1 (, arrow). PI-4Kβ1 was primarily membrane associated and was not detected in soluble protein fractions from whole plant tissue. Using immunofluorescence and confocal microscopy, we determined that PI-4Kβ1 localized primarily to the tips of root hairs and overlapped with EYFP-RabA4b–labeled compartments (). This tip-localized PI-4Kβ1 fluorescence was specific, and no tip-localized fluorescence was detected when anti–PI-4Kβ1 antibodies were left out (). Further, these EYFP-RabA4b and PI-4Kβ1 membranes were distinct from plant Golgi compartments in these cells, as no significant overlap was observed between PI-4Kβ1 and the plant Golgi marker EGFP-GmManI (; ). These data are consistent with reports from the similar PI-4Ks in yeast and mammals. Direct interaction between mammalian PI4KIIIβ and Rab11 has been reported (). Also in yeast, Ypt31 has been shown to interact genetically with Pik1p, although no physical interaction was demonstrated (). We then asked whether PI-4Kβ1 functions in the tip growth of root hair cells. We identified SALK T-DNA insertion lines for both PI-4Kβ1 and -4Kβ2 genes (). Plants homozygous for insertions in either the PI-4Kβ1 or -4Kβ2 genes alone showed no obvious phenotype (unpublished data), even though by RT-PCR the appropriate PI-4Kβ transcript was not detected in either insertion line (). Using Western blot () and immunofluorescence (), we were not able to detect a functional PI-4Kβ1 protein in the PI-4Kβ1/β2 double mutant. Surprisingly, the highly similar PI-4Kβ2 protein was not recognized by this antibody (see online supplemental material, available at ). However, based on the lack of PI-4Kβ2 transcript (), no PI-4Kβ2 protein should be present. Double-mutant plants were smaller than wild type (WT), and plants homozygous for the PI-4Kβ1 T-DNA insertion but heterozygous for the PI-4Kβ2 T-DNA insertion were intermediate in size (). The root hairs of PI-4Kβ1/β2 double mutants were shorter and were abnormal compared with WT root hairs (). The percentage of aberrant root hairs per root was much higher in PI-4Kβ1/β2 double mutants than in WT plants (). This suggested that membrane trafficking required for proper polarized growth is defective in the absence of PI-4Kβ1/β2 activity, and these effects are most pronounced in highly polarized cells, such as the root hair. Present models suggest that recruitment of PI-4P interacting proteins is essential for sorting and budding of transport vesicles from the Golgi/TGN (, ). The secretion of cargo from the yeast Golgi complex requires generation of PI-4P by Pik1p (; ). How does lack of PI-4Kβ1/β2 affect the organization of the TGN and the RabA4b compartment? We examined the morphology of TGNs in the PI-4Kβ1/β2 double mutants by electron microscopy. Compared with WT plants (, arrows), the TGN in PI-4Kβ1/β2 double mutants showed a lighter staining pattern and clustered budding profiles (, arrowheads). This phenotype was consistently observed in three independent samples from PI-4Kβ1/β2 double mutants. Over 60% of TGN profiles in PI-4Kβ1/β2 double mutants ( = 49) displayed this aggregated appearance, which was never observed in WT cells ( = 51). Using antibodies specific to RabA4b (), RabA4b labeled both WT TGN budding profiles (, arrows) and the aberrant and aggregated structures in PI-4Kβ1/β2 double mutants (, arrowheads). Additionally, even when aggregated structures were not apparent in the double mutant, we observed fewer TGN budding profiles associated with Golgi complexes (). Although WT Golgi had a range of budding profiles, the majority of profiles had seven to nine distinct budding profiles per sample. In contrast, the PI-4Kβ1/β2 double mutant usually displayed only one to three budding profiles per sample. Therefore, loss of PI-4Kβ1/β2 function resulted in morphologically altered RabA4b-labeled TGN compartments, consistent with the interference of proper targeting and delivery of cell wall material. From these results, we conclude that PI-4Kβ1/β2 activity is necessary for proper organization of the TGN and post-Golgi secretion. Finally, we examined the role Ca binding proteins play in activation of PI-4Kβ1. Pik1p, the yeast () orthologue of PI-4Kβ1, is required for vesicular trafficking in the late secretory pathway (; ). Pik1p enzymatic activity is stimulated upon binding of frequenin, an EF-hand–containing Ca binding protein (; ). Similar Ca sensors, AtCBLs, have been described in (). We tested four representative members of this family, AtCBL1, -2, -3, and -5 for interaction with PI-4Kβ1 by yeast two-hybrid analysis (). AtCBL1 interacted with the NH terminus of PI-4Kβ1. AtCBL2 interaction was also sometimes detected, but growth rates were significantly lower than those of AtCBL1 (unpublished data). The AtCBL1–PI-4Kβ1 interaction was selective, as AtCBL3 and -5 did not interact with PI-4Kβ1. This suggested that the AtCBL1–PI-4Kβ1 interaction is evolutionarily conserved and that AtCBL1, acting as a Ca sensor, may modulate PI-4Kβ1 activity. Proper tip growth in root hair cells requires a tip-focused Ca gradient (for review see ). The interaction between PI-4Kβ1 and AtCBL1 implicates a role for Ca in the regulation of PI-4Kβ1 activity. Therefore, we hypothesized that dissipation of the Ca gradient in root hairs would alter the proper localization of EYFP-RabA4b compartments. Treatment of root hairs with , a Ca ionophore results in rapid loss of the tip-focused Ca gradient (). When we treated growing root hair cells with , a rapid dispersal of tip-localized EYFP-RabA4b was observed, accompanied by inhibition of root hair growth (). These results support our hypothesis that localization of RabA4b compartments is dependent on proper recruitment and activation of PI-4Kβ1 activity. Initiation of tip growth in root hairs and pollen tubes is dependent on the formation of a tip-focused Ca gradient (for reviews see ; ; ). However, it is becoming increasingly clear that lipid-derived signaling molecules play important roles in establishing and maintaining these Ca gradients. We have shown that the NH terminus of PI-4Kβ1, including the LKU domain, is capable of binding to AtCBL1, a plant homologue of frequenin. In yeast, Pik1p activity is stimulated upon binding of frequenin to the LKU domain in a Ca-independent manner (). But Ca does enhance association of frequenin with membranes, which may promote the interaction of frequenin with Pik1p in vivo. If Ca binding in the root hair tip stimulates AtCBL1 recruitment to RabA4b membranes, this would ultimately result in increased PI-4P production by PI-4Kβ1 on this compartment. In summary, the recruitment and activity of PI-4Kβ1 on RabA4b-labeled membranes plays an important role during polarized expansion of root hairs. Previously, we showed that localization of RabA4b-labeled membranes at root hair tips is associated with tip-restricted expansion (). Our model hypothesizes that recruitment of PI-4Kβ1 and AtCBL1 to RabA4b-labeled membranes results in localized PI-4K activity and enrichment of PI-4P on these compartments. Enrichment of PI-4P may stimulate recruitment of PI-4P binding domain proteins (; ; ), or the PI-4P could be delivered to tip-localized plasma membrane domains via fusion of RabA4b-labeled secretory compartments. There it might be a targeting determinant itself, or plasma membrane–localized PIP-5Ks, recruited to the root hair and pollen tube tips by Rop GTPases (), may phosphorylate PI-4P to PI-4,5P. This is supported by the observation that PI-4,5P is primarily associated with plasma membranes in the tips of root hairs and pollen tubes (; ; ). Therefore, RabA4b-dependent recruitment of PI-4Kβ1 would integrate the perception of tip-focused Ca gradients and generation of phosphoinositide-derived signaling molecules for the organization of post-Golgi secretory compartments at the tips of growing root hairs. Full-length clones from TAIR of PI-3K (), PI-4Kβ1 (), and PI-4Kγ6 () in the pUNI51 vector were cloned into the pGAD vector for yeast two-hybrid analysis. The PI-4Kα1 sequence in the pFastBac HT vector (a gift from W. Boss, North Carolina State University, Raleigh, NC; ) was cloned into the pGAD vector. Pieces of PI-4Kβ1 were PCR amplified and cloned into either the pGAD or pGBK vector. The construction of AtCBL1, -2, -3, and -5 in the pAS vectors was described previously (a gift from Y. Guo, National Institute of Biological Sciences, Beijing, China, and J.K. Zhu, University of California, Riverside, Riverside, CA; ). The yeast strain AH109 (CLONTECH Laboratories, Inc.) was used for two-hybrid experiments. Using a LiAc transformation protocol (CLONTECH Laboratories, Inc.), the CD4-22 library (Arabidopsis Biological Resource Center; ) was screened (∼6 million transformants) and plasmids were rescued from transformants surviving high-stringency selection conditions (−AdeHisLeuTrp + 7.5 mM 3-AT). Plasmids from 127 putative positive yeast colonies were rescued and sequenced. Of these, 30 were tested for the ability to interact with the active or inactive forms of RabA4b, and 4 showed nucleotide specificity in the interaction. Drop assays were performed by allowing inoculated cultures to grow for 2 d and diluting them to an OD of 0.02, of which 10 μl drops were spotted on selective and nonselective medium. RabA4b was PCR cloned into pGEX-6 (GE Healthcare) and transformed into BL21 cells. GST-RabA4b protein was expressed, and active or inactive RabA4b affinity columns were prepared as previously described (). In vitro–translated proteins were generated using a TNT-coupled reticulocyte lysate system (Promega). 45 μl PI-4Kβ1 in vitro–translation product and 150 μl nucleotide stabilization buffer (NS; containing 1 mM GTPγS or GDP) were incubated with 20 μl (18 mg/ml) GST active or inactive RabA4b beads for 2 h at 4°C; washed twice with NS (10 μM nucleotide); washed once with NS (250 mM NaCl and 10 μM nucleotide); washed once with 20 mM Hepes, pH 7.5, 250 mM NaCl, and 1 mM DTT; and eluted with 40 μl elution buffer. 40 μl SB was added to the eluate and boiled for 5 min, and 30 μl was analyzed by SDS-PAGE followed by fluorography. 5% of total in vitro–translation product was also loaded. A peptide (CTRQYDYYQRVLNGIL) corresponding to the COOH-terminal AtPI-4Kβ1 sequence was synthesized (Sigma-Aldrich) and used to generate rabbit polyclonal antibodies. Anti-peptide antibodies were affinity purified using the peptide immobilized on Sulfolink beads (Pierce Chemical Co.). seedlings (10–14-d old) were ground in 20 mM Hepes, pH 7.5, 100 mM NaCl, 5 mM MgCl, 1 mM DTT, 2.5 mM GTP, and protease inhibitors (Roche) and then spun at 2,000 . The postnuclear supernatant was collected and spun at 100,000 for 1 h at 4°C. The supernatant (soluble fraction) was separated from the pellet (membrane fraction), and each fraction was analyzed by immunoblotting. EYFP-RabA4b (), EGFP-GmManI (a gift from A. Nebenfuhr, University of Tennessee, Knoxville, TN; ), and PI-4Kβ1/β2 double-mutant seedlings were processed for immunolocalization as described in and using a “freeze-shattering” method (). After primary and secondary antibody incubations, slides were mounted with MOVIOL (Calbiochem). Samples were observed on a confocal microscope (LSM 510; Carl Zeiss MicroImaging, Inc.). Z stacks of root hairs were taken and three-dimensional projections from these stacks were used for the final images. The PI-4Kβ1 and -4Kβ2 T-DNA insertion mutants were obtained from the SALK T-DNA collection (SALK_040479 and SALK_098069; Arabidopsis Biological Resource Center). Root hairs of WT and mutant plants were imaged using a confocal microscope. The T-DNA insertion site in each line was confirmed by sequencing (primers LBB1 5′-GCGTGGACCGCTTGCTGCAACT-3′, B1 5′-TCCAGGCCTTCCTCTCAAAG-3′, and B2 5′-AACCTACCAGGTTGGGACTTG-3′). root tips were loaded in 0.1 M sucrose, frozen in a high-pressure freezer (Baltec HPM 010; Technotrade), and transferred to liquid nitrogen. Substitution was performed in 0.1% uranyl acetate plus 0.2% glutaraldehyde in acetone at −80°C for 72 h and warmed to −50°C for 24 h. After several acetone rinses, samples were infiltrated with Lowicryl HM20 (Electron Microscopy Sciences) for 72 h and polymerized at −50°C under UV light for 72 h. Sections were mounted on formvar-coated nickel grids and blocked for 20 min with 5% (wt/vol) nonfat milk in PBST (0.1% Tween 20). Sections were incubated with primary antibody for 1 h at RT. The sections were rinsed with PBST (0.5% Tween 20) and transferred to the secondary antibody conjugated to 15 nm gold particles for 1 h. Controls were performed by omitting the primary antibody. Sections were stained with 2% uranyl actetate in 70% methanol for 10 min followed by Reynolds's lead citrate (2.6% lead nitrate and 3.5% sodium citrate, pH 12) and observed in a transmission electron microscope (CM120; Philips). Images of TGNs were used for quantification of budding profiles and overall abnormality. Abnormal TGNs were defined as having aggregated budding profiles and a lighter staining pattern. seedlings were grown, treated, and analyzed as previously described (). (Sigma-Aldrich) was dissolved in DMSO and added at a concentration of 2 nM in 0.25× MS to growing root hairs. Fluorescent signal located within the proximal 15% of the length of the root hair was defined as tip fluorescence, and this was presented as a percentage of the fluorescence detected in the entire root hair. Fig. S1 shows that purified anti-PI-4Kβ1 antibodies do not recognize PI-4Kβ2. The supplemental text describes how in vitro transcription/translation products of HA-tagged COOH-terminal domains or either PI-4Kβ1 or -4Kβ2 were separated and analyzed to test the specificity of the anti–PI-4Kβ1 antibodies. Online supplemental material is available at .

Early embryogenesis in many organisms, including , , and is characterized by rapid progression through the cell cycle (for review see ). Features of early embryonic cell cycles that distinguish them from somatic cycles include cell division in the absence of cell growth and a lack of Gap phases. Another important difference between somatic and embryonic cell cycles concerns the utilization of S phase checkpoint pathways. In somatic cells, the S phase checkpoint senses DNA damage and responds by delaying progression into mitosis (for reviews see ; ). The protein kinases ATR and Chk1 are central to S phase checkpoint signaling. DNA damage causes replication fork stalling, which in turn activates ATR and promotes the ATR-directed phosphorylation of Chk1. Activated Chk1 delays cell cycle progression through attenuation of core cell cycle regulators such as the Cdc25 protein phosphatase. Thus, in somatic cells, a major function of the ATR checkpoint is to delay cell cycle progression in response to DNA damage until replication can finish. In embryonic cells, the ATR checkpoint is activated by endogenous, developmentally programmed cues. The nature of these signals is not defined, but it is clear that developmental activation of the checkpoint is important for regulating the timing of cell division during early embryogenesis. Two examples highlight this importance. In , the (ATR) and (Chk1) genes affect a developmentally programmed slowing of the cell cycle that occurs at the midblastula transition (, ; ; ). Fly embryos perform 13 rounds of rapid and synchronous cell division before the midblastula transition. After cycle 13, the checkpoint is activated by an endogenous signal, and this slows the cell cycle down. Slowing of the cell cycle in turn allows for zygotic transcription to begin, and the control of cell division is thereby transferred from maternal to zygotic regulators. In or mutants, the cell cycle does not slow down, zygotic control of the cell cycle does not happen on schedule, and the embryo dies. Therefore, in , the checkpoint plays an important role in remodeling the cell cycle so that zygotic transcription can begin on schedule. Another example of DNA damage–independent utilization of the ATR checkpoint is found in The one-cell embryo, or P0 cell, divides asymmetrically to produce the smaller (P1) and the larger (AB) daughter cells. The next round of cell division is asynchronous: AB divides first, followed by P1 about 2 min later. This 2-min delay is controlled in part through differential activation of the S phase checkpoint in the P1 cell (). Developmental checkpoint activation in the early embryo requires the homologues of ATR () and Chk1 (). Checkpoint-mediated asynchrony in cell division is extremely important to embryonic patterning in . When asynchrony is reduced, through loss of , the germ line fails to develop and the animal is sterilized (; ). Extending the asynchrony also has deleterious consequences. Hypomorphic mutations in , a gene encoding DNA polymerase α, cause replication problems that result in inappropriate activation of the pathway (; ). The –mediated activation of extends the asynchrony in cell division, and this results in mislocalization of developmental regulators, embryonic patterning defects, and lethality (). From these examples it is clear that, although checkpoint activation is important for development, it must only occur in response to developmental signals and not in response to unscheduled events such as replication problems. A common source of replication problems in wild-type cells is DNA damage, and thus it would seem that early embryogenesis in would be particularly sensitive to DNA damage because of the deleterious consequences of unscheduled checkpoint activation. Paradoxically, this is not so, as previous work has shown that wild-type embryos are resistant to relatively high amounts of both UV light and the alkylating agent methyl methanesulphonate (MMS; ; ), two DNA-damaging agents that are known to cause replication problems and subsequent checkpoint activation (; ; ). We resolve this paradox by showing that the checkpoint is actively silenced during the DNA damage response in early embryos. We go on to define genetic requirements and the basis for checkpoint silencing. Our results identify a novel developmental mechanism that ensures that cell cycle progression is not attenuated by DNA damage, thus providing embryos with a chance of survival even when their chromosomes are heavily damaged. It was not known whether the checkpoint pathway can sense the types of DNA damage that cause replication stress, such as alkylation or UV light–induced damage. Previous work has shown that nuclei in the mitotic zone of the hermaphrodite gonad, a nonembryonic tissue, undergo checkpoint-dependent cell cycle arrest in response to replication blocks and ionizing radiation (; ). This arrest is reflected by a reduction in nuclei number and an increase in nuclear size. To see whether MMS and/or UV light induced checkpoint activation in the gonad, animals were exposed to 0.005% MMS and then fixed and stained with Hoechst 33258 to visualize nuclei. MMS reduced the number of nuclei within the mitotic zone from a mean of 35 to a mean of 22 (; and ). The effect of MMS in the germ line was reversed when the checkpoint gene , the worm orthologue of ATR (), was depleted by RNA interference (RNAi; and ). Similar results were obtained when animals were irradiated with 100 J/m of UV light (). We conclude that germ cells undergo checkpoint-dependent cell cycle arrest upon exposure to either MMS or UV light and that, therefore, the checkpoint can indeed sense MMS- and UV light–induced damage. The effect of MMS and UV light on cell cycle progression in the early embryo was examined next. shows the major events of the first mitotic interphase in the early embryo. After fertilization (step i), the female pronucleus migrates across the embryo, or P0 cell, where it meets and apposes the male pronucleus (steps ii–iv). DNA replication then finishes, and mitosis is initiated by nuclear envelope breakdown (step v). Previous work has shown that replication stress–induced checkpoint activation, as triggered by the replication inhibitor hydroxyurea (HU), occurs at the one-cell stage (). This checkpoint requires the gene () and prevents the transition from step iv to v in . Animals were exposed to MMS, UV light, or, as a positive control, HU, and the timing of the first cell cycle was determined by direct microscopic visualization of living embryos. As shown in , and as previously reported (), when embryos were treated with HU, there was a significant delay in progression through the P0 cell cycle. This delay was checkpoint dependent, as it was reversed after depletion of by RNAi. In contrast to HU, we did not detect a significant P0 cell cycle delay when the embryos were exposed to MMS or UV light (). We conclude that the amounts of MMS or UV light that are sufficient to activate a checkpoint response in the germ line (0.005% and 100 J/m, respectively) cannot do so in early embryos. We refer to this phenomenon as early embryonic checkpoint silencing. We next tested how much MMS embryos could endure before a delay in cell division was detected. For this, we timed the first cell division after exposure to a range of MMS concentrations and found that concentrations >0.005% caused both a delay in progression through S phase () and high levels of embryonic lethality (). These data show that the ability of embryos to avoid a checkpoint response to MMS is saturable. They also indicate that checkpoint silencing and survival of DNA damage are linked, and this is consistent with previous work showing that even modest perturbations in the timing of cell division are lethal to the developing embryo (). A simple explanation for checkpoint silencing is that embryos rapidly repair damaged DNA. However, extensive analysis of the kinetics of DNA repair in has been reported (; ; ; ), and these studies demonstrate that repair is unlikely to account for checkpoint silencing in the embryo. For example, >80% of photoproducts remain in embryos 3 h after a dose of 50 J/m of UV light is delivered (). The data in were collected 1 h after a dose of 100 J/m was delivered, and thus the embryo could not possibly have repaired even a modest percentage of the damage in that short a period of time. We conclude that cell cycle progression occurs unimpeded even when the level of damage present greatly exceeds the capacity of the embryo to repair it. If embryonic cell cycle progression is truly independent of repair, then mutant embryos that are deficient in DNA repair would nonetheless exhibit normal cell cycles after DNA damage. To test this, we examined cell cycle progression in early embryos. mutant embryos have a defect in excision repair and are consequently very sensitive to both MMS () and UV light (; ). The rate of repair in embryos has been determined and is threefold lower than wild type (). Despite this reduced capacity for repair, however, the timing of cell division in mutant embryos was indistinguishable from wild type after exposure to either UV light or MMS (). The dose of UV light used in the experiment is sufficient to kill 100% of the mutant embryos and <10% of wild-type embryos (unpublished data; see for UV sensitivity of mutants). Interestingly, another radiation- and MMS-sensitive mutant, , showed altered progression through the first cell cycle after exposure to UV or MMS (). mutant embryos delayed progression through the P0 S phase in a manner that was dependent on DNA damage and similar to wild-type embryos exposed to HU. The damage-induced delay in embryos was checkpoint dependent, as it was reversed when was depleted by RNAi (). Importantly, the rate of repair in mutants is indistinguishable from wild type (). Thus repair-deficient mutants have normal cell cycles after DNA damage, whereas repair-proficient mutants do not. This shows that a process that is independent of DNA repair is responsible for preventing checkpoint activation during the early embryonic cell cycle. Consistent with this, we also found that RNAi-mediated depletion of another excision repair gene, the homologue of the human XPF endonuclease (F10G8.7), renders embryos extremely sensitive to both UV light (Fig. S1, available at ) and MMS () yet had no affect on cell cycle progression in the early embryo (). We conclude that checkpoint silencing is independent of lesion repair (based on the results with and F10G8.7) yet nonetheless under genetic control (based on the results with ). The finding that checkpoint silencing is under genetic control prompted a search for genes that silence the checkpoint when DNA damage is present. The gene has not yet been cloned, and we are currently working toward accomplishing this. Recent work from our laboratory has shown that the gene, which encodes an E3 SUMO ligase related to yeast and human PIAS1, is an important participant in the embryonic DNA damage response in (). Depletion of by RNAi renders embryos sensitive to both MMS () and UV light (Fig. S1). MMS-exposed RNAi early embryos display abnormal nuclear morphology, characterized by fused nuclei and anaphase bridging (). These results suggested that is important for early embryonic cell cycle progression when damage is present and prompted us to examine the kinetics of cell division in RNAi early embryos. MMS exposure delayed progression through the P0 S phase in RNAi embryos. At 0.005% MMS, we observed an ∼440-s delay () and at 0.001% MMS the delay was ∼300 s (not depicted). These data demonstrate that S phase takes longer in RNAi embryos exposed to 0.001% MMS than it does in wild-type embryos exposed to 10-fold more MMS (the wild-type delay at 0.01% MMS was ∼120 s; ). To determine whether these MMS-induced effects were caused by activation of the checkpoint, we codepleted with As was the case with , codepletion of with reversed the MMS-induced delay in progression through the P0 cell cycle (). This result demonstrates that activity suppresses checkpoint activation in response to DNA damage in the early embryo. To see whether the effect of on checkpoint activation was specific for DNA damage, we next examined checkpoint activation in RNAi embryos in response to both HU and developmental signals. For the HU experiment, we used a lower concentration of HU than that used in (25 as opposed to 75 mM), and this resulted in a more modest delay in cell division in wild-type embryos (∼160 s delay after 25 mM HU in contrast to the ∼475-s delay after 75 mM; and ). If functions to suppress HU-induced checkpoint activity, we would expect this modest delay to be extended in RNAi embryos, but this did not occur (). Similar results (i.e., no difference between wild-type and RNAi embryos) were obtained when 75 mM HU was used to trigger a stronger checkpoint response (unpublished data). We conclude that although activity reduces the checkpoint response to DNA damage, it has no effect on checkpoint activation by HU. To examine the effect of loss of on checkpoint activation in response to developmental signals, we analyzed the second round of cell division in early embryos. As described in the Introduction, there is a checkpoint-dependent delay in division of the P1 cell relative to the AB cell during normal development (). The delay normally lasts 2 min; however, when is depleted by RNAi, it is reduced to ∼1 min (; ). If negatively controlled the checkpoint response to developmental signals, we would expect the delay to be extended in RNAi embryos, but this was not the case, as RNAi embryos showed the same delay as wild type (). When MMS was included, however, the lag was significantly extended in RNAi embryos and only very modestly extended in wild type (). We conclude that functions to suppress checkpoint activity specifically in response to DNA damage and not in response to HU-induced stalled replication forks or developmental signals. One explanation for the ability of to suppress damage-induced checkpoint activation is that it promotes the rapid replication of damaged DNA. In both and yeast, the checkpoint response to MMS-induced damage is known to require the stalling of replication forks (; ); thus, if prevents damage-induced fork stalling, then checkpoint activation would not be expected to occur. To directly assess a role for in the replication of damaged DNA, a previously described assay system was used to monitor DNA replication in the early embryo (; ). Egg shells from four-cell embryos were permeabilized and the samples treated with cytochalasin B to block cytokinesis. The embryos were then cultured for 1 h before fixation and DNA staining. Despite the block to cell division, the DNA replication cycle continues unabated, and after 1 h this results in embryos that contain multiple nuclei in each of the four cells (). The appearance of multiple nuclei is dependent on DNA synthesis because it does not occur in the presence of the replication inhibitors aphidicolin or HU (; ). MMS did not affect the appearance of multinucleated cells in wild-type embryos (, compare panels II and III). In contrast, the combination of MMS and RNAi caused a defect in DNA replication, as these embryos failed to produce multinucleated cells (, panel VI). This was not observed in undamaged RNAi embryos (, panel V), demonstrating that is required for the replication of damaged, but not undamaged, chromosomes. We also note that the replication defect in MMS-exposed RNAi embryos was uniform and occurred in all four cells of the embryo. As the checkpoint is only highly active in one of these cells (the P lineage cell; ) in intact embryos, this result suggests that the requirement for in the replication of damaged DNA is independent of the checkpoint status of the cell. It is possible, however, that permeabilization perturbs the asymmetric distribution of the checkpoint within the four-cell embryo. The double-strand break repair protein RAD-51 is known to accumulate in immunologically detectable foci when replication forks are stalled by DNA damage (; ). The results in and show that is required for S phase progression () and for DNA replication (), specifically when chromosomes are damaged. To determine whether loss of induces RAD-51 foci, we stained early embryos (<100 cells) with anti–RAD-51 antibodies (). In the absence of MMS, we did not detect RAD-51 foci in either wild-type or RNAi early embryos or in the mitotic zone of the hermaphrodite gonad (, panels I–III). In MMS-exposed animals, we could detect robust RAD-51 foci formation within the mitotic zone of the hermaphrodite gonad (, panel VI) but not in wild-type early embryos (, panel IV). In contrast to wild type, RNAi embryos readily formed RAD-51 foci in response to MMS (, panel V). These data indicate that stalled replication forks, as inferred by the presence of MMS-induced RAD-51 foci, form in cells where MMS triggers the checkpoint (wild-type germ lines and RNAi embryos) but not in cells where the checkpoint is silenced (wild-type embryos). RAD-51 foci were not observed in MMS-exposed RNAi embryos (unpublished data), indicating that attenuation of the checkpoint alone is not sufficient to explain damage-induced foci formation. These results provide further evidence that loss of causes replication fork stalling in MMS-exposed embryos. All organisms contain mechanisms for promoting the replication of damaged DNA in a manner that does not rely on physical repair of the lesion. These pathways, termed postreplication repair or lesion bypass, rely on either translesion synthesis or homologous recombination to rescue replication forks that stall at sites of damage (for review see ). To explore an involvement of lesion bypass pathways in embryonic checkpoint silencing, we determined the effect of inactivation of known lesion bypass components on P0 cell cycle progression after DNA damage. The role of homologous recombination was assessed by studying embryos derived from adults carrying homozygous deletion mutations in the essential recombination genes and . Neither mutant displayed a defect in P0 cell cycle progression after MMS exposure, showing that homologous recombination is not essential for checkpoint silencing (). This is consistent with a lack of RAD-51 foci in MMS-exposed wild-type embryos (). The other major lesion bypass pathway in eukaryotes is translesion synthesis, where specialized DNA polymerases are recruited to the replication fork to synthesize DNA across damaged bases on the template strand (for review see ). In yeast and human cells, access of translesion polymerases to sites of damage is thought to occur through -mediated ubiquitination of proliferating cell nuclear antigen (PCNA), a DNA replication clamp protein (; ; ). Yeast mutants are sensitive to MMS and UV light and do not show DNA damage–induced mutagenesis, a hallmark of translesion synthesis (for review see ). To determine whether the Rad6 pathway is responsible for checkpoint silencing in early embryos, we examined the MMS response in mutants. The gene represents the sole orthologue of budding yeast , and expression of the gene in yeast is sufficient to rescue the translesion synthesis defect (). Surprisingly, embryos derived from adults carrying a homozygous deletion of the gene did not display MMS sensitivity () and progressed normally through the P0 cell cycle after MMS exposure (). Thus, in embryos, the Rad6 orthologue is not important for the response to MMS-induced damage. It was possible that in translesions polymerases can access sites of damage in a Rad6/–independent manner. To pursue this hypothesis, we screened all five of the identifiable translesion polymerases present in by RNAi for MMS sensitivity in embryos. The genes that we screened included putative orthologues of human Polθ (W03A3.2), Polη (F53A3.2), Polκ (F22B7.6), Polζ (Y37B11A.2), and Rev1 (ZK675.2). The assignment of these genes to their putative human counterparts is based purely on sequence conservation, as information on the biochemical properties of the encoded proteins is not available. Of the five, we found that RNAi-mediated depletion of both the Polη orthologue and the Polκ orthologue caused MMS sensitivity in embryos (). Only RNAi, however, delayed progression through P0 S phase (). This delay was dependent on MMS and was reversed upon codepletion of (), demonstrating that like and , loss of allows checkpoint activation in the early embryo. Consistent with this, we observed that RAD-51 foci could be detected in RNAi embryos, in an MMS-dependent manner (). RAD-51 foci were not observed in MMS-exposed early RNAi embryos (unpublished data). These data indicate that –mediated translesion synthesis is the lesion bypass mechanism used by early embryos to silence the checkpoint during the DNA damage response. summarizes the findings reported here and integrates them with previous work on cell cycle control in the early embryo. Previous studies have shown that developmental signals, the nature of which are unknown, trigger checkpoint activation and that this contributes to the asynchrony in cell division that is required for developmental patterning and germ line formation (, shaded portion). Thus, developmental signals represent one class of input into the embryonic checkpoint pathway. Another type of input is stalled replication (, unshaded portion). Stalled replication has been induced in early embryos through mutations in () or through the use of HU (; this study). Embryonic sensitivity to stalled replication has been documented; it causes checkpoint activation and extends the natural asynchrony of cell division (; ). This in turn perturbs development and kills the embryo. The focus of the work presented here was on another inducer of stalled replication, DNA damage. We have found that early embryos do not stall replication when their chromosomes are damaged and that protection against damaged-induced stalled replication is conferred by , , and . These results explain how the checkpoint can be accessed by developmental signal–based inputs and insulated from DNA damage–based inputs. The checkpoint is not insulated from mutant or HU-based inputs, but these are conditions that are irrelevant to wild-type worms in their natural environments. Embryonic checkpoint control has been studied in other organisms, most notably and . In both of these organisms, checkpoints that respond to DNA damage are not evident until after the rapid cleavage cycles have ended, at the midblastula transition (; ). In , the lack of DNA damage checkpoint activation in early embryos is likely due to a low DNA/cytoplasm ratio, as it has been recently demonstrated that increasing the amount of damaged DNA in younger frog embryos results in a checkpoint-dependent delay in cell division before the midblastula transition (). The interpretation of this is that the checkpoint signal is not strong enough to neutralize the mitosis-promoting capacity of the cytoplasm until the proper ratio is achieved. Thus, in frog and fly embryos checkpoint avoidance occurs passively. This is in contrast to the active mechanism that we have discovered in , and the difference is likely due to when the checkpoint functions during development. In frogs and flies the checkpoint is not needed until the midblastula transition, whereas in worms it is used from the first division onward. Rapid embryonic cell cycles occur in all major animal phyla (for review see ). is no exception, as the early cycles last only 10–40 min. It is possible that the rapid cycling allows no time for lesion repair, and therefore lesion bypass may be the only viable option for embryos exposed to DNA-damaging agents. This is in contrast to the mitotic cells of the gonad that can survive delays in cell division and go on to divide normally. Indeed, we have shown that mitotic gonad cells arrest in a checkpoint-dependent manner upon MMS or UV exposure and RAD-51 foci are clearly evident ( and and ). Additionally, we were unable to detect any cell cycle arrest phenotype in the germ lines of RNAi animals after either high () or low (not depicted) MMS or UV exposure, suggesting that this pathway is not a major component of the germ line DNA damage response. Our results therefore demonstrate a distinct difference between embryonic and germ line responses to DNA damage that could be explained by embryonic sensitivity to the timing of cell division. The molecular basis for this difference is not yet known but likely involves differential expression and/or regulation of members of the – pathway. Both and are components of an embryonic checkpoint silencing pathway that bypasses MMS-induced lesions. The identification of as the primary polymerase required for progression through S phase in MMS-exposed early embryos is somewhat surprising, as Polη in yeast and human cells is primarily associated with UV light–induced damage (for review see ). However, budding yeast Polη () efficiently bypasses abasic sites (a major MMS-induced lesion) when coupled to PCNA in vitro () and is required for maximal abasic site bypass in vivo (). mutants are accordingly MMS sensitive (). These findings therefore suggest that a role for in responding to MMS-induced damage in could be explained by the ability of the enzyme to bypass abasic sites. Although the role of as a translesion polymerase is directly related to replicating damaged DNA, it is not clear what role the E3 SUMO ligase, , actually plays in this process. Recent work has shown that , which sumoylates PCNA, functions to ward off homologous recombination during lesion bypass through recruitment of the antagonist to the replication fork (; ). It is therefore possible that promotes translesion synthesis in embryos through negative regulation of recombination, although we do not favor this model, as loss of still negatively affects S phase progression in mutant embryos (). Thus, the elimination of recombination in mutants does not suppress the RNAi phenotype, and this argues against a role for in preventing recombination. One possibility is that functions in polymerase switching at sites of DNA damage, and biochemical analysis of the polymerase switch reaction in embryos will be required to determine whether this is so. We note that in yeast and mammalian cells polymerase switching is controlled by the Rad6 E2 ubiquitin-conjugating enzyme and the Rad18 E3 ubiquitin ligase; however, we have shown here that the Rad6 orthologue is not required for translesion synthesis in embryos (), and there is no recognizable Rad18 homologue present in the worm genome. Our results also shed light on the relationship between checkpoint activation and translesion synthesis, as they suggest that in the early embryo translesion synthesis trumps checkpoint activation to ensure that DNA damage does not slow the cell cycle down. How decisions are made at stalled replication forks to activate one pathway over another is not understood and is an active area of research. Our data show that, in the early embryo, translesion synthesis is so efficient that checkpoint activation fails to occur, even when relatively high levels of damage are present. This may reveal a general principle, in that during the DNA damage response the default response is to access the translesion synthesis pathway and that checkpoint activation can only occur at levels of damage that saturate translesion synthesis. Alternatively, embryo-specific factors may exist that allow translesion synthesis to supersede checkpoint activation. The use of the POLH-1 translesion polymerase to prevent fork stalling during the early embryonic cell cycles answers the question of how embryos bypass checkpoint activation and so survive exposure to DNA-damaging agents. Although this pathway allows embryonic cells to divide on schedule, translesion polymerases are notoriously error prone, and use of this pathway predicts that embryos likely trade survival for an increase in mutation frequency. This is especially true when abasic sites, which are noncoding forms of damage, are considered. Thus, it appears that during evolution there has been stronger selection for adherence to the schedule of cell division than for error-free replication during early embryogenesis, and understanding the basis for this preference will be the goal of future studies in this system. The N2 Bristol strain was used as wild type in all control experiments and for all RNAi experiments. SP482 (), SP488 (), TG5 (), VC531 (), and VC18 () strains were obtained from the Caenorhabditis Genetics Center. Animals were maintained as described previously (). RNAi by feeding was performed for F10G8.7, W03A3.2, Y37B11A.2, ZK675.2, , , and as described previously (). All bacteria were cultured for 24 h at 37°C in Terrific Broth containing 50 μg/ml ampicillin, seeded onto nematode growth media () plates containing 5 mM IPTG, and allowed to dry overnight. With the exception of RNAi, worms were grown for two generations on RNAi bacteria. F1 progeny of RNAi worms are sterile; therefore, RNAi was fed for one generation and analysis was performed on F-1 embryos. For codepletions, worms were first grown on RNAi bacteria for one generation and then moved as F1 L1s onto a plate containing a 1:1 mixture of the feeding vectors. and RNAi was accomplished by soaking (). codepletions were accomplished by first feeding worms bacteria and then soaking P0 L4s in double-stranded RNA. Worms were then plated onto regular media or media containing 0.05 mg/ml MMS (Sigma-Aldrich), both seeded with RNAi bacteria, and analysis was performed on their progeny. Worms were collected and placed in a drop of M9 buffer for dissection. Released embryos were then transferred to agarose pads (2% SeaKem Gold agarose in water) in a small volume of M9 and visualized under Nomarski optics on a microscope (BX51 TF; Olympus). Embryos exposed to MMS were timed after 20 h of exposure to plates containing 0.05 mg/ml MMS. Embryos exposed to HU (Calbiochem) were timed after 5 h of exposure to plates containing 75 mM HU. Embryos exposed to UV light were timed 1 h after irradiation. Irradiation was performed by placing an open dish of worms in a Stratalinker (Stratagene). To measure the P0 S phase, timing started when the female pronucleus passed the midline of the embryo. Timing continued until nuclear envelope breakdown had occurred, just before first mitosis. Because it is unclear when replication initiates, this represents the timing of a partial S phase (). The persistence of three-cell embryos was determined by timing the interval between cytokinesis of the AB cell and cytokinesis of the P1 cell. Worms were dissected on glass microscope slides and permeabilized by freeze cracking. Slides were fixed for 10 min in methanol/formaldehyde fixative at −20°C and washed in PBS Tween 20. Slides were then incubated with anti–RAD-51 antibody () at 1:200 overnight followed by a 2-h incubation in FITC-tagged anti-rabbit secondary antibody. DNA staining was accomplished by adding 10 μl of 10 μg/μl Hoechst 33258. To count nuclei in the mitotic zone of the gonad, adult worms were fixed and stained with Hoechst 33258. The distal tip of the gonad was then visualized using fluorescence microscopy, and the number of nuclei within a constant volume was counted. Embryos were prepared for culturing as described previously (). MMS exposure was accomplished by culturing worms for 20 h on 0.05 mg/ml MMS plates and then exposing permeabilized embryos to 0.2 mg/ml MMS in egg growth media. After incubation, embryos were stained with Hoechst 33258 and visualized on a microscope. Pictures were captured using a monochrome camera (SPOT RT; Diagnostic Instruments). L4 F1 worms grown on plates containing the appropriate bacterial expression vectors were transferred to plates containing 0.05 mg/ml MMS. Eggs laid by these worms were collected over time and scored for survival as described previously (). The images shown in , , , and were obtained as follows. All images were collected on a microscope. The type, magnification, and NA of the objective lenses were UPlanAPO, 60× oil, and NA 1.40, respectively. The experiments were performed at room temperature using Hoechst 33258 and FITC-labeled secondary antibodies as fluorochromes. Images were captured on a camera (model 2.1.1; Diagnostic Instruments) and processed using SPOT Advanced version 3.2.4 software (Diagnostic Instruments). Fig. S1 shows that F10G8.7 and embryos are sensitive to UV light. The supplemental text describes UV sensitivity assays and the observation that RNAi causes in UV sensitivity in embryos. Online supplemental material is available at .

Microtubules are intrinsically dynamic cytoskeletal structures. Rather than reaching a steady-state length, single microtubules switch stochastically between states of growth and shrinkage, and transitions between these states are known as catastrophe (growth to shrinkage) and rescue (shrinkage to growth), respectively. This dynamic instability allows quick reorganization of the microtubule cytoskeleton, such as from the interphase microtubule array to the mitotic spindle (for review see ). Microtubule-associated proteins (MAPs) regulate microtubule dynamics. Some MAPs increase the net microtubule polymer mass (stabilizers), whereas others decrease it (destabilizers). As microtubules grow and shrink at ends, targeting to ends is considered an important feature of most MAPs (for review see ). Three mechanisms of end targeting have been identified (for review see ; ) and described as transport, hitchhiking, and direct end binding. A transported MAP binds a motor, which carries it to the end of a microtubule. For example, in , Kar9 is transported to plus ends likely by associating with the kinesin motor Kip2 (; ). A hitchhiking MAP is recruited to ends by interacting with another MAP already bound there. For example, the dynein homologue Dyn1 interacts with end-bound Bik1 and Pac1 (; ). A direct end-binding MAP recognizes particular features of microtubule ends or copolymerizes with tubulin. Examples of direct end-binding MAPs are CLIP-170, mitotic centromere-associated kinesin (MCAK)/XKCM1, and members of the Dis1/XMAP215 family such as budding yeast Stu2p or human ch-TOG (; ; ; ; ; ). Once at microtubule ends, MAPs influence microtubule dynamics by a still largely obscure mechanism. The best understood factors in this regard are members of the kinesin-13/MCAK family, which destabilize microtubules by inducing catastrophes. Kinesin-13 family members bind directly to microtubule ends and transform the end structure into one that is normally found in microtubules undergoing disassembly (; ; ). A cycle of end binding, microtubule conformational change, and recycling of the MCAK family member requires the energy of ATP hydrolysis. Less is known about end binding and microtubule regulation by microtubule stabilizers. The microtubule stabilizer XMAP215 from increases microtubule growth rates in vitro. This property suggests that it could facilitate tubulin addition onto microtubule ends (; ). There is only anecdotal evidence, however, that XMAP215 binds preferentially to ends, and a direct interaction with unpolymerized tubulin has not yet been described. Better evidence for such an assisted tubulin addition comes from work on the in vitro properties of ch-TOG, the human XMAP215 homologue (; ). It is unclear, however, whether XMAP215 or ch-TOG stabilize microtubules by the same mechanism in vivo as they do in vitro. Indeed, XMAP215 apparently functions in vivo by opposing the catastrophe- promoting kinesin-13 member XKCM1 rather than by its direct influence on the microtubule growth rate (; ). An antagonistic relationship to the XKCM1 homologue MCAK has also been demonstrated for ch-TOG (; ). XMAP215 and ch-TOG are members of the Dis1/XMAP215 family. This is the only MAP family with representatives in fungi, plants, and animals. All of its members are essential for correct microtubule function during cell division (for review see and ), but the individual Dis1/XMAP215 family proteins exhibit a puzzling diversity of functional properties. Most family members have a microtubule-stabilizing activity (; ; ; ; ; ; ; ; ), but they can also show destabilizing activity in some experimental contexts or can act to make microtubules more dynamic. Examples of the former are Stu2p (; ), ch-TOG (), and XMAP215 (); examples of the latter are Stu2p () and msps, which acts as an antipause factor in vivo (). Conservation of primary structure among the Dis1/XMAP215 family members is variable and largely limited to the NH-terminal region, which contains two to five so-called TOG domains. The TOG domains, in turn, consist of degenerate HEAT repeats. Although HEAT repeats are generally described as protein interaction domains, the partners for these elements of Dis1/XMAP215 family members have yet to be identified (for review see ). In this study, we report the results of experiments designed to identify functions for the TOG domains of the Dis1/XMAP215 family member Stu2p from budding yeast. We provide evidence that Stu2p stabilizes microtubules in vivo and that its TOG domains have crucial roles in this process. Stu2p has two TOG domains, which we designate TOG1 and TOG2 (). The region containing TOG1 interacts in vivo and in vitro with dimeric tubulin. Removal of this region does not affect the end-binding properties of Stu2p either in vivo or in vitro, but it leads to shortened microtubules in vivo. Thus, end binding and microtubule regulation by Stu2p in vivo are separable processes. We have complemented these observations with a biochemical characterization of the Stu2p domains and their role in the formation of a complex with free tubulin dimers. We show that a Stu2p dimer undergoes extensive conformational changes upon tubulin binding. Our data support the following two-step picture for the mechanism of action of Stu2p in vivo. First, a Stu2p dimer forms a complex with a single tubulin heterodimer through contacts with its TOG1 domains. Second, this protein complex associates directly with microtubule plus ends, in part through the properties of TOG2 domains, where it facilitates the addition of tubulin to protofilaments at the microtubule ends. We propose that assisted plus end addition of tubulin, which is mediated by TOG domains, is the common function of Dis1/XMAP215 family MAPs. Stu2p is a dimer of 200 kD (; ; see and Fig. S3, available at ). To interpret any experiments in which domains of Stu2 were deleted, we first mapped the dimerization domain by coimmunoprecipitation experiments to the region between residues 599 and 774 (unpublished data), a segment that includes a predicted coiled coil (residues 658–761). We showed that this predicted coiled coil is indeed the dimerizing element by coexpressing differentially tagged forms of Stu2p and subsequent coimmunoprecipitation (). We demonstrated the importance of Stu2p dimerization in vivo using a strain in which endogenous Stu2p can be depleted by Cu addition. In this strain, plasmid-expressed intact Stu2p compensated for the loss of endogenous Stu2p, whereas plasmid expression of Stu2p lacking the dimerization domain led to a growth defect (). We also showed that the loss of dimerization substantially reduces Stu2p's affinity for microtubules. We coexpressed full-length HA-tagged Stu2p and myc-tagged Stu2p lacking the dimerization domain. To soluble extracts from these cells, we added increasing amounts of taxol-stabilized microtubules and separated microtubule-bound and unbound Stu2p by centrifugation. At ∼3 μM of microtubules, 50% of full-length HA-Stu2p was bound, whereas even at ∼33 μM of microtubules, binding of monomeric Stu2p-myc was less than half saturated (). Strains expressing truncated Stu2p lacking both TOG domains are not viable (unpublished data). Therefore, we focused on examining the effects of removing the NH-terminal first TOG domain (TOG1) in vivo. The following experiments show that TOG1 contributes to microtubule stabilization. We determined mean microtubule lengths in a strain expressing Stu2 lacking TOG1 (Stu2-ΔTOG1) as exclusive Stu2 copy (). In early S phase arrested and released cells (), the removal of TOG1 resulted in a strong decrease in the mean lengths of both cytoplasmic microtubules and mitotic spindles compared with the intact Stu2 control (). We infer that the presence of TOG1 enhanced microtubule stability. As expected from this inference, the ΔTOG1 mutant has an increased sensitivity to the microtubule-destabilizing drug benomyl (). We tried to measure the in vivo dynamics of microtubules in the Stu2 mutant lacking TOG1, but we found that microtubules in the mutant were too short and short lived to be traced reliably (unpublished data). Anaphase spindle elongation is a two-phase process in budding yeast (; ; ). A first, fast phase driven by the Cin8 motor is followed by a slower phase driven by the Kip1 motor (). We took time-lapse videos of spindle elongation in strains expressing GFP-tubulin and either intact Stu2 or Stu2-ΔTOG1 as the only source of Stu2p. Spindle lengths determined from projected images showed that the fast phase but not the slow phase was affected in the ΔTOG1 mutant (). The mean elongation rate in the fast phase was 0.57 μm/min (SD of 0.22 μm/min; = 15) in the ΔTOG1 strain compared with 0.92 μm/min (SD of 0.16 μm/min; = 10) in the intact Stu2 strain. The slow phase rates were essentially identical (0.25 and 0.23 μm/min, respectively; SD in both cases was 0.06 μm/min; = 15 and = 10, respectively; see Fig. S5 for details, available at ). Stu2p associates with microtubule ends both in vivo and in vitro (; ; ; ). Analyzing strains expressing either Stu2-GFP or Stu2-ΔTOG1–GFP () as exclusive Stu2 copies, we found by live cell imaging and indirect immunofluorescence that both forms localize correctly to microtubule ends (). Thus, destabilization of microtubules in the Stu2-ΔTOG1 mutant is not caused by a failure of the truncated protein to associate in vivo with microtubule ends. Using recombinant Stu2p and taxol-stabilized microtubules (see Materials and methods), we could likewise show that the deletion of TOG1 does not affect the direct association of Stu2p with microtubule ends in vitro (). However, the deletion of both TOG domains eliminates end binding in vitro () and results in microtubule wall binding instead. Thus, in the absence of TOG1, Stu2p with only the TOG2 domain can still mediate end binding in vitro. Because the deletion of TOG1 results in shortened microtubules and increased benomyl sensitivity, we tested whether increasing the number of TOG1 domains would lengthen microtubules and decrease benomyl sensitivity. Neither of these outcomes was observed. A strain with a doubled TOG1 domain of Stu2p does not have a lower sensitivity toward benomyl than the wild-type control (), nor is the mean length of cytoplasmic microtubules increased (unpublished data). We used affinity chromatography to identify proteins that interact with the TOG domains of Stu2p (). Yeast extracts were run over columns carrying covalently immobilized recombinant TOG1, TOG2, or BSA. Bound proteins were eluted with high salt and analyzed by SDS-PAGE and Coomassie staining. One band was found specifically in the eluate of the TOG1 column that was not present in the eluate of the BSA or the TOG2 column (). This band was identified by immunoblotting () and mass spectrometry (not depicted) as yeast tubulin. Thus, TOG1 but not TOG2 interacts tightly with tubulin heterodimer in vivo. We expressed and purified recombinant TOG1, TOG1–TOG2, and full-length Stu2p to characterize their interaction with tubulin in vitro. TOG1–TOG2 is monomeric in solution, as shown by analytical ultracentrifugation (see and Fig. S3). Its Stokes radius, estimated by size-exclusion chromatography (), is larger than expected for isometric molecules of corresponding mass, indicating an elongated shape (). We mixed TOG1 and TOG1–TOG2 with tubulin dimers in 1:1 molar ratios and analyzed the mixture by size-exclusion chromatography (). Both fragments formed complexes with tubulin. We estimated the contents of tubulin and TOG construct in each of the peak fractions by SDS-PAGE and quantitative densitometry. Coomassie staining ratios of the peak fractions suggest that both TOG1 and TOG1–TOG2 bind to tubulin dimers with a stoichiometry of 1:1 (Fig. S2 C, available at ). When TOG1 and tubulin were mixed in a 2:1 stoichiometry, additionalTOG1 did not bind tubulin and eluted alone, suggesting that TOG1 does not bind tubulin in a stoichiometry higher than 1:1 (Fig. S2 D). We have observed similar effects with TOG1–TOG2 (unpublished data). The similarity of the apparent mass of the TOG1 or TOG1–TOG2–tubulin complexes to their expected masses suggests that rearrangement to a more globular shape accompanies tubulin binding (). Electron microscopy of negatively stained preparations of the TOG1 and TOG1–TOG2 tubulin complexes supports this notion, showing globular particles, which are larger than free tubulin dimers (). Although the images of the particles are slightly heterogeneous, they are characterized by square shapes compared with the rectangular shape of the free tubulin dimer. Differences between the TOG1 and TOG1–TOG2 complexes were difficult to ascertain from the images. Although TOG1 and TOG1–TOG2 tubulin complexes remained intact on the time scale of the gel filtration experiment, they dissociated within hours after eluting from the column. This dissociation was also apparent in analytical ultracentrifugation experiments (unpublished data). We confirmed by analytical ultracentrifugation that purified Stu2p is a dimer ( and Fig. S3), which is consistent with earlier studies (; ). Its large apparent molecular mass and Stokes radius, like those of its fragments, indicate an elongated conformation in solution ( and ). Indeed, electron microscopy of negatively stained preparations of Stu2p reveals elongated, stringlike molecules of roughly constant width. Careful examination of the images suggests the presence of flexible joints (Fig. S4, available at ). The Stu2p molecules have an average contour length of 320 ± 35 Å and a width of ∼20–25 Å ( and Fig. S4). Based on X-ray structures of other HEAT repeat–containing proteins, a single TOG domain, which consists of six HEAT repeats, should have a length of 65–70 Å and a thickness of 20–25 Å. Therefore, a dimer of TOG1–TOG2 should be ∼300 Å long. Our measurements of electron microscopy image dimensions are consistent with these estimates. The coiled coil and COOH-terminal extension, which we would expect to project from the midpoint of the molecular contour, were not detected, probably because they are too thin to exhibit contrast in negative stain. When mixed in a 1:1 molar ratio with tubulin (two tubulin heterodimers per Stu2p dimer), Stu2p captured only about half of the tubulin added. SDS-PAGE and quantitative densitometry of the peak fractions from size-exclusion chromatography suggest that one tubulin heterodimer bound per Stu2p dimer ( and Fig. S2 C). A large molecular weight aggregate containing only tubulin also formed (). This aggregate was not present when we added at least one Stu2p dimer per tubulin dimer (unpublished data). Unlike the complexes of tubulin with TOG1 and TOG1–TOG2, tubulin complexes with dimeric Stu2p were stable for many days after assembly. Deletion of the COOH-terminal segment (residues 773–888) distal to the coiled coil to generate Stu2-ΔC did not affect the stoichiometry of the Stu2p–tubulin complex (Fig. S2, A and B). Deletion of the TOG1 domain (Stu2-ΔTOG1) interrupted the formation of a complex with tubulin (Fig. S2 E). Complexes of full-length Stu2p with tubulin tend to aggregate over time; the Stu2-ΔC–tubulin complex does not. Therefore, we used analytical ultracentrifugation to determine the molecular mass of a dimeric Stu2-ΔC–tubulin complex after purification to confirm the tubulin binding stoichiometry ( and Fig. S3). The mass difference between Stu2-ΔC bound to tubulin (268 ± 13 kD) and Stu2-ΔC alone (167 ± 7 kD) was exactly equivalent to the mass of a single αβ-tubulin dimer (∼101 kD). The relative homogeneity in the mass of the Stu2-ΔC–tubulin complex shows that binding of homodimeric Stu2p to a single tubulin subunit is a tight stoichiometric interaction. Electron microscopy of the negatively stained Stu2p–tubulin complex showed compact and relatively homogeneous structures ∼100 Å in diameter, although a few images of larger diameter were also evident (). Images of negatively stained Stu2-ΔC–tubulin complexes were very similar to those of complexes with the intact protein (Fig. S2 B), as expected from the absence of contrast from the COOH-terminal segment in images of free Stu2p. The circumference of the particles matches the average contour length of free Stu2p (320Å; and Fig. S4). Thus, when it binds tubulin, the Stu2p dimer undergoes a transition from an extended conformation to a more compact one, perhaps encircling the captured tubulin dimer. The apparent variation in Stu2p–tubulin particle images is probably the result of deformation by negative stain, incomplete immersion in stain, and interaction with the carbon surface, as the molecular mass is homogenous ( and Fig. S3). Some of the variation may also reflect different views of the particle. The Stokes radius of the complex (92 Å) derived from size-exclusion chromatography is nonetheless larger than the apparent radius of the negatively stained complexes (∼50 Å) because of additional frictional drag from the projecting coiled coil and COOH-terminal segments not clearly contrasted by negative stain. The coiled coil alone should be ∼150 Å long and relatively stiff. In the experiments reported here, we have attempted to dissect the complex contributions of Stu2p to microtubule dynamics both in vivo and in vitro by examining the effects of altering one or more of its key structural elements and by studying the molecular properties of its complexes with tubulin. Our data assign a function to the evolutionally conserved TOG domains of the Dis1/XMAP215 family of MAPs and strongly suggest an in vivo mechanism of action for the budding yeast member of this family. Stu2p, like XMAP215, is an elongated molecule (; ; ). From the properties of Stu2p in solution and the images of negatively stained molecules in electron microscopy, we can picture it schematically, as shown in (top): a pair of TOG domain elements in each Stu2p subunit are connected through a linker to a dimerizing coiled-coil domain. The individual TOG domains, like other HEAT repeat–containing proteins, are probably stiff, but the segment between them may flex and so may the longer basic linker segment between TOG2 and the coiled coil. This linker was termed the microtubule-binding domain by . The basic linker is also found in Dis1 of and was shown to be involved in microtubule binding (). However, the results in show that it does not direct end-specific binding. We suggest that its strongly positive charge leads to an affinity for microtubule walls. Therefore, we have labeled this segment as basic linker in , , and and Fig. S1 (available at ); its sequence is not conserved among homologous proteins from other fungi. Stu2p binds microtubules (), associating preferentially with microtubule ends (; ; ; ). The data shown here suggest that TOG1 contributes to microtubule stability in vivo, as the removal of TOG1 decreases microtubule length without affecting end association (). This result also implies that Stu2p end association is not strictly coupled to the stabilization of microtubules. The mechanism by which Stu2p alters microtubule dynamics in cells is, therefore, likely to be more complex than simple end capping or distortion of the microtubule end structure. How, then, does Stu2p stabilize microtubules? The mechanism may be related to our observation that Stu2p sequesters free tubulin in vitro. The open, flexible Stu2p dimer undergoes a substantial compaction when it binds tubulin, and the simplest interpretation of the electron microscopy images is that its two “wings” collapse or wrap around a single αβ-tubulin heterodimer (, bottom). The length of the TOG regions of Stu2p particle electron microscopy images (∼320 Å) matches the circumference of Stu2p–tubulin complexes (∼315 Å). Full affinity and a well-defined structure depend both on the two TOG domains and on Stu2p dimerization. TOG1 probably mediates the interaction with tubulin; monomeric TOG1 and TOG1–TOG2 bind tubulin dimers tightly, but TOG2 alone does not (). The central role of TOG1 in tubulin binding in vitro and microtubule stabilization in vivo thus suggests that both activities might be linked. A recent in vivo imaging study () suggested that Stu2p can be transported toward the plus end. The arrival of Stu2p at an end coincides with a microtubule rescue event (i.e., transition from shrinkage to growth). After reaching their maximum length, microtubules subsequently shrink, and the Stu2p signal slowly fades, presumably as the molecule diffuses away from microtubule ends. Combining these observations with the results reported in this study, we propose the following model for the mechanism of Stu2p action (). We envisage that Stu2p captures a tubulin heterodimer in the cell through its two sets of TOG domains, which are linked by the dimerizing coiled coil (). Once transported to microtubule plus ends, Stu2p might associate directly with the microtubule, at least in part, through one or both TOG2 domains in the Stu2p–tubulin complex (, I). Stu2p could then position its bound tubulin heterodimer at the plus end of a protofilament (, II), dissociate from that dimer, and return to its open conformation (, III–I V). This active tubulin-loading process could enhance tubulin assembly at growing plus ends or rescue shrinking ends. It could be part of a mechanism of microtubule rescue observed in vivo. The two activities of Stu2p, microtubule end binding and tubulin dimer sequestration, which are linked to the full assembly of TOG1 and TOG2 domains in the context of the Stu2p dimer, may well account for the complexities of its effects on microtubule dynamics under different circumstances in vivo and in vitro (; ; ; ). Stu2p dimerization through its coiled coil is necessary for its correct function. Deletion of the coiled coil leads to a growth defect in vivo () and produces defects in microtubule binding in vitro (). Our in vitro results also show that monomeric Stu2p would not retain its bound tubulin dimer tightly enough for reliable delivery to microtubule plus ends, and both TOG2 domains of a Stu2p dimer may also be necessary for stable plus end association (). In contrast to its apparent activity in vivo, Stu2p destabilizes microtubules in vitro (). This destabilization correlates with end binding, as only Stu2p fragments that associate preferentially with microtubule ends are able to destabilize microtubules (unpublished data). Why are the in vitro and in vivo effects of Stu2p on microtubule stability different? One explanation might be that under the in vitro conditions used for stability assays, Stu2p has an altered affinity for a reaction intermediate at microtubule ends, driving the equilibrium toward destabilization in vitro and stabilization in vivo (, reverse arrows from III to I). For example, a higher affinity for free tubulin in vitro than in vivo could pull tubulin from microtubule ends rather than add it to them. A similar mechanism might explain why the microtubule-stabilizing Stu2p homologue XMAP215 can act in vitro as a destabilizer as well (). Finally, posttranslational modifications might account for differences in the observed activities. We found that Stu2p is multiply phosphorylated in vivo and that this phosphorylation changes in a cell cycle–dependent manner (unpublished data). A symmetric Stu2p dimer associates with an asymmetric tubulin heterodimer. The two Stu2p subunits must recognize different tubulin surfaces, and one contact could be substantially stronger than the other. The TOG2 region, which does not interact strongly with tubulin dimers but plays a structural role in the complex, might recognize protofilament ends on its own. We have not yet been able to capture the end-bound state by electron microscopy. Homologues of Stu2p like XMAP215 and DdCP224 are monomers (; ; ) and lack the dimerizing coiled-coil domain, but they contain three more TOG domain segments (five in XMAP215 and DdCP224; for review see ). The first two of these TOG domains have weak sequence similarities with TOG1 and TOG2, respectively (), and the differential function of the two domains in binding free tubulin dimers and microtubule ends, respectively, may be preserved in the longer XMAP215 and ch-TOG proteins. The property conferred to Stu2p by dimerization may be mimicked in the monomeric homologues by differentiation among the additional TOG domains in those proteins. The starting plasmid for in vivo Stu2p constructs was a gift of T. Huffaker (Cornell University, Ithaca, NY) and carried (under its own promoter region) full-length Stu2 COOH-terminally tagged with HA (parent pRS315). A HindIII restriction site in the upstream promoter region was removed by ApaI cleavage, and the vector was religated. Stu2 was removed from the vector by cleavage with HindIII–SphI (cutting at 84–88 bp within the coding region and downstream of the stop codon, respectively), and the corresponding Stu2 constructs were reintroduced into the HindIII–SphI sites. The Stu2 fragments were cloned as HindIII–EcoRI pieces lacking a stop codon and ligated together with an EcoRI–SphI piece carrying the myc or EGFP tag followed by a stop codon. For the cloning of Stu2 carrying an extra first TOG domain, the parent vector was cut with HindIII (cutting at codon 29), and a three-way ligation was performed using a HindIII–XhoI piece from codon 29–325 (end of the spacer between TOG1 and TOG2) and a XhoI–HindIII piece from codon 1–29. For the cloning of Stu2 lacking the coiled-coil domain, a four-way ligation was performed using a HindIII–XmaI piece from codon 29–657 (last codon of basic linker; previously termed microtubule-binding domain by ), a codon before the start of the coiled-coil domain, an XmaI–EcoRI piece from codon 761 (last codon of the coiled coil)–888 (last codon before the stop), and an EcoRI–SphI piece carrying the myc tag followed by a stop codon. In vivo constructs used in this study are also shown in Fig. S1: Stu2 (residues 1–888), Stu2-ΔTOG1 (residues 1–30 and 304–888), Stu2-2TOG1 (residues 1–325-LE-1-888), and Stu2–Δcoiled coil (residues 1–657-PG-761-888). In vitro constructs used in this study are also outlined in Fig. S1 and included: Stu2p (residues 1–888), Stu2-ΔTOG1 (residues 266–888), Stu2-ΔTOG1+2 (residues 551–888), Stu2-ΔC (residues 1–772), TOG1–TOG2 (residues 1–587), TOG1 (residues 1–306 or residues 1–317 for gel filtration and electron microscopy), and TOG2 (residues 307–550). Stu2-ΔTOG1 or Stu2-ΔTOG1+2 were prepared by cloning the corresponding coding sequences as an NcoI–XhoI fragment into pFASTBac HTa (Invitrogen). Sequences corresponding to TOG1 and TOG2 were cloned as EcoRI–XhoI pieces into pGEX-6P-1 in frame with an NH-terminal GST tag. Sequences of TOG1 and TOG1–TOG2 were also cloned in pET28a using NcoI–XhoI fragments in frame with a COOH-terminal His tag (). Baculoviruses expressing wild-type Stu2p, Stu2p-ΔTOG1, or Stu2p-ΔTOG1+2 were generated by transforming corresponding transfer vectors into the DH10bac strain. Bacmids were prepared and transfected into cells according to standard procedures. Recombinant baculoviruses were amplified three times before the large-scale infection of SF+ cells. Wild-type Stu2p was purified as previously described (Stu2 baculovirus was a gift of P. Sorger, Massachusetts Institute of Technology, Cambridge, MA; ) or by using Ni–nitrilotriacetic acid (NTA) chromatography followed by gel filtration using a Superdex-200 column and by ion exchange chromatography using a Hitrap-SP column (GE Healthcare). Insect cell–expressed Stu2p-ΔTOG1 and Stu2p-ΔTOG1+2 were prepared by successive steps of ammonium sulfate precipitation, metal affinity chromatography using Talon resin (CLONTECH Laboratories, Inc.), and ion exchange chromatography using a 1-ml Mono-S column (GE Healthcare) followed by gel filtration on a Superose 6 column run in BRB80 (80 mM Pipes, pH 6.8, 1 mM EGTA, and 1 mM MgCl), 100 mM KCl, 0.25% Brij-35, 5% glycerol, and 1 mM DTT. TOG1, TOG1, and TOG1–TOG2 protein were expressed in BL21 (DE3) codon+ RIL cells (Stratagene). Cultures were grown at 29°C in LB to 0.5–0.6 OD and induced either for 2–3 h with 0.1–0.3 mM IPTG at 29°C or induced at 16°C for 5 h with 0.2 mM IPTG. Cells were harvested by centrifugation and cold lysis buffer (10 mM K-Pipes, pH 6.8, 1 mM MgCl, 3 mM EGTA, 0.25% Brij-35, 5% glycerol, 500 mM KCl, 2 mM EDTA, and 1 mM DTT). Lysis buffer was supplemented with complete protease inhibitor tablets (Roche), and lysozyme (Sigma-Aldrich) was added to the pellets followed by resuspension. The solution was sonicated using a sonicator (450-D; Branson) for several cycles. Lysates were spun at 70 krpm on a rotor (MLA80; Beckman Coulter) for 10 min at 4°C, and the supernatants were incubated with either 2–3 ml of lysis buffer–equilibrated glutathione–Sepharose beads (GE Healthcare) for GST-tagged constructs or with 4–5 ml Ni-NTA resin (QIAGEN) for His-tagged constructs. For GST-tagged constructs, the glutathione beads were incubated at 4°C for 1 h and were washed with lysis buffer and with precision cleavage buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM DTT, and 0.01% Tween-20). The glutathione beads were then mixed with 150–250 μl GST-tagged precision protease (a gift of D. Drechsel, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany). Cleavage was performed for 8–16 h at 4°C. The bead solutions were spun through a frit, and the flow-through fractions containing cleaved protein were collected. His-tagged proteins were purified from Ni resin by washing with wash buffer (50 mM Tris, 250 mM NaCl, pH 7.5, and 5 mM β-mercaptoethanol) and were eluted with elution buffer (wash buffer + 320 mM imdazole) in 1-ml fractions. Fractions were evaluated by SDS-PAGE, and aliquots were snap frozen in liquid nitrogen. The NH-terminal Stu2p antibody was raised against Stu2p-TOG1 (aa 1–306) by Elevage Scientific des Dombes. The COOH-terminal Stu2p antibody was raised against a COOH-terminal peptide of Stu2p (Eurogentec). Both antibodies were affinity purified. FACScan was used as described previously (). Strains expressing tandem affinity purification (TAP)–tagged Stu2p together with myc-tagged Stu2p or myc-tagged Stu2p–Δcoiled coil were grown in synthetic medium (synthetic drop-out media [−Leu] containing 2% glucose) to midlogarithmic phase and were harvested by centrifugation. Lysis buffer BRB80 (80 mM Pipes, pH 6.8, 1 mM EGTA, and 1 mM MgCl), 5% glycerol, 100 mM KCl, 0.25% Brij-35, and 2 mM DTT supplemented with complete protease inhibitor tablets (Roche) was added 1:1 (vol/vol), and cells were lysed by bead beating. Lysates were spun three times (2 min at 14 krpm, 5 min at 100 krpm, and 2 min at 14krpm; all at 4°C), and 150 μl of cleared lysates were added to 10 μl of equilibrated IgG Sepharose 6 Fast Flow slurry (GE Healthcare). Reactions were incubated on a rotating wheel for 1 h at 4°C. After six 0.2-ml washes with lysis buffer, beads were eluted with laemmli buffer. Eluates were analyzed by immunoblotting using an excess of rabbit antibody (to block protein A of the TAP tag) and the anti-myc antibody 9E10. After stripping of the blot, protein A detection reagent (Sigma-Aldrich) was used to detect the TAP tag. Yeast strain TH1029 expressing HA-tagged Stu2p and myc-tagged Stu2p–Δcoiled coil was grown in synthetic medium (synthetic drop-out media [−Leu] containing 2% glucose) to midlogarithmic phase. Cells were harvested by centrifugation. Lysis buffer supplemented with 20 μM taxol was added 1:1 (vol/vol), and cells were lysed by bead beating. Lysates were cleared twice by centrifugation (5 min at 100 krpm and 4°C) and were respun (14 krpm for 10 min at 4°C). The protein concentration was 18 mg /ml. 5 μl taxol-stabilized microtubules were added to 20 μl of the supernatants (0–32.9-μM final concentrations). Reactions were incubated for 7 min on ice, and 20 μl were spun for 10 min at 70 krpm through a 50-μl glycerol cushion (50% in BRB80) in TLA100 tubes (Beckman Coulter) at 4°C. Equivalent amounts of supernatants and pellets were separated by SDS-PAGE followed by immunoblotting using the anti-myc antibody 9E10, the antitubulin antibody DM1α (Sigma-Aldrich), and a polyclonal anti-HA antibody (Santa Cruz Biotechnology, Inc.). Blot quantification was performed using the plot profile function of Image 1.62a software (Scion). A best-fit curve was created only for optical display using the data points and the function y = y + (Bmax − y) × x/(Kd + x), where x = the concentration of microtubules, y = the percent binding of Stu2p, y = the percent binding of Stu2p at x = 0, Bmax = the percent binding of Stu2p at x = ∞, and Kd = the dissociation constant. Bmax, y, and Kd were determined from the fit. Cells were grown at 30°C in YPD to an OD of 0.4 in a 50-ml culture and were arrested in early S phase by the addition of 0.6 g hydroxyurea for 3.5 h. Cells were washed twice with 50 ml YPD by centrifugation and released at 30°C by resuspension in 50 ml YPD. Samples were taken at 10-min intervals for immunofluorescence and FACS analysis. The remainder of the culture was used for extract preparation and immunoblotting. For immunofluorescence, cells were fixed with formaldehyde and processed according to standard yeast techniques. Antibodies were monoclonal antitubulin antibody DM1α (Sigma-Aldrich) and a directly Cy3-labeled anti–mouse antibody (Jackson ImmunoResearch Laboratories). DNA was visualized using DAPI. Imaging of immunofluorescence was performed using a 100× NA 1.35 objective on a microscope (DeltaVision; Olympus) using SoftWorx (Applied Precision) and taking stacks (30 images with 0.2-μm z spacing) for each field using a camera (CoolSNAP HQ; Roper Scientific). Cytoplasmic microtubule lengths were measured from projected images from the microtubule tip to the spindle pole body (SPB) as multiple segments (SoftWorx), taking the samples from time point zero. The average cytoplasmic microtubule length for Stu2-ΔTOG1–myc was 0.99 ± 0.06 μm (±SEM; α = 0.05; = 366) and for Stu2-myc was 1.37 ± 0.06 μm (±SEM; α = 0.05; = 590). Spindle lengths were measured as multiple segments along elongated spindles from SPB to SPB using projected images of the time points at which the peak of spindle elongation frequency occurred (50–70 and 50–80 min after release for Stu2-myc and Stu2-ΔTOG1–myc, respectively). Elongated spindles were defined as those along which the DNA masses were completely segregated as judged by DAPI staining. The average spindle length for Stu2-ΔTOG1–myc was 5.36 ± 1.27 μm (±SD; = 178) and for Stu2-myc was 8.03 ± 2.56 μm (±SD; = 195). Images depicted in this paper were projected using SoftWorx and were processed using Adobe Photoshop. 5 μl of yeast cultures were spotted at the same OD and in serial dilutions of 1:10, 1:100, and 1:1,000 on YPD plates containing 10, 12.5, or 15 μg/ml benomyl or on the DMSO control plates. The plates were subsequently incubated at 30°C for the same amount of time. Agar pads were prepared using molten synthetic drop-out media (−Leu) containing 2% glucose and agar and depression slides. 4 μl of midlogarithmic yeast cultures concentrated by centrifugation (at 3 krpm for 20–40 s at room temperature) were deposited onto the pads, which were then covered by 25 × 25-mm coverslips, and the edges were sealed with Vacuum Grease (Beckman Coulter). Imaging was performed using a 100× NA 1.35 objective with a slightly closed numeric aperture ring on a microscope (DeltaVision; Olympus) using SoftWorx. Stacks (five images with 0.6-μm z spacing) were taken for each time point with a 0.2-s exposure per image; 100 stacks were recorded with a camera (CoolSNAP HQ; Roper Scientific) with a time lapse between stacks of 20 s. The temperature as measured by a thermometer attached to the stage ranged from 24.5 to 25.8°C. Spindle lengths were measured from projected stacks using multiple segments along the spindles (SoftWorx) from SPB to SPB. In case of unclear SPB position, no spindle length for this time point was measured. Spindle average elongation speed was calculated from the start of elongation to a visible drop in the elongation speed (fast phase) and from this point to the end of spindle elongation (slow phase). If no drop in elongation speed could be assigned, the total speed of spindle elongation was taken for both the fast and the slow phase. Average fast spindle elongation speeds for Stu2-myc was 0.92 ± 0.16 μm/min (±SD; = 10) and for Stu2-ΔTOG1–myc was 0.57 ± 0.22 μm/min (±SD; = 15). The slow spindle elongation speed for Stu2-myc was 0.23 ± 0.06 μm/min (±SD; = 10) and for Stu2-ΔTOG1–myc was 0.25 ± 0.06 μm/min (±SD; = 15). Images were processed using Photoshop. 1 ml of midlogarithmic yeast culture was concentrated by brief centrifugation and directly imaged in the GFP channel using a microscope (DeltaVision; Olympus) as described below. Another 1 ml of the culture was fixed using a formaldehyde/PFA mixture and processed for immunofluorescence using standard yeast techniques. Antibodies were a sheep anti-GFP antibody (), DM1α mouse antitubulin antibody (Sigma-Aldrich), and directly labeled secondary antibodies (AlexaFluor594-labeled donkey anti–sheep; Invitrogen) or FITC-labeled donkey anti–mouse (Jackson ImmunoResearch Laboratories). DNA was visualized with DAPI. Imaging was performed using a 100× NA 1.35 objective on a DeltaVision microscope using SoftWorx. Stacks (30 images with 0.2-μm z spacing) were taken for each field using a camera (CoolSNAP HQ; Roper Scientific). Images were projected using SoftWorx and processed using Photoshop. The remainders of the cultures were used for extract preparation and immunoblotting using a polyclonal goat anti-GFP antibody (gift of D. Drechsel). In this assay, –derived Stu2-ΔTOG1 but not –derived Stu2-ΔTOG1+2 behaved differently from the insect cell–derived equivalent protein used in this study. In contrast to insect cell–derived Stu2-ΔTOG1, –derived Stu2-ΔTOG1 shows no preferential end binding but binds laterally along the microtubules and behaves differently in a sucrose gradient and in size-exclusion chromatography. As –derived full-length Stu2 could not be made, we assume that –derived Stu2-ΔTOG1 is improperly folded. Thus, for consistency, the in vitro end-binding analysis presented in this study has been based on insect cell–derived proteins. Recombinant protein was immobilized on a 1-ml -hydroxysuccinimide (NHS)–activated HiTrap column according to standard procedures (GE Healthcare). Yeast extracts were prepared from a haploid W303 strain. Cultures were grown in YPD at 30°C to ∼1.0 OD, spun, and pellets were washed twice with cold HO and weighed. 1 vol/wt of ice-cold lysis buffer was added (100 mM K-Hepes, 50 mM NaF, 50 mM β-glycerophosphate, 10 mM EDTA, 10 mM EGTA, 0.5% Brij-35, 10% glycerol, and 2 mM DTT, pH 7.6, supplemented with complete protease inhibitor tablets [Roche]), and the cells were lysed by bead beating at 4°C. Lysates were cleared by centrifugation, and the supernatant were snap frozen in aliquots in liquid nitrogen. 2-ml aliquots were thawed, spun briefly, and the supernatant was injected at a flow rate of 0.1 ml/min onto the affinity column equilibrated and run in 50 mM Na-Hepes, 20 mM NaCl, and 1 mM DTT, pH 7.6, using an FPLC (GE Healthcare). After 30 min, the flow rate was increased to 1 ml/min, the column was washed to ∼0.08 OD, and bound protein was eluted with 50 mM Na-Hepes, 1 M NaCl, and 1 mM DTT, pH 7.6, taking 250-μl fractions. The peak fractions were pooled and snap frozen. Similar protein elution amounts were loaded and separated on 4–12% gradient gels. Full-length Stu2p, TOG1, or TOG1–TOG2 protein was loaded at 20 μM onto a 10/5 Superose 6 gel filtration column preequilibrated with gel filtration buffer at 4°C (25 mM Tris, 200 mM NaCl, 1 mM MgCl, and 1 mM EGTA, pH 7.5) and eluted in 0.4-ml fractions. For tubulin-binding experiments, Stu2p or TOG constructs were first dialyzed against running buffer and were then mixed with an equal amount of recycled bovine brain tubulin dimer. Stu2p–tubulin (or TOG–tubulin) complexes were allowed to form for 30 min at 4°C before loading onto a Superose 6 10/300 GL gel filtration column (GE Healthcare). Fractions were evaluated by SDS-PAGE. The protein content of each band was determined from scans of the gels using Image J 1.32j software (National Institutes of Health). Apparent molecular masses and Stokes radii of the Stu2p constructs and tubulin complexes were determined by calibrating the size-exclusion column with the following set of protein standards: thyroglobulin, ferritin, catalase, aldolase, albumin, ovalbumin, chymotrypsinogen, and ribonuclease A. Stokes radii were calculated as averages of the Porath and Laurent-Killander Stokes' radii, which matched closely. Sedimentation equilibrium experiments of wild-type Stu2p, TOG1–TOG2, and Stu2-ΔC with and without tubulin were conducted at 4°C in an analytical ultracentrifuge (Optima XLA; Beckman Coulter). Samples corresponding to an absorbance at 280 nm of 0.25–0.75 were prepared in 50 mM Hepes, pH 7.5, 180 mM KCl, 1 mM MgCl2, and 1 mM Tris-(2-carboxyethyl) phosphine. Data were acquired as an average of two absorbance measurements at a nominal wavelength of 280 nm and a radial spacing of 0.001 cm at the rotor speeds indicated for each experiment. In each speed, equilibrium was achieved within 40-h global analyses in terms of a single ideal solute by multiple fit alignment in the XL-I software package (Beckman Coulter) using data from three speeds and protein concentrations to obtain the buoyant molecular mass M(1-vρ). Early fractions of each of the gel filtration experiments were evaluated using negative stain electron microscopy. Fractions were diluted to 0.1 mg/ml and incubated on glow discharged continuous carbon support film on 400 copper mesh grids for 5 min. Samples were washed with gel filtration buffer, stained with 0.5% uranyl formate, and dried. Samples were imaged at 52,000× using a microscope (Techni-12; Phillips) operated at 120 kV with a low dose kit. Images were collected on a CCD camera (model 894; Gatan) using Digital Micrograph software (Gatan). Contour lengths of images of free Stu2p dimer along which particle paths could be traced clearly (Fig. S4) were measured using the measuring tool of Image J 1.32 software. Fig. S1 shows a schematic of all constructs used in this study. Fig. S2 shows the size-exclusion chromatography and electron microscopy of the Stu2-ΔC–tubulin complex, quantitative densitometry of Stu2p to tubulin in complexes, titration of TOG1 binding to tubulin, and shows that Stu2-ΔTOG1 does not bind tubulin. Fig. S3 shows analytical ultracentrifugation data of Stu2p and tubulin complexes as well as their curve fits. Fig. S4 shows the heterogeneous conformations of Stu2p in electron microscopy images. Fig. S5 shows data spread and statistics for the slow and fast anaphase spindle elongation velocities shown in . Online supplemental material is available at .

The plasma membrane of polarized epithelial cells is divided by tight junctions into apical and basolateral domains, which display different protein and lipid compositions that are required for a range of specialized functions. This asymmetric distribution is achieved by continuous sorting of newly synthesized components and their regulated internalization (; ). Evidence derived from biochemical and live imaging studies have shown that apical and basolateral membranes segregate into different vesicles upon exit from the TGN (; ; ; ), supporting the hypothesis that the TGN is the major sorting station during exocytosis of newly synthesized proteins. More recent work has shown that protein sorting could also occur in recycling endosomes (REs) after their exit from the TGN (; ), similar to what happens in yeast (). Intracellular sorting of newly synthesized proteins at the TGN or in REs is based upon recognition by the sorting machinery of specific apical and basolateral sorting signals () that mediate their incorporation into apical and basolateral sorting vesicles (; ; ). As a result of these events, proteins can be transported to the surface after either a direct or an indirect (transcytotic) route that first passes through the opposite membrane domain. Utilization of the direct or indirect pathways seems to be both cell and protein specific (). For example, in liver cells, the majority of proteins follow an indirect pathway, whereas in Caco-2 intestinal cells, both pathways are used, and the choice is likely to be protein specific. In MDCK (kidney) and FRT (Fischer rat thyroid) cells, the direct pathway is more commonly used, whereas the transcytotic pathway seems to be more specific for basolateral proteins and transmembrane receptors (; ), where specific signals for this route have been found (; ; ). In these cells, apical proteins mainly use a direct pathway (), and the indirect pathway has been shown mainly during the establishment of the polarized epithelium (; ). Basolateral sorting is mediated by discrete domains in the cytosolic protein tail frequently containing tyrosine or dileucine motifs (), which are recognized by the clathrin adaptor complex (; ). However, the situation is more complicated for apical proteins because lumen-localized domains, transmembrane domains, and membrane-binding features have all been shown to be important for apical sorting (, ; ; ; ). A case in point are the glycosylphosphatidylinositol (GPI)-anchored proteins (GPI-APs) that have been shown to be directly targeted to the apical domain after lateral segregation from basolateral cargo in the TGN () because of their incorporation into sphingolipid- and cholesterol-rich microdomains (rafts), which were assayed by their insolubility in cold detergents (; ). However, it was recently demonstrated that association with detergent-resistant microdomains is not sufficient to determine apical sorting of GPI-APs but that stabilization into rafts, promoted by their oligomerization (which most likely leads to the coalescence of more rafts), is required (). It was also found that oligomerization of GPI-APs begins in the medial Golgi and is concomitant with association in detergent-resistant microdomains. Therefore, this data suggest that sorting of apical GPI-APs occurs during passage through the Golgi apparatus, presumably at the TGN and before reaching the plasma membrane. This hypothesis has been supported by several biochemical experiments showing that after their segregation at the TGN, GPI-APs are directly delivered from the Golgi to the apical plasma membrane (; ). The aforementioned model has been challenged recently by , who, by using state of the art technologies for imaging in living cells, have shown that a GPI-linked protein (GFP-GPI) was apically targeted after a transcytotic route, thus suggesting that apical sorting of GPI-APs occurs after membrane delivery. However, the experiments showing either direct or indirect sorting of GPI-APs in living cells have been performed in conditions that were not fully polarized because of the difficulty of imaging polarized cells growing on a filter in three dimensions and in transiently transfected cells where biochemical experiments are difficult to perform. Therefore, to understand the routing and sorting site of GPI-APs, we directly followed the sorting and trafficking of three different GPI-APs to the apical surface of stably transfected clones of MDCK cells by using both biochemical and live imaging approaches in fully polarized cells. We demonstrate that all proteins are directly delivered to the apical surface and that the impairment of basolateral fusion/endocytosis does not affect apical transport, thus suggesting that sorting occurs intracellularly before reaching the plasma membrane. The mechanism of sorting and the route that GPI-APs follow to reach the apical membrane are still debated. We have recently shown that both raft association and protein oligomerization are required for apical sorting of GPI-APs. These two events occur concomitantly in the Golgi apparatus, suggesting that the sorting of GPI-APs occurs in the Golgi (). In direct contrast to these data, a recent study using live imaging techniques showed that a chimeric GPI protein (GFP-GPI) is apically targeted by a transcytotic pathway in MDCK cells (). We decided to directly investigate the pathway and the sorting site of GPI-APs by using both biochemical and imaging approaches and three different GPI model proteins stably transfected in MDCK cells: GFP-GPI, encoded by the same cDNA used in the live study of ; placental alkaline phosphatase (PLAP), a native GPI-AP; and neurotrophin receptor (NTR)–PLAP, in which the ectodomain of p75 is fused to the GPI attachment signal of PLAP (; ). To analyze the pathway followed by the newly synthesized GPI-APs, we performed a biotin-targeting assay (). Fully polarized MDCK cells grown on filters for 4 d were pulse labeled with [S]methionine for 15 min and chased for the indicated times (). At each chase time, biotin was added to the apical or basolateral surface to catch the arrival of the protein at the plasma membrane. After 15 min of chase, GFP-GPI was detected on the apical surface, and it progressively accumulated on this domain, whereas very low amounts also accumulated with similar kinetics on the basolateral plasma membrane (). These results suggest that GFP-GPI follows a direct route to reach the apical surface and that small amounts of the protein are missorted to the basolateral domain. To rule out the possibility that this behavior was specific for GFP-GPI, we performed the same experiment using PLAP. Newly synthesized PLAP was initially detected on the apical surface after 60 min of chase and accumulated there for up to 2 h (). A small amount of the protein was missorted to the basolateral domain at all chase times, as was found for GFP-GPI, indicating that PLAP also followed a direct apical route. Similar results were obtained with NTR-PLAP in 4-d-old monolayers (Fig. S1 A, available at ). It was previously shown that during the establishment of a polarized monolayer in filter culture, apical proteins that are normally sorted via a direct route can use the transcytotic pathway (). Because of this, we repeated the same targeting assay in MDCK cells grown on filter for just 1.5 d (). Also in these conditions, we found that both GFP-GPI and PLAP were directly targeted to the apical surface, although a higher amount of both proteins was missorted to the basolateral surface in comparison with the fully polarized 4-d-old cultures. Interestingly, in one case, we found that a small amount of GFP-GPI accumulated on the basolateral surface before reaching the apical membrane (Fig. S1 B), which is a classic behavior for a transcytotic protein. Thus, to establish the influence of time in culture for the initial establishment of the apical delivery of GFP-GPI, we repeated the targeting assays after 1, 2.5, and 4 d in culture. By quantifying different experiments, we could show that the basolateral delivery of GFP-GPI observed after 1 d of culture progressively disappeared over several days () and that the apical polarity of GFP-GPI increased from 50% on the first day to 80% on the fourth day of culture (). These results confirmed our previous findings that an indirect route could be used during the establishment of a polarized monolayer to salvage missorted proteins and reestablish the correct polarity (). This behavior appears to be protein specific () because it was observed for GFP-GPI and NTR-PLAP (Fig. S1, A and B) but never for PLAP. The biotin-targeting assay analyzes only newly synthesized proteins. Because it has been shown that GPI-APs can recycle intracellularly through the Golgi apparatus (), it is possible that the newly synthesized and recycled proteins follow different pathways and that their sorting sites are, therefore, different. To analyze exclusively the targeting of GPI-APs from the Golgi apparatus, we set up a cold targeting assay using a Golgi temperature block combined with protease digestion. To eliminate the protein already present on the plasma membrane, 25 μg/ml trypsin was added for 1 h to the apical and basolateral sides of filter-grown MDCK cells expressing GFP-GPI. This treatment was able to remove almost all GFP-GPI present at steady state on the surface (Fig. S2, A and B; available at ). Conversely, by confocal microscopy, we did not detect any GFP signal on the surface to which the enzyme was added (Fig. S2 A). By Western blotting, we found the protein only in the media of trypsin-treated cells (Fig. S2 B), thus indicating that trypsin does not pass trough the monolayer but digests only the proteins present on the side exposed to it (Fig. S2, A and B). After clearing the surface by trypsin treatment, fully polarized monolayers were subjected to a temperature block by incubation at 19.5°C in the presence of cycloheximide to accumulate the protein in the Golgi and to eliminate newly synthesized proteins. Cells were then warmed at 37°C for the different indicated times in the presence of cycloheximide, and the arrival of proteins at the cell surface was assayed both by confocal microscopy () and by domain-selective biotinylation at different chase times (). For microscopy, cells were fixed and imaged by collecting z axis stacks to visualize all of the protein inside the cells. Xz reconstructions are shown for different times in . At time = 0, GFP-GPI was present almost exclusively in the Golgi apparatus, as was previously demonstrated (; ; ). The ring sometimes observed at the top of the cells at time = 0 in the xy section (also see and ) represents the edge of the apical surface, possibly as a result of incomplete trypsin digestion. During the chase times, GFP-GPI progressively accumulated on the apical surface, as clearly shown both in xz microscopy sections and in xy sections collected from the top of the cells as well as by quantification of the fluorescent signal in selected regions from a single z plane at different heights through the cell (). Consistent with the confocal data, we found by surface biotinylation that after 15 min at 37°C, GFP-GPI began to arrive at the apical surface and accumulated there for up to 60 min (). Thus, both these approaches confirmed that GFP-GPI previously accumulated in the Golgi is directly targeted to the apical surface. To rule out the possibility that we were missing a fast passage of proteins through the basolateral surface, we used two approaches. In one, we added trypsin to the basolateral surface to remove any protein that was passing through the basolateral domain, and, in the other, we inhibited basolateral traffic using tannic acid as previously described (). In the first approach, to detect proteins passing through the basolateral surface, we adapted an assay that was used before to demonstrate transcytosis of the polymeric Ig receptor in MDCK cells (). We added 25 μg/ml trypsin to the basolateral medium during the chase time course at 37°C after the temperature block so that the molecules passing through the basolateral surface would be proteolysed and should not be recovered at the apical surface. To the contrary, molecules directly delivered to the apical surface would not be exposed to trypsin and would be detected normally. In the presence of trypsin, GFP-GPI was still apically delivered to the same extent as in untreated cells (), and only the portion of protein that was missorted to the basolateral surface was recovered in the basolateral media as a faster migrating, partially digested form (). It is to be expected that the β barrel of GFP will be more resistant to enzymatic activity, whereas the loops outside the barrel would be digested. On the other hand, the GFP-GPI that was detected in the apical media after 60 min of chase time both in control and in treated cells likely represents the proportion of protein that is normally shedded from the cell surface, and it runs on the gel at its expected molecular weight. These experiments further confirm that GFP-GPI does not need to travel through the basolateral surface before reaching the apical plasma membrane. As an alternative to trypsin, we used tannic acid, a cell-impermeable fixative that cross-links cell surface carbohydrate groups and does not diffuse across tight junctions (; ). When added selectively to the different domains of the plasma membrane of filter-grown cells, tannic acid would be expected to inhibit plasma membrane fusion and internalization only in the specific domain to which it was added. To check whether the tannic acid impairs internalization in MDCK cells, we followed transferrin endocytosis (Fig. S3, available at ). In control cells, transferrin was exclusively localized in intracellular spots, whereas in cells treated with tannic acid, it was blocked at the cell surface after 30 min of internalization (Fig. S3, top). We also show that tannic acid does not affect binding but only affects the internalization of transferrin by allowing first its binding for 5 min at 37°C and then the internalization for 30 min in the presence or absence of basolateral tannic acid (Fig. S3, bottom). Also in this case, in the presence of tannic acid, transferrin was exclusively found at the basolateral membrane. Furthermore, these results showed that a pretreatment with tannic acid for 30 min impairs plasma membrane internalization at least for the next 30 min. Thus, it appeared to be a convenient tool to analyze whether a protein follows a transcytotic pathway via the basolateral surface because this process requires the protein to first be endocytosed (). To analyze the effect of tannic acid on GFP-GPI trafficking, we repeated the same cold targeting assay as described in . After the trypsin treatment, cells grown for 4 d on filters were incubated at 19.5°C in the presence of cycloheximide for 2 h to accumulate the protein in the Golgi. During the last 30 min at 19.5°C, 0.5% tannic acid was added to the basolateral medium to block basolateral fusion. After washing to remove tannic acid, the monolayers were then incubated at 37°C in the presence of cycloheximide for the different indicated times to follow the arrival of the protein at the surface (). We found both by confocal microscopy () and by surface biotinylation () that in the presence of tannic acid, GFP-GPI still reached the apical surface directly with similar kinetics to the untreated cells (). However, by confocal microscopy, it appears that in tannic acid–treated cells, trafficking of the protein is slightly slower, and the newly arrived protein appears to fuse or accumulates mainly at the edge of the apical surface above the tight junctions. Unexpectedly, a small portion of GFP-GPI was detected on the bottom confocal sections or by biotinylation of the basolateral surface. This basolateral signal could result either from the fusion of some GFP-GPI–containing carriers with the basolateral surface or from diffusion of the protein from the apical domain, therefore implying either an incomplete effect of tannic acid in blocking basolateral fusion or an effect of tannic acid on the integrity of the tight junctions. Nonetheless, both the tannic acid and the trypsin experiments showed that GFP-GPI does not need to reach the basolateral surface to be apically sorted but is directly targeted to the apical membrane. We reasoned that if the low level of basolateral missorting observed in the aforementioned experiments was caused by an initial effect of tannic acid on the integrity of the monolayer, it should increase with the time of the treatment. Therefore, we incubated cells grown on filter for 4 d with basolateral tannic acid during the last 10 min of the temperature block and during all chase times at 37°C as shown previously by . The cells were then fixed and imaged by confocal microscopy or subjected to surface biotinylation (). Also in this case, at time = 0, GFP-GPI is present in the Golgi apparatus. However, after 30 min of chase, it is present both on the apical and on the basolateral surface and continues to be completely mislocalized after 60 min (). These results were confirmed by the surface biotinylation assay that showed a complete mislocalization of the protein (). These data suggest that a long treatment with tannic acid leads to monolayer depolarization. To address whether tannic acid could alter the integrity of the cell monolayer, we evaluated the functional state of tight junctions during tannic acid treatment by different methods (). First, we investigated the morphology of the tight junctions by staining them with an antibody against PATJ (Pals1-associated tight junction protein), a component of tight junctions (; ). To verify that the basolateral domain maintains its physical integrity, we labeled the cells with an antibody against Na,K-ATPase (). In control cells, PATJ staining results in a chicken wire–like pattern (typical of tight junction markers) that was detected exclusively in one z plane (at 2.5 μm from the out-of-focus top signal) where the signal for the Na,K-ATPase was almost completely out of focus. On the other hand, a distinct signal for Na,K-ATPase was detectable starting at 6.5 μm from the top out-of-focus signal all along the lateral membrane (; ). In contrast, in cells treated with tannic acid, the PATJ signal was found also in the bottom planes at the same level where the Na,K-ATPase–derived signal was in focus. This effect was more evident after 60 min of treatment. These results indicated that tight junctions were impaired by tannic acid treatment. To then check the integrity of the monolayers, we measured both the passage of electric current and the flux of a tracer across monolayers, which are the two most commonly used methods to analyze the integrity of tight junctions (). Transepithelial resistance (TER) of fully polarized monolayers was measured before and after basolateral incubation with tannic acid for 30–60 min. Although in control cells TER is stable during the time course of the measurements, in treated cells, it decreased progressively (). After 60 min of tannic acid treatment, the TER decreased to the value of empty filters. Finally, we performed a permeability assay using [C]inulin as previously described (). Cells grown on filter for 4 d were treated with tannic acid on the basolateral side for 30 or 60 min. At the same time, [C]inulin (1 μCi per filter) was added to the apical medium. Subsequently, apical and basolateral media were recovered, and C was quantified. The extent of inulin passage was quantified by dividing the amount of [C]inulin found in the basal chamber by the total [C]inulin added to each sample. We found that in the control cell monolayer, only 0.25 ± 0.01% and 0.49 ± 0.03% of apically added [C]inulin for 30 or 60 min, respectively, was found in the basal well, whereas in the cells treated for the same length of time with basolateral tannic acid, 1.2 ± 0.01% and 3.75 ± 0.09% was collected after 30 and 60 min, respectively (). These results clearly show that the permeability of the cell monolayer is strongly altered by tannic acid treatment in a time-dependent manner (fivefold more after 30 min and eightfold more after 60 min). Thus, the nonpolarized targeting results shown in are explained by the fact that after long tannic acid treatment, although the intracellular sorting might still occur, the polarity is lost because tight junctions are impaired and the monolayer permeability is altered, thus allowing both apical/basolateral protein diffusion and passage of the biotin between the apical and basolateral compartments. Although the aforementioned confocal and biochemical experiments clearly showed a direct apical arrival of GFP-GPI, they represent images taken at specific times of chase corresponding to times used in the biochemical assays. The most direct method to analyze the targeting pathway of GPI-APs would be to follow the route of GFP-GPI all the way from the Golgi to the surface in living cells using an imaging approach. This has never been possible before in fully polarized cells grown on filters because of the limits of the imaging systems for acquisition along the z axis, so all the dynamic studies on protein sorting to date have been performed in cells grown on coverslips in semipolarized conditions (; ; ; ). We used a confocal imaging system based on spinning disc technology that has the advantage of being faster and causes less photobleaching in comparison with laser scanning–based confocal methods. This system has been previously used to follow apical and basolateral cargos for short times in cells grown to confluency on coverslips (), whereas we have used it to monitor the arrival of GFP-GPI from the Golgi to the plasma membrane in live cells grown on filters in fully polarized conditions (). After incubation at 19.5°C, cells grown on filters were warmed at 37°C on the microscope stage, and images were taken every 3 min for a total of 60 min (). For each time point, an average of 20 z planes were collected to detect the signal from the whole cell volume. In , selected frames of the time lapse show a gradual brightening only of the apical surface as a result of GFP-GPI arrival both in control and tannic acid–treated cells. However, from the side view images (, bottom), it is possible to note an extension of the fluorescent signal toward the lateral side of the cells in the presence of tannic acid. Videos 1 and 2 (for control and treated cells, respectively; available at ) obtained after volume rendering clearly showed that the fluorescent signal of GFP-GPI, which is present at the beginning of the experiment at the TGN, gradually brightened the apical surface, whereas no transient brightening was observed at the basolateral surface (as indicated by rotation of the entire cell monolayer). Furthermore, Videos 3 and 4 (for control and treated cells, respectively) mounted with a rotation of 45° with respect to the z axis showed clearly that the signal reached the apical surface progressively, which was well defined by its typical dome aspect. Finally, by observing the cells from the top (Videos 5 and 6 for control and treated cells, respectively), the progressive accumulation over time of GFP-GPI exclusively on the top of the cell was particularly evident. A better resolution was obtained by collecting three-dimensional stacks every 60 s, thereby reducing the possibility of missing a rapid passage through the basolateral surface ( and Videos 7 and 8). Furthermore, to avoid both the light absorption and the background from the filter, we seeded cells on filters upside down. In these growing conditions, GP114 and Na,K-ATPase, which are endogenous apical and basolateral markers, respectively, are correctly localized as shown by immunofluorescence (Fig. S4 A), indicating that cells were well polarized. To monitor the arrival of GFP-GPI at the surface, we had to take images every 1 or 3 min for up to 60 min at low magnification, which did not allow the direct visualization of post Golgi carriers. To visualize the trafficking of GFP-GPI containing carriers in cells grown on filters, we imaged the cells every 2–3 s for short time intervals (3–5 min) at high magnification. A range of 4–6 μm in depth of the cell from the bottom to the top of the Golgi apparatus was collected every 2–3 s. The time lapses showed GFP-GPI–containing membranes budding from the Golgi complex as tubular processes and post-Golgi intermediates moving toward the surface (Video 9, available at ). These three-dimensional images were comparable in resolution with the ones taken before in single plane in nonpolarized cells (; ). To determine the site of segregation of apical and basolateral cargos, we transiently cotransfected the CFP and YFP variants of GFP-GPI and GFP-PIT (a fusion construct between the transmembrane and cytosolic domains of the low density lipoprotein receptor and GFP, which is basolaterally localized as shown in Fig. S4 B), respectively. As previously described (; ; ), we also found that GFP-GPI is partially segregated from a basolateral protein (YFP-PIT) in the TGN (Fig. S4 C, top) and that these proteins appear to be in distinct carriers after 10 min of release of the block (Fig. S4 C, bottom). Thus, our data demonstrate that in fully polarized monolayers, the apical delivery of GPI-APs occurs via a direct pathway, and it supports the notion that their sorting occurs intracellularly before arrival at the plasma membrane. Several studies have shown that apical and basolateral proteins segregate into different vesicles upon exit from the TGN (; ; ). GPI-APs are delivered to the apical membrane via a raft-mediated mechanism (; ) and have been thought to be sorted in the TGN (). Imaging studies in living cells have shown that GPI-APs segregate from basolateral proteins in separate carriers in the TGN, and, from there, they reach the apical surface directly (). Notwithstanding, a recent study based on imaging in living cells of one GPI protein (GFP-GPI) has proposed a mechanism whereby GPI-APs may use a transcytotic route to the apical surface of MDCK cells and, therefore, that their sorting occurs after arrival to the basolateral plasma membrane (). These data are in direct contrast with our model of the mechanism of GPI-AP sorting, which is based on segregation and stabilization of GPI-APs in rafts at the level of the Golgi apparatus. We believe that these two events would lead to the budding of an apical carrier that excludes basolateral proteins (). Because also observed a physical segregation of the apical and basolateral cargo in the TGN, the question arises as to why this segregation would occur in regions of the Golgi membrane if the proteins are transported in the same carrier to the basolateral surface where sorting would then occur. We directly investigated the pathway followed by three different GPI-APs to the apical surface in MDCK cells both by biochemical and imaging approaches in fully polarized cells that were grown on filters for 4 d. By using pulse-chase experiments combined with selective surface biotinylation, we found that newly synthesized proteins, PLAP, GFP-GPI, and NTR-PLAP progressively reach the apical plasma membrane without passing through the basolateral surface ( and S1 A). Indeed, we never observed accumulation on the basolateral surface before the apical appearance of these proteins. We obtained the same results when we examined GFP-GPI transport to the surface that previously accumulated in the TGN after a temperature block ( and ). Furthermore, by using a spinning disc confocal-based instrument, we could analyze the arrival of GFP-GPI from the Golgi to the surface in living cells grown on filters ( and Videos 1–9). This was not previously possible because of the thickness of polarized monolayers and the speed and sensitivity of commercially available imaging systems. The results obtained by the biochemical and imaging methods concordantly demonstrate that both newly synthesized and recycled GPI-APs are directly delivered from the TGN to the apical surface, suggesting that they are sorted intracellularly. These results were confirmed by the fact that neither basolateral cleavage by trypsin () nor addition to the basolateral domain of tannic acid (), which inhibits basolateral transferrin endocytosis (Fig. S3), affect the apical delivery of GFP-GPI. This demonstrates that we did not miss any rapid or transient passage through the basolateral surface that could have been undetectable in the less sensitive biotinylation-based targeting assays. The discrepancy between the results of and our own most likely lies in the cell culture conditions and in the different methods used. The experiments were performed in MDCK cells grown on filters for only 2–3 d, when the cells are not, in fact, fully polarized. When we performed our experiments in similar nonpolarized conditions, we found that both GFP-GPI and PLAP were missorted to the basolateral surface () and that this missorting decreased with the time the cells were in culture (). Interestingly, in one experiment after 1.5 d in culture, we observed a transient peak of GFP-GPI at the basolateral surface before it was apically delivered (Fig. S1 B), which might suggest that it uses a transcytotic pathway in semipolarized conditions. However, as previously demonstrated, the transitory use of the transcytotic pathway appears to be a protein-specific feature (), and, in this specific case, it was observed only for GFP-GPI and for NTR-PLAP (Fig. S1 A) but not for PLAP. In addition, we demonstrate that 30 min of tannic acid is sufficient to block transferrin internalization, but it does not impair the direct apical delivery of GFP-GPI from the TGN in fully polarized cells (). However, when we used the same conditions as , who treated the cells with tannic acid for a longer time period (45–60 min treatment), we found that GFP-GPI becomes progressively depolarized (). Indeed, a prolonged treatment with tannic acid causes a redistribution of tight junction proteins and alters the integrity of the monolayer (), thus permitting protein diffusion from one domain to the other. Strangely enough, did not report a basolateral GFP-GPI signal but described only accumulation in intracellular compartments. In some cells, we also observed some intracellular patches below the plasma membrane, as they described (, bottom; 60 min), but never saw an impairment of apical delivery ( and and Videos 1–8). This difference could derive from the fact that they used transiently expressing cells, whereas we have used stable clones. It is well known that transiently transfected cells overexpress proteins, and overexpression is known to saturate the sorting mechanisms and to lead to missorting or the use of different pathways not used by the cells in normal conditions (). Because, in our hands, apical surface delivery of GFP-GPI occurs rapidly (apical GFP-GPI was detected as early as 15 min after warming cells at 37°C after temperature block; and ), it is unlikely that the transcytotic route determines its steady-state distribution. In hepatocyte 5′ nucleotidase, a GPI protein following a transcytotic route to the apical surface requires 3.5 h to reach steady-state apical distribution, and this rate of transcytosis is much slower than that observed for transmembrane proteins (). GPI-APs are slowly endocytosed from both sides (; ), and they also recycle back to the surface three- to fourfold slower than other recycling membrane components (; ). Furthermore, also found that YFP-GPI needs 2 h to be transcytosed from the basolateral side. Therefore, this discrepancy in the timing of kinetics of the arrival to the surface and the transcytotic process is inconsistent with the fact that the basolateral fraction of GPI-APs is a precursor to the apical pool. We directly ruled out this possibility by measuring the amount of missorted basolateral GFP-GPI that was internalized from the basolateral membrane in 1.5-d-old monolayers. In these conditions, a small portion (∼6–8%) of GFP-GPI appeared inside the cells after 30 min of internalization at 37°C (Fig. S5, available at ), suggesting a slow internalization rate inconsistent with the use of a transcytotic pathway. However, in agreement with previous data (), this experiment also showed that some of the protein that missorted to the basolateral membrane in cells that were not fully polarized can reach the apical surface via a transcytotic pathway, as we observed the appearance of GFP-GPI at the apical surface 60 and 120 min after internalization (Fig. S5). Our data also support the model that sorting of GPI-AP occurs intracellularly. We have recently proposed that GPI-AP oligomerization or clustering in high molecular weight complexes is the prime mechanism determining the apical sorting of GPI-APs (). This is because it leads to the stabilization of proteins into rafts and to the coalescence of more rafts with the consequent formation of larger functional rafts, which provide the platform for the budding of an apical carrier. Oligomerization of GPI-APs begins in the medial Golgi and is concomitant with raft association (), therefore supporting the hypothesis that apical GPI-AP sorting occurs at the TGN. In this study, we show that segregation between GFP-GPI and a basolateral marker (GFP-PIT) occurs during the block in the TGN and that the two proteins appear in distinct carriers after 10 min of chase at 37°C. Nonetheless, we cannot exclude the possibility that GPI-APs require travel through some endosomal compartments before reaching the apical membrane. Recently, it has been found that the inactivation of REs affect the transport of VSV-G and of an apically targeted mutant from the Golgi to the plasma membrane, suggesting a role of REs for polarized sorting in endocytic and secretory pathways (). A presurface post-Golgi compartment could act as a main sorting station or as a second step after sorting at the TGN. Alternatively, it could be a pathway of salvage in the case of partial polarity or of saturation of the normal sorting machinery. For basolateral proteins, it has indeed been shown that the same sorting signals function at the level of the TGN and REs. Further studies focused and optimized to visualize the trafficking between these compartments in fully polarized cells are necessary to elucidate these different possibilities. Cell culture reagents were purchased from Invitrogen. Antibodies were purchased from the following companies: polyclonal anti-GFP from CLONTECH Laboratories, Inc.; monoclonal anti-GFP and Cy-5 transferrin from Invitrogen; anti-PLAP from Rockland Biosciences; and anti–Na, K-ATPase from Upstate Biotechnology. The antibodies against PATJ, GP114, and p75 were gifts from A. le Bivic (Faculté des Sciences de Luminy, Marseille, France). Biotin and streptavidin beads were obtained from Pierce Chemical Co. HRP-linked antibodies and streptavidin were purchased from GE Healthcare. All other reagents were purchased from Sigma-Aldrich. MDCK cells were grown in DME containing 5% FBS. Stable clones expressing GFP-GPI and PLAP were previously obtained (). In particular, the GFP-GPI construct previously used by us and others (; ; ) was a gift from S. Lacey (Southwestern University, Georgetown, TX). The fusion was constructed in the eukaryotic expression vector pJB20. It has an EcoRI site at the 5′ end, a HindIII site at the 3′ end, and a PstI site that separates the ecto- and anchor domain. TER was measured using a meter (Millicell-ERS; Millipore). Unless specifically indicated, MDCK cells were grown on Transwell filters for 3.5–4 d, washed with PBS containing CaCl and MgCl, fixed with 4% PFA, and quenched with 50 mM NHCl. Depending on the experiment, cells were permeabilized with 0.075% saponin. Primary antibodies were detected with FITC or TRITC-conjugated secondary antibodies. Images were collected using a laser scanning confocal microscope (LSM 510; Carl Zeiss MicroImaging, Inc.) equipped with a plan Apo 63× NA 1.4 oil immersion objective lens (Carl Zeiss MicroImaging, Inc.). As previously described (), the quantification of mean fluorescence intensities in selected regions of interest were performed using a laser scanning microscope (LSM 510; Carl Zeiss MicroImaging, Inc.). In particular, the fluorescence intensities of areas of equal size in a single z plane through the cell monolayer (from a range of 1–3, 4–6, and 8–12 μm starting from the top of the cell for apical, Golgi, and basolateral signals, respectively) were measured and corrected for background. Time-lapse images were collected using a confocal microscopy system (UltraView ERS; PerkinElmer) equipped with a microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.). A plan Apo 63× or a 100× NA 1.4 oil immersion objective lens was controlled by a piezoelectric z stepper. To image living cells, we mounted the filter on an optical glass in a petri dish containing CO independent medium (150 mM NaCl, 5 mM KCl, 1 mM CaCl, 1 mM MgCl, and 20 mM Hepes, pH 7.4). Cells were imaged using a thermostatic chamber mounted on the microscope set at 37°C. Cells were imaged every 3 min for a total of 50–60 min or every 1 min for 10–15 min, collecting 15 or 20 z slices (confocal depth of 0.5 μm) each time. For short time lapses (3–5 min), cells were imaged every 2 s, collecting four to six z slices (confocal depth of 1 μm) at high magnification using a plan Apo 100× NA 1.4 oil immersion objective lens and an optovar 1.6× lens. Images and videos were volume rendered by Volocity software (PerkinElmer). ext-link #text C e l l s g r o w n o n f i l t e r w e r e i n c u b a t e d w i t h 2 5 μ g / m l t r y p s i n i n D M E w i t h o u t s e r u m f o r 3 0 m i n t w i c e . T h e y w e r e t h e n s u b j e c t e d t o t e m p e r a t u r e b l o c k a s d e s c r i b e d b e l o w a n d w a r m e d a t 3 7 ° C f o r d i f f e r e n t t i m e s i n c u l t u r e m e d i u m c o n t a i n i n g 1 5 0 μ g / m l c y c l o h e x i m i d e . A t t h e e n d o f e a c h t i m e p o i n t , s u r f a c e p r o t e i n s w e r e s e l e c t i v e l y b i o t i n y l a t e d , i m m u n o p r e c i p i t a t e d w i t h a n a n t i b o d y a g a i n s t G F P , a n d r u n o n S D S - P A G E . B i o t i n y l a t e d p r o t e i n s w e r e r e v e a l e d b y H R P - c o n j u g a t e d s t r e p t a v i d i n . Q u a n t i f i c a t i o n s w e r e p e r f o r m e d w i t h I m a g e s o f t w a r e ( N a t i o n a l I n s t i t u t e s o f H e a l t h ) . To achieve an almost complete protein block in the TGN, we used a previously published protocol (). Filter-grown cells were incubated at 19.5°C for 2 h in areal medium (F12 Coon's modified medium without NaHCO and with 0.2% BSA and 20 mM Hepes, pH 7.4). In the last hour at 19.5°C, they were treated with 150 μg/ml cycloheximide. 1 μCi [C]inulin (MP Biomedicals) diluted in 500 μl DME was added to the apical chamber of the Transwell filters, and cells were incubated for 30 or 60 min at 37°C. The amount of [C]inulin that remained in the apical well, traversed the filter to the basal well, or remained cell associated was quantified in a liquid scintillation counter. Fig. S1 shows pulse chase and biotinylation targeting assays. In fully polarized cells (A), NTR-PLAP is directly targeted to the apical surface, whereas it undergoes transcytosis in unpolarized cells. In not fully polarized cells (B), a portion of GFP-GPI can be indirectly sorted to the apical membrane. Fig. S2 shows by immunofluorescence (A) and Western blotting (B) that trypsin is a suitable tool to remove GFP-GPI without impairing the integrity of the cell monolayer. Fig. S3 shows that tannic acid impairs transferrin internalization but does not affect surface binding. Fig. S4 A shows that apical and basolateral markers are correctly localized in cells grown on filters upside down, indicating that cells are well polarized. It also shows that YFP-PIT is basolaterally localized, and it is partially segregated from CFP-GPI at the TGN and is in different post-TGN carriers (B and C, respectively). Fig. S5 shows that ∼6–8% of GFP-GPI is slowly internalized from the basolateral surface. Videos 1–9 show the arrival of GFP-GPI from the TGN to the plasma membrane. They clearly show the direct appearance of GFP-GPI fluorescence at the apical surface both in control and tannic acid–treated cells. All z planes were volume rendered by Volocity software. Online supplemental material is available at .

Epithelial cells are characterized by the asymmetric distribution of plasma membrane proteins to form apical and basolateral domains. Many membrane proteins are thought to reach their respective domains by intracellular sorting events mediated by distinctive targeting elements (; ; ). For example, many apical proteins, including those with glycosylphosphatidylinositol (GPI) membrane anchors, are sorted into clustered lipid rafts and traffic to the apical surface (; ; ). In contrast, many basolateral proteins, such as low density lipoprotein receptor and vesicular stomatitis virus glycoprotein (VSVG), are sorted as a result of the recognition of signals localized in the cytoplasmic tail (). The protein components responsible for at least some basolateral sorting events, such as the AP–1B clathrin adaptor complex, are beginning to be discovered (; ; ). However, the actual pathways of polarized membrane protein trafficking, including the identity or location of sorting sites, the itinerary taken by individual proteins, and the spatial distribution of plasma membrane delivery sites, remain largely unresolved. Polarized membrane protein sorting on the secretory pathway has long been presumed to occur at the TGN (; ), although the closely associated recycling endosomes also appear to play an important role (; , ; ). Recently, however, questions have emerged concerning the extent to which polarized sorting even occurs intracellularly (). In one study, the transport of polarized membrane proteins was assayed in MDCK cells whose apical or basolateral surface had been selectively inactivated by tannic acid fixation. Striking images suggested that a well-characterized reporter of apical proteins, a GPI-anchored GFP (GPI-GL-YFP), was first inserted into the basolateral surface and transcytosed across the cell to reach the apical domain (). Although inconsistent with some classic biochemical studies (; ; ), this indirect pathway is reminiscent of what is thought to occur for many apical proteins in hepatocytes (). In MDCK cells, however, apical proteins bearing conventional membrane anchors were suggested to take the classic direct route (; ; ; ). It remains similarly unclear whether membrane proteins insert at specific sites on the apical or basolateral surfaces. Junctional complexes have been suggested to define a spatially restricted insertion site for basolateral traffic based on the distribution of tethering complexes (Sec6–Sec8) involved in vesicle docking at the basolateral membrane (). However, the only direct evidence from live cell imaging experiments thus far is that basolateral proteins do not first appear at the adherent surface of coverslip-grown MDCK cells (). Some transport vesicles were seen to enter the junctional complex region, but definitive evidence for vesicle fusion to the plasma membrane or a quantitative assessment of vesicle traffic to this domain of the lateral surface was not achieved. Such considerations have emphasized the need for visualizing the biosynthetic pathway of membrane proteins in filter-grown and fully polarized epithelial cells. Although this approach has thus far proved challenging because of the cells' z-axis height (∼10 μm; ), we have now been able to image and quantify the features of secretory membrane traffic in live, unperturbed MDCK cells. By combining laser scanning confocal microscopy with rapid data acquisition and computation, we were able to visualize and quantitate polarized membrane protein traffic in filter-grown MDCK cells for a period of ∼30 min. As a test, we first studied the intracellular sorting and insertion of a basolateral membrane protein, ts045 VSVG-YFP (). VSVG-YFP was transfected into MDCK cells expressing CFP-tagged TGN38 (TGN38-CFP; ) as markers for the TGN. VSVG-YFP was accumulated in the ER by incubation at 40°C overnight and was released and chased to the TGN by a 1-h incubation at 20°C (). Synchronous cell surface transport was imaged by transferring the filters to the permissive temperature (31°C) on a temperature-controlled stage, mounted as shown in . The preparations remained viable for >12 h. YFP and CFP fluorescent signals in single MDCK cells were collected in multiple z stacks over a 30–40-min time course and combined to produce three-dimensional (3D) videos (Video 1; available at ). 3D stacks were acquired at 1-min intervals to minimize photobleaching. At the start of the 31°C chase, most of the VSVG-YFP (, red) colocalized with TGN38-CFP (, green) as expected (, 0 min). VSVG-YFP then gradually exited from the TGN and appeared at the lateral membrane (, 17 and 34 min). As found previously for MDCK cells on glass coverslips (), VSVG was not observed reaching the basal (filter attached) domain of the plasma membrane; this domain was thus excluded from further analysis. Although the time resolution was not enough to allow tracking of individual transport vesicles, we noted only two significant accumulation sites of VSVG: the TGN region and the lateral membrane, suggesting that after exit from the TGN, there were no rate-limiting steps before plasma membrane appearance. We next quantified the amount of VSVG-YFP in TGN and basolateral regions in the 3D renderings at each time point. Because all signals were within the linear range of the detector, the amount of VSVG-YFP was proportional to the fluorescence signal in any given region. Measurements were converted into percentages of total signal in the cell to compensate for photobleaching (determined separately for each probe; Fig. S1, available at ). The amount of VSVG in the TGN region decreased concomitantly with the increase in the amount of VSVG at the lateral membrane (). TGN exit could be approximated by a linear decay curve, suggesting that the exit rate was independent of VSVG concentration. The rate of VSVG appearance at the lateral membrane increased after a brief lag (, red), suggesting at least one intermediate step before VSVG's plasma membrane arrival. To compute the kinetics of arrival at the lateral domain, we constructed a simple three-state model that described the localization of VSVG in its three definable compartments: TGN, the lateral plasma membrane, and transport carriers between the TGN and plasma membrane, the latter representing cytoplasmic volume that was not associated with the TGN or the lateral membrane (, open circles; the line is nonlinear because of its multicomponent nature). As illustrated in , two forward rate constants were used to fit the data (model 1): k, the rate of exit from the TGN, and k, the rate of exit from the transit compartment. k and k were determined by measuring the rate of exit from the TGN and arrival at the lateral domain, which was defined to include the cell margin and an intracellular volume ∼1.5 μm from the cell margin. Thus, k describes the duration of VSVG in the transit compartment; i.e., the time required to transfer from the TGN to the plasma membrane but not docking and fusion to the membrane. The kinetic analysis indicated that exit from the TGN was rate limiting (k = 0.019 min), with 1.9% of the VSVG-YFP exiting the TGN per minute, which is approximately equal to the first-order rate measured using a linear fit to the TGN exit curve. k was ∼10 times faster than TGN exit (; = 12 cells), indicating that once in transit, VSVG was rapidly transferred to the lateral domain. The Golgi exit rate was comparable with that measured in nonpolarized COS cells by . Previous studies have suggested that basolateral traffic is initially targeted to a site near the junctional complex and possibly even subjected to sorting from apical proteins upon delivery (; ; ). Therefore, we analyzed our dataset to determine whether there was evidence for a transient accumulation of VSVG-YFP at the junction-associated region or any other subdomain of the lateral membrane. We modified the kinetic model to include a new rate k that would describe possible processes restricted to the junctional complex region (, model 2). Such events might include transport vesicle docking, vesicle fusion, sorting of apical versus basolateral proteins, or exit from the junction-associated region (via diffusion or vesicular transport). At steady state, the amount of VSVG that accumulated in the junction-associated region would thus depend exclusively on k and k. If this junctional region was the unique site to which post-TGN carriers were targeted, we might expect a transient accumulation of VSVG in that segment of the lateral domain. To test this hypothesis, we measured the appearance of VSVG-YFP in three equal vertical, circumferential segments of the lateral membrane. The most apical section contained the junctional complex, as confirmed by staining with the tight junction marker ZO-1 (unpublished data). Contrary to expectations, however, VSVG-YFP did not appear preferentially in this upper segment. Fluorescence instead appeared in the bottom two lateral segments at equivalent rates (, red and blue dots). The same trend was observed for the apical-most segment (, yellow dots), although the amount of signal was reduced because this segment likely also contained a portion of the apical domain (above the junctional complex). Nevertheless, if all of the VSVG represented by the yellow dots () reflected accumulation in the junction-associated region, the quantity was far less than predicted for preferential targeting (, dashed line). To fit the actual data to model 2 (), k would need to be ≥1.6 min (corresponding to t of <26 s). k is unlikely to be that fast because our previous study using glass-grown cells found that the docking of post-TGN VSVG-containing transport vesicles is considerably slower than 1.6 min (k = 1.1 min; t of 39 s; ). Thus, despite the possible restriction of vesicle-tethering complexes close to the junctional complex (), VSVG-YFP is delivered to a broad area of the lateral domain or, alternatively, the docking and fusion of vesicles at a restricted region is far faster than observed for other closely related events. Next, we studied the intracellular transport of a newly synthesized apical membrane protein, GPI-GL-YFP, which is the same construct also used by . Newly synthesized GPI-GL-YFP was accumulated in the TGN by a 1-h temperature block at 20°C (), released at 31°C, and imaged at 1-min intervals in three dimensions (Videos 2 and 3, available at ). GPI-GL-YFP (, red) migrated out from the TGN region (, green) and appeared at the apical surface, which is defined as red pixels that were within ∼2 μm of the apical cell margin (; 0, 9, and 18 min of chase; see Image and data processing). Transport was largely complete within 20 min. Importantly, no GPI- GL-YFP was observed visually at the lateral membrane at any time point. There was, however, an intermediate time when GPI-GL-YFP appeared as a region or compartment juxtaposed apically to the TGN (Videos 4 [∼12 min] and 5). This region could correspond to a subdomain of the TGN or a subapical compartment containing presorted GPI-anchored molecules (). In any event, the video data strongly suggested a simple model in which GPI-GL-YFP migrates upward out of the TGN through an intermediate subapical domain and reaches the apical surface without ever appearing basolaterally. We then quantified the kinetics of GPI-GL-YFP transport from the TGN to the apical regions. As shown in , the amount of GPI-GL-YFP in the TGN region decreased in accordance with the increase in the amount of GPI-GL-YFP at the apical membrane (green and red dots, respectively). The simplest three-state model (, model 1) accurately represented the transport data (, solid green and red lines). Again, the rate-limiting step along the apical pathway was exit from the TGN. At 0.023 min (2.3% GPI-GL-YFP transferred out of the TGN per minute), this rate was similar to that derived for the TGN exit of VSVG-YFP (; = 9 cells). To test quantitatively whether GPI-GL-YFP was delivered to the junctional region or the lateral domain before appearing apically (, inset; models 2 and 3), we determined the distribution of GPI-GL-YFP at each time point in the reconstructed 3D datasets. We first compared the appearance of GPI-GL-YFP in the apical membrane (choosing a segment ∼2.5 μm from the cell lateral margin to avoid any contribution of the lateral domain; , inset; red) versus a perimeter segment including the edge of the apical membrane, the junctional region, and the apical-most segment of the lateral membrane (, inset; yellow). GPI-GL-YFP accumulated in both regions with similar kinetics (, red and yellow dots), but far more GPI-GL-YFP accumulated in the delimited apical region despite being a smaller volume. As with our analysis of VSVG-YFP transport, the only way model 2 could be consistent with the data is if the junctional residence of GPI-GL-YFP was so short as to be experimentally undetectable (k of >4 min). This is also unlikely because the newly delivered GPI proteins are immobile on the apical surface () and should diffuse from a lateral insertion site slowly, enabling their detection. Thus, it seems unlikely that GPI-GL-YFP is inserted at the apical margin or the junction-associated region. Although not visually obvious, a small fraction of GPI-GL-YFP was quantified as residing at the lower two segments of the lateral membrane at the start of the 31°C chase; this amount decreased with the time of chase (, blue and pink dots). However, this behavior was not consistent with a major indirect route to the apical membrane ( and , inset; model 3). Such a mechanism was calculated to require a flat or slightly increasing curve for the amount of lateral GPI-GL-YFP during the 20-min time course used here, which is inconsistent with the data. We suspect that the lateral GPI-GL-YFP reflected a partial missorting as expected (; ). Although it is possible that the lateral GPI-GL-YFP proceeded by transcytosis to the apical side, which is consistent with previous results (), the kinetic analysis is inconsistent with it being an obligatory intermediate on the apical pathway for the entire newly synthesized cohort. At all times of chase, ∼20% of GPI-GL-YFP fluorescence was localized neither to the TGN region nor to the apical (or lateral) membrane. Instead, it was distributed intracellularly underneath the apical membrane (, open circles). This signal may represent GPI-GL-YFP in equilibrium with endosomes in the apical cytoplasm. Therefore, it was important to determine whether the GPI-GL-YFP that appeared in the apical region had actually fused with the apical membrane. We repeated the experiment with fluorescent anti-GFP antibody in the medium (accessible to both apical and basolateral membranes). The insertion of GPI-GL-YFP will expose the extracellular YFP domain to the medium, enabling concentration of anti-GFP at the plasma membrane. Indeed, as GPI-GL-YFP was sorted away from TGN38-CFP and delivered toward the apical domain, it was accompanied by a steadily increasing amount of anti-GFP labeling on the apical membrane (). No significant anti-GFP labeling in the basolateral domain was detected, again indicating that little GPI-GL-YFP was inserted in the basolateral membrane. The total amount of antibody on the apical membrane increased with time, and the rate of insertion in the apical domain was in agreement with the kinetic data derived from quantifying GPI-GL-YFP fluorescence (). In MDCK cells, the TGN may exhibit an overall expansion in surface area as a result of the 20°C block (; ). Because of this, upon shifting to 31°C, the TGN returns to its original size, perhaps slowing the initial rates of exit. Therefore, we measured the TGN exit rate of VSVG-YFP in the absence of a 20°C block. VSVG-YFP was accumulated in the ER at 40°C and was released by shifting directly to 31°C. The colocalization of VSVG-YFP with TGN38-CFP was used to estimate entry and exit from the TGN. Indeed, the rate of exit under these conditions (0.008 ± 0.002 min; = 2) was somewhat slower as compared with the exit rate after the 20°C block (). This suggests that after the 20°C block, VSVG exit from TGN may occur normally, with the faster rate perhaps reflecting the accumulation of VSVG in the TGN (). To further characterize the effect of the 20°C block on VSVG-YFP exit from the TGN, rates were calculated from the first and last 10 min of our complete dataset ( = 12). On average, the rate calculated for the last 10 min was 17% slower, which is also inconsistent with the possibility that VSVG exit is slowed after 20°C block. When this analysis was performed for GPI-GL-YFP exit, we found that the initial rate was somewhat slower in three of nine cells (e.g., the initial lag in ; green line), perhaps reflecting the time required for partitioning of GPI-anchored proteins into lipid microdomains before TGN exit (; ). To avoid the issue of temperature blocks entirely, we developed an assay to monitor transport kinetics at steady state and to assess independently the contribution of basolateral to apical transcytosis to GPI-GL-YFP targeting. For this purpose, we determined the binding and transport of anti-GFP antibody that was vectorially added to the basolateral or apical media of filter-grown cells. GPI-GL-YFP at both surfaces was first blocked with nonfluorescent anti-GFP IgG for 2 h. Cells were then incubated with anti–GFP-AlexaFluor647 in the basolateral medium for 30 min at 37°C (, diagram; blue antibody). Blue antibody detected by 3D imaging at the apical surface would, therefore, reflect GPI-GL-YFP transcytosis. At the end of the chase (at 0°C), anti–GFP-AlexaFluor594 (, diagram; red antibody) was added to the apical medium to detect GPI-GL-YFP molecules that had failed to bind the blue antibody (i.e., the population of GPI-GL-YFP that had not been exposed to the basolateral medium). As expected, the apical membrane was heavily labeled with AlexaFluor594 (red) antibody and was only slightly labeled with AlexaFluor647 (blue) antibody (). Thus, the majority of GPI-GL-YFP was apparently transferred directly from the TGN to the apical surface, as it did not bind anti-GFP in the basolateral medium. For reference, the TGN was indicated by TGN38-CFP (, green). The digital reconstructions were used to quantify the relative contributions of direct versus transcytotic transport by determining the ratio of the two antibodies at the apical surface. To eliminate any bias caused by the differences in detection or binding of the two antibodies, we also collected datasets in which the antibodies were reversed. In either event, <5% of GPI-GL-YFP on the apical membrane came via the transcytotic pathway (i.e., had been tagged by the basolateral antibody; ). The result was independent of the antibody concentration, indicating that anti-GFP saturated the membrane-bound GPI-GL-YFP. The same result was obtained if the cells were blocked for 1 h at 20°C, as in the imaging experiments (). As a positive control to show that established transcytosis markers could be labeled on the basolateral surface while en route to their final destination (the apical surface), we tested cells expressing NgCAM-GFP (). Cells were incubated with anti–NgCAM-AlexaFluor647 (, blue antibody) in the basolateral medium for up to 90 min at 37°C (transcytosis of NgCAM takes ∼90 min, which is slower than the apical appearance of GPI-GL-YFP; ). The cells were cooled to 0°C, and anti–NgCAM-AlexaFluor594 (, red antibody) was then added to the apical medium to detect any NgCAM that had been directly inserted apically. The blue antibody labeled both the apical and lateral surfaces as well as a variety of intracellular endosomal compartments, which is consistent with transcytosis (). Only a small amount of red antibody bound to the apical surface. Quantitation of red versus blue revealed that >95% of the apical NgCAM-GFP had captured antibody in the basolateral medium (). 3D imaging of living filter-grown MDCK cells has finally permitted a quantitative analysis of the rate and spatial restriction of domain-specific membrane protein transport in polarized epithelial cells. To image entire pathways, however, it was necessary to collect data over an ∼30-min time course, requiring that frames be taken at relatively long (1 min) intervals to minimize photobleaching. Although interstack time intervals were too long to monitor individual transport vesicles, we did demonstrate unambiguously that at least two well-described apical and basolateral proteins were sorted intracellularly and inserted with a high degree of fidelity directly into their respective domains. Our visual and computed results confirm several early biochemical studies on traffic in polarized MDCK cells (; ; ; ; ; ), conclusions that were recently questioned based on immunofluorescence studies of vectorially fixed (and therefore highly perturbed) cells, using the same constructs as used here (; ). Some interesting differences were revealed between the apical and basolateral pathways. Although VSVG-YFP and GPI-GL-YFP had comparable exit rates from the TGN, the overall rate of exit from the transit compartment and apical delivery was demonstrably slower in the case of GPI-GL-YFP. One reason for this may be the appearance of a subapical intermediate compartment that accumulated GPI-GL-YFP at equilibrium. Similar compartments have been previously noted and termed apical endosomes or vacuolar-apical compartments, and they do appear to play a role in the endocytosis and recycling of apical components or as a biosynthetic transport intermediate (; ; ; ). No such kinetic intermediate was observed on the basolateral pathway. Given the limited time resolution used in this study, we would not have expected to detect the transit of VSVG or GPI-GL-YFP through recycling endosomes () unless this step was similarly rate limiting (i.e., <0.23 min). It apparently was not. The apical endosomal compartment may accumulate GPI-GL-YFP because of its ability to partition into relatively immobile lipid microdomains that may be abundant in these compartments (; ; ). The processes of vesicle docking and fusion were not studied in detail. For VSVG, we could not distinguish whether VSVG transporters were docked versus fused with the lateral membrane. Increased time resolution alone would certainly permit the tracking of individual vesicles, as accomplished by . However, from the point of view of establishing kinetics, such an analysis would not help. At present, there is no way to judge definitively that fusion of a single vesicle has occurred at the lateral surface, which cannot yet be studied by total internal reflectance microscopy. In our dataset, appearance at the lateral membrane represents the aggregate rate of docking and fusion followed by rapid diffusion of VSVG in the lipid bilayer. As a result, we cannot definitively rule out the possibility that laterally targeted vesicles dock and fuse at a site close to the junctional complex. Because of the high mobility of VSVG on the lateral membrane, even if insertion is strictly limited to the junctional region, VSVG-GFP would appear evenly distributed on the lateral membrane in our experiments. However, the data do predict that if insertion at the junctional region is obligatory, the mean residence time at the junction (including docking, fusion, and diffusional exit) must be 26 s or less. Longer residence times, such as those measured for post-TGN VSVG-containing vesicles docking at the plasma membrane of glass-grown cells (t of 39 s; ), would have been detected. Although our approach is certainly just a methodological beginning to the problem of elucidating the kinetic and spatial control of membrane traffic in complex cells, it has permitted a definitive resolution of the vectorial nature of biosynthetic protein transport in MDCK cells, which is one of the oldest and most fundamental problems in the field. Clearly, the approach will lend itself to other cargos and cell types, especially cells that have been manipulated to exhibit the reduced expression of candidate genes involved in polarity or transport. A confocal laser scanning inverted microscope (LSM-510; Carl Zeiss MicroImaging, Inc.) with a temperature-controlled stage set at 31°C was used for image acquisition. To minimize refractive index mismatch, we used a 40× NA 1.2 water immersion lens (Carl Zeiss MicroImaging, Inc.) with temperature and coverslip thickness correction. To achieve quantitative 3D live cell imaging of the entire sorting pathway that takes ∼30 min, with minimum photobleaching of the fluorescence, the excitation laser power was adjusted to instrument minimum (0.1%), and the maximum laser scanning speed was used. It took 2–30 s (typically 15 s) to complete 6–10 z stacks of a single cell at ∼1 μm per image, depending on the number of fluorescent channels to scan (typically four), the size of the scanning areas (typically 50–100 μm), and the number of scanning repeats (typically two). Pinhole size was increased up to 4 Airy units to increase light collection without significant background noise from stray light. Image resolution is independent of pinhole size under these conditions. Total photobleaching of our CFP and YFP probes (see next section) under these conditions was ≤40% (see Fig. S1). The detectors (photomultiplier tubes) were adjusted to maximally amplify but not saturate the signal. The collected digital signals were proportional to the photon input (LSM-510; Carl Zeiss MicroImaging, Inc.), which was, to the first order, proportional to the amount of fluorescent proteins being illuminated in a given area. To compensate for the slow z drift during the course of recording, the Zeiss software package was used to automatically correct the focus every 3 min according to the z position of the filter on which the MDCK cells were grown. Standard MDCK II cells or MDCK II that stably expressed human transferrin receptor () were cultured under standard procedures (); i.e., incubated at 5% CO in DME with 10% FCS and 1× glutamine. TGN38-CFP, VSVG-SP-YFP (abbreviated here as VSVG-YFP), GPI-GL-YFP, NgCAM-GFP constructs, and anti-NgCAM antibody were described previously (; ; ; ). Anti-GFP antibody labeled with AlexaFluor594 and 647 was obtained from Invitrogen. Anti-NgCAM was labeled with AlexaFluor568 and 647 with the Zenon Antibody Labeling Kit (Invitrogen). For microscopy experiments, MDCK cells were seeded at 2 × 10 cells/well on polycarbonate transwell filters (clear type; Corning) in six-well plates. 56 h later, cells were transfected with 2–3 μg DNA per well using LipofectAMINE 2000 (Invitrogen) according to the manufacturer's instructions and incubated at 40°C (except for transcytosis assays, which were performed at 37°C). The 40°C overnight incubation blocked the temperature-sensitive (ts045) VSVG-YFP in the ER but not GPI-GL-YFP or NgCAM-GFP. Imaging was performed ∼16 h after transfection. MEM medium without phenol red (Invitrogen) was supplemented with 10 mM Hepes and used for all microscope experiments (MEM-H). A 35-mm petri dish (Matek) containing a coverslip bottom was precleaned with ethanol. A piece of filter was cut off (together with a patch of cells grown on top) and placed on the coverslip with cells facing down the coverslip. A custom-made three prop weight was put on top of the filter–cell complex for sample stabilization while enabling free medium exchange for the cells through the space between the cells and the coverslip (). Cells were viable for >12 h. All of the cells recorded (total number indicated in ) showed behavior and kinetics similar to those shown in and . For all assays (except the transcytosis assay), the incubation medium was MEM-H. When needed, the medium was supplemented with 33 ug/ml transferrin-AlexaFluor568 and 250 μg/ml dextran-AlexaFluor647 or 3 μg/ml anti-GFP antibodies. Transferrin and dextran were used to label the recycling endosomes and lysosomes, respectively, and VSVG transport rates in MDCK cells expressing human transferrin receptors was indistinguishable to normal MDCK cells. Cells were incubated at 40°C with MEM-H for 1 h and blocked at 20°C for 1 h to accumulate cargo (VSVG-YFP and GPI-GL-YFP) in the TGN. GPI-GL-YFP that was already on the apical surface was removed by 0.8 ug/ml phosphatidylinositol-specific PLC (Sigma-Aldrich) treatment for 30 min at 20°C in PBS supplemented with 100 μg/ml Ca and Mg. The time from the placement of the cells in 31°C medium to the start of the image acquisition was recorded (t) and was typically between 3–20 min. Live cell microscope experiments (except for the transcytosis assay) were performed at 31°C with 20 μg/ml cycloheximide to inhibit new protein synthesis (). Cells were incubated with 50 μg/ml of unlabeled antibody in both the apical and basolateral medium for 2 h at 37°C, washed, and chased with 10 μg/ml AlexaFluor594 (or AlexaFluor568)-labeled antibody in basolateral medium. The chase time was 30 min for the GPI-GL-YFP experiment and 90 min for the NgCAM-GFP experiment to account for the slow transcytosis of NgCAM (). Cells were washed, and a 10× dilution of the Zenon Kit (Invitrogen) blocker reagent was added to the apical medium and incubated for 10 min. Cells were washed, and 10 μg/ml AlexaFluor647-labeled antibody was added to the apical medium and incubated for 5 min. Cells were then incubated in antibody-free medium for 10 min at 0°C, and the filter–cell complex was cut off and imaged at <20°C without fixation. The duration of the microscope experiment was <10 min, and the pattern of the antibody labeling was unchanged during the time. Similar results were obtained with 4% PFA-fixed cells except that the cells slightly deformed in the axial direction as a result of fixation, and the autofluorescent background was higher. For the GPI-GL-YFP transcytosis assay, the antibody concentration described was referred to as 3× in the text. The experiment with the AlexaFluor647 antibody used in the basolateral medium and AlexaFluor594 (or AlexaFluor568) in the apical medium were performed simultaneously under otherwise identical conditions. Images were processed and quantified with LSM-510 software (Carl Zeiss MicroImaging, Inc.), and Volocity with the visualization and classification modules (Improvision). Slight submicrometer spatial spectral shifts were sometimes observed and corrected using the software shift function. Background noise (e.g., the fluorescent signal in neighboring cells not expressing GFP or its color variant) was minimal when optimal gain/offset settings for the detectors were used. No nonlinear algorithms were used to alter the quantification of the fluorescent signals, so the final signals were proportional to the amount of fluorescent proteins. In some cases, the reflection of the laser from the filter creates wide-spectrum noise at the z sections close to the filter, and the affected sections were removed from further processing. The definition of cellular structures are as follows: the TGN region was defined by the 3D TGN38-CFP image through the classification process in Volocity. Specifically, an intensity threshold was first applied to eliminate background fluorescence followed by a medium spatial filter to smooth the image. Background signal was further rejected using particle size threshold to exclude particles that were <0.06 μm. The resulting fluorescent volume was used to represent the TGN structure; its fluctuation was <20% over 20 min. The boundary of the cell of interest (that expresses both CFP and YFP) was determined by the increased background fluorescence level in the target cell. The apical domain was defined as the top 25% of the cell image minus the TGN region if there was any overlap. The lateral domain was defined as the volume from the cell's lateral edge to approximately two thirds of the distance from the cell's center vertical axis minus the TGN and apical regions. For kinetic measurements, at every time point, the total fluorescent signal of the transport cargo proteins (GPI-GL-YFP and VSVG-YFP) in the defined structures (such as the TGN) was quantified, and the final signal was calculated by the fluorescent signal in that structure divided by total fluorescent signal in the whole cell. In the transcytosis assay, we measured total fluorescence on the apical domain because of antibody added to the apical (F) or basolateral medium (F) and calculated the ratio k′ = F/F. Two sets of the experiments—one with the AlexaFluor647 antibody in the basolateral medium and AlexaFluor594 (or AlexaFluor568) in the apical medium and the other with the antibody color swapped—were performed side by side. The geometric mean of the k′ measured from the two sets of experiments was calculated (k). The fraction of new apical proteins that captured antibody at the basolateral surface was expressed by k/(1 + k). The protein transport steps that we observed fit well with a simple two-step process (, model 1) in which the intermediate transit of proteins was detected as significant intracellular fluorescence outside of the TGN. Protein quantity in the TGN under the experimental condition P = P × [1 − k × (t + t)], where P is the initial protein quantity and t is the duration between the start of the 31°C chase and the start of the recording, was described in Time-series microscope experiment. Protein quantity at the destination plasma membrane is P = P × [k × (t + t) − k/k × [1 − e]] and in the intermediate transit is P = P × k/k × (1 − e) + D, where D is the fixed amount of intracellular labeling as a result of various background noise. Thus, the total fluorescence is P = P + P + P. P/P and P/P were globally fitted to the experimental curves such as those in and (SigmaPlot software; Systat). D/P from all experiments was ≤35% (typically ∼15%). In models 2 and 3, three steps with rates k, k, and k are used. The last two states were first combined, and such a two-step process was fitted to the data to obtain k and k as described above. For model 2, P = k/(k − k) × [P − P × k/k × [1 − e]] and P = P × k × (t + t) − P − P. P = P × (1 − g) and P = P + P × g, where g is the proportion of the lateral membrane included in the yellow section ( and ), were fitted to the fluorescence in the blue + purple lateral area and the yellow area, respectively. For model 3, P = k/(k − k) × [P − P × k/k × [1 − e]] and P = P × k × (t + t) − P − P. In fitting the data to model 1, the degree of freedom-adjusted R was 0.97 ± 0.03 for all of the nonlinear regression fits to the VSVG-YFP data and 0.92 ± 0.05 for all GPI-GL-YFP data. R from fitting data to model 1 (two-step model) was statistically indistinguishable from the R for fitting to models 2 and 3 (three-step models), suggesting that the simplest two-step model can explain the data as well as the three-step model; thus, introducing another parameter, k, can be unnecessary. Fig. S1 shows the quantitation of total fluorescence signal (, dataset) used to correct for loss as a result of photobleaching. Video 1 is an animated 3D video of . Videos 2–4 are animated 3D videos of . Video 5 shows an illustration of the post-TGN subapical accumulation of GPI-GL-GFP. Online supplemental material is available at .

Genetic studies in yeast have identified a subset of () mutants, which are called the class E mutants. These mutants display an exaggerated prevacuolar/late endosome compartment, called the class E compartment, which is caused by defects in multivesicular body (MVB) sorting (). There are 17 soluble and 1 membrane class E proteins in yeast, including Vps27p, the ESCRT-I, -II, and -III complexes, Vps4p, Bro1/Vps31, Vta1, Vps60p/MOS10, and Did2/Fti1 (). A recent model has proposed that monoubiquitinated receptors and cargo proteins are first recognized by Vps27p and Hse1p, which results in the sequential recruitment of three distinct multiprotein complexes, i.e., ESCRT-I, -II, and -III, to endosomal membranes from the cytosol (; ,; ). Although the details are unclear, these complexes are required for the sorting of monoubiquitinated cargo for inclusion in MVBs, as well as the formation of the MVBs themselves (for reviews see ; ). The final step in the membrane invagination that forms MVBs may be specifically associated with the ESCRT-III complex and its ability to interact with the AAA-ATPase Vps4p (). Doa4p is a deubiquitinating enzyme that recycles ubiquitin by releasing monoubiquitin moieties before the incorporation of proteins into the internal membranes of the MVB (). Several in vitro studies have investigated the protein– protein interactions of the class E Vps proteins in yeast and mammalian cells using yeast two-hybrid and GST pull-down assays (; ; ). Through a coherent protein network, the ESCRT-I, -II, and -III complexes and associated proteins form a large MVB-sorting complex on endosomal membranes. The ESCRT-III complex is likely composed of two functionally distinct subcomplexes—a membrane-associated subcomplex (Vps20p–Snf7p) and a cytosolic subcomplex (Vps2p–Vps24p; ). Recently, Did2/Fti1 and Vta1p were found to interact with Vps60p/MOS10, Vps4p, and the ESCRT-III complex, suggesting that together with Vps60p/MOS10 they play a role in regulating the activity of Vps4p and ESCRT-III (; ). Several mammalian homologues of yeast Vps proteins have been identified (). Many of these share common yeast homologues, implying a greater degree of complexity in the mammalian MVB-sorting pathway (). This may not be surprising, given the greater functional diversity and specialization that exists in animal cells. Hrs is homologous to the yeast Vps27p and recognizes ubiquitinated receptors through a conserved ubiquitin-interacting motif (UIM), which is essential for MVB sorting to degradative pathways (). Tsg101 is homologous to yeast Vps23p, which is an ESCRT I component, and down-regulates growth factor signaling through its interaction with Hrs (; ). The AAA-ATPase SKD1 (Vps4B) is homologous to the yeast Vps4p and regulates the association/dissociation of the MVB-sorting complex in a manner that is dependent on its ATPase activity (; ). Human Vps34 is a phosphoinositide 3 kinase that is required for internal vesicle formation within MVBs (), and human Vps28 directly interacts with Tsg101 and is recruited to human Vps4 (E235Q)-positive endosomal membranes (). 10 human charged MVB protein (CHMP) family proteins, which are structurally related cytosolic proteins containing coiled coil domains and are homologous to six yeast ESCRT-III components, have been identified. They have been suggested to play important roles in the final step of MVB-sorting pathways, namely the invagination of internal vesicles in MVBs, which is regulated by Vps4p (). CHMP proteins also function in HIV budding, a process that is topologically similar to MVB sorting (; ); however, the mechanism, function, and significance of CHMP proteins in mammalian MVB-sorting pathways remain to be defined. We originally isolated CHMP5 as a protein that copurified with cytosolic NF-κB–IκB complexes from rabbit lung extracts. Microsequencing of this 32-kD copurifying protein revealed that it was identical to CHMP5, a CHMP family member and the mammalian homologue of yeast VPS60/MOS10 (). Phylogenetic analyses indicate that CHMP5 is a unique CHMP protein that is quite divergent from other CHMP family proteins (unpublished data). We began our analysis of CHMP5 by cloning and sequencing the murine and human cDNAs, which were obtained by screening a mouse liver cDNA library and by PCR from HeLa cell cDNA, respectively. CHMP5 is highly conserved through evolution and its homologues can be found in , , , and yeast (). CHMP5 is also ubiquitously expressed in embryonic and adult mouse tissues (not depicted and , respectively). mice. genomic DNA (29.8 kb) was isolated from a 129/SvJ murine genomic DNA library using cDNA as a probe. The gene contains eight small exons and is separated by only 414 bp from the gene (). These two genes are encoded by different DNA strands, and their 5′ ends are positioned head- to-head (). To generate a null mutation of , exons 3–7 of the gene were replaced by a loxP-flanked neomycin-resistant gene cassette. mice with mice (; ). The resulting heterozygous mice were phenotypically normal, but the homozygous mice died at approximately embryonic day 10 (E10). Most wild-type embryos and mutant littermates did not exhibit any gross morphological difference until E7.5, after which the mutant embryos displayed severe developmental abnormalities in the ventral region (). At E8.75, mutant embryos were severely disorganized, with abnormal neural tube formation, allantois-chorion fusion, and somite segmentation, although embryonic axes and structures are normal in mutant embryos (). embryonic structure, we performed histological analysis of E8.5 wild-type embryos and mutant littermates (). Consistent with , severe developmental abnormalities of allantois, head fold, heart, and somite, and an apparent defect of ventral folding morphogenesis, were detected in the mutant embryos. To characterize the mutant phenotype, we performed whole-mount in situ hybridization with as a marker of heart formation (, top) and TUNEL assay to assess cell death in E8.5 mutant embryos (, bottom). Remarkably, mutant embryos exhibited the formation of two independent hearts (cardia bifida), accompanied by massive cell death in the ventral region. These phenotypes are similar to those of murine and embryos lacking Hrs (; ; ). These findings suggest that CHMP5, like Hrs, may play a role in regulating the endocytosis or lysosomal transport of receptors involved in signal transduction and, therefore, is indispensable for early embryonic development. However, a role for a putative ESCRT-III complex in receptor trafficking in mammalian cells is yet to be demonstrated. To evaluate the involvement of CHMP5 in endocytosis, we isolated and cultured primary embryonic cells derived from E8.5 embryos. cells contained enlarged vacuole-like structures (), a phenotype similar to primary embryonic cells (). To characterize these structures, we performed immunofluorescence analysis with markers specific for different subcellular compartments. Immunostaining with transferrin receptor (early and recycling endosomes), EEA1 (early endosomes), TGN38 (trans-Golgi network), and Rab8 (recycling endosomes and trans-Golgi network) did not reveal any gross differences between wild-type and mutant cells ( and not depicted). cells were positive for CI-M6PR, LBPA, and LAMP1 (, b and c). Indeed, the degree of colocalization was far more pronounced in the mutant cells as compared with wild-type cells in which, as expected, distributions of CI-M6PR and LAMP1 and LBPA and LAMP1 overlapped only partially. Structures positive for CI-M6PR and lgp/LAMP1 or for LBPA and LAMP1 are generally defined as being late endosomes and as distinct from lysosomes which do not contain appreciable levels of CI-M6PR and LBPA (; ; ). Thus, CHMP5 deficiency was accompanied by an accumulation of structures that appear to be late endosomes. cells, we performed immunostaining for major histocompatibility class (MHC) II. The majority of newly synthesized MHC II molecules are diverted from the secretory pathway upon exit from the trans-Golgi network and are directly targeted to endosomes/lysosomes (). Mouse embryonic fibroblasts (MEFs) derived from E9.0 wild-type embryos were stimulated with IFN-γ to induce MHC II expression (; ). MEFs, stimulated with IFN-γ, were then stained with anti-MHC II (I-Aβ) antibody (). cells exhibit enlarged endosomes positive for MHC II molecules, whereas MHC II–positive endosomes are scattered throughout the peripheral cytoplasm in wild-type cells. Thus, CHMP5 deficiency led to the accumulation of MHC II–positive endosomes/lysosomes. We then used transmission electron microscopy to examine MVBs in wild-type and mutant embryos. Strikingly, MVBs in the endodermal cells from the mutant embryos were abnormally enlarged and heavily packed with internal vesicles and electron-dense content, as compared with the analogous structures in cells from wild-type embryos (). cells was ∼1.5-fold that in wild-type cells (), confirming quantitatively that the endosomes were enlarged in the mutant cells. Interestingly, this phenotype was most clearly manifested in endodermal cells, where endocytosis is most active during embryogenesis. Collectively, these observations suggest that CHMP5 deficiency caused a global disruption in protein sorting to, or the formation of, lysosomes (M6PR/LBPA/LAMP1) from late endosomes/MVBs (M6PR/LBPA/LAMP1), rather than a specific deficit in the sorting of select proteins into lysosomes. Surprisingly, although CHMP5 deficiency affected the ability of target proteins to traffic to lysosomes, it did not prevent MVB formation. cells, we derived embryonic stem (ES) cells from the inner cell mass of blastocysts obtained by breeding heterozygotes. ES cells are equivalent to E3.5 embryonic cells and the mutant phenotype was therefore less dramatic in ES cells than in E8.5 embryonic cells. Both wild-type and mutant ES cells were able to internalize FITC-conjugated dextran at similar rates (unpublished data), indicating that CHMP5 was not required for fluid phase endocytosis by itself. ES cells exhibited a greatly reduced capacity to degrade material after internalization. After a 1-h pulse of the fluid phase marker HRP, wild-type ES cells degraded nearly all of the internalized protein after 9 h of chase. ES cells, as determined by immunoblotting () and flow cytometry (). ES cells than in wild-type cells (). Thus, although HRP reached canonical late endosomes/MVBs, it was degraded less effectively in the absence of CHMP5. Defective degradation of lysosomal contents could reflect either a disruption in the transport of internalized proteins to lysosomes or a lower overall proteolytic capacity. ES cells (). Thus, it appeared more likely that the defective degradative phenotype reflected inefficient delivery of endocytosed proteins to compartments containing active lysosomal hydrolases. Such a defect might also affect the normal down-regulation of signaling receptors, suggesting that the embryonic lethality of deletion might reflect hyperactive signal transduction. Therefore, we examined the dynamics of TGFβ receptor II (TβRII), which has a well characterized role in early embryogenesis (). The early embryonic lethality of mutant embryos prevented us from obtaining appropriate mutant cells that could be used to study receptor turnover and signaling in vitro. We therefore used small interfering RNA (siRNA) technology to generate CHMP5 knockdown cells. Two target sequences of the gene (Sh1 and Sh2; murine CHMP5) were cloned into a pSUPER.retro vector along with an H1-RNA promoter, which directs synthesis of short hairpin RNA (shRNA) that can be converted into an siRNA capable of knocking down target gene expression (). Transient transfection of the shRNAs showed that Sh2 was capable of dramatically reducing the level of transiently expressed mouse CHMP5, whereas Sh1 was inactive in this assay. To determine if CHMP5 deficiency affects the postendocytic fate of TβRII, we transfected NIH3T3 cells with HA-tagged TβRII (HA-TβRII) in the absence or presence of Sh2 (). Receptor endocytosis was triggered by the addition of TGFβ, and receptor fate was monitored using an anti-HA monoclonal antibody that was added simultaneously to the medium, as previously described (; ). Receptor distribution was analyzed relative to LAMP1 staining. In control cells, HA-TβRII was found largely in LAMP1-negative endosomes scattered throughout the peripheral cytoplasm, reflecting internalization of the receptor and its degradation in lysosomes. cells. At least at the level of immunofluorescence, the receptor appeared to reside within the endosomal lumen as well as on the limiting membrane. Consistent with this, a recent report has also shown that GFP-tagged CHMP5 protein, which is a putative CHMP5 dominant-negative mutant, leads to accumulation of ligand-bound EGF receptor (EGFR) in enlarged perinuclear vesicles and a delay in EGFR degradation (). To determine localization of TβRII in these endosomes, we performed a protease protection assay on subcellular fractions of cells expressing COOH-terminal HA-TβRII along with control vector or human CHMP5 shRNA (Sh3; ). Percoll gradients allowed a separation of a light fraction containing the plasma membrane, early/recycling endosomes, and a population of Rab7 or LAMP1 endosomes from denser fractions containing Rab7 late endosomes, as well as LAMP1 lysosomes (). In control cells, the majority of the HA-TβRII was, as expected, detected in the light fraction (i.e., at the plasma membrane) in the absence of TGFβ stimulation. After TGFβ addition for 5 h, the total amount of receptor was greatly decreased because of degradation, but some of the remaining HA-TβRII was detected in the dense fractions, consistent with ligand-induced internalization and delivery to lysosomes. The majority of HA-TβRII, particularly in the light-density plasma membrane– containing fractions, was sensitive to proteinase K treatment. A different picture emerged in cells lacking CHMP5. First, consistent with our analysis of steady-state () and cell surface expression () of TβRII receptors, HA-TβRII levels were significantly increased in CHMP5 knockdown cells. Importantly, after TGFβ treatment the HA-TβRII that appeared in the denser fractions was now protected from degradation by proteinase K. Thus, at least a portion of the HA-TβRII that accumulates after ligand-induced receptor internalization appears to be localized within the lumen of Percoll-dense compartments, likely MVBs. Together with our morphological experiments, these data suggest that CHMP5 deficiency leads to the accumulation of internalized TβRII, at least, in part, on the internal membranes of late endosomal MVBs. The increased staining of the HA-TβRII suggested that CHMP5 knockdown interfered with the normal down-regulation of the receptor after ligand binding. To test this hypothesis, we transfected NIH3T3 cells with HA-TβRII, together with either control vector or Sh2, and assessed receptor turnover using pulse-chase analysis (). Consistent with a previous study showing that chloroquine significantly increases the stabilization of endogenous TGFβ receptors (), treatment of cells with chloroquine blocked degradation of HA-TβRII receptors in cells expressing either control vector or mouse Sh2. In cells expressing the control vector, the half-life of HA-TβRII receptors was ∼1.5 h, whereas in cells where CHMP5 was knocked down, the half-life was extended to ∼3.5 h. To further characterize the effect of CHMP5 expression on TGFβ receptor turnover, we determined receptor levels in human embryonic kidney 293 (HEK293) cells that had been transfected with vectors encoding either wild-type CHMP5 or Sh3. The resulting cells slightly overexpressed or dramatically diminished the expression of CHMP5 protein, respectively (). Consistent with the decreased rate of degradation in NIH3T3 cells, knockdown of CHMP5 greatly increased the steady-state levels of both TβRI and TβRII (, bottom). In contrast, overexpression of CHMP5 slightly decreased the amount of receptor protein, relative to nontransfected controls. Similarly, using flow cytometry we found that CHMP5 knockdown significantly increased steady-state levels of HA-TβRII receptors on the cell surface, relative to the control vector (). Together, these findings suggest that the loss of CHMP5 slows the degradation of internalized TGFβ receptors by preventing ligand-induced down-regulation. Like internalized HRP, the receptors accumulated intracellularly in late endosomes/MVBs, which are structures normally associated with receptor degradation. To investigate whether the defect in receptor degradation might reflect a failure of late endosomes/MVBs to fuse with preexisting lysosomes, we performed an endosome–lysosome fusion assay using shRNA to transiently knockdown CHMP5 expression. To label preexisting lysosomes, NIH3T3 cells pretreated with protease inhibitors were incubated with biotin-conjugated HRP (biotin-HRP) for 3 h, and then transfected with Sh2. 48 h later, the cells were incubated with streptavidin for 10 min, and the formation of streptavidin–biotin-HRP complexes was determined after various chase intervals. Vector and control shRNA (Sh1)–expressing cells showed increasing complex formation over time, whereas complex formation was dramatically inhibited in Sh2-expressing cells (). Therefore, streptavidin cannot easily reach previously formed lysosomes in the absence of CHMP5, implying that CHMP5 is required for efficient fusion of late endosomes/MVBs with preexisting lysosomes. If, in fact, an alteration in receptor trafficking or degradation is responsible for the developmental phenotype associated with CHMP5 deletion, one would expect that knockout embryos would exhibit defects in TGFβ receptor signal transduction. There is evidence that TGFβ receptor signaling is regulated by endocytosis, with internalization being required for the receptor to reach an essential adaptor component, Smad anchor for receptor activation (SARA), in EEA1-positive endosomes, and to be subject to normal degradation and down-regulation after signal transmission (; ; ). Moreover, signaling by FGF and TGFβ family members plays an essential role in embryogenesis at the stage when CHMP5 deficiency causes lethality (; ). embryos by examining the phosphorylation status of relevant signaling proteins. We made whole cell lysates from E8.5 wild-type and embryos and immunoblotted these extracts with phosphospecific antibodies against different proteins that are phosphorylated upon signaling. embryos. cells, unlike wild-type cells in which phosphorylated Smad2 is rarely observed (). embryos and it is likely that such dysregulation leads to embryonic lethality. Consistent with this hypothesis, TGFβ-mediated gene induction is significantly inhibited by CHMP5 overexpression and is increased upon CHMP5 knockdown (). cells. cells. cells, although TβRII and SARA are relatively enriched in the TGN and cytosol of wild-type cells (). This is consistent with flow cytometric analyses, which indicated CHMP5 knockdown increases cell surface expression of HA-TβRII, relative to control (). Therefore, these results suggest that CHMP5 deficiency increases the steady-state levels of TGFβ receptor expression at the cell surface by decreasing degradation of internalized receptors. cells. We further demonstrated a generalized role for CHMP5 in down-regulation of receptor signaling pathways by monitoring NF-κB activity. embryos, and the inability to generate knockout mouse embryonic fibroblasts prevented us from assaying NF-κB activation in the knockout cells themselves. Therefore, we tested the role of CHMP5 in NF-κB activation in cell lines using a reporter assay system. As shown in , CHMP5 overexpression inhibits NF-κB activation induced by the inflammatory cytokines TNFα and IL-1β, as well as CD4-TLR4, a dominant active chimera of the toll-like receptor for the bacterial component lipopolysaccharide (). This result suggests that, similar to growth factor and TGFβ signaling, signaling to NF-κB is also regulated by endocytosis and, hence, that CHMP5 plays a role in the down-regulation of multiple signaling pathways. sup #text NIH3T3 (murine fibroblast cells) and NMuMG (mammary gland cells; CRL-1636) cell lines were purchased from the American Type Culture Collection. The antibodies used were anti–phospho-Erk1/2 (Cell Signaling Technology), anti–phospho-Smad2 (Cell Signaling Technology), anti-Smad4 (Santa Cruz Biotechnology, Inc; H552), anti-Smad2 (Cell Signaling Technology), anti-M6PR (Affinity BioReagents, Inc.), anti-EEA1 (BD Biosciences), anti-TGN38 (BD Biosciences), anti–mouse transferrin receptor (BD Biosciences; C2), anti–mouse LAMP1 (BD Biosciences; ID4B), and anti-GAPDH4 (Fitzgerald Industries International). An anti-CHMP5 polyclonal antibody was generated by immunizing rabbits with a 17-aa synthetic peptide corresponding to the COOH-terminal sequence of mouse CHMP5 and affinity purified with immobilized antigen. Antibodies specific to LBPA, Rab8, and MHC II (I-Aβ) were gifts from T. Kobayashi (Institute of Physical and Chemical Research, Wako, Saitama, Japan) and I. Mellman (Yale University School of Medicine, New Haven, CT). Plasmids encoding HA-TBI and HA-TBII were a gift from J.L. Wrana (University of Toronto, Toronto, Canada). CHMP5 was copurified with IκBα from rabbit lung tissue extract. After separation on SDS-PAGE and silver staining, a 32-kD band was excised from the gel and trypsinized, and five peptides were microsequenced. One mouse EST (available from GenBank/EMBL/DDBJ under accession no. ) matched these peptides and was used as a probe to screen a mouse liver cDNA library. One clone containing the open reading frame was obtained and an additional 5′ sequence was cloned by 5′ rapid amplification of cDNA ends. Four genomic clones were isolated from a 129/SvJ mouse genomic DNA library (Stratagene). The targeting vector consists of a 5.3-kb 5′ homology region and a 4.7-kb 3′ homology region. A loxP-PGKneo-loxP cassette was inserted between the two regions and a PGK-TK cassette was placed upstream of the 5′ homology region, resulting in a vector designed to delete exons 3–7 of . Linearized targeting vector was electroporated into TC1 ES cells. Clones resistant to G418 and gancyclovir were selected, and homologous recombination was confirmed by Southern blotting. Two targeted clones were injected into C57BL/6 blastocysts and both produced germline chimeras. Chimeras were mated with C57BL/6 females, and heterozygous male offspring were bred with female mice to delete the floxed neo cassette. Their offspring were screened for the targeted allele without the neo cassette and in the absence of the transgene. Positive mice were interbred and maintained on a mixed 129 × C57BL/6 background. ES cell lines, mice were mated, blastocysts were collected, and ES cell lines were isolated from the inner cell mass and cultured as previously described (). The genotypes of ES cell lines were determined by Southern blotting and immunoblotting with anti-CHMP5 antibody. Whole-mount TUNEL assay and in situ hybridization using digoxigenin UTP-labeled riboprobes were performed as previously described (). Embryos were placed on 60-mm Petri dish containing 1% agarose that was filled with PBS and oriented as indicated. The embryos were imaged on a dissecting microscope (model Stemi 2000-C; Carl Zeiss MicroImaging, Inc.) using 16×/16, NA 2.5, objectives (AxioCam; Carl Zeiss MicroImaging, Inc.). Image data was acquired and stored as TIFF files using AxioCam software. E8.5 embryos from breeding of mice were dissected free of maternal tissues and had their Reichert's membrane removed, after which they were washed with PBS and incubated with 0.1% collagenase (Sigma-Aldrich) and Trypsin-EDTA (Invitrogen) for 30 min at 37°C. The cell suspension was plated in 24-well plates precoated with 0.2% gelatin (Sigma-Aldrich) and cultured in DME supplemented with 15% FBS (Sigma-Aldrich). cells were identified by immunostaining with CHMP5 antibody or PCR from extraembryonic tissues. E8.0–8.25 embryos were dissected and washed with PBS, and then fixed with 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.2, for 1 h at room temperature. Further procedures were performed by standard protocols. For immunofluorescence analysis, cells were grown on 0.2% gelatin-coated coverslips, washed with PBS, fixed with 4% paraformaldehyde for 30 min at room temperature, and permeabilized with permeabilization buffer (0.05% saponin, 1% FBS, 10 mM Hepes, and 10 mM glycine in PBS, pH 7.5) for 30 min at room temperature. Cells were incubated with the indicated primary antibodies for 30 min at room temperature and then incubated with either goat anti–rabbit or anti–mouse antibodies conjugated with FITC or Texas red under identical conditions. Subsequently, cells were washed three times with PBS, mounted in GEL/MOUNT (Biomeda), and examined under either a Plan Apochromat 63× oil objective or a Plan Neofluar 40× oil objective on a fluorescence microscope (Axioplan 2; all Carl Zeiss MicroImaging, Inc.) equipped with a charge-coupled device camera (Orca ER; Hamamatsu). Image data was acquired and stored as TIFF files using OpenLab software (Improvision, Inc.). shRNAs with murine CHMP5 target sequence Sh2 (5′-CCTGGCCCAACAGTCCTTT-3′) and murine CHMP5 control sequence Sh1 (5′-AAGCGAAACCCAAGGCTCC-3′), or human CHMP5 target sequence Sh3 (5′-AAGGACACCAAGACCACGGTT-3′) were produced by chemically synthesized DNA oligonucleotides and cloned into pSUPER.retro vector following the manufacturer's instruction (OligoEngine). To test ShRNAs, Flag-tagged cDNA and CHMP5 ShRNAs were cotransfected into COS1 cells. 48 h after transfection, cells were lysed and immunoblotted with antibodies specific for Flag and GAPDH4. NIH3T3 cells were cultured in medium containing protease inhibitors to inhibit protein degradation in lysosomes throughout the experiment. Cells were incubated with biotin-conjugated HRP for 3 h to label lysosomes before being transfected with CHMP5 shRNA. 48 h after transfection, cells were incubated with streptavidin for 10 min, washed, chased for the indicated time points, and lysed with biotin-containing lysis buffer. Lysates were applied to antistreptavidin-coated plates, incubated for 1 h, and washed; HRP enzymatic activity was measured by colorimetric assay and is expressed as arbitrary units. HEK293 cells were transfected with COOH-terminal HA-TβRII in the presence or absence of human CHMP5 shRNA (Sh3). 48 h after transfection, cells were treated with 10 ng/ml TGFβ for 5 h or left untreated and fractionated using Percoll gradients as previously described (). In brief, cells were washed and resuspended in homogenization buffer (10 mM triethanolamine, 10 mM acetic acid, 1 mM EDTA, and 0.25 M sucrose, pH 7.4) and disrupted with 20 strokes in a dounce homogenizer. Microscopic analysis assured that the cell breakage was nearly complete. The homogenate was centrifuged at 1,000 for 5 min at 4°C to remove nuclei and unbroken cells. The postnuclear supernatant was mixed with Percoll (Sigma-Aldrich) in homogenization buffer to give a final concentration of 27%, and the mixture was underlaid with a 27.6% nycodenz solution (Sigma-Aldrich). The gradients were centrifuged in a SW41Ti rotor at 17,500 rpm for 1 h at 4°C, and 14 fractions were collected from the top of gradient. Protease protection assays have traditionally been used to determine the sidedness of proteins relative to a sealed membrane compartments (; ). These experiments typically involve the complete digestion of exposed domains of proteins on the outside of a sealed compartment and the protection of those domains or proteins that reside on the inside of the compartment (; ). To apply the protease protection assay to our system, we modified the original protocol of the protease protection assay. For protease protection assay, 25-μl aliquots of each fraction were treated with 2–5 μg proteinase K for 30 min at room temperature or left untreated. The reaction was stopped by adding SDS loading buffer, and proteinase K was inactivated by boiling at 90°C. The samples were subjected to 10% SDS-PAGE, followed by immunoblotting with the indicated antibodies.

Upon recognizing pathogen-associated molecular patterns, the family of 11 Toll-like receptors (TLRs) provides the initial activation signal to the immune system, resulting in costimulatory molecule expression and cytokine secretion (; ). These cytokines can modulate the adaptive immune response to eliminate particular classes of pathogens by polarizing CD4 T cells to either a Th1 or -2 phenotype (). However, an unbalanced or sustained Th1 or -2 response can lead to diseases such as rheumatoid arthritis and asthma (). The pathogen-associated molecular pattern of unmethylated bacterial DNA is specifically recognized by TLR9, and its immunomodulatory effects can be mimicked by oligodeoxynucleotides (ODNs) containing unmethylated deoxycytidyl-deoxyguanosine (CpG) motifs (CpG-ODN; ). TLR9, expressed by B cells, macrophages, and dendritic cells (DCs), recognizes CpG in the acidic environment of the endosome (). Variant sequences flanking the stimulatory CpG core motif have been described for optimal TLR9 activation expressed by specific cell types. Depending on the cytokines secreted, those ODNs have been categorized as A, B, or C class ODNs; however, it remains unclear why specific CpG sequences trigger different biological effects (). Studies using fluorescently labeled ODNs indicate that both stimulatory and nonstimulatory ODNs are internalized nonspecifically, but only stimulatory ODNs activate TLR9 in endosomes, where both ligand and receptor colocalize (; ; ; ). Uptake of ODN is dependent on dose, time, and temperature but independent of the CpG motif (). Upon recognition of CpG-rich sequences in the endosome, TLR9 initiates a conserved TLR family signaling cascade that begins with the recruitment of the adaptor protein MyD88 via the Toll/interleukin (IL) 1 receptor domain (). MyD88 then recruits IL-1 receptor–associated kinase (IRAK) 1 and 4 (). When phosphorylated by IRAK-4, IRAK-1 interacts with TRAF6 (TNF receptor–associated factor 6; ; ) and disengages from the receptor. A complex consisting of TRAF6, TAK1 (TGFβ-activated kinase 1), and TAB (TAK1 binding protein) 1 and 2 goes on to activate the IκB kinase complex, resulting in nuclear factor (NF) κB translocation to the nucleus (; ; ). CpG-induced NF-κB activation initiates the up-regulation of costimulatory molecules and the secretion of proinflammatory cytokines, such as TNFα and IL-6. The adaptor protein MyD88 plays a crucial role in transducing signals from TLR family members. Upon binding to TLR4, bacterial lipopolysaccharide not only activates NF-κB via the MyD88–IRAK serine/threonine kinase pathway but also induces tyrosine phosphorylation in macrophages (). Two Src family kinases (SFKs), Hck and Lyn, are responsible for initiating this pathway (; ). Hck has been shown to regulate differentiation and several actin-dependent processes such as F-actin–based membrane protrusions (), monocyte chemotaxis (; ), phagocytosis (; ), and cellular adhesion (; ). In the well-characterized system of Fcγ receptor signaling, Hck and Lyn are the initiating tyrosine kinases that phosphorylate immunoreceptor tyrosine-based activation motifs (ITAMs) in the signaling chains of the receptor complex, which then serve as docking sites for a second tier of tyrosine kinases such as Syk (). Both Src and Syk family kinases phosphorylate several downstream signaling proteins and thus initiate multiple signaling events. It has recently been reported that CpG-ODN stimulation induces tyrosine phosphorylation of the GTP exchange factor Vav1 (). The TLR9 inhibitor chloroquine and analogues such as quinacrine inhibit immune stimulation by bacterial DNA and CpG-ODN (; ). The inhibition of TLR9 occurs at endosomes where chloroquine was shown to block CpG and TLR9 interaction (). Chloroquine () has also proven effective in the treatment of inflammatory diseases such as systemic lupus erythematosus and rheumatoid arthritis (; ; ; ). We have investigated the role of SFKs upon CpG-induced TLR9 stimulation. Two SFKs, Hck and Lyn, are phosphorylated upon monocyte stimulation with CpG-ODN but not when cells are treated with nonstimulatory deoxyguanosine-deoxycytidyl (GpC) ODN. The SFK-initiated tyrosine phosphorylation cascade activates signaling proteins implicated in reorganization of the actin cytoskeleton, resulting in cell spreading, adhesion, and motility. Surprisingly, the TLR9 inhibitor chloroquine failed to block SFK activation and any of the signaling pathways downstream of tyrosine kinase activation. Macrophages from TLR9 and MyD88 mice exhibited normal SFK activation, indicating that the SFK pathway is TLR9 independent. Unexpectedly, this pathway is triggered upstream of the traditional TLR9 pathway, as indicated by SFK inhibitors, which blocked CpG-induced phosphorylation of the NF-κB regulator IκB, IL-6 secretion, and up-regulation of CD40 and CD69. Both pathways intersect at the level of TLR9, which is tyrosine phosphorylated upon CpG stimulation and interacts with the tyrosine kinase Syk. Furthermore, CpG-coated beads induce strong actin polymerization at the plasma membrane–bead contact area of macrophages in a chloroquine-insensitive manner. Together, these findings indicate that the SFK-driven tyrosine phosphorylation pathway is an early CpG-induced event that is initiated at the cell surface and is upstream of and required for TLR9/MyD88 activation in endosomes. The role of SFKs after CpG-ODN treatment is unclear. To study the role of tyrosine kinases in TLR9 signaling, we stimulated the human monocyte cell line THP-1 with a control GpC and an activating CpG-ODN. Cell lysates were then analyzed by immunoblotting with an anti-phosphotyrosine (pTyr) antibody. CpG treatment induced rapid tyrosine phosphorylation of multiple proteins, with maximum phosphorylation achieved 3 min after stimulation (, lanes 4–6). The control non–CpG-ODN did not trigger tyrosine kinase activation (, lanes 1–3). To specifically identify phosphoproteins, candidate proteins were immunoprecipitated and then immunoblotted with anti-pTyr antibody. Major phosphoproteins of 47, 52, 70, 95, 120, and 125 kD were identified in this manner (). The three most prominent phosphoproteins identified were the cytoplasmic tyrosine kinase Syk (70 kD), the hematopoietic-specific exchange factor Vav1 (95 kD), and the adaptor protein Cbl (120 kD). Two SFK members expressed in monocytic cells, Lyn (49 kD) and Hck (52 kD), were also phosphorylated upon CpG treatment. In addition, the hematopoietic-specific focal adhesion kinase Pyk2 (125 kD), talin (200 kD), paxillin (68 kD), and vinculin (117 kD) were also phosphorylated. However, other prominent B cell and macrophage tyrosine substrates, such as B cell linker protein and Bruton's tyrosine kinase, were not phosphorylated upon CpG stimulation (unpublished data). Because our results indicated that two SFKs, Hck and Lyn, were phosphorylated upon CpG treatment, we wanted to test whether an SFK inhibitor could block tyrosine phosphorylation. Pretreatment of THP-1 cells with PP2, a potent SFK inhibitor, completely blocked CpG-induced phosphorylation (, lanes 10–12). Pretreatment with PP3, an inactive version of PP2, had no effect on CpG-induced tyrosine phosphorylation (, lanes 13–15). Pretreatment with two well-described inhibitors of TLR9-induced cytokine production, chloroquine and quinacrine, also had no effect on CpG-induced tyrosine phosphorylation (, lanes 4–9). SFK activation results in cytoskeleton changes and the transcription of effector genes (). Six of the proteins that we found to be phosphorylated upon CpG stimulation, Pyk2, Vav, Cbl, talin, paxillin, and vinculin, along with the activation of PI3K, have been described as mediators of actin cytoskeleton reorganization (; ; ; ; ; ). These events promote cell spreading, adhesion, and motility, resulting in monocyte recruitment to areas of infection. When cells were stained for F-actin fibers, we found that CpG treatment induced strong actin polymerization in both mouse macrophages and human monocytes (). These rearrangements translated to rapid cell spreading, conspicuous morphological changes, and increased cell size. In the case of THP-1 cells, CpG treatment induced the formation of actin-rich lamellipodia structures. Similar to CpG-induced tyrosine kinase phosphorylation (), the TLR9 inhibitor chloroquine was unable to block CpG- induced actin reorganization (). SFK-initiated actin reorganization can also facilitate cell adhesion to promote the recruitment of cells to tissues infected by pathogens. Because CpG induced cytoskeleton reorganization, we next tested the ability of CpG to induce monocyte adhesion. When human monocytes were stimulated with CpG, they rapidly adhered in culture, with maximal adhesion detected at 30 min (). A negative control GpC-ODN failed to induce monocyte adhesion. Interestingly, a time-course analysis of CpG-induced adhesion indicated that this event was transient, with cells detaching after 2 h (). Overnight culture of CpG-stimulated monocytes revealed that cells eventually reattach, an event that we hypothesize is driven by secondary CpG-induced effects, such as chemokine secretion, because it can be blocked by chloroquine (unpublished data). Next, we further characterized the ability of different CpG-ODNs to induce adhesion. We found that the human CpG-ODN that induced robust tyrosine phosphorylation also induced adhesion in a concentration-dependent manner (). Although 2 μg/ml CpG produced the highest level of adhesion, 0.5 μg/ml was sufficient to induce significant adhesion (). A GpC control ODN or a mouse-selective CpG-ODN was unable to induce significant cell adhesion at any concentration tested. Interestingly, when we compared the performance of an A class ODN with that of a B class ODN (ODN 2006), we found that the latter yielded a less pronounced adhesion response (). This also correlated with a poor induction of the tyrosine phosphorylation cascade (unpublished data). Although only immunostimulatory CpG promotes cytoskeleton rearrangement, it has been reported that both immunostimulatory and nonimmunostimulatory ODNs are internalized at the same rate in a time-, dose-, and temperature-dependent manner (; ; ; unpublished data), indicating that CpG-induced SFK activation does not enhance ODN uptake. This is consistent with our observations that PP2 treatment had no effect on CpG internalization (unpublished data). Our previous experiments indicated that inhibition of SFKs blocked the phosphorylation of proteins that promote cellular adhesion, such as Pyk2 and Vav1 (). Based on these data, we hypothesized that the SFK inhibitor PP2 would also block monocyte adhesion. When we tested this hypothesis, we found that PP2 potently blocked CpG-induced adhesion, even at concentrations below the micromolar range (). As expected, the control inhibitor PP3 had no effect on monocyte adhesion. Herbimycin A, another SFK inhibitor, also blocked monocyte adhesion at submicromolar concentrations (). Actin-driven morphological changes are dynamic processes mediated by the polymerization and depolymerization of actin fibers. Therefore, we examined the effects of the actin microfilament–disrupting drugs cytochalasin D, latrunculin A, and jasplakinolide. All three inhibitors potently blocked monocyte adhesion at the nanomolar range (Fig. S1, available at ). Together, these data indicated that CpG-triggered SFK activation initiates a tyrosine-kinase signaling cascade in monocytes that results in actin-driven cytoskeleton changes that promote cell adhesion. Chloroquine blocks CpG interaction with TLR9 at the endosome (), and our previous findings indicated that although CpG-induced tyrosine kinase activation was blocked by the SFK inhibitor PP2, it was insensitive to chloroquine (). We therefore tested the ability of chloroquine and quinacrine to modulate CpG-induced adhesion. Even at high concentrations that completely inhibit TLR9-induced cytokine production (10 μM for chloroquine and 1 μM for quinacrine), both inhibitors failed to block CpG-induced adhesion (). To corroborate our findings using primary cells, we repeated these studies using immature human DCs. Similar to a human monocytic cell line, CpG induced adhesion of human monocyte–derived DCs, whereas GpC did not (). In immature DCs, we found that SFK inhibitors are able to effectively block CpG-induced adhesion (). Although chloroquine and quinacrine blocked IFN-α secretion from peripheral blood mononuclear cells (PBMCs) at concentrations of <10 and 1 μM, respectively (), they were unable to block CpG-induced adhesion at similar concentrations (). Another prominent cytoskeleton-dependent process is cell motility. All of the identified CpG-induced phosphoproteins have been implicated in driving or regulating cell motility (). As previously described (), we found that monocytes migrate toward a CpG gradient (Fig. S2, available at ). As in the case of adhesion, chloroquine and quinacrine failed to block monocyte migration in response to CpG (Fig. S2). The secretion of CpG-induced inflammatory cytokines is dependent on NF-κB translocation to the nucleus (). The negative regulator of NF-κB activation, IκB-α, is phosphorylated and degraded after 5 min and remains strongly phosphorylated 45 min after CpG treatment (). As expected, the TLR9 inhibitor chloroquine blocked IκB-α phosphorylation and degradation in monocytes (). Our previous results () indicated that CpG-induced SFK activation precedes IκB-α phosphorylation and degradation, which are undetectable at 2.5 min as compared with maximal CpG-induced tyrosine phosphorylation at 3 min after CpG treatment. Given the rapid kinetics of SFK activation and insensitivity to chloroquine treatment, we hypothesized that CpG-induced tyrosine phosphorylation is upstream of IκB-α phosphorylation. When we tested this hypothesis in macrophages, we found that the SFK inhibitor PP2 effectively blocked IκB-α phosphorylation upon CpG stimulation (). Like chloroquine, the SFK inhibitor PP2 also inhibited CpG-induced IFN-α secretion in immature human plasmacytoid DCs (pDCs; ). Furthermore, PP2 effectively blocked CpG-induced IL-6 secretion () and the up-regulation of B cell activation markers such as CD40 and CD69 (). To complement our pharmacological inhibitor data, we established a small interfering RNA (siRNA) system to specifically knock down the expression of Lyn and Hck, two leukocyte SFKs that we identified as being activated by CpG. Immunoblot analysis of siRNA-transfected cell lysates revealed that an 80% knockdown of Lyn and a >90% knockdown of Hck was achieved (). When these cells were used in a CpG-induced adhesion assay, single knockdown of Lyn or Hck resulted in a significant decrease in cellular adhesion (). A combined knockdown of Lyn and Hck further decreased CpG-induced adhesion by ∼60%. The residual adhesion is most probably mediated by other SFKs, such as Blk, Fyn, and Src. Moreover, analysis of cytokine production revealed a similar trend, where combined knockdown of Lyn and Hck reduced CpG-induced TNFα production by ∼60% (). Together, both the pharmacological inhibitor and genetic data indicated that early activation of SFKs plays a key role in CpG-induced activation of monocytes. The SFK pathway remains intact even when the TLR9– MyD88 signaling pathway is blocked using chloroquine. Next, we used an alternative method to pharmacological inhibitors to confirm our results. Because the adaptor protein MyD88 is required for signal transduction downstream of TLR9, we decided to test whether the CpG-induced tyrosine phosphorylation cascade is intact in the absence of MyD88. The same intensity and kinetics of phosphorylation was observed in both control and MyD88 macrophages stimulated with CpG-ODN (). Consistent with our previous results, chloroquine treatment did not block CpG-induced tyrosine phosphorylation in either wild-type or MyD88 macrophages (). Although the SFK pathway was intact in MyD88 macrophages, CpG treatment was unable to induce translocation of the p65 subunit of NF-κB to the nucleus (). Consequently, NF-κB–driven responses such as the up-regulation of CD40 and CD69 and the secretion of IL-6 were impaired in cells from MyD88 mice (Fig. S3, available at ). Because our data indicated that the SFK pathway was activated at the cell surface, we decided to directly test this hypothesis by stimulating cells with CpG-ODN that are unable to be internalized. 10-μm red fluorescent polystyrene beads that are large enough to avoid internalization by macrophages were coated with CpG or control GpC-ODN. Upon incubating these beads with mouse macrophages for 20 min, we observed strong actin polymerization localized specifically at the plasma membrane of the cell–bead contact area (). Beads coated with CpG, but not with the control GpC-ODN, specifically induced this actin polymerization. Consistent with our previous findings, actin reorganization was chloroquine insensitive (). It has recently been reported that CpG-induced Akt phosphorylation occurs via a TLR9-independent pathway (). It is possible that the chloroquine-insensitive SFK pathway that we describe originates at the plasma membrane in a TLR9-independent manner, similar to that described for Akt phosphorylation. Once CpG is internalized into endosomes, it would interact with TLR9 and initiate the chloroquine-sensitive TLR9–MyD88 pathway. To test this hypothesis, we stimulated TLR9 macrophages with CpG and obtained a pattern of tyrosine phosphorylated proteins similar to that of wild-type macrophages, indicating that CpG-induced SFK activation is intact in cells lacking TLR9 (). Furthermore, the downstream actin cytoskeleton rearrangements were also unperturbed in cells from both TLR9 and MyD88 mice stimulated with CpG-coated beads (). These data indicate that this pathway is both TLR9 independent and initiated at the cell surface before CpG endocytosis. Although CpG recognition can happen before CpG internalization, as we show here, it has been demonstrated that CpG internalization is required for cell activation and cytokine production (; ; ). In accordance with these findings, plate-bound CpG was unable to induce IL-6 secretion (). Although the TLR9–MyD88 pathway was impaired when CpG was immobilized, confocal analyses indicated that CpG-induced cytoskeletal rearrangements were intact (). Macrophages stimulated with CpG immobilized on plastic generated large, actin-rich lamellipodia. Similar actin rearrangements were observed in macrophages from MyD88 and TLR9 mice (). Next, we examined whether the CpG-induced SFK pathway is upstream of the previously described TLR9-independent phosphorylation of Akt (). Activation of SFKs promotes actin reorganization through the participation of PI3K, an upstream regulator of Akt (). Treatment of cells with the PI3K inhibitor wortmannin inhibited both CpG-induced cell adhesion () and phosphorylation of Akt (). Furthermore, use of the SFK inhibitor PP2 blocked CpG-induced Akt phosphorylation (), indicating that the SFK pathway is upstream of the previously reported TLR9-independent phosphorylation of Akt. Given that the CpG-induced SFK pathway originates at the plasma membrane and TLR9 is itself located in endosomes, we next wanted to investigate how these two pathways intersect. Recently, Syk, a CpG-activated kinase that we identified (), was reported to be a key component in the innate response to zymosan (). Syk was shown to couple to the β-glucan receptor Dectin-1 and work synergistically with TLR2 to mount an immune response to zymosan. Using a THP-1 cell line expressing a Flag-tagged TLR9, we found that upon CpG stimulation Syk coimmunoprecipitates with TLR9 (). This interaction provides a link between the SFK pathway and the MyD88 pathway. This interaction is unperturbed in the presence of the TLR9 inhibitor chloroquine, suggesting that it occurs before the generation of a signal through TLR9. Both the SFK inhibitor PP2 and a specific Syk inhibitor blocked the interaction of Syk with TLR9. To further examine the association of Syk with TLR9, we examined whether TLR9 itself was tyrosine phosphorylated. CpG treatment did indeed induce tyrosine phosphorylation of TLR9 (). Consistent with our previous experiments, this event was chloroquine insensitive but completely blocked by PP2 and partially blocked with a Syk inhibitor. CpG-containing ODNs activate TLR9 in endosomes and trigger the secretion of inflammatory cytokines by a mechanism that is dependent on the adaptor protein MyD88 (; ). We demonstrate that cellular activation by immunostimulatory ODNs is more complex than previously described and requires a TLR9-independent SFK-driven pathway. This SFK-dependent signaling cascade is not parallel but upstream of the chloroquine-sensitive TLR9–MyD88 pathway. We first provide evidence that stimulation with CpG, but not the control GpC, induces a complex pattern of tyrosine phosphorylation in monocytes. While characterizing several of these phosphorylated proteins, we identified two SFKs expressed in myeloid cells, Hck and Lyn. Treatment of monocytes with the SFK inhibitor PP2 inhibited the CpG-induced tyrosine phosphorylation cascade. Interestingly, treatment with chloroquine and quinacrine, two known inhibitors of TLR9 signaling, failed to block tyrosine phosphorylation. In addition to the kinases Hck and Lyn, we found that CpG also induced the phosphorylation of Pyk2, Cbl, and the previously described CpG-activated Vav (). These proteins are required for adhesion and migration responses that involve rearrangement of the actin cytoskeleton (; ). For example, the phosphorylation of Pyk2 by SFKs, along with the participation of other focal adhesion proteins, promotes actin cytoskeleton rearrangements that induce cell adhesion, spreading, and lamellipodia formation (). Cbl, which interacts with Pyk2, Vav, and SFKs, has been associated with increased motility of macrophages (). Vav, a multidomain signal integrator, transduces signals to cytoskeleton-dependent pathways, which include the PI3K pathway and the activation of extracellular signal–regulated kinase and NF-κB (). The phosphorylation of Pyk2, Cbl, and Vav upon CpG stimulation suggested that actin reorganization plays an important role upon CpG stimulation. Indeed, our experiments confirmed that CpG triggered the reorganization of the actin cytoskeleton, promoting cell adhesion and migration. We found that CpG treatment induced a transient, SFK-dependent cell adhesion in human monocytes but not when cells were stimulated with either control GpC-ODN or a mouse-selective CpG-ODN. Again, the TLR9 inhibitor chloroquine was unable to block CpG-induced adhesion and migration of THP-1 monocytes and adhesion on human monocyte–derived DCs; however, those cellular responses were completely blocked with an SFK inhibitor. CpG-induced SFK signaling was upstream and required for NF-κB activation and subsequent cytokine secretion. SFK inhibitors blocked CpG-induced secretion of IL-6 in splenocytes, IFN-α production in human pDCs, and the up-regulation of activation markers on murine B cells. Furthermore, we found that SFK activation was intact in MyD88 mice, indicating that this pathway is both upstream and independent of MyD88. The rapid kinetics of tyrosine phosphorylation suggested that this event was initiated at the cell surface and independent of the endosomal localized TLR9. The use of CpG-coated beads, too large to be internalized, revealed that CpG induced robust, rapid actin reorganization at the bead–cell contact area, further suggesting that these events did not require internalization of CpG. The use of TLR9 and MyD88 monocytes indicated that these cellular events were independent of TLR9 and MyD88 in addition to being chloroquine insensitive. Together, these studies indicated that CpG-induced SFK activation at the plasma membrane drives cytoskeletal reorganization upstream and independently of the TLR9–MyD88 pathway that could result in cell adhesion and migration. In support of our findings, it was recently reported that, upon DNA viral infection of mice, leukocyte recruitment to the liver is not altered in TLR9 and MyD88 mice, whereas the secretion of cytokines in vivo was dramatically reduced (). Our findings indicate that the recognition of CpG-ODN is more complex than previously described, but as reported (; ), this TLR9-independent pathway cannot induce cytokine secretion without ODN internalization. Although plate-bound CpG induced the formation of large lamellipodia in macrophages from wild-type, TLR9, and MyD88 mice, it was unable to induce IL-6 secretion. The TLR9 receptor responds differently to the three identified ODN classes (A, B, and C). The mechanism by which these diversified responses are generated remains unresolved (). It has been hypothesized that TLR9 requires additional coreceptors/cofactors, as is the case for TLR4 (), for the recognition of CpG–A class ODNs (). Our findings support this hypothesis because SFK activation and downstream cellular events are most strongly triggered by CpG–A class ODN when compared with B class ODN. Based on our preliminary characterization of this pathway, we hypothesize that a CpG-sensing coreceptor/cofactor is localized at the plasma membrane, which could also potentially be internalized with CpG-ODN, resulting in endosomal TLR9 activation (). Until this pathway is further characterized, it remains possible that this coreceptor/cofactor is also localized in the endosome in resting cells. The role of PI3K family on TLR signaling is poorly understood. PI3K interacts with the cytosolic Toll/IL-1 receptor domain of TLR2 () and has been found activated downstream of IRAK1 in response to IL-1β (). Recently, Akt, a downstream effector of the SFK–PI3K pathway (), was shown to be phosphorylated in response to CpG by a TLR9-independent mechanism that involves DNA-dependent protein kinase (). In our experimental system, Akt phosphorylation was intact in chloroquine-treated cells, indicating that the MyD88 cascade is not involved in Akt activation. However, Akt phosphorylation was completely blocked by both PP2 and the PI3K inhibitor wortmannin. These data indicate that the TLR9-independent tyrosine phosphorylation cascade that we describe is upstream of Akt activation. Finally, we provide evidence that the SFK pathway does intersect with the TLR9–MyD88 pathway. TLR9 is tyrosine phosphorylated upon CpG stimulation, and this event is independent of chloroquine but blocked by PP2. Furthermore, Syk coimmunoprecipitates with TLR9 upon CpG stimulation, and this interaction can be blocked by PP2. Because Syk is normally recruited to ITAM-containing receptors, which is not the case of TLR9, it is likely that the TLR9–Syk association is indirect and that other proteins participate in the formation of a complex. Two potential ITAM chains, the γ chain of the Fc receptor and DAP12 (DNAX-activating protein of 12 kD) were examined, but no phosphorylation was detected upon CpG stimulation (unpublished data). Although our experiments do not reveal whether Syk associates directly with TLR9, coimmunoprecipitation after CpG stimulation is not affected by chloroquine, indicating that this event is driven by the initial SFK pathway independently of TLR9. This link between tyrosine kinase signaling and TLRs is further supported by a recent study involving TLR3, a receptor for double-stranded RNA. Like TLR9, TLR3 is tyrosine phosphorylated, and this event is linked to PI3K-mediated enhancement of the interferon regulatory factor 3 pathway (). Unlike our findings with TLR9, TLR3 phosphorylation does not regulate NF-κB activation. In conclusion, we describe a TLR9-independent, SFK-driven signaling cascade induced by CpG. This cascade is upstream of the MyD88-dependent endosomal pathway and is initiated at the plasma membrane. This chloroquine-insensitive pathway initiates complex cytoskeletal rearrangements necessary for cell biological events such as adhesion and migration. Moreover, these two pathways interact, as inhibition of the SFK pathway blocks activation events downstream of endosomal TLR9/MyD88 such as NF-κB activation and cytokine secretion. Our findings suggest that a potential CpG-sensing receptor is localized at the plasma membrane and might interact with CpG before the activation of endosomal TLR9. Future studies are required to identify candidate proteins that would perform this function. The identification of a TLR9 coreceptor would potentially provide a new target for pharmaceutical intervention that could result in new treatments for autoimmune diseases. THP-1 and RAW264.7 (RAW) cells were obtained from American Type Culture Collection and cultured in RPMI 1640 medium (Invitrogen) supplemented with 10% FCS (Invitrogen), 2 mM -glutamine, and antibiotics. Flag-TLR9 stable transfected THP-1 were selected after transfecting THP-1 with NH-terminal–tagged version of hTLR9 cloned in p3× Flag-CMV-9, where original TLR9 leader peptide was replaced by the one provided in the vector. TLR9 and MyD88 mice were provided by S. Akira (Osaka University, Suita, Japan). Single-cell spleen suspensions from wild-type (C57BL/6 background) and MyD88 mice were obtained after mincing spleens and lysing erythrocytes. Cells were then cultured in RPMI 1640 medium supplemented with 10% FCS. Human monocyte–derived DCs were derived from human peripheral blood as previously described (). Human pDCs were isolated from PBMC using the BDCA-4 isolation kit (Miltenyi Biotec). Purified pDCs were cultured for 16–20 h in RPMI 1640 supplemented with 10% FCS, GlutaMAX, kanamycin, and Na-pyruvate (all from Invitrogen). IL-3 (R&D Systems) was added to a final concentration of 20 ng/ml. Unless indicated in the figure legends, type I interferon levels in culture supernatants were measured by evaluating the inhibition of Daudi cell proliferation with reference to a standard IFN-α curve (). Murine B cells were obtained by negative selection, removing attached cells and then using anti-Thy1.2 magnetic beads (Dynal). Murine bone marrow–derived DCs were generated from bone marrow progenitors as described by . In brief, freshly prepared bone marrow cells were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated FCS, 2 mM -glutamine, 10 mM Hepes buffer, 50 μg/ml penicillin, and nonessential amino acids in the presence of 200 U/ml GM-CSF (granulocyte-macrophage colony-stimulating factor; Sigma-Aldrich). Cultures were supplemented with GM-CSF on days 3 and 8. Murine-elicited macrophages were obtained by peritoneal lavage 4 d after peritoneal injection of 2 ml of thioglycollate broth (Sigma-Aldrich). They were washed three times in FCS-supplemented serum and then plated. 1 h later, unattached cells were discarded. For biochemistry, cells were cultured for 24 h in DME supplemented with 0.5% FCS, 2 mM -glutamine, 50 ng/ml IFN-γ, and antibiotics and then stimulated as indicated in figure legends. Anti-p65, anti–IκB-α, anti-Cbl, anti-Vav, anti-Syk, anti-Syk anti-Hck, and anti-Lyn were obtained from Santa Cruz Biotechnology, Inc. Akt and P-Y-Akt were from Cell Signaling. Monoclonal anti-Flag (M2) was purchased from Sigma-Aldrich. Anti-Pyk2 was purchased from BD Biosciences, and anti-pTyr (clone 4G10), anti-paxillin, anti-vinculin, and anti-talin antibodies were purchased from Upstate Biotechnology. Hoescht 33342, FITC-phalloidin, FluoSpheres polystyrene microspheres (10 μm for RAW cells and 15 μm in diameter for peritoneal macrophages), red fluorescent (580/605), and TOPRO-3 were obtained from Invitrogen. Herbimycin, Syk inhibitor, PP2, and PP3 were purchased from Calbiochem. Chloroquine, quinacrine, lipopolysaccharide, and red blood cell lysis buffer were obtained from Sigma-Aldrich. Oligonucleotides were obtained from GenBase and, unless indicated otherwise, were used at 1 μg/ml. The human stimulatory sequences used were as follows: human ODN2216, referred to as CpG gggggACGATCGTCgggggg, and a control ODN, GpC gggggAGCATGCTgggggg. We used the B class CpG oligo 2006 tcgtcgttttgtcgttttgtcgt. For stimulating murine macrophages and DCs, we used as stimulatory CpG ODN1585 ggGGTCAACGTTGAGggggg, and as control, GpC ggGGTCAAGCTTGAGggggg. We also used ODN1668 tccatgacgttcctgatgct. For mouse B cell stimulation, we used mouse-selective CpG gccatgacgttgagct. Capital letters for ODN sequences indicate regular phosphodiester-linked nucleotides, and lowercase letters indicate phosphorothioate-linked nucleotides. Human Hck and Lyn siRNA SMARTpools were obtained from Dharmacon. THP-1 cells were transfected with 100 nM of each siRNA or in combination using a Nucleofector and Kit V according to the manufacturer's recommendations (Amaxa Biosystems). Cells were cultured for 48 h and then analyzed in adhesion and cytokine assays. Cells were lysed in RIPA buffer for 30 min on ice (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% DOC, 0.1% SDS, protease inhibitor tablet [Roche], 1 mM NaF, 1 mM NaVO, and 1 mM PMSF). After centrifugation (20,000 , 10 min, 4°C), supernatants were analyzed by SDS-PAGE. For anti-paxillin, anti-talin, and anti-vinculin immunoprecipitations, cells were lysed directly in sample buffer containing 1% SDS, boiled, and diluted 10-fold with 1% Triton X-100 containing lysis buffer. Cells were stimulated as indicated in figure legends and then fixed with 4% paraformaldehyde in PBS, washed twice with PBS, and blocked with 2% BSA in PBS for 2 h. All incubations and washes were performed in 0.5% BSA and 0.1% Triton X-100 in PBS. Antibody staining was performed at 37°C in a moist chamber for 1 h. F-actin was stained with FITC-phalloidin for 20 min, cells were washed twice in PBS, and nuclei were stained with TOPRO-3 for 20 min. Samples were mounted, and images were captured with a confocal microscope (IX70; Olympus) using the uPlanApo 100×/1.35 lens and Fluoview 2.0 (Olympus) program. For immobilized ODN microsphere cell binding, 10 microsphere beads were coated with a 1-mg/ml solution of GpC or CpG-ODN overnight and then washed four times in culture medium. Coated microspheres (10-μm diameter for RAW cells or 15 μm for peritoneal macrophages) were added to cells in culture at a microsphere/cell ratio of 3:1. Cells were fixed after an incubation of 20 min. 10 cells/well were plated in 96-well plates. After the indicated treatments, cells were fixed with 4% paraformaldehyde in PBS for 5 min. Unattached cells were removed by washing six times with PBS. The nuclei of attached cells were stained with Hoechst dye for 10 min. Cells were then visualized using a 5× magnification on a ArrayScanII (Cellomics), and the number of nuclei per field were counted using the manufacturer's software. For final analysis, a mean of 36 fields from at least four different wells were counted. Single-cell splenocyte suspensions or human pDCs (5 × 10 cells/well) were stimulated for 24 h with 1 μM ODN. Supernatants were collected, and IL-6 or IFN-α (R&D system) were measured by ELISA according to the manufacturer's instructions. Splenocytes were stained and analyzed on a FACScalibur (Becton Dickinson). For measuring up-regulation of B cell activation markers, cell surface expression of CD40 and CD69 were analyzed on gated B-220–positive cells. The cells were stained with FITC-conjugated anti-CD40, anti-CD69, and allophycocyanin-conjugated anti-B220 from BD Biosciences. Fig. S1 shows the effect of the microfilament-disrupting drugs cytochalasin D, latrunculin A, and jasplakinolide on CpG-induced adhesion. Fig. S2 shows that chloroquine and quinacrine are unable to block CpG-induced cell motility. Fig. S3 shows that MyD88 mice are unresponsive to CpG as measured as the secretion of IL-6 or the up-regulation of the plasma membrane level of CD40 and CD69. Online supplemental material is available at .

Integrins are adhesion receptors that connect components of the extracellular matrix or molecules borne by cells or microbial pathogens to intracellular signaling pathways (; ). Integrins are subjected to bidirectional signaling; ligand binding can be regulated by inside-out signaling, and it also triggers a variety of downstream pathways, including the physical association of integrins to the actin cytoskeleton and cytoskeletal remodeling (e.g., during focal adhesion formation and cell motility; ; ; ). Signaling to and from integrins is mainly regulated by the short cytoplasmic tails of α and β subunits, which lack catalytic activity. In contrast to the well conserved cytoplasmic domains of β subunits, α subunit cytoplasmic domains share few sequence similarities (), which has led to the idea that the two integrin tails play distinct roles and to the suggestion that α tails could confer specificity to integrins sharing identical β chains (; ). α and/or β cytoplasmic tails have been involved in several aspects of integrin biology and function (e.g., inside-out signaling, localization to focal adhesions, and cell adhesion; ; ; ; ). It is thought that both inside-out and outside-in signaling are regulated through the binding of cytoskeletal and regulatory proteins to integrin cytoplasmic regions. Over 20 such proteins have been described to date (; ; ; ). Small GTP-binding proteins also play crucial roles in integrin function; Rap1 activity controls inside-out activation of β, β, and β integrins (; ), whereas Rho family proteins control integrin signaling to the cytoskeleton (). However, the molecular mechanisms underlying the cross-talk between integrins and small GTPases remain unclear. β integrins, which are mutated in type-1 leukocyte adhesion deficiency patients, control the majority of immune cell functions, including cell migration, immunological synapse formation, and phagocytosis (). Accordingly, immune cells isolated from mice that were deficient in one or more β integrins manifested a range of defects in cytoskeletal-based functions (). The αβ integrin (which is also known as complement receptor 3 [CR3] or CD11b/CD18) is the main phagocytic receptor for particles opsonized with the complement fragment C3bi; as such, it plays a major role in the phagocytosis of microorganisms and apoptotic cells (). α and β are both needed for the efficient binding and ingestion of C3bi-opsonized particles (), and, similar to αβ (; ; ; ), αβ function is regulated by small GTP-binding proteins (). Interestingly, unlike other modes of uptake, αβ-dependent phagocytosis relies exclusively on Rho function (). RhoA, but not Rac1, accumulates at sites of particle binding, where it colocalizes with F-actin. A dominant-negative allele of Rho (N19), overexpression of a kinase-dead Rho kinase mutant, or treatment with the Rho kinase inhibitor Y-27632 blocks the recruitment of the actin nucleator/organizer Arp2/3 complex at nascent phagosomes, whereas blockers of Rac and Cdc42 function have no effect (; ; ). RhoA and Rho kinase activities are thus specifically required at phagosomes to drive Arp2/3-dependent actin polymerization and uptake. These results also indicate a striking similarity between the signaling pathways elicited during αβ-driven phagocytosis and other processes that are mediated by integrins, Rho, and Rho kinase (; ; ; ; ). Nonetheless, the mechanisms involved in Rho regulation downstream of integrin ligation are totally unknown. We show that the binding of C3bi-opsonized targets to αβ activates Rho, but not Rac, in macrophages and αβ-transfected Cos-7 cells. Moreover, the cytoplasmic domain of β, but not of α, controls Rho function during phagocytosis. Remarkably, two distinct regions within the β cytoplasmic tail control Rho activation and stable recruitment. Finally, we show that recruitment of active Rho to nascent phagosomes is sufficient to induce αβ-dependent uptake. αβ-mediated phagocytosis requires Rho, but not Rac, activity (). To investigate if the levels of active, GTP-bound Rho increase during phagocytosis, we performed pull-down experiments using the Rho-binding domain of rhotekin (RBD) fused to GST, as previously described (). J774.A1 macrophages were pretreated with the phorbol ester PMA, to activate αβ binding ability through inside-out signaling (), and challenged with C3bi-opsonized sheep red blood cells (RBCs). At different time points, cells were lysed and levels of active Rho were determined. Progression of phagocytosis correlated with a transient increase in GTP-loaded Rho, peaking 20 min after RBC challenge. A significant increase (up to sevenfold) in GTP-Rho occurred only after stimulation with both PMA and RBCs. The levels of endogenous active Rac1, as determined in the p21-activated kinase (PAK)–CRIB pull-down assay (; ), did not change significantly during phagocytosis (). These results show that αβ ligation by C3bi-opsonized particles (i.e., outside-in signaling) leads to a specific increase in Rho activity. To characterize the mechanism of Rho regulation during CR3-mediated phagocytosis, we turned to Cos-7 cells, which do not contain any endogenous α or β and offer a robust alternative system for the study of phagocytosis, as shown by several studies (; ; ). As shown in , cells expressing α bound RBCs poorly, and expression of β alone conferred no binding ability. However, as previously reported (), coexpression of α and β induced efficient binding and phagocytosis. PMA treatment increases RBCs binding by twofold in transfected COS7 cells, as it does in macrophages and neutrophils. This is not because of increased αβ expression, but, rather, because of the activation of regulators of inside-out signaling such as PKC and Rap1 (; unpublished data). Nevertheless, there are clearly enough active integrins in transfected COS cells to bind and zipper around RBC, thereby allowing binding and phagocytosis in the absence of PMA. Coexpression of dominant-negative Rho decreased uptake by 70%, but had no effect on RBC binding (). Furthermore, phagocytosis is also accompanied by an increase in the levels of GTP-Rho in transfected Cos-7 cells (). No changes in Rac-GTP levels were detected in these cells. A mutagenesis strategy was undertaken to identify which regions of αβ are responsible for signaling to Rho and phagocytosis. We deleted 14 (αΔ1139) or 23 (αΔ1130) aa of the α cytoplasmic domain or deleted the whole β cytoplasmic domain (βΔ724; ) and cotransfected each mutant with its wild-type (wt) counterpart (). At steady-state, αΔ1139/β and αβΔ724 were expressed at wt levels, as determined by flow cytometry; however, αΔ1130β did not reach the plasma membrane (unpublished data). The membrane-proximal 10 residues of α, but not the β cytoplasmic tail, are thus important in directing heterodimerization and/or surface expression of the αβ integrin. We next compared the capacity of wt and mutant αβ integrins to bind and internalize C3bi-opsonized RBC. The combination αΔ1139/β led to control levels of binding and uptake. In contrast, deleting the cytoplasmic domain of β (βΔ724) increased binding, as described previously for αβ (), but completely abolished phagocytosis (). The β cytoplasmic domain is therefore necessary for αβ-mediated uptake. Accumulation of RhoA at nascent phagosomes can be visualized by expressing a tagged wild-type RhoA construct (). In cells expressing full-length αβ, Rho-GFP was enriched at 60 ± 11% of the phagosomes (). Interestingly, when coexpressed with wt β, αΔ1139 was as able to recruit Rho around RBC as wt α. In contrast, deletion of the β cytoplasmic tail abolished Rho recruitment, reducing it to the background levels observed with GFP alone (). Thus, Rho recruitment to early phagosomes during CR3-mediated phagocytosis is controlled by the cytoplasmic domain of β. Rho proteins are thought to cycle between an active GTP-bound, membrane-bound state and an inactive cytosolic state, which suggests that the recruitment we observe might result from RBC-induced (outside-in) activation of transfected RhoA and local enrichment underneath RBC. To determine which of the cytoplasmic domains of αβ controlled Rho activation during phagocytosis, we performed rhotekin pull-down assays in cells transfected with the truncation mutants. None of the cytoplasmic deletions influenced Rho-GTP levels in resting cells (, open bars). When coexpressed with αΔ1139, β was still able to increase the levels of active Rho in response to RBC (, shaded bars). Strikingly, deletion of the β cytoplasmic tail abrogated the RBC-induced increase in GTP-Rho, strongly suggesting that the β cytoplasmic domain controls both αβ-induced Rho activation and phagocytosis. Additional truncations and point mutations () were engineered into the β cytoplasmic tail to analyze in detail the regions required for phagocytosis. Ftlow cytometry analysis showed that all the combinations of β mutant chains with wt α were expressed at the cell surface at the same levels as wt β, except βF766A and βAAA, which were consistently expressed at 80 and 50% of the control levels (unpublished data). Interestingly, only three mutants (βΔ732, βΔ767, and βΔN) showed binding indices similar or superior to wt values. This suggests that, apart from βAAA, whose reduced binding ability () correlated with reduced surface expression, all other mutations influence integrin activation and/or recycling. Importantly, the phagocytic capacity of Cos-7 cells transfected with all but two β mutants was low (), indicating a possible link between β mutation and the abrogation of Rho function. The exceptions were βΔ767, which behaved like wt not only for particle binding but also for phagocytosis, and βF766A, which was still able to engulf bound particles efficiently. Interestingly, the ability of β mutants to mediate phagocytosis was correlated with their ability to trigger the accumulation of F-actin at sites of particle binding (). We next determined whether any of these β mutants affected RBC-induced Rho activation. Importantly, none of the mutants influenced Rho-GTP levels in resting cells (unpublished data). The deletion mutants βΔ747 and βΔ755 induced significant Rho activation after phagocytic challenge (). In contrast, βΔ732 was unable to increase Rho-GTP levels in response to RBC. To confirm the requirement of the 732–747 region of β in Rho activation, we engineered a mutant (βΔN) lacking the NH-terminal half of the cytoplasmic tail (residues 724–755), thereby missing the putative RhoA activation domain. βΔN was also unable to increase Rho-GTP levels upon RBC binding. Together, these results strongly suggest that the 732–747 region of the β tail is necessary to promote Rho activation during αβ-mediated phagocytosis, but not sufficient to trigger actin polymerization and uptake. We next investigated the impact of β mutations on Rho recruitment to nascent phagosomes using confocal microscopy. COOH-terminal deletions of the β cytoplasmic domain had varying impacts on Rho recruitment (), suggesting that one or more aa within the 756–767 region of β are required for a detectable accumulation of Rho at phagosomes. Interestingly, several residues within this 12-aa region of β are conserved in other β chains and proposed to regulate integrin function. In particular, a cluster of threonine residues (758–760) is phosphorylated upon phorbol ester stimulation of leukocytes () and required for the modulation of leukocyte adhesiveness (; ). Also, mutations in the conserved Asn-Pro-Xaa-Tyr/Phe motif (NPxΦ, with Φ at positions 754 and 766 in human β) perturb the interaction of integrins with binding partners and integrin signaling (; ; ). Thus, we engineered mutants βF766A and βAAA (threonines 758–760 mutated to alanine) and studied their impact on the recruitment of Rho-GFP to forming phagosomes. Flow cytometry analysis revealed that the βF766A mutant was expressed at 80% of the control value at steady-state. This mutant also showed a markedly reduced ability to bind opsonized RBC (), but phagocytosis () and Rho recruitment () were both measurable. However, although the βAAA mutant was still able to bind particles, it was unable to recruit Rho to nascent phagosomes. In contrast, βAAA showed wt binding to a GFP–talin head fragment whose interaction with the β integrin depends on residue F754 (). We concluded that the threonine residues at position 758–760 are necessary for Rho recruitment during CR3-mediated phagocytosis. Despite being unable to detectably recruit Rho, βΔ755 and βAAA were still able to up-regulate Rho activity upon RBC binding ( and not depicted). This suggests that Rho activation and visible accumulation at sites of particle binding are independent steps and that Rho activation precedes its stable recruitment. In line with this, inactive (N19) Rho is not recruited beneath RBC bound to αβ (). To obtain further evidence of the mechanism involved, we analyzed the recruitment of a constitutively active form of Rho (L63) in cells cotransfected with α and β with either wt, AAA, or ΔN (lacking the Rho activation domain but comprising the three threonine residues). Although βΔN cannot activate Rho (), coexpression with L63Rho allowed the recruitment of GFP-tagged, active Rho underneath RBC. However, when the experiment was repeated with βAAA, L63Rho was no longer recruited to nascent phagosomes (), confirming our hypothesis that Rho is locally recruited to forming phagosomes in its active, GTP-bound form, in a manner that is dependent on threonines 758–760. To investigate the consequences of the local recruitment of active Rho on phagocytic uptake, we analyzed the capacity of βΔN and βAAA to internalize RBC when coexpressed with wt α and L63Rho-GFP. As shown in , coexpression of L63Rho rescued the phagocytic defect of βΔN-expressing cells without affecting their ability to bind RBC or the phagocytic ability of αβ-expressing cells. Therefore, expression of active Rho completely bypassed the need for the 732–747 region, which is normally necessary for Rho activation, actin polymerization, and phagocytosis (; and ), and allowed both Rho accumulation and RBC uptake by cells expressing the internally truncated mutant. In contrast, cells expressing βAAA were not rescued for phagocytosis by L63Rho-GFP (). Together, these results indicate that threonines 758–760 control the local accumulation of active Rho, which is sufficient to affect CR3-mediated phagocytosis. The αβ integrin, which is mainly expressed in macrophages and neutrophils, mediates phagocytosis of microbes and apoptotic cells. αβ-mediated phagocytosis requires Rap1, RhoA, and Rho kinase activity and is dependent on the Arp2/3 complex and actin polymerization (; ). Because of its spatially localized signaling, phagocytosis offers an ideal system to understand how extracellular (particularly integrin) ligands regulate the function of small GTP-binding proteins to remodel the actin cytoskeleton. We show that outside-in signaling from the αβ integrin requires the cytoplasmic tail of β, but not α. Specifically, two distinct subregions within the β tail control the recruitment and activation of RhoA. Interestingly, Rho recruitment to sites of particle binding was dependent on prior activation of the G protein and sufficient for phagocytosis. Our data shed light on the mechanisms by which integrin ligation is coupled to the activation of small GTPase function. Unlike other modes of phagocytosis or bacterial invasion, αβ uptake is dependent on RhoA signaling, but does not require Rac or Cdc42 activity (). Until now, the underlying mechanism had remained unclear. We show that the levels of active, GTP-bound RhoA increase transiently during αβ phagocytosis, whereas the levels of active Rac show insignificant variation, suggesting that ligation of αβ is sufficient to activate RhoA specifically. In line with this, antibody cross-linking of αβ in J774.A1 macrophages also leads to RhoA activation (unpublished data). Whereas αβ is so far the only known phagocytic receptor that is exclusively coupled to the activity of the RhoA protein, several studies have previously linked integrin signaling to the up-regulation of Rho activity. For example, a transient up-regulation of RhoA activity was described in human neutrophils freshly plated on immobilized anti-β antibodies (). Although integrin-mediated spreading is associated to Rac1 activation (; ), integrin ligation or overexpression have already been linked to increased RhoA activity in a variety of cell types (; ; ). Mutagenesis revealed that the cytoplasmic domain of β is the main regulator of αβ during outside-in signaling to phagocytosis. Maximal cytoplasmic deletions in α that were compatible with surface expression did not alter any of the heterodimer properties that we examined (such as particle binding and phagocytosis and Rho recruitment and activation). In contrast, the phagocytic function of αβ was critically dependent on the integrity of the β cytoplasmic domain. Truncation of the entire β tail increased the association of C3bi-opsonized RBC, as previously described (), which is consistent with the idea that disruption of intrachain interactions between α and β cytoplasmic domains results in constitutively active integrins (). Except for βAAA, for which reduced surface expression correlated with reduced binding, our results suggest that mutations that decrease RBC binding significantly result from an impact of these mutations on integrin inside-out activation or recycling. Indeed, in αβ and other integrins, several of the residues between 732 and 766 have been involved in talin binding and activation by Rap1, two important mechanisms in inside-out activation (; ; unpublished data) and in recycling or stimulated internalization (; ; ). Remarkably, deletion of the entire β tail abolished both RhoA recruitment and activation and also abrogated phagocytosis. Added to the fact that blocking Rho function inhibits RBC uptake without affecting binding, this strongly suggests that Rho activity is specifically required during outside-in signaling from αβ. This seems to conflict with several papers claiming that RhoA is needed for β activation in lymphocytes and thymocytes (; ; ). Although it is conceivable that RhoA could regulate either inside-out or outside-in signaling for different integrins or in different cell types, it seems more likely that the discrepancy comes from a different understanding of the readouts used in these studies. Indeed, adhesion to ligand-coated surfaces is not simply a consequence of pure inside-out signaling, but rather a reflection of both inside-out and outside-in signaling. Nevertheless, our results show clearly that the intracellular domain of the β integrin chain controls RhoA activation. β subunits consist of a large extracellular domain, a transmembrane segment, and a relatively short cytoplasmic tail (20–70 aa, 46 in human β, except for the much longer β integrin). Truncation of the intracellular region of β, β, and β abrogates their ability to localize to focal adhesions, and prevents activation of signaling molecules and downstream cytoskeletal remodeling (; ; ; ; ). Reciprocally, isolated β and β tails are sufficient to activate focal adhesion kinase and can regulate cell cycle progression and actin cytoskeleton assembly (; ; ). Although there is no crystal structure available for any isolated integrin cytoplasmic domain, several groups used nuclear magnetic resonance to investigate the three-dimensional organization of αβ intracellular domains (; ; ). We engineered our truncated β mutants based on their results, which suggested that, in both tails, the membrane-proximal regions were α-helical and interacted with each other, whereas the more distal regions are disordered in aqueous environment. Interestingly, we found that two distinct cytosolic regions within β control RhoA activation and recruitment to forming phagosomes. The region controlling Rho activation maps to the helical domain, whereas Rho recruitment is controlled by the TTT motif within the disordered region. Importantly, Rho activation can occur independently of a measurable recruitment. However, the reverse is not true, as nascent phagosomes are negative for RhoA enrichment unless Rho can be activated. There is no Rho recruitment in cells expressing wt αβ and N19RhoA or in cells coexpressing wt RhoA and a β mutant harboring an internal deletion of the Rho activation domain, therefore suggesting that only active RhoA can accumulate where RBCs bind. Therefore, there might be a structural basis for the mechanisms underlying Rho recruitment and activation. We also speculate that recruitment of active Rho may play a role in stabilizing the disordered domain of β and possibly of other β chains. How could Rho recruitment and activation be coordinately regulated? We describe a specific, transient up-regulation of RhoA activity upon challenge with RBC. The simplest explanation is that, as has been proposed for bacterial invasion, FcγR-mediated phagocytosis, and various extracellular stimuli, particle binding triggers the local activation of guanine-nucleotide exchange factor molecules (; ; ). Because N19Rho, which is thought to titrate endogenous exchange factors, is not recruited to nascent phagosomes, our data might suggest that Rho is activated away from ligated integrins. However, it is hard to envisage how the membrane-proximal RhoA activation domain could direct activation of the GTPase at a distance. The alternative possibility is that RhoA is activated locally. We are currently investigating the possible role of guanine-nucleotide exchange factors, RhoGDI, and, indirectly, GTPase-activating proteins in Rho activation during αβ phagocytosis. To explain the local activation of RhoA in the absence of detectable recruitment, we favor the hypothesis that our experimental approach does not allow us to visualize the interaction between overexpressed RhoA and the truncated mutants (e.g., βΔ747). This could happen either because the TTT motif is necessary to stabilize the interaction of β with active RhoA, with a regulator of Rho activation, or, finally, because a crucial regulator of Rho recruitment to the truncated mutants is missing in our overexpression setting. We also found that three threonine residues in the β cytosolic region governed the recruitment of active Rho to nascent phagosomes. Interestingly, the corresponding three-residue sequence, which is flanked by two NPxY/F motifs (), has been associated to the regulation of the adhesive function of β, β, β, and β integrins. However, whether this short sequence is involved in inside-out or outside-in signaling has remained controversial (; ; ; ; ; ). Our results strongly suggest that residues 758–760 regulate the recruitment of active Rho to the integrin complex in response to ligand binding (outside-in signaling). Phorbol esters were shown to induce phosphorylation of these threonine residues in β (; ). As PKC was shown to interact with β tails (), it is tempting to imagine a regulatory mechanism whereby PKC would interact with β and phosphorylate the three threonine residues, allowing the docking of active RhoA (binding either directly or indirectly). Although PKC activity is required for αβ phagocytosis (), a role upstream of Rho is hard to reconcile with the observation that mutations in the putative PKC-binding sites on β () do not affect Rho recruitment in our model system (unpublished data). Moreover, the β–PKC interaction has been connected to increased motility rather than increased adhesion (). Alternatively, it is possible that another as yet unidentified serine/threonine kinase mediates the phosphorylation of these threonine residues. Importantly, several binding partners, including ICAP-1, filamin, and 14-3-3 proteins interact in a phosphorylation-dependent manner with the TTT region of various β chains (; ; ; ; ). Unfortunately, none of these molecules has been implicated in αβ-mediated phagocytosis. Interestingly, however, they are all related to small GTPase signaling (; ; ; ); 14-3-3 and filamin offer the most direct connection with RhoA signaling, respectively through Rho kinase and binding/activation, via LIM-kinase-induced phosphorylation of cofilin (; ; ). Whether any of these proteins are involved in the stable recruitment of Rho to β integrins during phagocytosis will form the basis of future studies. In conclusion, we show that ligand binding to αβ specifically activates Rho, but not Rac. Rho function is regulated in two steps, which are governed by distinct regions of the β cytoplasmic tail. Whereas a membrane-proximal α-helical region controls the activation of Rho, its detectable recruitment involves a distal TTT motif. We demonstrate that stable enrichment of activated Rho at the plasma membrane is controlled by the threonines 758–760 in β-cytoplasmic domain and is sufficient to promote αβ-dependent phagocytosis, a process known to require, successively, Rho kinase, myosin light chain kinase, myosin II, the Arp2/3 complex, and actin polymerization. Because of the conservation of signaling pathways downstream of integrin β chains, we propose that similar mechanisms operate in related cellular activities, e.g., rear detachment during cell motility. The antibodies used in this study were mouse anti-Rac1 (clone 23A8; Upstate Biotechnology), anti-GFP (JL-8; CLONTECH Laboratories, Inc.), anti-myc (9E10; Santa Cruz Biotechnology, Inc), anti-β (BD Biosciences), and rabbit anti-RhoA (119; Santa Cruz Biotechnology, Inc). Conjugated secondary antibodies were obtained from either Jackson ImmunoResearch Laboratories (fluorescence) or GE Healthcare (Western blots). Eukaryotic expression vectors (pRK5) encoding human wt β, wt α, and myc-tagged N19Rho constructs were previously described (), as were pGEX-2T vectors encoding the RBD and the Rac/Cdc42-binding domain of PAK-CRIB (; ). The pRC/CMV plasmid encoding the βΔ732 mutant was provided by Y. van Kooyk (VU University Medical Center, Amsterdam, Netherlands). GFP-tagged small GTPase constructs were generated by subcloning wt, N19, and L63Rho from pGEXRho () or pRK5-myc into pEGFP-C1 (CLONTECH Laboratories, Inc.). GFP-talin head was a gift of D.R. Critchley (University of Leicester, Leicester, UK). α mutants were engineered via PCR by introducing a premature stop codon within the cytosolic tail at positions 1130 (αΔ1130) and 1139 (αΔ1139) using pRK5-α as a template; the sense primer 5′-GCTCTAGACTCTCGAGTACGTGCCACACC-3′, which incorporates an internal XhoI site; and the antisense primers 5′-CTTGAAAAGCTTCTA CTTGTACAGCGCGGCGGT-3′ for αΔ1130 and 5′-TTCACTAAGCTTCTACTTGTATTGCCGCTTGAA-3′ for αΔ1139. β mutants were generated in an analogous manner. Because of unfavorable restriction sites, some required initial subcloning of an Xbal–HindIII fragment of pRK5-β into pUC19. Mutants were cloned by replacing a SacII–HindIII fragment of pUC19-β with different SacII–HindIII PCR fragments, which were amplified using pUC19-β as a template, the sense primer 5′-GCTCTAGAACCCCGCGGCGTGTTGAGTG-3′, an incorporated internal SacII site, and one of the following antisense primers: 5′-CAGGTGAAGCTTCTACTTCCAGATGACCAGCAG-3′ (βΔ724); 5′-CCACTGAAGCTTCTACTTCTCCTTCTCAAAGCG-3′ (βΔ742); 5′-CGTCGTAAGCTTCTACTTGAAAAGGGGATTATC-3′ (βΔ755); 5′-CCCCTAAAGCTTCTAACTCTCAGCAAACTTGGGGTTCATGACCGTCGTGGTGGCGCTCTTCCAGATGACCAG-3′ (βΔN); 5′-CCCCTAAAGCTTCTAACTCTCAGCAAACTTGGGGTTCATGACCGCCGCGGCGGCGCTCTTGAAAAG-3′ (βAAA). The βΔ767, βΔ747, and βF766A mutants were constructed by PCR using wt β as a template, the sense primer 5′-GGGGGGTCTAGAATGCTGGGCCTGCGCCCCCCACTG-3′, which incorporates an internal XbaI site, and the following antisense primers: for βΔ767, 5′-GGGGGGAAGCTTCTAAGCAAACTTGGGGTTCAT-3′; for βΔ747, 5′-GGGGGGAAGCTTCTA CCACTGGGACTTGAGCTTCTC-3′; and for βF766A, 5′-GGGGGGAAGCTTCTAACTCTCAGCAGCCTTGGGGTTCAT-3′. The aa sequences of the cytoplasmic tails of all α and β mutants used in this study are given in . Cells from the murine macrophage J774.A1 and Cos-7 cell lines, which were obtained from the American Type Culture Collection, were maintained, seeded, and transfected as previously described (). For pull-down assays, 9 μg DNA was used to transfect 3 × 10 cells on a 10-cm dish, using SuperFect (QIAGEN) as recommended by the manufacturer. After 4 h, fresh medium was added for 16 h and cells were then serum starved overnight. Transfected Cos-7 cells were washed with cold wash buffer (0.5% BSA and 0.02% azide in PBS) and stained to detect surface β, using mouse monoclonal and FITC-conjugated goat anti–mouse antibodies successively. The relative fluorescence of gated Cos-7 cells was analyzed using a FACSCalibur analyzer (Becton Dickinson). C3bi-opsonized RBCs were essentially prepared and used to challenge phagocytes as previously described (), using 0.5 and 20 μl of fresh RBC per coverslip/per dish for immunofluorescence and/or pull-down assays, respectively. For efficient binding and phagocytosis of C3bi-opsonized erythrocytes, macrophages require preactivation (; ); we routinely pretreated J774.A1 cells with 150 ng/ml PMA in buffered serum-free DME for 15 min at 37°C. PMA-treated cells were washed with warm PBS to remove unbound RBC and fixed in cold 4% paraformaldehyde for 10 min at 4°C. Transfected Cos-7 cells were first stained for surface β. For differential staining of adherent and internalized C3bi-opsonized RBC, cells were incubated with Texas red–conjugated goat anti–rabbit antibodies, permeabilized with 0.2% Triton X-100, and incubated with FITC-conjugated goat anti–rabbit antibodies. Coverslips were finally mounted in Mowiol (Calbiochem) containing -phenylenediamine (Sigma-Aldrich) as an antifading reagent and analyzed using a fluorescence microscope. Only β-positive cells were analyzed; internalized particles (which were red) were easily distinguishable from extracellular RBCs (which were yellow), a property that was used to score phagocytosis. The binding and phagocytic indices are defined as the number of targets bound to and engulfed by 100 phagocytes, respectively. The recruitment of Rho during CR3-mediated phagocytosis was studied by scoring local enrichment in the GFP signal at sites of RBC binding by confocal microscopy (model LSM510; Carl Zeiss MicroImaging, Inc.). A minimum of 20 transfected cells were analyzed per condition. The percentage of bound RBC showing a discrete local enrichment in GFP signal was scored. Phagosomes were scored as positive when at least a quarter of the underlying/surrounding area showed significant GFP enrichment compared with the neighboring areas. All of the overexpressed GFP constructs showed a low level of plasma membrane localization. Rhotekin RBD and PAK-CRIB fused to GST were prepared as previously described (; ). J774A.1 or Cos-7 cell lysates were incubated for 45 min at 4°C with 15 μg of a 50% slurry of glutathione Sepharose 4B beads coupled to GST-RBD-rhotekin or GST-PAK-CRIB to precipitate GTP-RhoA or GTP-Rac. Beads were washed three times in cold lysis buffer (10% glycerol, 1% NP-40, 50 mM Tris, pH 7.6, 200 mM NaCl, 2.5 mM MgCl, 1 mM PMSF, and protease inhibitor cocktail). Equal volumes of beads and total lysates were analyzed by SDS-PAGE and Western blotting, which were performed as previously described (). Intensities of the RhoA and Rac1 signals were determined by densitometric analysis and related to the levels of total RhoA or Rac1 in the lysates, using the ImageJ software.

Neuronal apoptosis is induced by numerous stressors and underlies many human neurodegenerative disorders, such as Alzheimer's and Parkinson's disease. Under such apoptotic conditions, several neurotrophic factors such as glial cell line–derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF) can activate the antiapoptotic process to rescue neurons from death. However, the signaling pathway leading to cell survival is not yet completely understood. GDNF was reported to exert a potent survival-promoting activity in neurons (; ; ) and to reduce neuronal death induced by various toxic challenges both in vitro () and in vivo (; ). Recent evidence suggests that a part of molecular mechanisms for GDNF-induced cell survival relates to an increase in intracellular Ca concentration, and it subsequently activates some survival pathways such as the phosphatidylinositol 3-kinase (PI3-K)–Akt pathway (). Ca is the most versatile and important intracellular messenger in neurons, regulating a variety of neuronal processes such as neurotransmission and signal transductions. The various actions of Ca are mediated by a large family of EF-hand Ca-binding proteins, which may act as Ca sensors or Ca buffers. One of them, neuronal Ca sensor-1 (NCS-1; mammalian homologue of frequenin), was originally identified in in a screen for neuronal hyperexcitability mutants (). Overexpression of NCS-1 has been shown to enhance evoked neurotransmitter release and exocytosis (; ). NCS-1 directly interacts with phosphatidylinositol 4-hydroxykinase (PI4-K; ; ) and enhances neuronal secretion by modulating vesicular trafficking steps in a phosphoinositide-dependent manner (). We have previously demonstrated that NCS-1 modulates the voltage-gated K channel Kv4 (). Subsequently, certain voltage-gated Ca channels have also been reported to be regulated by NCS-1 (; ; ). Furthermore, NCS-1 enhances the number of functional synapses (), potentiates paired pulse facilitation (), and may be involved in associative learning and memory in (). Despite the participation of NCS-1 in a wide range of biological functions, however, the role of NCS-1 in neuronal survival under pathophysiological conditions or the involvement of NCS-1 in neurotrophic factor–mediated neuroprotection are unknown. Because we found that the expression levels of NCS-1 is significantly higher in immature brain () and a remarkable similarity exists between immature and injured neurons during the development and regeneration process, respectively (), these findings prompted us to study the expression level and the functional roles of NCS-1 in damaged neurons. In this study, we found that NCS-1 is a survival-promoting factor, which increases the resistance of neurons to several kinds of stressors. In addition, NCS-1 is up-regulated in response to axonal injury in adult motor neurons, and this protects cells from apoptosis. Furthermore, NCS-1 mediates GDNF-induced neuroprotection via activation of Akt pathways. This is the first study demonstrating a novel role of NCS-1 on neuronal survival. To examine the expression level of NCS-1 in injured neurons, we performed unilateral vagal axotomy (transaction of nerves) on adult rats. 1 d to 2 mo after the in vivo axotomy, brainstems, including the bilateral dorsal motor nucleus of the vagus (DMV) neurons, were isolated. Immunohistochemical staining and computerized image analysis of frozen sections revealed that axotomy significantly (more than threefold) increased the expression level of NCS-1 in the DMV when compared with those on the control side at 1 wk after the surgery (). NCS-1 immunoreactivity was mainly expressed in cell bodies of neurons, as shown using hematoxylin counterstaining to identify the nuclei (, brown staining accompanied with blue staining; depicted by arrows). The increase in NCS-1 level started at 1 d after axotomy, reached a peak at 1 wk, and gradually decreased to control levels over the next 2 mo (). We also conducted quantitative immunoblot analysis on tissue samples from DMV neurons 1 d and 1 wk after axotomy, expressing NCS-1 density relative to levels of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The results confirmed the immunohistochemistry experiments, with levels of NCS-1 protein in ipsilateral DMV being increased significantly (by about threefold) by 1 wk after axotomy (). Up-regulation of NCS-1 protein was also observed using a different type of stressor. Continuous treatment of neurons with colchicine for 4 d, which disrupts tubulin polymerization and blocks axonal transport, also increased NCS-1 expression levels (1.40 ± 0.03-fold above control levels; P < 0.05; = 4), indicating that NCS-1 is up-regulated in vivo in response to two different kinds of stressors—one being mechanical and the other being chemical injury. To study the physiological role of NCS-1 in damaged neurons, we next examined the effect of NCS-1 overexpression on the susceptibility of cells to several kinds of stressors. PC-12 cells stably transfected with either the NCS-1 expression vector (NCS-1/PC-12) or the vector alone (vector/PC-12) were differentiated into neuronlike cells, and the resistance to HO toxicity was compared between these two groups. As shown in the immunofluorescent micrographs and immunoblot in , the expression level of NCS-1 was found to be significantly higher in NCS-1/PC-12 cells compared with vector-transfected cells, although these cells also had some endogenous NCS-1. Treatment with a relatively high dose (300 μM) of HO for 3 d in the absence of pyruvate resulted in severe cellular damage in vector/PC-12 control cells; most cells were rounded up and detached from the substratum (). In contrast, the same treatment caused only a little damage to cells overexpressing NCS-1 (), indicating that the expression of NCS-1 rendered PC-12 cells more tolerant to HO toxicity. The expression of NCS-1 reduced cell death caused by treatment with up to 1,000 μM HO (). The aforementioned results were obtained from three cell lines transfected with NCS-1 and with corresponding vector-transfected control. A similar beneficial effect of NCS-1 on cell survival in response to 300 μM HO was seen in two PC-12 cell lines that were not treated with NGF (not depicted). To further confirm the involvement of NCS-1 in neuronal survival, we overexpressed NCS-1 or its mutant E120Q in primary cultured embryonic rat cortical neurons that express endogenous NCS-1. The E120Q mutant possesses an amino acid substitution within the third EF-hand Ca-binding motif, which impairs Ca binding () but preserves the interaction with target proteins and, thereby, exerts a dominant-negative effect by disrupting the function of endogenous NCS-1 (). We used an adenoviral transfer system to transiently deliver the cDNA encoding NCS-1 together with EGFP (using an internal ribosome entry site–containing vector) and its E120Q mutant form into neurons cultured for 5 d in neurobasal medium containing B27 trophic supplements. As indicated by cells with EGFP fluorescence and nuclei stained with Hoechst 33258, nearly 70% of neurons were successfully infected with each virus at 3 d after infection (). We examined the effects of overexpression of wild-type and dominant-negative NCS-1 on neuronal survival under stress caused by B27 withdrawal, which has been reported to induce neuronal apoptosis (; ). As shown in , B27 withdrawal promoted cell death in vector-treated control neurons (left; also compare the vector groups with and without B27 in ). Overexpression of NCS-1, on the other hand, significantly rescued cells from death (, middle). In contrast, the expression of E120Q resulted in more severe cell death accompanying bleb formation (, right). To quantitatively analyze the time course for the changes in cell viability, the total number of surviving cells from the same field was counted daily by phase-contrast microscopy during 9 d (see Materials and methods). The results show that high cell viability was preserved upon expression of the wild-type NCS-1, whereas cell viability was reduced after the expression of E120Q; i.e., the number of days required to reach 70% cell viability were 5, 8, and 3 d for vector, NCS-1, and E120Q groups, respectively (). The expression levels of NCS-1 in each group of neurons before and after adenovirus infection in the absence of B27 trophic supplements are shown in the immunoblot (). Essentially the same results were obtained by counting neurons with condensed nuclei using Hoechst staining (), thus reinforcing the finding that the expression of NCS-1 protects neurons from cell death under apoptotic conditions. Furthermore, when B27 trophic supplement was kept in the culture medium (which is a less stress condition), the dominant-negative effect of E120Q was more clearly observed when compared with the vector control group (; also see A, where some blebs were observed in the neurons infected with E120Q mutant). In other preliminary experiments (not depicted), although the time course of the loss in cell viability was variable, overexpression of NCS-1 consistently delayed the loss of cell viability when B27 supplements were omitted, and the expression of E120Q always increased the rate of cell death when B27 was present. These results suggest that endogenous NCS-1 is playing an important role in keeping the long-term survival of cultured neurons under normal conditions in addition to the protective role from stress under apoptotic conditions. A large body of evidence suggests that neuronal survival is promoted by neurotrophic factors such as BDNF and GDNF (). Because the long-term application of GDNF has been reported to enhance the expression of NCS-1 in motor neurons (), we attempted to clarify the role of NCS-1 as a downstream mechanism of GDNF-induced cell survival in rat cortical neurons. When primary cultured cortical neurons were treated with 10 ng/ml GDNF for 2 d after the withdrawal of B27 supplements, neuronal survival was significantly enhanced when compared with time-matched control (, A [top] and B). Interestingly, the expression of NCS-1 mimicked the survival-promoting effects of GDNF; i.e., NCS-1 exerted a robust survival effect even in the absence of GDNF (, A [middle] and B). Most strikingly, the expression of the dominant-negative NCS-1 mutant E120Q largely prevented cell survival induced by GDNF (, A [bottom] and B). Immunoblot analysis revealed that the application of 10 ng/ml GDNF resulted in a significant increase in the expression level of endogenous NCS-1 within 10 min, which further increased at 40 min in these neurons (). The amount of NCS-1 remained elevated through 2 d of exposure to GDNF (not depicted). These results show that the treatment of GDNF increases the expression level of NCS-1, which subsequently promotes neuronal survival, suggesting that GDNF-induced neuroprotection is at least in part mediated by NCS-1. In neurons, GDNF has been reported to promote cell survival via activation of signaling cascades involving the PI3-K–Akt pathway (; ). In accordance with these studies, we also observed that exposure of primary cultured cortical neurons to GDNF resulted in a large increase in phospho-Akt levels (, left). Therefore, it was of interest for us to test whether NCS-1 also activates this kinase. We examined the effect of NCS-1 expression or its dominant-negative form on Akt phosphorylation in the presence or absence of B27 trophic supplements. When B27 trophic supplements were absent, the expression of NCS-1 significantly enhanced the phosphorylation of Akt, whereas expression of the dominant-negative mutant E120Q had little effect when compared with control vector-infected neurons (, A [middle] and B). On the other hand, when B27 supplements were present, a relatively high level of phosphorylated Akt was observed in the vector-treated control group (, A [right] and B). Additional expression of exogenous NCS-1 further increased the Akt phosphorylation level, whereas expression of the dominant-negative mutant suppressed phosphorylation (, A [right] and B). Thus, the phosphorylation levels of Akt in each group of neurons were well correlated with their viabilities, as shown in . In addition, pretreatment of cultured cortical neurons with LY294002, an inhibitor of PI3-K, completely abolished both GDNF- and NCS-1–induced neuronal survival, whereas PD98059, an inhibitor of MAPK kinase (MEK), did not (). These results suggest that the NCS-1–induced survival-promoting effect is mediated via the PI3-K–Akt pathway but not the MAPK pathway in cultured cortical neurons. To further understand the upstream mechanism of the NCS-1–induced activation of Akt, we next examined the effect of overexpression of NCS-1 and E120Q on the subcellular localization of Akt/PKB in living cells. Akt/PKB is known to be translocated to the plasma membrane when it is fully activated upon phosphorylation and bound with its substrates PtdInsP and PtdInsP (). We constructed the GFP-tagged pleckstrin homology (PH) domain of Akt/PKBα (EGFP-Akt/PKB-PH) and transiently cotransfected it into CCL39 cells, which express a small amount of endogenous NCS-1, together with NCS-1, E120Q, or empty vector. 2 d later, the subcellular localization of EGFP-tagged Akt/PKB-PH was assessed on a confocal microscope. Akt/PKB-PH was diffusely localized in the cytosol of vector-transfected control cells (). Interestingly, Akt/PKB-PH became localized in the peripheral region of cells when NCS-1 was coexpressed, but this peripheral localization was abolished when E120Q was coexpressed (). Qualitatively similar results were also obtained when primary cultured cortical neurons were treated with the same vectors; i.e., Akt/PKB-PH was localized in the peripheral regions of neurons when NCS-1 was overexpressed, but a more diffuse localization pattern was observed when E120Q was overexpressed (). The distribution pattern of Akt/PKB-PH in vector-transfected neurons was similar to that of NCS-1–overexpressing cells (not depicted). These results strongly demonstrate that NCS-1 increases the levels of plasma membrane PtdInsP and PtdInsP and, thus, activates Akt/PKB in living cells. We examined the effects of the overexpression of NCS-1 and its dominant-negative mutant on the survival of these neurons to clarify the physiological role of NCS-1 in injured motor neurons in vivo. One side of vagus nerves of adult rats were axotomized as previously described () and infected with adenoviral vectors encoding NCS-1, E120Q, or EGFP vector alone, and neuronal degeneration was evaluated by histological analysis. 1 wk after axotomy, nearly 30% of nerve cells were found to be EGFP positive in the injured side (). There were clear differences in the staining pattern between control and injured sides for all groups, probably because the regeneration process, such as activation of the surrounding glial cells, was ongoing on injured sides. However, the number of surviving motor neurons stained with hematoxylin were not significantly decreased at the injured side for vector-treated DMV sections (, B and D; examples of counted neurons are indicated by black arrows in C). This would probably be the result of natural antiapoptotic mechanisms induced by injury, which exist in mature neurons as previously reported (). Because the expression level of NCS-1 was significantly increased in response to in vivo axotomy (), we hypothesized that NCS-1 may be involved in this antiapoptotic mechanism. If so, blocking of endogenous NCS-1 would reduce this beneficial effect. As expected, the dominant-negative E120Q mutant resulted in a significant loss of neurons in the injured side (), and some TUNEL-positive nuclei were also detected only in this group (, arrows; and its magnified image in F). Considering that the infection efficiency was only ∼30% in these experiments, a large majority of neurons successfully infected with E120Q appear to have undergone apoptosis. Infection of neurons with the functional NCS-1 adenovirus only had a modest effect on neuronal survival. This probably results from both the low infection efficiency and the high levels of endogenous NCS-1 expression in axotomized neurons () because the NCS-1 effects are already close to maximum. Thus, the dominant-negative mutant E120Q inhibited the survival of adult DMV neurons from axotomy-induced injury, strongly suggesting that NCS-1 is one of the important factors mediating neuronal survival after in vivo axotomy. Numerous stressors, including physical or chemical injury and genetic abnormalities, lead to neuronal degeneration by programmed cell death along an apoptotic pathway. Under these conditions, some intrinsic and extrinsic factors, including neurotrophic factors, are known to activate the antiapoptotic process to rescue neurons from death. However, the signaling pathway leading to cell survival is not yet completely understood. In this study, we identified a novel function for the Ca-binding protein NCS-1, which promotes the long-term survival of cultured neurons via PI3-K–Akt signaling pathways; mediates, at least in part, GDNF-induced neuroprotection; and is up-regulated in response to axonal injury and plays an important role in the antiapoptotic mechanism in injured motor neurons. E120Q NCS-1 point mutant was generated with a conventional PCR protocol using the wild-type rat NCS-1 (GenBank/EMBL/DDBJ accession no. ) as a template and was sequenced to confirm the mutation. Akt/PKBα cDNA was cloned from the human kidney cDNA library (CLONTECH Laboratories, Inc.), and NH-terminally tagged fluorescent protein EGFP-Akt/PKB-PH was constructed incorporating a fragment of 750 bp, encoding the first 250 amino acids of PKBα (containing the PH domain) into EGFP-vector as described previously (). Adenovirus containing wild-type NCS-1 and the E120Q mutant inserts were generated by cotransfecting either of these plasmids and pBHG11 (Microbix Biosystems, Inc.) into HEK 293 cells. Viral DNA was isolated from the supernatant in the wells displaying the cytopathic effect. Replication-incompetent virus containing DNA inserts were plaque-purified twice and grown on HEK 293 cells to produce large amounts of adenovirus. Tissue culture supernatant containing adenovirus was concentrated by centrifugation over cesium chloride. The titers of viral stocks were 2.2 × 10 pfu/ml for EGFP–NCS-1, 1.1 × 10 pfu/ml for EGFP-E120Q, and 2.2 × 10 pfu/ml for EGFP-vector. PC-12 cells stably transfected with vector alone or vector containing cDNA coding for the wild-type NCS-1 (several clones) were grown onto collagen-coated (500 μg/ml of type I; Sigma-Aldrich) culture dishes in growth medium (DME containing 10% horse serum, 5% FBS, and 400 μg/ml geneticin and gentamicin) as described previously (). When cells became 80% confluent, they were switched to the differentiation medium (growth medium with half serum) supplemented with 100 ng/ml NGF-7S (Invitrogen). Primary culture of cortical neurons was performed using the cortex from Sprague-Dawley rats at embryonic day 18. In brief, cortical tissues were isolated from whole brain, minced into small pieces, and digested for 10 min at 37°C in a 20-U/ml papain solution containing 0.002% DNase I (Worthington Biochemical Corp.). After titration of the enzymatic activity, cells were mechanically dissociated by several passages through pipette tips. After centrifugation, cells were resuspended in neurobasal medium supplemented with B27 trophic factors (both from Invitrogen), whose compositions were reported previously (). They were then plated onto culture dishes coated with 0.1% polyethylenimine at a density of 2.5–5 × 10 cells/cm for cell survival assay and 10 cells/cm for immunoblot analysis. CCL39 cells and primary cultured rat cortical neurons were plated onto collagen-coated glass coverslips and cultivated for 1 d. They were then transiently transfected with the EGFP-Akt/PKB-PH construct together with either NCS-1, E120Q, or pCDNA3 (1:3 ratio) using LipofectAMINE 2000 (Invitrogen) and were subjected to immunocytochemistry. In brief, cells were fixed with 4% PFA, permeabilized with 0.2% Triton X-100, and blocked with 5% BSA. They were then incubated for 1 h with anti–NCS-1 antibody (1:200) followed by incubation with secondary antibodies (FITC- or rhodamine-conjugated goat anti–rabbit IgG; 1:200; Jackson ImmunoResearch Laboratories). After extensive wash with PBS, cell images were scanned on a laser confocal microscope (MRC-1024K; Bio-Rad Laboratories) or obtained with conventional epifluorescence illumination (BX50WI; Olympus) with a cooled CCD camera (CoolSNAP; Photometrics) using a 0.9-W 60× water immersion objective lens. Immunocytochemistry for PC-12 cells were also performed in the same way. The primary cultured neurons were infected with viruses at a multiplicity of infection of 100 pfu/cell at 5 d after plating. Under this condition, we found that nearly 70% of the neurons were infected by monitoring EGFP fluorescence (). The number of living neurons was counted within the fixed area of images taken by a digital camera (Coolpix 4500; Nikon). To count the number of cells always within the same area, a grid seal with numbering (Asahi Techno glass) was stuck on the bottom of each culture dish. The number of living neurons remaining at each day was expressed as a percentage of the initial number. Neurons showing the degenerating stage characterized by nuclear condensation, membrane blebbing, or extensive neurite fragmentation were excluded. Four different regions were selected from one dish, and six separate experiments were performed for each condition. To identify and quantify apoptotic neurons, cells were fixed with 4% PFA and were stained with Hoechst 33258. Coverslips were mounted onto glass slides, and cells were observed under epifluorescence illumination on an inverted microscope (1X71; Olympus) using a 40× NA 1.35 oil immersion objective lens (Olympus). Cells were considered apoptotic if their nuclear chromatin was condensed or fragmented, whereas cells were considered viable if their chromatin was diffusely and evenly distributed throughout the nucleus (). DMV tissue samples were obtained by scratching the DMV neurons from several frozen sections of brainstem (described in the next section) using pulled glass capillary under the light microscope. These tissue samples or cultured cells (PC-12 cells and cortical neurons) were then solubilized in SDS-PAGE sample buffer containing protease and phosphatase inhibitors and subjected to immunoblot analysis using image density software (Scion Image; Scion Corp.) as previously described (). Primary antibodies used were anti–NCS-1 antibody (1:1,000), which was previously described (), and publicly available antibodies: monoclonal anti-GAPDH antibody (1:1,000) obtained from Chemicon as well as antiphospho-Akt antibodies (detectable for the phosphorylation of Thr308 and Ser473; 1:1,000) and anti-Akt antibody (1:1,000; both from Cell Signaling Technology). Secondary antibodies used were HRP-conjugated anti–rabbit and anti–mouse antibodies or a combination of biotinylated anti–rabbit (or mouse) antibodies (Zymed Laboratories) and HRP-conjugated streptavidin (Zymed Laboratories). The method of vagus axotomy was described previously (). In brief, 4–6-wk-old Sprague-Dawley rats were deeply anesthetized with 50 mg/kg pentobarbital, and axotomy of the vagus motor neurons was performed with fine scissors at the unilateral vagus nerve at the neck. Injured neurons were confirmed by detecting the fluorescence of Di-I in the DMV, which had been placed at the proximal cut site of the nerve bundle (). To test the effects of colchicine, an implantable polymer containing 10% (wt/vol) colchicine was made by mixing colchicine with ethylene-vinyl acetate copolymer (Elvax) followed by drying as described previously (). Solid slices (∼1 mm) were placed around the unilateral vagal nerve to allow the continuous release of colchicine from slices. The skin incision was closed, and rats were returned to the cage after awaking from the anesthetic. Neuronal degeneration was evaluated by counting surviving neurons as described previously () as well as by TUNEL staining using the apoptag peroxidase in situ Apoptosis Detection Kit (Chemicon). In brief, in vivo axotomy was performed as described above, and, at the same time, adenoviral vectors carrying EGFP only, EGFP plus NCS-1, or E120Q (10 pfu each) was injected into the stump of the nerve using a 34-gauge needle. 1 wk after axotomy, paraffin-embedded serial sections (3–4 μm) were made from the brainstem. After they were deparaffinized, sections were directly stained with hematoxylin/eosin to visualize the structure of the DMV region. TUNEL staining was performed in accordance with the manufacturer's method. The sections were lightly counterstained with methyl green. Control sections were treated similarly but incubated in the absence of TdT enzyme. To confirm whether the adenoviral vectors were transferred to the DMV neurons, another set of animals were treated in the same way. 8-μm frozen sections were cut 1 wk after operation, and EGFP signals were viewed under a fluorescence microscope (1X71; Olympus). All image acquisitions were performed at room temperature, and images were subsequently processed using Adobe Photoshop (version 7) and Adobe Illustrator (version 10) software. All experiments conformed to the Guiding Principles for the Care and Use of Animals approved by the Council of the Physiological Society of Japan. All efforts were made to minimize the number of animals used and their suffering. Comparisons between two groups were performed using the paired or unpaired test. Values of P < 0.05 were considered statistically significant. All summarized data are expressed as means ± SEM.

It is estimated that each year >7 million people develop chronic nonhealing wounds, including pressure, leg, and diabetic ulcers and burns, in the United States. These wounds require long-term care that is labor intensive and costly. Delayed wound healing among the elderly in the United States, for instance, is estimated to cost >$9 billion each year (). Although tremendous efforts were made on the development of recombinant growth factors (GFs) and organotypic skin equivalents, the overall outcomes of GF treatments or the use of skin substitutes, such as xenografts, have not generated satisfactory cost-effective benefits (; ). Few of the GFs have ultimately received approvals from the Food and Drug Administration. Therefore, there is a pressing need to better understand the fundamentals of the skin wound-healing processes. Skin wound healing is a complex process involving collaborative efforts of multiple types and lineages of skin cells, ECMs, and soluble GFs. Inflammation, reepithelialization, tissue formation, and tissue remodeling are proposed sequential events to heal skin wounds (; ). Abnormalities in any of the events could result in nonhealing wounds or healed wounds with hypertrophic scars (). Throughout these processes, cell motility control is critical. The epidermal cells, largely keratinocytes, laterally migrate across the wound bed from the cut edge to resurface the wound in the process known as reepithelialization. The human dermal cells, including dermal fibroblasts (DFs) and dermal microvascular endothelial cells (HDMECs), move into the wound to produce and deposit large amounts of matrix proteins, to contract and remodel the wound, and to build new blood vessels. Thus, it is critical to understand what cells move into the wound first, second, or third and what mechanism orchestrates the order of the multitype skin cell motility during wound healing. In unwounded skin, the resident skin cells are nourished by a filtrate of plasma. When skin is wounded, the resident cells in the wound encounter an acute transition from an initial stage of plasma to a stage of serum for the first time. As the wound heals and subsequent wound remodeling initiates, the resident cells experience a transition from plasma back to serum. In fact, the plasma→serum→plasma transition coincides with the classical phases of skin wound healing, as mentioned in the previous paragraph. There have been few studies that define the physiological function of this transition in the wound repair. In addition, the full ingredients in wound fluid may be more complex than those in plasma or serum. For instance, it should also contain released factors from inflammatory leukocytes and even from the resident skin cells (). In particular, the inflammatory cells and factors have long been proposed to play important roles in the repair process. However, recent studies suggest that inflammation, which is a necessary mechanism of defense in adults, is not only dispensable for wound healing but rather harmful to the purposes of fast healing and less scaring. First, embryos, in which no inflammation takes place, heal wounds perfectly without a scar (). Second, Smad3 and Pu.1 knockout mice cannot mount an inflammatory response; however, the reepithelialization and wound healing occur faster than their wild-type littermates and show less scaring (; ). We recently reported that human serum, but not human plasma, promotes human keratinocyte (HK) migration (). This suggested, for the first time, that the plasma to serum transition differentially regulates skin cell motility. In the present study, we studied the effects of plasma versus serum on the motility of three primary human skin cell types: DFs, HDMECs, and HKs. Our results suggest that the plasma→serum→plasma transition serves as a “traffic control” for the dermal and epidermal cell motility, in which TGFβ3 in serum acts as the “traffic controller” and the cell surface levels of type II TGFβ receptor (TβRII) operate as the “sensor” to determine the order of skin cell migration. Human DFs, HDMECs, and HKs are the three major types of skin cells involved in wound healing. During the wound healing, either human plasma or serum represents the main source of soluble GFs in the wound fluid. Therefore, we studied the migratory responses of these three cell types to human plasma and serum. The individual cell motility-based colloidal gold motility assay () and a modified in vitro wound-healing assay () were performed side by side. In the colloidal gold motility assay, cells are allowed to attach and migrate on an ECM-precoated colloidal gold surface in the presence of a cell proliferation inhibitor. As the cells move, they leave behind migration tracks that can be visualized under a microscope and added as a migration index (MI) by computer-assisted analysis (). As shown in , in the absence of plasma or serum, all three types of skin cells modestly migrated on collagen (a, d, and g) and generated MIs (see Materials and methods) from 4 to 8. This is consistent with our previous study reporting that ECMs alone can initiate cell migration in the total absence of soluble GFs (). The addition of human plasma strongly stimulated DF and HDMEC migration (), generating MIs >25. The mean size migration tracks were marked with open circles. In contrast, human serum showed little stimulation of DF and HDMEC migration () with MIs between 7 and 10. However, as previously described (), plasma and serum showed opposite effects on HK migration. Plasma had a weak stimulation of HK migration (), with MIs of 9–11. In contrast, serum robustly stimulated HK migration, which left behind long and linear tracks () with MIs of 28–30. The results of the in vitro wound-healing assay, which measures the averaged and directional migration of a cell population, confirmed the findings. The unclosed distance in multiple wounded areas in each cell culture well was measured and quantified as an average gap (AG). As shown in , plasma but not serum stimulated wound closure by DFs () and HDMECs (). In contrast, human serum but not plasma stimulated complete closure of the wounded area of HKs (). Therefore, the aforementioned data suggest that: plasma contains a promotility factor for DFs and HDMECs but not for HKs; as plasma converts to serum, serum acquires new factors that specifically stimulate HK migration; and serum stops stimulation of the dermal cell migration by losing the plasma promotility factors, by gaining antimotility factors, or both. Therefore, we, investigated these possibilities. It is known that in human plasma, the levels of all three TGFβ isoforms are either low, such as TGFβ1, or undetectable, such as TGFβ2 and TGFβ3. In human serum, however, TGFβ1, TGFβ2, and TGFβ3 are all elevated to mean values of 30, 1–1.5, and 1–2 ng/ml, respectively (; ; ; ; ). We tested whether depletion of the TGFβ function in serum would convert serum to plasma, regaining its promotility effect on the dermal cells. For simplicity, only the MIs of the colloidal gold cell migration experiments are presented. As shown in , plasma (bar 2) but not serum (bar 3) stimulated DF migration over the serum-free control (bar 1). However, the addition to serum of increasing amounts of pan–anti-TGFβ–neutralizing antibody, which neutralizes all three human TGFβ isoforms, completely converted the serum to a promotility stimulus of DF migration just like plasma (, bars 4–6 vs. bar 3). To determine further which TGFβ isoform was specifically responsible for the inhibitory effect in serum, we made use of three isoform-specific neutralizing antibodies under the concentrations that block a similar amount of TGFβ1, TGFβ2, or TGFβ3 but show little cross-reacting activities. To our surprise, the anti-TGFβ1 antibody failed to eliminate the inhibitory effect of serum on DF migration even at 2.0 μg/ml, which is capable of blocking >50 ng/ml TGFβ1 (, bars 7–9 vs. bar 3). Just like the anti-TGFβ1 antibody, the addition of an anti-TGFβ2 antibody did not relieve the serum's inhibitory effect on DF migration (, bars 10–12 vs. bar 3). In contrast, the addition of an anti-TGFβ3 antibody completely converted serum to a plasmalike promotility stimulus for DFs (, bars 13–15 vs. bars 2 and 3). The importance of TGFβ3 was confirmed by in vitro wound-healing assays. As shown in , the addition of increasing amounts of anti-TGFβ3 antibody to serum eliminated its inhibitory effect on DF (c–e vs. b) and HDMEC (h–j vs. g) migration in a dose-dependent manner, resulting in closure of the wounded areas. In contrast, the addition of the antibodies to serum had neither stimulatory nor inhibitory effects on the closure of the wounded HK cell monolayer (). Quantitation of the wound closures is shown below the corresponding images as AGs. These data indicate that it is the increased TGFβ3 after plasma to serum transition that selectively stops dermal cell migration. To gain a reciprocal support of this finding, we compared the effect of recombinant TGFβ3 on GF-induced migration of DFs, HDMECs, and HKs. As shown in , TGFα-stimulated HK migration (TGFα is used to stimulate HK migration because plasma does not stimulate HK migration and serum already contains TGFβ3) was not affected by the addition of TGFβ3 (). In contrast, plasma-stimulated migration of DFs and HDMECs was blocked by the addition of TGFβ3 in a dose-dependent manner (). Therefore, we conclude that along the steps of the plasma→serum→plasma transition during wound healing, TGFβ3 acts as an on-and-off switch to separate the migration of dermal cells from that of epidermal cells. To investigate the molecular basis by which the migratory responses of dermal and epidermal cells naturally divide them into either TGFβ-sensitive or TGFβ-insensitive cells, we examined the profiles of three TβRs—TβRI/ALK5, TβRII, and TβRIII/betaglycan—in DFs, HDMECs, and HKs as well as in melanocytes (MCs). As shown in , all four cell types expressed comparable levels of TβRI (a), with a similar level in HKs and DFs (lanes 1 and 3) and a slightly lower level in MCs and HDMECs (lanes 2 and 4). Among the four cell types, only DFs showed a detectable level of TβRIII/betaglycan (; lane 3). Anti–β-actin antibody blotting of the lower part of the same membranes () was used as the control for equal sample loading and the references for densitometry scan quantitation (ratio of TβR/β-actin). It can be seen that there is no apparent correlation between TβRI and TβRIII expression (note that HDMECs show no expression of TβRIII and yet are sensitive to TGFβ) and TGFβ sensitivity between the dermal and epidermal cells (the migration of MCs responds to plasma and serum just like HKs; not depicted). In contrast, there is a clear correlation between the levels of TβRII expression and the sensitivity of dermal and epidermal cell migration to TGFβ. The TGFβ-sensitive DFs and HDMECs exhibited 5–12-fold higher levels of TβRII than TGFβ-insensitive HKs and MCs (; lanes 3 and 4 vs. lanes 1 and 2) based on their anti–β-actin antibody blotting controls (). To confirm the TβR expression patterns in human skin, sections of frozen normal human skin were subjected to indirect immunofluorescence staining with the corresponding anti-TβR antibodies, DAPI staining (to strain the nuclei); and phalloidin (to stain F-actin) or IgG control (control for anti-TβR antibodies). As shown in , the control IgG antibodies showed little staining in either the epidermal or the dermal areas (e, j, and o). In contrast, anti-TβRI antibody staining showed the expression of TβRI in both the epidermis and dermis (). This staining colocalized with phalloidin (membrane and cytosol) but not DAPI (nuclei), resulting in the yellow/orange-colored area when the images were merged (). Consistent with the Western blot data, the anti-TβRII antibody staining clearly showed stronger expression of the TβRII in the dermis than the epidermis (). The staining colocalized with phalloidin, resulting in a merged yellow/orange color in the dermis and less in the epidermis (). As expected, anti-TβRIII antibodies almost entirely stained the dermis (). No merged yellow/orange color was detected in the epidermis (). We questioned whether the difference in TβR profiles affects the TGFβ-stimulated phosphorylation of Smad between dermal and epidermal cells. As shown in , the dose-dependent analyses showed little differences in TGFβ-induced Smad phosphorylation (ser-465/467) in all three cell types, in which the induced Smad phosphorylation could be clearly detected after stimulation with 0.1–0.5 ng/ml TGFβ3 (lanes 3 and 4). Moreover, as shown in , we also detected similar kinetics of TGFβ3-induced Smad phosphorylation in both epidermal (HKs) and dermal cells (DFs and HDMECs). Close analyses of the data from multiple experiments revealed a relatively sustained Smad phosphorylation in HKs and a transient Smad phosphorylation, which declined after 45 min of TGFβ3 stimulation in DFs and HDMECs (). Therefore, the insensitivity of HKs to the antimotility effects of human serum or TGFβ3 was not caused by a complete lack of or a significant difference in TGFβ3-stimulated Smad phosphorylation. The critical question was whether the differences in TβR expression, especially TβRII, account for the differential migratory responses of the dermal and epidermal cells to plasma and serum. We took the following approaches: to increase the TβR levels in epidermal cells to similar levels in the dermal cells and to down-regulate the TβR expression in the dermal cells to similar levels seen in the epidermal cells. Then, we tested whether these changes would convert the TGFβ-insensitive epidermal cells to TGFβ-sensitive dermal cells and vice versa. To overexpress the TβR of interest, we chose to use lentiviral vector pRRLsinhCMV (,). Using the EGFP gene as the marker, as shown in , we provide evidence that this gene delivery system offers >90% gene transduction efficiency in all four human skin cell types. To study the role of TβRII, we focused on HKs, which showed a much lower level of TβRII than dermal cells, DFs, and HDMECs. As shown in , HKs express a lower level of TβRII than DFs or HDMECs (a; lane 2 vs. lanes 1 and 3). However, after pRRLsinhCMV-TβRII virus infection, the TβRII expression in HKs was increased to a similar level seen in DFs and HDMECs (, lane 5 vs. lanes 1 and 3). The kinase-defective TβRII mutant (TβRII-KD) was included as the control for specificity. As shown in , the exogenously expressed TβRII-KD (lane 3) was equivalent to the exogenously expressed wild-type TβRII (lane 2). As expected, overexpression of TβRII-KD but not wild-type TβRII blocked TGFβ3-stimulated Smad phosphorylation (, lanes 5 and 6 vs. lanes 3 and 4 and lanes 1 and 2). To study TβRIII, we infected HKs with pRRLsinhCMV-TβRIII virus. As shown in , the TβRIII expression in HKs was elevated from a nondetectable level in uninfected HKs (lane 3) to a similar level observed in DFs (lane 4 vs. lane 2). DFs are the only skin cell type expressing a detectable level of TβRIII. These TβRII and TβRIII expression-equalized HKs were subjected to migration assays in response to plasma, serum, or TGFα in the absence or presence of TGFβ3. As shown in , migration of the parental HKs was unchanged by plasma (bar 2 vs. bar 1), dramatically stimulated by serum or TGFα (bars 3 and 4), but not blocked by the addition of TGFβ3 (bar 5 vs. bar 4). In contrast to the parental HKs, the TβRII-overexpressing HKs responded to serum as a strong antimotility reagent (, bar 8 vs. bar 3). Moreover, TGFβ3 completely blocked the TGFα-stimulated migration of these cells (, bar 10 vs. bar 9). However, overexpression of the TβRII-KD mutant did not show the same effect. The migration of TβRII-KD–HK cells, like the parental HKs, could still be stimulated by serum (, bar 13 vs. bar 3), and the TGFα-stimulated migration remained refractory to the presence of TGFβ3 (, bar 15 vs. bar 4). The migration of TβRIII-overexpressing HKs was slightly inhibited by serum (, bar 18) in comparison with their parental counterpart (, bar 3; P < 0.001). TGFβ3 also partially blocked TGFα-driven migration of the cells (, bar 20 vs. bar 5; P < 0.005). It is possible that TβRIII overexpression allows HKs to have more access to the TGFβ3 that is bound to the exogenously overexpressed TβRIII. Based on our previous study that TGFβ selectively blocks the proliferation but not migration of HKs in response to GFs (), the aforementioned finding indicates that it is the lower TβRII level that determines HKs' sensitivity to the antiproliferation but not the antimotility signal of TGFβ3. To further verify the key role of TβRII, we generated TβRII- as well as TβRI-overexpressing MCs because the parental MCs express the lowest or undetectable level of TβRI and TβRII among the four human skin cell types (). As shown in , the parental MCs expressed a lower level of TβRI (lane 1) than both HKs (lane 3) and DFs (lane 4). However, the pRRLsinhCMV-TβRI virus-infected MCs showed an increased expression of TβRI (, lane 2) similar to the TβRI expression in HKs and DFs. Similarly, as shown in , parental MCs expressed low or undetectable levels of TβRII (, lane 1). pRRLsinhCMV-TβRII virus infection increased the TβRII expression in MCs (, lane 2) to a similar level seen in DFs (, lane 3). We then tested the migratory response of these cells to plasma and serum. As shown in , parental MC migration was equally stimulated by plasma and serum (bars 2 and 3 vs. bar 1) and was not sensitive to TGFβ3 (bar 4). The increased expression of TβRI in MCs had no effect on cell migration in response to plasma (, bar 6), serum (, bar 7), or plasma plus TGFβ3 (, bar 8). In contrast, the migration of TβRII-overexpressing MCs became converted from TGFβ insensitive to TGFβ sensitive, in which serum strongly inhibited migration of the cells (, bar 11 vs. bar 3). In addition, TGFβ3 completely blocked the plasma-induced migration (, bar 12 vs. bar 4). To reversely confirm the determining role for TβRII levels of expression, we wanted to down-regulate the endogenous TβRII expression in DFs and HDMECs to the similar level observed in HKs and tested whether these TβRII–down-regulated dermal cells would become refractory to the antimotility effect of serum and TGFβ3 just like the epidermal HKs and MCs. The FG-12 system, a lentiviral backbone carrying a short inhibitory RNA (siRNA) expression cassette (), was used to transduce siRNA against human TβRII into DFs and HDMECs. To select the optimal virus titer, we infected DFs and HDMECs with various dilutions (vol/vol) of the original FG-12–TGFβ-RII–siRNA virus stock (∼4–7 × 10 transduction units/ml). As shown in , infection of DFs (a) with 50% of the original virus stock (lane 3) reduced the TβRII expression in DFs (lane 6) to the similar level observed in HKs (lanes 1). Similar results were obtained in HDMECs with 50% of the virus stock (, lane 4 vs. lane 1). The dilutions of the virus stocks did not cause any significantly compromised transduction efficiency. As shown in , the transduction efficiencies for 50% (1:1) or less dilutions were all >90%, as indicated by the coexpressed GFP for both HDMECs (a–c) and DFs (e–g). Only 25% (1:4) or higher diluted viruses showed a significant decrease in the transduction efficiency (). Therefore, we chose 50% diluted virus stock to infect DFs and HDMECs for further migration assays. As shown in , in the vector- and control lac-Z–siRNA-infected DFs, plasma stimulated cell migration (bars 2 and 6), whereas serum and TGFβ3 inhibited cell migration (bars 3, 4, 7, and 8). However, in TβRII–down-regulated DFs, neither serum nor TGFβ3 was able to inhibit the cell migration (, bars 11 and 12 vs. bars 3 and 4). Instead, the cells responded to both plasma and serum as promotility agents (, bars 11 and 12). Similarly, in HDMECs, down-regulation of TβRII converted the cells from serum- and TGFβ3-sensitive cells to TGFβ3-insensitive cells just like epidermal HKs and MCs. As shown in , migration of the HDMEC–RII–siRNA cells became insensitive to the antimotility effect of serum or TGFβ3 (, bars 11 and 12 vs. bars 7 and 8). Moreover, both plasma (, bar 10) and serum (, bar 11) became promotility agents for the cells. In conjunction with our previous study reporting that TGFβ blocks both proliferation and migration of DFs (), these findings again indicate that the levels of TβRII determine the uptake of the antimotility signals, the antiproliferation signals, or both of TGFβ3. This study is a significant departure from previous wound-healing studies using human skin cells (for review see ). First, those studies used bovine pituitary extract (BPE) or FBS as the stimuli of human skin cell migration and proliferation. However, human skin cells are never in contact in real life with BPE or FBS. In fact, neither BPE nor FBS resembles human serum to differentially affect human dermal and epidermal cells. Second, because previous studies often focused on a single type of human skin cell or cell line, little could be learned about coordinated migration of the multitype skin cells, which are simultaneously present in the wound bed in vivo. Because of these new approaches, the results of this study unveil a previously unrecognized function for the naturally occurring plasma→serum→plasma transition along the classical four phases of skin wound healing. A schematic representation of the key findings is shown in . When skin is acutely wounded, whereas plasma is still not fully converted to serum, DFs and HDMECs may first migrate into the wound over provisional ECM. In contrast, the epidermal cell migration (i.e., reepithelialization) has yet to take place because of the lack of HK promotility activity in plasma (step 0). Then, the transition to serum quickly stops dermal cell migration by increasing the TGFβ3 level (first step) and, at the same time, promotes HK migration by newly acquired HK promotility factors, predominantly TGFα (second step; ). The selective inhibition of dermal cell migration by TGFβ3 from serum results from the higher TβRII expression levels in these cells, and, in contrast, the escape of inhibition by the epidermal cells is granted by either the lower (HKs) or undetectable (MCs) TβRII levels in the epidermal cells. After the wound heals, serum transforms back to plasma, which resumes DF and HDMEC migration into the newly healed wound for the final phase of wound healing, remodeling, and angiogenesis (third step). Therefore, this study has provided a mechanism by which the orderly migration of three major skin cell types take place during wound healing. The main TGFβ in human serum was thought to come from platelet degranulation, in which TGFβ1 is the only isoform present in the α granules of platelets (; ). However, human serum but not plasma was reported to contain TGFβ3 at a concentration range of 1–2 ng/ml (; ), which likely comes from the blood leukocytes (; ; ). An increase in TGFβ2 in serum has also been reported (). Because of technical difficulties, there is still a lack of consensus in the exact levels of TGFβ in human plasma and serum (; ; ). First, it is extremely difficult to collect platelet-free plasma and to completely prevent degranulation of the contaminated platelets because platelets tend to resuspend during plasma preparation (; ). Second, a wide range of methods for measuring TGFβ (mostly for TGFβ1) has been used. The bioassay-based measurements selectively measure the amount of the biologically active TGFβ, and, therefore, they provide data on the “operational” TGFβ (; ). On the other hand, the antibody-based ELISA assays detect the total amount of free TGFβ. In the latter case, during the dilution steps of these assays, some TGFβ peptides dissociate from their noncovalently bound complexes and become detectable. This dissociation could also cause nonlinear curves of the assays (). Third, because the half-life of an active TGFβ peptide is only 2–3 min, it is difficult to determine whether the detected amount of TGFβ represents the steady-state level or in vitro broken down level of TGFβ (; ). Finally, the wide variations on age, gender, or racial background of the selected donors for the human subjects also contributed to the variations in detected TGFβ. The three mammalian TGFβ isoforms TGFβ1, TGFβ2, and TGFβ3 bind to and transmit signals via the heteromeric complex of TβRII and TβRI/activin receptor-like kinase (Alk) serine/threonine kinases (; ; ). However, the three mammalian TGFβ isoforms diverge significantly in their potency as growth inhibitors in vitro as a result of differences in receptor recognitions and binding to extracellular antagonists (; ). In vascular endothelial cells and hematopoietic progenitor cells, TGFβ1 and TGFβ3 showed stronger growth-inhibitory effects than TGFβ2 (; ). showed that TGFβ3 was 10-fold more active than TGFβ2 in mesoderm induction in , whereas TGFβ1 had little effect. showed that TGFβ3 but not TGFβ1 and TGFβ2 induced the expression of presenilin-1, a familial Alzheimer's disease–linked gene, in neurons and astrocytes. Moreover, ex vivo wound-healing studies showed that TGFβ1 and TGFβ2 cause a fibrotic scarring response and that TGFβ3 elicits scar-free or regenerative healing responses (for review see ). However, discrepancies have also been reported (; ). Consistent with the divergence in vitro, knockout studies in mice showed that TGFβ1, TGFβ2, and TGFβ3 are not functionally redundant (for review see ; ; ). Unfortunately, skin wound-healing studies using these TGFβ knockout mice were compromised by severe developmental defects in the mice. TGFβ2 and TGFβ3 knockout mice die during or shortly after birth (; ; ). While one half of the TGFβ1 knockout mice are born alive, they undergo early postnatal death (3 wk) as a result of a massive infiltration of inflammatory lymphocytes and macrophages onto several organs. The other half dies in utero because of defects in vasculogenesis and hematopoiesis (; ). Nonetheless, reported little impairment in wound healing in <10-d-old neonatal TGFβ1 knockout (β ) mice and in 30-d-old β mice treated with rapamycin, an immune suppressant. In contrast, reported a 1-wk delay in each of the major phases of wound healing in immunodeficient TGFβ1 knockout mice (β/ mice; ). The reason for the discrepancy is unclear. Our study showed that the TβR profiles, especially the differences in the levels of TβRII, in three major human skin cell types as well as MCs determine their migratory responses to plasma or serum, respectively. How the quantitative differences in the TβRII levels are translated into the distinct signaling outcomes remain to be studied. We detected little differences in Smad activation between the dermal and epidermal cell types in response to various doses of TGFβ3. However, a difference in the kinetics of Smad activation between TGFβ-sensitive DFs and HDMECs and TGFβ-insensitive HKs was reproducibly observed. TGFβ3 appeared to stimulate a more sustained phosphorylation of Smad in the epidermal cells in comparison with the TGFβ3-stimulated kinetics of Smad phosphorylation in the dermal cells. Furthermore, we found that the difference in the kinetics of Smad activation was caused by the difference in TβRII levels between dermal and epidermal cells (unpublished data). Whether this difference accounts for the distinct physiologic responses of dermal versus epidermal cells remains to be studied. It is equally possible that TβRII might mediate the antimotility effect of TGFβ3 via a Smad-independent signaling pathway (). i m a r y h u m a n n e o n a t a l D F s , H D M E C s , H K s , a n d M C s w e r e p u r c h a s e d f r o m C l o n e t i c s . H K s w e r e c u l t u r e d i n t h e E p i L i f e m e d i u m s u p p l e m e n t e d w i t h t h e H K G S K i t ( C a s c a d e B i o l o g i c s ) . M C s w e r e m a i n t a i n e d i n M e d i u m 1 5 4 s u p p l e m e n t e d w i t h h u m a n M C g r o w t h s u p p l e m e n t ( C a s c a d e B i o l o g i c s ) . D F s w e r e c u l t u r e d i n D M E s u p p l e m e n t e d w i t h 1 0 % F B S . H D M E C s w e r e c u l t u r e d i n G F - s u p p l e m e n t e d M e d i u m 1 3 1 ( C a s c a d e B i o l o g i c s ) . T h i r d o r f o u r t h p a s s a g e s o f t h e c e l l c u l t u r e s w e r e u s e d i n a l l e x p e r i m e n t s . H u m a n p l a s m a a n d s e r u m c o l l e c t e d f r o m a v a r i e t y o f d o n o r s w e r e p u r c h a s e d f r o m S i g m a - A l d r i c h . H u m a n r e c o m b i n a n t T G F β 1 , T G F β 2 , a n d T G F β 3 w e r e p u r c h a s e d f r o m R & D S y s t e m s . A n t i – h u m a n T G F β ( 1 , 2 , a n d 3 ) p a n - n e u t r a l i z i n g a n t i b o d y a n d n e u t r a l i z i n g a n t i b o d i e s s p e c i f i c a l l y a g a i n s t h u m a n T G F β 1 , T G F β 2 , o r T G F β 3 w e r e o b t a i n e d f r o m R & D S y s t e m s . H u m a n T β R ( I a n d I I ) c D N A s w e r e o b t a i n e d f r o m T . I m a m u r a a n d K . M i y a z o n o ( J a p a n e s e F o u n d a t i o n f o r C a n c e r R e s e a r c h , T o k y o , J a p a n ) . T β R I I I c D N A w a s a g i f t f r o m J . M a s s a g u é ( M e m o r i a l S l o a n - K e t t e r i n g C a n c e r C e n t e r , N e w Y o r k , N Y ) . T β R I I - K D c D N A w a s a g i f t f r o m K . L u o ( U n i v e r s i t y o f C a l i f o r n i a , B e r k e l e y , B e r k e l e y , C A ) . A n t i b o d i e s a g a i n s t T β R I / A l k 5 , T β R I I , a n d T β R I I I / b e t a g l y c a n w e r e e i t h e r g i f t s o f J . L i ( H o u s e E a r I n s t i t u t e , L o s A n g e l e s , C A ) o r w e r e o b t a i n e d f r o m C e l l S i g n a l i n g T e c h n o l o g y , S a n t a C r u z B i o t e c h n o l o g y , I n c . n d R & D S y s t e m s , r e s p e c t i v e l y . D A P I a n d r h o d a m i n e - c o n j u g a t e d p h a l l o i d i n w e r e p u r c h a s e d f r o m S i g m a - A l d r i c h . T h e F G - 1 2 s y s t e m w a s p r o v i d e d b y I . C h e n ( U n i v e r s i t y o f C a l i f o r n i a , L o s A n g e l e s , L o s A n g e l e s , C A ) . R a t t y p e I c o l l a g e n w a s p u r c h a s e d f r o m B D B i o s c i e n c e s . A n t i p h o s p h o - S a m d 2 ( s e r - 4 6 5 / 4 6 7 ) a n t i b o d y ( A B 3 8 4 9 ) w a s p u r c h a s e d f r o m C h e m i c o n I n t e r n a t i o n a l . A n t i – β - a c t i n a n t i b o d y w a s o b t a i n e d f r o m C e l l S i g n a l i n g T e c h n o l o g y , a n d t h e P l a s m i d M i d i K i t w a s p u r c h a s e d f r o m Q I A G E N . X L - 1 0 G o l d u l t r a c o m p e t e n t c e l l s w e r e p u r c h a s e d f r o m S t r a t a g e n e . A l l o t h e r r e a g e n t s a n d s u p p l i e s , u n l e s s i n d i c a t e d , w e r e o b t a i n e d f r o m V W R o r S i g m a - A l d r i c h .

Proper segregation of chromosomes during mitosis is essential for the accurate transmission of genetic material. Each chromatid of a replicated chromosome assembles a kinetochore, which forms a dynamic interface with microtubules of the mitotic spindle (). To facilitate chromosome segregation, sister kinetochores must attach to and regulate the assembly properties of microtubules emanating from opposing spindle poles. This process, called chromosome biorientation, requires the integrated activities of multiple kinetochore proteins. New insights have come from the discovery that many kinetochore proteins and their associated complexes are widely conserved (). For example, purifications from , humans, and fission yeast have identified a protein network that includes Mis12, KNL-1, and the Ndc80 complex (; ; ). Work in has demonstrated that depletion of a functional class of proteins that includes MIS-12 disrupts kinetochore assembly, resulting in alignment and segregation defects (). In fungi, Mis12 functions in a complex with Nnf1, Nsl1, and Dsn1 (for review see ). Initial studies of the Mis12 orthologue in human cells suggested that it is required for proper chromosome alignment and segregation (). Here, we present functional analyses of three human kinetochore proteins (hNsl1, hNnf1, and hDsn1) that are part of the conserved network. These three proteins function in a discrete complex with hMis12 at inner kinetochores and are required for proper chromosome alignment and cell cycle progression as a result of a central role in kinetochore assembly. Tandem affinity purifications of hMis12 from human cells previously isolated a network of 10 interacting kinetochore proteins including hMis12, hKNL-1, the Ndc80 complex, Zwint, and three additional proteins: Q9H410, PMF1, and DC31 (; ). In contrast, purification of Mis12 from budding yeast identified a complex with just three other interacting proteins (Dsn1, Nsl1, and Nnf1; for review see ). This disparity suggested that hMis12 and the three additional proteins might form a stable subcomplex analogous to the budding yeast complex. To test this possibility, we generated antibodies (Fig. S1, available at ), fractionated mitotic HeLa extracts using gel filtration chromatography, and analyzed their migration relative to hMis12 (). Under low salt conditions, hMis12, PMF1, Q9H410, and DC31 were distributed throughout the column fractions (). Under higher salt conditions, the majority of all four proteins cofractionated with a Stokes radius of ∼4.6 nm. To determine whether the stable association of hMis12, PMF1, Q9H410, and DC31 accounted for this cofractionation, we coexpressed all four proteins in bacteria using a polycistronic system (). The four proteins copurified to near homogeneity through nickel affinity and gel filtration chromatography, indicating that they form a stable complex (). The recombinant complex cofractionated as a single peak nearly coincident with that observed in HeLa extracts at 600 mM KCl (). We estimate that the Svedberg coefficient (S) value is ∼5.8 (), resulting in a native molecular mass of ∼112 kD. The combined subunit molecular mass is 119.7 kD, suggesting that the complex contains a single molecule of each protein. We conclude that hMis12, PMF1, DC31, and Q9H410 are the human equivalent of the Mis12 complex in budding yeast. Nnf1 and PMF1 show weak sequence similarity (unpublished data), and the position of coiled coils and limited regions of identity suggest that Q9H410 and DC31 are analogous to Dsn1 and Nsl1 (). Therefore, we refer to these four proteins as the hMis12 complex and to the three new subunits as hDsn1 (Q9H410), hNnf1 (PMF1), and hNsl1 (DC31). Immunolocalization studies demonstrated that the three new hMis12 complex subunits localize coincidently with centromere protein (CENP) A at inner kinetochores and internally to Ndc80 at outer kinetochores (), which is consistent with previous studies of hMis12 (; ). Punctate nuclear localization was observed in only a subset of interphase cells, suggesting that the hMis12 complex is not constitutively present at centromeres throughout the cell cycle (not depicted). All of the hMis12 complex proteins localized to kinetochores at constant levels throughout mitosis, which is consistent with this complex being a stable component of the mitotic inner kinetochore (not depicted). To analyze the consequences of hMis12 complex inhibition, we transfected HeLa cells with gene-specific siRNAs. Quantification of kinetochore fluorescence intensities indicated that hMis12 complex constituents could be depleted by 73–93% (Table S1, available at ). Because of variable depletion levels from cell to cell in a population, immunoblot analyses indicated that hDsn1 and hNnf1 were depleted by >80–90%, whereas hMis12 and hNsl1 were depleted by ∼50% (Fig. S1). To generate an alternative genetic means to examine vertebrate Mis12 depletion, we also obtained a conditional loss-of-function mutant for Mis12 in chicken DT40 cells (Fig. S2). Cell growth was sustained in chMis12-deleted cells by the expression of chMis12 cDNA under the control of a tetracycline (tet)-repressible promoter. chMis12 protein was significantly reduced 18 h after the addition of tet and was undetectable by 24 h (Fig. S2). RNAi-mediated depletion of hMis12, hDsn1, hNnf1, or hNsl1 resulted in a dramatic reduction in the kinetochore levels of the other three proteins (). This reduction was caused by alterations in both kinetochore targeting and, in some cases, stability of the assayed protein (). Controlled depletion of chMis12 also resulted in the delocalization of chDsn1, chNnf1, and chNsl1 (). Thus, the vertebrate Mis12 complex proteins are interdependent for kinetochore localization and exhibit interdependent stability, which is consistent with their function as a complex at inner kinetochores. Reduced levels of Mis12 complex proteins resulted in chromosome alignment defects in both human and chicken cells, although spindle bipolarity was not perturbed (). To analyze mitotic chromosome dynamics, we transfected siRNAs into HeLa cells stably expressing YFP-histone H2B () and performed live imaging. Cells depleted of hDsn1, hNnf1, hNsl1, or hMis12 remained in mitosis for a mean of 8 h, whereas control cells completed mitosis in 1 h (; and Video 1, available at ). Previous studies indicated that chromosome misalignment also led to a mitotic delay after hDsn1 depletion () but not after hMis12 depletion (). Although chromosomes were observed making multiple attempts to congress, they never fully aligned, and some cells underwent aberrant segregation or exhibited abnormal chromatin morphology (; and Videos 2 and 3). Similarly, the depletion of chMis12 resulted in a significant accumulation of cells with unaligned chromosomes (Fig. S2). Kinetochore levels of the spindle checkpoint protein BubR1 were reduced by 42% in depleted HeLa cells treated with nocodazole ( and Table S1). These findings demonstrate that the Mis12 complex is required for normal chromosome alignment and segregation. Although the spindle checkpoint is active after hMis12 complex inhibition, the decreased BubR1 levels and occasional chromosome missegregation indicate that the checkpoint may be partially compromised. To characterize the chromosome alignment defects after hMis12 depletion, we performed RNAi in HeLa cells stably expressing YFP-labeled CENP-A () and imaged kinetochore motility at high temporal resolution. In control cells, paired sister kinetochores congressed to the metaphase plate and oscillated until anaphase onset (; and Video 4, available at ). In depleted cells, some kinetochore pairs remained at the spindle poles for the duration of the analysis. Strikingly, sister kinetochores were frequently observed moving back and forth between spindle poles and the metaphase plate (; and Video 5). These defects were observed in every cell depleted of any subunit of the hMis12 complex (Videos 5–8). In total, these findings indicate that upon inhibition of the hMis12 complex, sister kinetochores are defective in both the establishment and maintenance of bioriented kinetochore microtubule attachments, which is consistent with previous studies in fungi and (; ; ; ). These defects in biorientation prompted us to analyze the stability of kinetochore microtubule attachments. Cells were incubated in ice-cold media before fixation to depolymerize all microtubules except stable kinetochore fibers (). Compared with controls, the kinetochore fibers in depleted cells were significantly diminished (). In addition, chromosomes located near the spindle poles lacked any associated microtubule staining, indicating that no stable end-on kinetochore microtubule attachments were present on these chromosome populations (). Overall, the reduction in kinetochore fiber stability following inhibition of the hMis12 complex was less severe than that following inhibition of the Ndc80 complex (; ). As an additional assessment of kinetochore microtubule attachments, we measured the distance between sister kinetochores, which correlates with the tension exerted across the centromere (). We restricted our analysis to aligned bioriented chromosomes with end-on attachments. We found that the mean interkinetochore distance was decreased from 1.42 μm in control metaphase cells to 1.11–1.20 μm in depleted cells (P < 0.001; > 100 chromosomes from five cells per condition; ). This 30–40% reduction in centromere stretch relative to the rest length of 0.64 μm (measured in control prophase cells) suggests that the net force produced by kinetochore microtubule attachments is diminished upon depletion of hMis12 complex proteins. Our previous study in suggested that MIS-12 regulates outer kinetochore assembly (). Thus, we examined localization of the Ndc80 complex, which physically associates with the Mis12 complex (; ) and is a major constituent of the outer plate (). In cells with greatly reduced levels of Mis12 complex subunits, the kinetochore localization of endogenous Ndc80 () and stably expressed GFP-hNuf2 (not depicted) was significantly lower than in control cells. Measurements of kinetochore fluorescence intensity indicated that after the depletion of any of the four Mis12 complex components, Ndc80 targeting was reduced a mean of 61–82% (Table S1). Chicken DT40 cells depleted of Mis12 also showed a significant reduction in kinetochore-localized Ndc80 (Fig. S2). These results suggest that the outer kinetochore is disrupted in cells depleted of the Mis12 complex. In support of this idea, CENP-E targeting was diminished in depleted HeLa cells, and expansion of the fibrous corona into crescent-shaped collars, which is normally observed upon nocodazole treatment, did not occur (). Because CENP-E and the Ndc80 complex both contribute to the formation of stable, bioriented kinetochore microtubule attachments, the role of the hMis12 complex in chromosome biorientation may be explained, in part, by its requirement for the recruitment of outer kinetochore components. To determine whether inner kinetochore assembly is also affected in cells depleted of hMis12 complex subunits, we quantified the levels of CENP-A and CENP-H, two DNA-proximal kinetochore components (). We found that the levels of both were reduced in cells depleted of hDsn1: CENP-A was decreased by 49–54%, and CENP-H was decreased by 70% (; and Table S1). Previous data indicated that CENP-A was unaffected, whereas CENP-H was decreased upon the depletion of hMis12 (). Because the targeting of CENP-H requires CENP-A (), the reduction in CENP-H could be explained, in part, by the lower CENP-A levels we observed. The functional consequence of this CENP-A reduction is unclear. Reduction of CENP-A to ∼30% by direct siRNA does not prevent mitotic progression and does not cause chromosome alignment and segregation defects as severe as those seen in hMis12 complex–inhibited cells (Black, B., D. Foltz, and D. Cleveland, personal communication). Consequently, we favor the idea that outer kinetochore defects are primarily responsible for the dramatic phenotypes observed in hMis12 complex depletions. It should also be noted that hMis12 was previously suggested to localize to kinetochores in a manner that is partially independent of CENP-A (). Exploring the role of the hMis12 complex in this key early step of kinetochore assembly is an important future direction. The data presented in this study demonstrate conservation of a four-subunit Mis12 complex in human cells. Functional analyses revealed defects in chromosome alignment and segregation, kinetochore fiber stability, and chromosome biorientation. These defects are likely explained by a central role for this complex in kinetochore assembly. Both inner and outer kinetochore protein targeting is defective in cells depleted of the hMis12 complex. The reduction in CENP-A levels may reflect a role for the hMis12 complex in CENP-A loading/stabilization or could be an indirect consequence of defective chromosome segregation in a prior cell division. The effect on Ndc80 complex targeting is likely direct given the conserved physical interaction between the Mis12 and Ndc80 complexes. In general, the defects in chromosome movement and kinetochore fiber formation in hMis12 complex–inhibited cells were less severe than those observed upon the direct depletion of the Ndc80 complex. This difference could arise from technical variations in RNAi efficacy; however, studies in fungi and have revealed that the Ndc80 complex functions mostly downstream of the Mis12 complex, yet its inhibition results in more severe kinetochore microtubule attachment and chromosome segregation defects (; ; ; ; ; ; ; ). The similar trend in human cell studies (; this study) leads us to speculate that additional mechanisms contribute to Ndc80 complex function in chromosome segregation. For example, Ndc80 complex targeting is also mediated by CENP-H (), which was still present at ∼30% of wild-type levels in hMis12 complex–depleted cells. HeLa cells and clonal lines derived from HeLa cells stably expressing GFP-Mis12 (), YFP–CENP-A (a gift from D. Foltz, Ludwig Institute for Cancer Research [LICR], La Jolla, CA), and YFP-histone H2B (a gift from J. Shah, Harvard Medical School, Boston, MA; ), or GFP–CENP-H and GFP-hNuf2 (this study) were maintained in DME supplemented with 10% FBS, penicillin/streptomycin, and -glutamine (Invitrogen) at 37°C in a humidified atmosphere with 5% CO. Cells were plated on 12- or 22-mm glass coverslips coated with poly--lysine (Sigma-Aldrich) for immunostaining or 35-mm glass bottom microwell dishes (MatTek) for time-lapse imaging. Predesigned siRNAs targeting hNsl1 (Ambion), hDsn1, hNnf1, hMis12 (Dharmacon), hNuf2 (a gift from J. DeLuca, University of North Carolina, Chapel Hill, NC; ), or nonspecific control siRNAs (Dharmacon) were transfected according to manufacturer's directions using Oligofectamine and serum-free OptiMEM (Invitrogen). FBS was added to 10% to the transfection reaction after incubation on cells for 5–6 h. Cells were assayed 48 h after transfection for hDsn1, hNnf1, or hNsl1 and 72 h after transfection for hMis12. Chicken Mis12 target disruption constructs for each gene were generated such that genomic fragments encoding an entire coding region in a single exon were replaced with a histidinol or puromycin resistance cassette under control of the β-actin promoter. Target constructs were transfected with a Gene Pulser II electroporator (Bio-Rad Laboratories). Chicken DT40 cells were cultured and transfected as described previously (, ). All DT40 cells were cultured at 38°C in DME supplemented with 10% FCS, 1% chicken serum, and penicillin/streptomycin. To repress the expression of the tet-responsive transgenes, tet (Sigma-Aldrich) was added to the culture medium to a final concentration of 2 μg/ml. Chicken homologues of Dsn1, Nsl1, and Nnf1 were cloned and fused with GFP under control of the cytomegalovirus promoter. These fusion constructs were transfected into chMis12 conditional knockout cells to examine their localization. For gel filtration and sucrose gradient analysis, HeLa cells were grown until they were ∼80% confluent, and 100 ng/ml nocodazole was added for 14 h to enrich for mitotic cells. Cells were harvested and washed into lysis buffer (50 mM Hepes, pH 7.4, 1 mM EGTA, 1 mM MgCl2, 100 mM KCl, and 10% glycerol). Cell extract was prepared as described previously () and fractionated on either a Superose 6 gel filtration column or a 5–20% sucrose gradient at 50 krpm for 8 h in either lysis buffer or lysis buffer with a total of 600 mM KCl. Coexpression of Mis12 complex proteins was conducted by cloning the cDNAs for each protein into pST39 (a gift from S. Tan, Pennsylvania State University, University Park, PA; ). carrying the plasmid were induced with 0.1 mM IPTG, and the proteins were purified using nickel affinity chromatography. The eluate from the nickel column was further fractionated on gel filtration and sucrose gradients as described above. Affinity-purified rabbit polyclonal antibodies were generated against hDsn1 (residues 181–356), hNsl1 (residues 22–281), hNnf1 (full length), hMis12 (residues 105–205), chMis12 (full length), and chNdc80 (residues 465–640) as described previously (). For analysis of checkpoint proteins BubR1 and CENP-E, cells were treated with 10 μg/ml nocodazole diluted in media for 1 h before processing for immunofluorescence. For analysis of kinetochore proteins, cells were rinsed in PBS (12 mM PO , 137 mM NaCl, and 3 mM KCl, pH 7.4), extracted for 5 min in warmed PHEM (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, and 2 mM MgCl, pH 6.9) plus 1% Triton X-100, and fixed at room temperature for 20 min in PHEM plus 4% formaldehyde. For analysis of spindle microtubules, cells were not extracted before fixation. Treatment and fixation of cells for the cold-stable kinetochore fiber assay were performed as described previously (). All fixed cells were rinsed with TBS-TX (20 mM Tris, 150 mM NaCl, pH 7.5, and 0.1% Triton X-100) and blocked in AbDil (2% BSA and 0.1% NaN in TBS-TX) for 30–60 min. For fluorescence of microtubules, DM1α (Sigma-Aldrich) was used at 1:500. For visualization of kinetochore proteins, mouse anti-HEC1 (9G3; Abcam) was used at 1:1,000, and human anticentromere antibodies (ACAs; Antibodies, Inc.) were used at 1:200. YFP–CENP-A, GFP–CENP-H, or GFP-Mis12–expressing HeLas were counterstained with goat anti-GFP (a gift from D. Drechsel, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany) at 1:500. Mouse anti–CENP-A (a gift from K. Yoda, Nagoya University, Nagoya, Japan), rabbit anti–CENP-E (a gift from D. Cleveland, LICR), and mouse anti-BubR1 (a gift from S. Taylor, University of Manchester, Manchester, UK) were used at 1:200. Polyclonal antibodies against hDsn1, hNsl1, and hNnf1 were used at 1 μg/ml. Cy2-, Cy3-, and Cy5-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:100, although in some cases, directly labeled antibodies against hDsn1 and hNnf1 were used without secondaries. DNA was visualized using 10 μg/ml Hoechst in TBS-TX. Coverslips were mounted using 0.5% p-phenylenediamine and 20 mM Tris-Cl, pH 8.8, in 90% glycerol. For DT40 immunofluorescence, cells were collected onto slides with a cytocentrifuge, fixed in 3% PFA in 250 mM Hepes, pH 7.4, for 15 min at room temperature, permeabilized in 0.5% NP-40 in PBS for 10 min at room temperature, rinsed three times in 0.5% BSA, and processed for indirect immunofluorescence. Images of fixed cells were acquired on a deconvolution microscope (DeltaVision; Applied Precision) equipped with a CCD camera (CoolSNAP; Roper Scientific). 30–80 z sections were acquired at 0.2-μm steps using a 100× NA 1.3 U-planApo objective (Olympus) with 1 × 1 binning. For analysis of microtubule attachments, images were deconvolved using the DeltaVision software (Applied Precision). Measurements of the intensity of kinetochore localization were conducted on nondeconvolved images. All images for a specific experiment used identical exposure settings and scaling. For live cell imaging, medium was replaced with CO-independent medium supplemented with 10% FBS, penicillin/streptomycin, and -glutamine (Invitrogen) and was covered with mineral oil immediately before analysis. Cells were maintained at 35–37°C using a heated stage. Images of cells expressing YFP-histone H2B were collected on an inverted microscope (Eclipse 300; Nikon) and a 60× NA 1.4 planApo objective. Acquisition parameters, including exposure, focus, and illumination, were controlled by MetaMorph software (Universal Imaging). Single focal plane images were collected by a camera (CoolSNAP HQ; Photometrics) at 2-min intervals with an exposure time of 80 ms with 2 × 2 binning. Images of cells expressing YFP–CENP-A were collected on a spinning disc confocal (McBain Instruments) mounted on an inverted microscope (TE2000e; Nikon) using a 60× NA 1.40 planApo objective plus 1.5× auxiliary magnification. Five z sections were acquired at 1-μm steps at 10-s time intervals with exposure times of 100 ms and 2 × 2 binning. Z stacks were projected in MetaMorph. All subsequent analysis and processing of images were performed using MetaMorph software. Images of DT40 cells were collected with a cooled CCD camera (CoolSNAP HQ; Photometrics Image Point) mounted on an inverted microscope (IX71; Olympus) with a 60× NA 1.40 planApo objective lens together with a filter wheel. Images were analyzed with an IPLab software (Signal Analytics) or DeltaVision deconvolution system (Applied Precision). All distance and fluorescence intensity measurements were made using MetaMorph software. Analysis of kinetochore movements was performed by measuring the distance from the center of sister kinetochores to a point on the spindle equator along the trajectory of chromosome movement. The bulk of metaphase-aligned chromosomes were used as a reference point for the spindle equator, which occasionally shifted during our analysis. Interkinetochore distances were measured using the centers of the paired CENP-A dots. Kinetochore fluorescence intensities were determined by measuring the integrated fluorescence intensity within a 7 × 7 pixel square positioned over a single kinetochore and subtracting the background intensity of a 7 × 7 pixel square positioned in a region of cytoplasm lacking kinetochores. Maximal projected images were used for these measurements. To determine significant differences between means, unpaired tests assuming unequal variance were performed; differences were considered significant when P < 0.05. Fig. S1 shows immunoblots of HeLa cell lysates probed with antibodies to hDsn1, hNnf1, or hNsl1 as well as immunoblots after RNAi of each of these proteins. Fig. S2 illustrates the disruption of chMis12 in DT40 cells, the cell cycle consequences of chMis12 knockdown, and the concomitant decrease in kinetochore-localized chNdc80. Table I summarizes the quantification of kinetochore fluorescence intensities of targeted Mis12 complex subunits and CENP-A, CENP-H, Ndc80, and BubR1. Videos 1–3 show mitosis in HeLa cells expressing YFP-histone H2B after transfection of control siRNA (Video 1), siRNA targeting hDsn1 (Video 2), or siRNA targeting hNnf1 (Video 3). Videos 4–8 show mitosis in HeLa cells expressing YFP–CENP-A after transfection of control siRNA (Video 4), siRNA targeting hMis12 (Video 5), siRNA targeting hDsn1 (Video 6), siRNA targeting hNnf1 (Video 7), and siRNA targeting hNsl1 (Video 8). Online supplemental material is available at .

The archetypical protein modifier ubiquitin is a ubiquitously expressed, highly conserved polypeptide best known as a marker for intracellular protein turnover (). Proteasomal degradation of proteins is generally preceded by covalent tagging of proteins with a ubiquitin polymer (). Ubiquitin tagging is the result of an enzymatic cascade executed by a ubiquitin-activating enzyme (E1), ubiquitin-conjugating enzymes (E2), and ubiquitin-ligating enzymes (E3; ). The E1, E2, and some E3 enzymes form a thiolester linkage with ubiquitin, which is eventually conjugated by an isopeptide bond either to an internal lysine residue or to the free NH terminus of a target protein. Ubiquitylation plays a critical role in many other cellular events as well (). Histones were the first ubiquitin-modified proteins to be identified and are the predominant ubiquitin targets in the nuclei of metazoans (). Ubiquitylated histone H2A (uH2A) is required for gene silencing (; ; ). The internalization of receptors and the delivery of proteins to the multivesicular bodies are also dependent on ubiquitylation (). Although the roles of ubiquitin in these processes have been studied in detail, the dynamic exchange of ubiquitin between these different systems is less well understood. We followed the dynamics of fluorescently tagged ubiquitin in living cells and showed that histones and other ubiquitin substrates compete for a limited pool of free ubiquitin. This links ubiquitin-dependent processes, coupling protein degradation to chromatin remodeling, and adds a dynamic dimension to ubiquitin as a general regulator of the cellular proteome. We generated a construct encoding wild-type ubiquitin with an NH-terminal GFP tag. It has been recently shown that GFP–ubiquitin (GFP-Ub) fusions are functionally conjugated to substrates and show similar localization as endogenous ubiquitin (). A similar fusion was made with a conjugation-deficient mutant ubiquitin lacking all internal lysine residues and the COOH-terminal glycine residue (GFP-Ub). Western blot analysis of the total lysates of human melanoma Mel JuSo cells stably expressing these fusions confirmed that GFP-Ub was present both as free monomers (∼33 kD) and in large ubiquitin conjugates, whereas GFP-Ub was exclusively found as free monomers (, left). Importantly, comparing the signals that were obtained when both the parental and stable Mel JuSo cell lysates were probed with the antiubiquitin antibody showed that GFP-Ub and GFP-Ub were expressed in minute amounts compared with endogenous ubiquitin (, right). Under nonreducing conditions, the levels of free GFP-Ub and ubiquitin were lower, suggesting that a major fraction of the monomeric GFP-Ub and ubiquitin is not free, but covalently linked by reducible thiolester linkage to ubiquitylation enzymes (). We consistently found that fewer ubiquitin conjugates were recovered under nonreducing conditions, which may be caused by poorer solubility of the conjugates in the absence of reducing agents. Microscopic analysis of living cells showed that GFP-Ub was present in both nucleus and cytosol. Although GFP-Ub was equally distributed throughout the cytosolic and nuclear compartments, GFP-Ub levels were highest in the nucleus, where it displayed a punctuate staining with irregular granular dots, and was lower in the nucleoli (). In the cytosol, GFP-Ub was distributed in a diffuse pattern and associated with a large number of mobile punctuate structures (, top), of which many appear to be lysosomes (, bottom). The staining that was obtained with a ubiquitin-specific antibody matched the GFP fluorescence in GFP-Ub–expressing Mel JuSo cells (). Notably, because GFP-Ub forms only a small fraction of the total ubiquitin pool in these cells, GFP-Ub apparently distributes like endogenous ubiquitin. The ubiquitin–proteasome system was functional in the presence of the GFP–Ub fusions because the cell cycle distribution pattern and the cell surface expression of stable major histocompatibility class I molecules (Fig. S1, available at ), two events that strongly depend on ubiquitylation machinery that is intact, were not affected. Our biochemical analysis () and that of others (; ) suggested that cells contain only a limited pool of free ubiquitin. To test this in living cells, we took advantage of the fact that the molecular mass of free monomeric GFP-Ub is 33 kD, which allows passive diffusion through the nuclear pore (), unless it is incorporated into larger complexes. We photobleached GFP-Ub and GFP-Ub in the cytosol or nucleus and quantified the redistribution of fluorescence from the nonbleached compartment in a fluorescence loss in photobleaching protocol, basically using the nuclear pore as a molecular sieve to distinguish free from conjugated GFP-Ub molecules (). GFP-Ub displayed biphasic redistribution between the two compartments, with a fast component in the first minute followed by a major slow component (). The presence of a small fraction that is rapidly exchanged during the first minute is in agreement with a small amount of free monomeric GFP-Ub, as detected biochemically (). The slow exchange persisted with similar kinetics throughout the recording, suggesting the continuous generation of freely diffusing GFP-Ub. Rapid redistribution was observed with the GFP-Ub, with a complete exchange of fluorescence within 6 min confirming that this monomeric form efficiently diffuses through the nuclear pore (). The slow exchange of GFP-Ub between the nuclear and cytosolic compartments suggests that the vast majority of ubiquitin is incorporated into large complexes that cannot pass the nuclear pore. We performed FRAP analysis, which allows determination of protein diffusion and mobility rates (). The Brownian motion of particles is related to their size, and large polyubiquitin complexes are thus expected to diffuse considerably slower than free ubiquitin. For comparison, we included a Mel JuSo cell line expressing a GFP-tagged α3 subunit of the proteasome, which is a large, freely diffusible complex (). Coimmunoprecipitation and sucrose gradient experiments confirmed that the α3 subunit is incorporated into the proteasome particle (unpublished data). FRAP analysis revealed both the diffusion rate and the fraction of mobile proteins. A large portion was mobile in the cytosol, unlike GFP-Ub in the nucleus, which is where the majority of GFP-Ub was immobile (). Quantitative analysis of the FRAP data revealed a much larger fraction of immobile nuclear GFP-Ub, as compared with the cytosolic GFP-Ub (). An immobile GFP-Ub fraction in the cytosol is likely to be a consequence in part of the role of ubiquitin in membrane trafficking (). Moreover, ubiquitylated proteins can bind to cytoskeletal-associated proteins () and form cytosolic clusters (). Some 70% of GFP-Ub is immobile in the nucleus, which supports the notion that a major fraction of GFP-Ub is conjugated to histones (see ). The monomeric GFP-Ub diffused rapidly through the cytosol and nucleus, whereas the GFP-tagged proteasome moved relatively slow in both compartments, in line with their size differences. Consistent with the notion that ubiquitin is incorporated in large ubiquitin chains (), the GFP-Ub pool had a surprisingly slow diffusion rate in the nucleus and cytosol, especially when compared with the proteasome (). Biochemical analysis has revealed that proteasome inhibitor treatment and heat shock can deplete histones from ubiquitin (; ). To reveal the dynamics of this process, we monitored GFP-Ub in living cells after the administration of the proteasome inhibitor MG132. A rapid accumulation of GFP-Ub in the cytosol and the formation of aggresomes in the perinuclear region were observed within 2 h, which was accompanied by a profound loss of nuclear GFP-Ub (). Staining of fixed cells with the ubiquitin-specific FK2 antibody revealed a similar redistribution of endogenous ubiquitin (; unpublished data). During the 2-h inhibitor treatment, we observed a steady and gradual decline in nuclear GFP-Ub, coinciding with an increase in cytosolic fluorescence (). FRAP analysis demonstrated that the mobile pool of GFP-Ub in the nuclear and cytosolic compartment was further decelerated by the inhibitor treatment (), which correlated with an accumulation of ubiquitin conjugates, as well as with a shift of the conjugates to higher molecular masses (). Both an increase in the amount of polyubiquitylated proteins, as well as an increase in the size of the polyubiquitin changes, is likely responsible for the reduced velocity of ubiquitin in MG132-treated cells. Notably, in the presence of MG132, the diffusion was reduced to velocities that were in the same range as proteasomes, emphasizing the considerable size of these polyubiquitin complexes or direct association with proteasomes (compare and ). The putative GFP-Ub–modified histones were only found in the nucleus and strongly declined during inhibitor treatment (). A similar reduction in the GFP-Ub–histone band was observed during heat shock, which is another form of proteotoxic stress, although under this condition polyubiquitylated material primarily accumulated in the nucleus (). A gradual redistribution of endogenous ubiquitin from the nuclear to the cytosol compartment was also evident when lysates of cells harvested at various times after inhibitor administration were probed with a ubiquitin-specific antibody (Fig. S2, available at ). Notably, proteasome inhibitor treatment reduced the nuclear immobile pool of GFP-Ub, which is in line with a reduction in histone-conjugated ubiquitin (). Western blot analysis confirmed a decrease in endogenous uH2A levels under these stress conditions that was analogous to GFP-Ub–histone (). To further test whether GFP-Ub correctly reflected the behavior of endogenous ubiquitin in the process of MG132-driven histone deubiquitylation, cells were incubated with MG132 and histones were analyzed at various periods after proteasome inhibition. Both GFP-Ub–histone and uH2A were quantified and followed similar kinetics of deubiquitylation (). Half of the histones had released ubiquitin or GFP-Ub ∼30 min after proteasome inhibition. Chromatin of proteasome inhibitor–treated and heat-shocked cells was less sensitive to staphylococcal nuclease (Fig. S3), suggesting a general condensation of nucleosomes that is similar to what has been observed previously for cells subjected to heat shock (). To gain insight into the mechanism responsible for depletion of uH2A, we followed the redistribution of ubiquitin during proteotoxic stress in living cells. The GFP in the fusion constructs was replaced by a photoactivatable GFP (PAGFP; ), and PAGFP-Ub was photoactivated in a confined region in the nucleus ( and Video 1, available at ). Although most of the fluorescence was maintained in the photoactivated region, a small fraction of the photoactivated PAGFP-Ub immediately diffused to other regions in the nucleus, probably because of rapid redistribution of a small pool of free PAGFP-Ub (). Subsequently, fluorescence slowly appeared in the cytosolic compartment coinciding with a gradual decrease in nuclear fluorescence. The fluorescent PAGFP-Ub distributed homogenously in the cytosol and on intracellular punctuate structures. In line with the notion that the vast majority of nuclear PAGFP-Ub was conjugated to histones, PAGFP-Ub only slowly disappeared from the photoactivated region in the nucleus. We monitored the disappearance of the immobile PAGFP-Ub as a measure for histone deubiquitylation. Administration of proteasome inhibitor did not affect the rate of disappearance of the immobile nuclear PAGFP-Ub from the photoactivated region (), which suggests that the rate of histone deubiquitylation is not altered by proteasome inhibition. Alternatively, the redistribution of ubiquitin may be the result of competition of two classes of ubiquitin substrates, i.e., proteasome substrates and histones, for the rate-limiting pool of free ubiquitin. If the loss of histone-conjugated ubiquitin in the nucleus is the result of limiting free ubiquitin levels, experimental introduction of another ubiquitin competitor should have a similar effect. Indeed, microinjection of a GFP-specific antibody in the cytosol of GFP-Ub–expressing cells caused the accumulation of GFP-Ub in the cytosol and the depletion of nuclear GFP-Ub, which is very similar to proteotoxic stress (). An irrelevant antibody did not affect the distribution of GFP-Ub (). These data show that changes in the ubiquitin equilibrium can dramatically affect various ubiquitin-dependent processes. Our data reveal a new dimension of ubiquitin-dependent regulation as the result of a delicate ubiquitin equilibrium (). This ubiquitin equilibrium may be a reflection of the constraints of the heavily used ubiquitylation system by various ubiquitin-dependent processes. Alternatively, changes in the cellular proteome as a consequence of the depletion of ubiquitylated histones may aid the cellular stress response. It has been shown that the decrease in the levels of ubiquitylated histones during proteotoxic stress causes major changes in gene expression (; ). In fact, the depletion of ubiquitylated histones is a rapid response, and the first changes can already be observed within 5 min. Cellular stress is apparently rapidly translated into chromatin alterations, which are likely to affect gene expression. Cross-talk between these ubiquitin-dependent processes by means of limiting free ubiquitin levels may be of functional significance, as it may integrate diverse mechanisms in the combined effort to adapt the cellular proteome to the altering intracellular environment. Wild-type Ub and the Ub mutant were cloned into EGFP-C1 vector (CLONTECH Laboratories, Inc.) and PAGFP-C1 vector (gift from J. Lippincott-Schwarz, National Institutes of Health, Bethesda, MD) and transfected into the human melanoma cell line Mel JuSo. Stable cell lines were generated under the selection of 1 mg/ml neomycin containing Iscove's DME supplemented with penicillin/streptomycin and 8% FCS (Invitrogen). For live cell imaging, cells were either cultured on 24-mm glass coverslips or cultivated in 0.17-mm Delta T dishes (Bioptechs). Before microscopic analysis, the culture medium was covered with a thin layer of mineral oil (Sigma-Aldrich) to prevent evaporation of the medium during recording. Lysosomes were stained by incubating cells with 50 nM LysoTracker red (Invitrogen). The proteasome inhibitor MG132 (Sigma-Aldrich) was dissolved in DMSO and used at a 25-μM concentration, unless otherwise stated. Heat shock was induced by incubating the cells for 3 h at 42°C. Cells were cultured on 15-mm glass coverslips, fixed with 3.7% formaldehyde for 10 min at room temperature, permeabilized with 0.5% Triton X-100 for 2 min, and immunostained in phosphate-buffered saline with 0.5% bovine serum albumin. FK2 antibody (Affinity BioReagents, Inc.) was used at a ratio of 1:1,000. 2 ng/ml DAPI (Sigma-Aldrich) was added during secondary antibody incubation with goat anti–mouse-TxR (Invitrogen). Parental GFP-Ub, stable GFP-Ub, and GFP-Ub Mel JuSo cells were washed with phosphate-buffered saline and trypsinized. Cells were lysed in SDS-PAGE sample buffer. Proteins were separated by SDS-PAGE, transferred onto nitrocellulose or PVDF membranes, and probed with two different rabbit polyclonal antibodies against GFP (Invitrogen; ) or a rabbit polyclonal antibody against ubiquitin (DakoCytomation and Sigma-Aldrich, respectively). The filters were reprobed with a mouse monoclonal antibody against glyceraldehyde-3-phosphate dehydrogenase (Fitzgerald Industries, Intl.) as a control for equal protein loading. After incubation with peroxidase-conjugated secondary antibodies, the blots were developed by enhanced chemiluminescence (GE Healthcare). For separation of nuclei and cytosol, cells were scraped in a buffer containing 100 mM NaCl, 300 mM sucrose, 3 mM MgCl, 50 mM Hepes, pH 7.0, 1 mM EGTA, and 0.2% Triton X-100 supplemented with protease inhibitors and 50 mM -ethylmaleimide. Cells were lysed for 10 min, and nuclei were pelleted by centrifugation for 5 min at 1,000 . The supernatant is the cytosolic fraction; nuclei were resuspended in a buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.1% SDS supplemented with 50 mM -ethylmaleimide and sonicated on ice to disrupt DNA. For fluorescence loss in photobleaching experiments, Mel JuSo cells were cultured in 0.17-mm Delta T dishes (Bioptechs). Confocal laser scanning microscopy was performed with an LSM 510 META with a Plan-Apochromat 63× oil objective, NA 1.4 (both Carl Zeiss MicroImaging, Inc.), equipped with a cell culture stage (Bioptechs) at 35°C. After photobleaching of the GFP fluorescence by exposure of selected regions to 488-nm laser with 100% intensity for 30 iterations, images were obtained every 10 s during a time frame of 4 min, followed by 10 images during a time frame of 10 min. Images were processed using the LSM software. Fluorescence intensities were measured using ImageJ software (National Institutes of Health). The relative fluorescence ratio between the nucleus and cytoplasm was averaged from three recordings. For line-scan FRAP experiments, we used a confocal system (TCS SP2; Leica) equipped with an external bleaching laser and a heating ring to keep the cells at 37°C. PAGFP-Ub was transiently expressed in Mel JuSo cells. In the photoactivation step, PAGFP was activated by applying a single pulse to a small region in the cell with 405-nm laser light at full intensity. For photoactivation experiments, we used a TCS SP2 AOBS system equipped with HCX PL APO and HCX PL APO lbd.bl 63× objective lenses, both with an NA of 1.4 (all Leica). Quantification was done with physiology software version 2.61 (Leica). FRAP data was analyzed as previously described (). For antibody injection, cells were seeded on 15-mm glass coverslips. Cells were microinjected with a mixture containing 1 mg/ml lysine-fixable 70-kD Dextran–Texas red (Invitrogen) and 1 mg/ml of purified polyclonal rabbit anti-GFP antibody () or purified polyclonal rabbit anti-mCD27 (gift from J. Borst, The Netherlands Cancer Institute, Amsterdam, Netherlands). Microinjections were done on an inverse epifluorescence microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.) equipped with a manipulator 5171/transjector 5246 system (Eppendorf) and a 37°C heated ring. After microinjection, cells were cultured for another 2 h and fixed with 3.7% formaldehyde for 10 min at room temperature. Fig. S1 shows the analysis of functionality for the ubiquitin–proteasome system, Fig. S2 shows changes in ubiquitin distribution during proteasome inhibitor treatment, Fig. S3 shows changes in nucleosome condensation during proteotoxic stress, and Video 1 shows the distribution of PAGFP-Ub after photoactivation. Online supplemental material is available at .

The formation of heritable haploid gametes from diploid parental cells requires meiosis, a special type of cell division. After premeiotic DNA replication, homologous chromosome pairing and genetic recombination occur during prophase I with formation of bouquet arrangement conserved among species (for review see ). During prophase I of fission yeast, horsetail nuclear movement occurs. It starts when all the telomeres become bundled at the spindle pole body (SPB; ). Subsequently, the nucleus undergoes a dynamic oscillation, resulting in elongated nuclear morphology (). This event has been proposed to facilitate the pairing of homologous chromosomes because it aligns the chromosomes in the same direction (through the bundling of their telomeric ends) and then shuffles them around each other (; ; ; ). It has been proposed that horsetail nuclear movement is predominantly due to the pulling of astral microtubules that link the SPB to cortical microtubule-attachment sites at the opposite end of the cell; the pulling force is believed to be provided by cytoplasmic dynein (; ; ) and the dynactin (; ) complex. However, the details of these mechanisms are still almost completely unknown. A key to understanding nuclear oscillation is the identification and characterization of the anchor protein at the cell cortex that binds to dynein. (from meiotic coiled-coil protein) encodes a protein, Mcp5 (), that functions as this anchor during the meiosis of . Recently, we found that it is also registered as Mug21 (). Mcp5/Mug21 is homologous to Num1 (nuclear migration) of , ApsA (anucleate primary sterigmata) of , and Ami1 (anucleate microconnidia) of , but these Num1-like proteins function during vegetative growth rather than in meiosis (; ; ; ), unlike Mcp5. We demonstrate a contribution for such an anchoring mechanism in the process of homologue pairing during meiosis and establish a link between meiotic and mitotic nuclear behavior, albeit in different organisms. We first found that Mcp5 shows meiosis-specific expression and is essential for sporulation as shown in Fig. S1 (available at ). Because Num1 is involved in nuclear migration during the budding process, we investigated the function of Mcp5 during the horsetail nuclear movement of , which is the nuclear migration event in this species and which is specifically observed in meiotic prophase. We subjected cells expressing DNA polymerase α-GFP to time-lapse observation under a microscope. In wild-type (WT) cells, sequential oscillatory nuclear movement is observed from karyogamy to slightly before meiosis I ( and Video 1). In contrast, such oscillatory nuclear movement was not observed throughout prophase I in cells ( and Video 2). This abnormal oscillation was also observed in the azygotic meiosis of cells (Videos 3 and 4). These data suggest that Mcp5 plays a significant role in nuclear oscillation. In cells, which show impaired horsetail movement (), the rates of pairing (Fig. S2, A and B; and Videos 5–8; available at ) and recombination (Fig. S2, C and D) of homologous chromosomes are also reduced. This is similar to what has been observed for the (; ), (), mutant (), (), and () cells, in which horsetail movement is also inhibited. Ectopic recombination rates are also increased in cells (Fig. S2 E), as has also been observed for the () and () cells. We also found that the period from karyogamy to meiosis I in zygotic meiosis was 1.5 times longer in cells than in WT cells () and that the timing with which cells containing two or four nuclei peaked was similar in azygotic meiosis in both strains of background (Fig. S2 F). These results suggest that the delay of prophase I in the zygotic meiosis of cells is due to a delay in the progression of karyogamy. Both phenotypes of deficiency in nuclear movement and delay in karyogamy of are similar to those of the mutant (). We examined the subcellular localization of Mcp5 by constructing a GFP-Mcp5–expressing strain in the genetic background. Because the frequency of recombination of the cells was at the WT level (unpublished data), we judged the function of GFP-Mcp5 to be intact in this strain. The cells were induced to enter zygotic meiosis, observed by living and methanol fixation, and stained with Hoechst 33342 to detect the nucleus and with anti–α-tubulin antibody to detect the microtubules. As shown in (top), no GFP signal was detected during the vegetative phase. During the horsetail phase, however, the GFP-Mcp5 signals appeared in the cytoplasm with a gradient pattern, namely, seeping from the arm to the middle and from the arm to the cell end. Furthermore, some bright dots were detected at the cell cortex. Significantly, astral microtubules were observed to be elongating in the direction of some GFP-Mcp5 dots (). Notably, dynein heavy chain (Dhc) 1, one of the +TIPs (microtubule plus-end tracking proteins), colocalized with Mcp5 at a point of cell cortex during horsetail phase (). These cortical dots disappeared after the horsetail phase, and the signal was not detected again through meiosis I and II (). When Mcp5-GFP was overproduced in vegetative cells, Mcp5-GFP localized at the cell cortex, suggesting that other meiosis-specific proteins are not required for the cell cortex localization of Mcp5 (). To examine the effect of Mcp5 on the sliding of microtubules at the cell cortex during the horsetail phase, WT (YY105) and (ST218-1) cells were chemically fixed 6 h after meiosis was induced by nitrogen starvation and then immunostained with anti-Sad1 antibody, and the tubulin-GFP image was observed under a microscope (). We counted only the cells at horsetail phase, at which point the microtubules exhibit a filamentous structure that extends from the SPB to the edge of the cell. We denoted the microtubules as “sliding” or “not sliding,” depending on whether the microtubules curved or did not curve along the cell cortex, respectively. This analysis revealed that ∼50% of WT cells (, A [top] and B [left]) but only 12.8% of cells (, A [bottom] and B [right]) exhibited sliding microtubules. This indicates that cells are defective in the cortical sliding of their microtubules. It is known that Mcp6/Hrs1 plays an important role in organizing astral microtubule arrays at the SPB during the horsetail phase (; ). However, cells show normal localization of Mcp6/Hrs1 (unpublished data). Thus, it appears that although Mcp5 is not required for the proper formation of astral microtubules, it is needed for the proper sliding of the microtubules along the cell cortex. For Dhc1 to generate pulling force on the microtubule tip and so pull the SPB to the cortex, it must be fixed at the cell cortex. Dynein has both microtubule-binding and minus end–directed motor activity, and proposed that a dynein anchor is required for dynein to accomplish these functions. To assess the role Mcp5 may play in this process, we investigated the localization of Dhc1-GFP during the horsetail phase in live observations (). In WT cells, the Dhc1-GFP dot was observed at the SPB and cell cortex along with the microtubules (, C [top] and D [left]; and Video 9, available at ). In contrast, most cells did not exhibit an intense Dhc1-GFP dot at the cell cortex (, C [bottom] and D [right]; and Video 10). When we counted the number of cortical Dhc1-GFP dots, we found that 81% of WT but only 14% of cells harbored these dots (, colored bars). Moreover, these dots localized more frequently at the leading edge (57%) than on the tail side (24%) in WT cells, whereas the cells showed a reversed trend (3 vs. 11%). These results suggest that horsetail movement is linked to the accumulation of cortical Dhc1-GFP dots and indicate that Mcp5 is required for anchoring dynein at the cell cortex, thus allowing the proper U-turn movement of the nucleus. To determine the function of the PH domain of Mcp5, we constructed the PH domain deletion mutant Mcp5 (PH). When we overproduced the Mcp5 (PH)-GFP fusion protein in WT cells (CRL266) using high-copy plasmid pREP1 driven by its promoter (), the cortical localization of Mcp5 was lost (). These results indicate that the PH domain is required for the cortical localization of Mcp5. Interestingly, most of the Mcp5 (PH)-GFP signals were detected in the nucleus, and only a small amount of signal was observed along the microtubule (like Dhc1; ) during both karyogamy and the horsetail phase (). It is possible that the PH-deleted Mcp5 mutant is transported to the nucleus through its nuclear localization signal (NLS). Furthermore, the Dhc1-like linear localization of Mcp5 (PH) could signify that Mcp5 (PH) is transported to the microtubules as a cargo protein of dynein. To determine the role of the coiled-coil motif of Mcp5 plays, we prepared a mutant strain that expresses Mcp5 protein lacking two coiled-coil motifs, namely, Mcp5 (C-CΔ). When we overproduced the Mcp5 (C-CΔ)–GFP fusion protein in WT cells (CRL266), it localized at the cell cortex (). These results indicate that the coiled-coil domain is not required for the cortical localization of Mcp5. strain during meiosis and found that it showed aberrant nuclear oscillation (), reducing the recombination rate (Fig. S2 C) and the abnormal spore formation (not depicted). These results indicate that the coiled-coil domain of Mcp5 plays a pivotal role in its full function and suggest that an unidentified partner that associates via the coiled-coil domain is involved in regulating nuclear movement. Mcp5 localizes at the cell cortex via its PH domain () and colocalizes with dynein (Dhc1) or tubulin-like Num1 of (). Given that Num1 forms a transient complex in vivo with dynein and tubulin (), it may be that Mcp5 also associates transiently with dynein and tubulin in vivo. illustrates our working hypothesis about the role Mcp5 plays in anchoring dynein and thereby generating pulling force. When the SPB (, gray circle) and the horsetail nucleus reaches at one end of the cell cortex, the microtubules carrying the SPB and nucleus extend toward the opposite end (). Pairing of homologous chromosomes may occur inefficiently in the round nucleus at the cell end. When the dynein complex (, green triangle) on the microtubules is trapped and anchored by Mcp5 (red circle), the minus-end motor activity may be stimulated (). Thus, the dynein starts to pull microtubules by its minus-end motor activity, which causes the cortical sliding of the microtubules. Then, the nucleus turns around and is pulled back toward the opposite cell end. Homologous chromosome pairing may be achieved widely in the elongated nucleus, which may induce the recombination of the homologous chromosomes. When the SPB arrives at the cell cortex where the dynein is anchored by Mcp5, the pulling force is gradually weakened and the microtubules behind the horsetail nucleus will extend to the opposite cell end, returning to stage A (). These stages are repeated several times during 2–3 h of prophase I. The magnified view of the cortical site on stage B is shown in . Like Num1 (), Mcp5 is localized at the cell cortex, probably through its PH domain–mediated interaction with phosphatidylinositol bisphosphate. Mcp5 on the membrane helps the cortical sliding of microtubules through anchoring the dynein–dynactin complex. In conclusion, the homologous proteins Mcp5 and Num1 play similar roles in dynein-dependent nuclear migration (for review see ; ), even though they are used in different cellular events, namely, Mcp5 in meiosis during prophase I and Num1 in mitosis during the budding (cell division) phase (). The strains used in this study are listed in Table S1 (available at ). The complete media YPD or YE, the synthetic Edinburgh minimal medium 2 (EMM2), and the sporulation medium molt extract or EMM2-nitrogen (EMM2-N; 1% glucose) were used. We used high-copy plasmid pREP1 driven by its promoter for overproduction experiments (). For time-lapse observations, cells expressing GFP-tagged Polα (CRL026-1, ST205, and ST349) or lacI-NLS-GFP and the lacO repeat integrated at the locus (AY174-7B and ST240) or locus (NP43-20 and ST268) were cultured in 10 ml EMM2 + supplements until they reached midlog phase at 28°C. They were then induced to enter meiosis by incubation in EMM2-N at 28°C. After 5 h of nitrogen starvation, the cells were put on a glass-bottomed dish whose surface was coated with 0.2% concanavalin A, and images under a fluorescence microscope (IX71; Olympus) were recorded every 2.5 min (1 s of exposure time) after the initiation of karyogamy. For observation of lacI-GFP dots, images were taken with a 0.3-s exposure at 5-min intervals with 10 optical sections made at 0.5-μm intervals for each time point. Projected images obtained with MetaMorph software (Universal Imaging Corp.) were analyzed. For NH-terminal tagging to generate GFP-tagged , we synthesized the following six oligonucleotides and used them as primers: MCP5-5F, 5′-AAAACTACGGCGGATTAAGGGTAG-3′; MCP5-5R, 5′-GGG-TTTGAAGGATCAGTATATGTTTTG -3′; Xho-mcp5N, 5′-CCGCGGATGGAGAAAAAACAAGATAACG-3′; Sma-mcp5C, 5′-TCCTTAAATGCCGGGCTTAGTATC-3′; MCP5-3F, 5′-TCCGCATGTTTCAGTCATTGC-3′; and MCP5-3R, 5′-CTCGAGCTTACAAAATAAAAATGCTG-3′. The underlined sequences denote the artificially introduced restriction enzyme sites for PstI, NdeI, XhoI, SmaI, SmaI, and SacI, respectively. These PCR products were inserted into the pRGT1 vector via the PstI–NdeI, XhoI–SmaI, and SmaI–SacI sites, respectively. strain ST256. After growth on EMM2 + 5-fluoroorotic acid plates, the ura transformants were screened by PCR analysis to identify the ( )( ) replacement. , we followed the previously described method (). gene. For this purpose, we synthesized the following two oligonucleotides and used them as primers: MCP5N, 5′-CGGCGCGCCGGAGAAAAAACAAGATAA-3′, and MCP5C, 5′-GTACTCGAGGCGGAATGCCGGGCTTAGTATCAA-3′. The underlined sequences denote the artificially introduced restriction enzyme sites for NdeI and NotI, respectively. To obtain the 3′ downstream region, we used the primers described in the previous paragraph that produce the Smal–Sacl fragment. These PCR products were inserted into the pTT (FLAG)-Lys3 vector (), which is designed to allow one-step integration. This plasmid construct was digested with NsiI. The resulting construct was introduced into HM105 ( ). We then screened the Lys+ transformants and confirmed the precise integration of the constructs by PCR. Fluorescent microscopic observations were performed as described previously (). Cells from a single colony were cultured in 10 ml EMM2 with supplements until they reached midlog phase at 28°C. The cells were collected by centrifugation, washed three times with 1 ml EMM2-N, and induced to enter meiosis by incubation in EMM2-N at 28°C for 6 h. Fig. S1 shows that Mcp5 is a meiosis-specific coiled-coil protein that belongs to Num1 family proteins found specifically among fungi. Fig. S2 shows characterization of cells. Video 1 shows a wild-type cell expressing GFP-labeled DNA polymerase α during zygotic meiosis. Video 2 shows an cell expressing GFP-labeled DNA polymerase α during zygotic meiosis. Video 3 shows a wild-type cell expressing GFP-labeled DNA polymerase α during azygotic meiosis. Video 4 shows an cell expressing GFP-labeled DNA polymerase α during azygotic meiosis. Video 5 shows a wild-type cell whose loci are labeled by GFP-LacI-NLS fusion protein. Video 6 shows an cell whose loci are labeled by GFP-LacI-NLS fusion protein. Video 7 shows a wild-type cell whose loci are labeled by GFP-LacI-NLS fusion protein. Video 8 shows an cell whose loci are labeled by GFP-LacI-NLS fusion protein. Video 9 shows a wild-type cell expressing GFP-labeled Dhc1 during horsetail phase. Video 10 shows an cell expressing GFP-labeled Dhc1 during horsetail phase. Table S1 shows a list of strains used in this study. Online supplemental material is available at .

RNase mitochondrial RNA processing (MRP) is an essential ribonucleoprotein endoribonuclease that cleaves RNA substrates in a site-specific manner and is highly conserved in eukaryotes in sequence and structure (). In the yeast , RNase MRP consists of an RNA core, encoded by the nuclear gene , and at least 11 protein subunits (; ). Nine of these proteins are shared with a second highly conserved endoribonuclease called RNase P, which is involved in pre-tRNA processing (). The two protein subunits unique to RNase MRP are an RNA binding protein, encoded by the gene , and a recently identified protein, Rmp1p (; ). In , as in mammalian cells, RNase MRP RNA is found in at least two subcellular organelles. In mitochondria, RNase MRP cleaves RNA transcripts complementary to the origin of replication, forming the RNA primer required for transcription-driven mitochondrial DNA replication (). In the nucleolus, RNase MRP specifically cleaves 27SA preribosomal RNA at the A3 site, forming the 5.8S(s) ribosomal RNA (rRNA; ; ). More recent research has uncovered a novel function for RNase MRP in cell cycle regulation (). Mutations in RNase MRP components cause a cell cycle delay in late mitosis characterized by large budded cells, a dumbbell-shaped nucleus, and extended spindles. Analyses determined that RNase MRP directly cleaves the 5′-untranslated region (UTR) of the yeast B-type cyclin, mRNA, allowing for rapid degradation by the 5′ to 3′ exoribonuclease Xrn1p (). Cleavage of the 5′-UTR of mRNA and subsequent degradation by Xrn1p is a unique mode of mRNA turnover in . The cell cycle delay observed in RNase MRP mutants can be explained by the elevated mRNA levels causing sustained levels of the Clb2p past anaphase, the stage at which Clb2p is normally degraded through ubiquitination by the anaphase promoting complex (APC; ). This results in prolonged Clb2p/Cdk activity, delaying the completion of mitosis. Degradation of the mRNA by RNase MRP fine tunes the system, allowing for rapid simultaneous degradation of both the mRNA and the protein. It is thought that RNase MRP is predominantly localized to the nucleolus; however, degradation of mRNA is presumed to occur in the cytoplasm. In addition, it is unknown how RNase MRP activity against the mRNA is regulated, as the enzyme is required to be active throughout the cell cycle to process rRNAs. To investigate this, we examined the in vivo localization of RNase MRP in yeast. We find that its localization is cell cycle controlled, exiting the nucleolus during mitosis and localizing to a single punctate cytoplasmic foci in daughter cells. We propose that it is in these foci that the mRNA is being degraded. In previous experiments, a Pop3p-GFP fusion was used to visualize the localization of RNase MRP. This produced a diffuse nuclear staining pattern indicative of poor association with the RNase MRP complex (), and similar results were seen with an Snm1p-GFP fusion (unpublished data). So that RNase MRP localization could be more clearly defined, a GFP-tagged version of the Pop1p subunit under control of the actin promoter (pTD125 GFP-Pop1p; ) was introduced as a reporter for RNase MRP. Because Pop1p is a protein component of both RNase MRP and RNase P, both complexes are visualized with this reporter. A wild-type strain carrying this plasmid had a strong fluorescent GFP signal, the majority of which localized in the nucleus with occasional punctate cytoplasmic staining. Within the nucleus, a more concentrated GFP fluorescence was observed that was presumed to be the nucleolus. This was confirmed by introduction of a second plasmid producing a DsRed1 fusion to the Nop1 protein (; ). Nop1p is a protein known to localize to the nucleolus (). This plasmid was used in all subsequent experiments to identify the nucleolus. Merging the GFP and DsRed images indicates that GFP-Pop1p and DsRed-Nop1p colocalized (), consistent with RNase MRP and RNase P localizing to the nucleolus (). However, cytoplasmic GFP signal was never seen to overlap with the Nop1p signal. The GFP-Pop1p fusion was able to complement the lethality of a deletion of the gene. Indeed, yeast carrying this as their only version of the gene grew at normal rate at a variety of temperatures and had no apparent defect in rRNA processing or tRNA processing. In addition, visualization of the GFP-Pop1p was identical with or without a wild-type copy of being present. This indicates that the GFP-Pop1 is fully functional and is assembled into active RNase MRP and P complexes. Several fusions to the carboxy terminus of the Pop1 coding region were tested, but all of them were unable to complement a deletion of the gene and were hence inappropriate for these studies (; unpublished data). Because Pop1p is also a subunit of RNase P, it was important to demonstrate that RNase MRP localized to all of the sites seen, as opposed to only a subset of them. To accomplish this, GFP was fused to Rmp1p, a protein component unique to RNase MRP. The resulting fluorescent Rmp1p localized to the nucleus, the nucleolus, and the punctate cytoplasmic spots in a fashion identical to the GFP-Pop1p (). However, as the fluorescent signal rapidly bleached, the GFP-Pop1p was used for all subsequent localization experiments. Like the GFP-Pop1p fusion, the Rmp1p fusion was found to fully complement a strain deleted for the corresponding wild-type gene (unpublished data). Nucleolar localization of RNase MRP is consistent with its known role in rRNA processing (). However, knowing that RNase MRP cleaves the 5′-UTR of the mRNA to promote its rapid degradation, it was difficult to speculate how this was occurring in the nucleolus (). Because mRNA only needs to be degraded at a certain time in the cell cycle, we examined whether the localization of RNase MRP changes as a cell divides. Wild-type cells (TLG205) at various stages of the cell cycle were examined for their localization pattern (). The stages of the cell cycle were defined by bud size and location of the nucleus relative to the bud neck. RNase MRP remained in the nucleolus until the nucleolus began to move into the bud neck. At that time, RNase MRP localization appears homogeneous throughout the nucleus. After the nucleus had completely divided and the mitotic spindle had disassembled, RNase MRP moved back into the nucleolus, as indicated by colocalization with the DsRed-Nop1p. While following RNase MRP localization throughout the cell cycle, we observed the consistent presence of a punctate spot localized in the cytoplasm of many, but not all, cells in an asynchronous culture. Hundreds of midlog, asynchronous, wild-type cells were examined, which revealed a pattern as to the appearance and localization of this foci. To rule out the possibility that the spots were random GFP aggregates, the localization of RNase MRP to cytoplasmic foci was quantitatively followed throughout the cell cycle. Cells in G1 phase, S phase, metaphase, anaphase, early telophase, and late telophase were scored for the presence of a cytoplasmic spot and the presence of the spot in the mother or daughter cell. The results, summarized in , confirm the observation that the localization of RNase MRP to cytoplasmic foci was a temporally regulated event. RNase MRP localization to a cytoplasmic spot was also seen in wild-type cells carrying the Rmp1p reporter for RNase MRP. The spot does not appear until the nucleus starts to move through the bud neck (metaphase) and can be seen through the end of the cell cycle. The spot was never seen in unbudded (G1 phase) or small budded (S phase) cells or cells arrested with hydroxyurea in S phase. Only a single spot was seen in nearly all cells. More than two spots were never seen in a wild-type strain. Interestingly, the spot localized to the daughter cell in >94% of cells counted. Because the RNase MRP–containing spot appears in a cell cycle–controlled manner and only in the daughter cell, we have named it the TAM body, for temporal asymmetric MRP body. Snm1 is a unique protein component of RNase MRP that is not shared with RNase P. Previous work has identified several temperature-conditional mutations in this essential gene (). Two of the mutations, and , have been shown to have several genetic interactions with mitotic exit and cyclin mutations (). In addition, these were the first RNase MRP mutants found to display a cell cycle delay in telophase (). To determine whether Snm1 has a role in localization of the complex to TAM bodies, we investigated the localization of RNase MRP in these mutations. As can be seen in , the mutant is defective in localization of the complex to both the TAM body and in concentration of the complex in the nucleolus. In this mutant, only a homogenous nuclear signal was seen and punctate cytoplasmic staining was never seen. This is in contrast to the mutant, which displayed a pattern indistinguishable from wild type. Loss of TAM body localization in the mutant also indicates that RNase P does not localize with RNase MRP to these cytoplasmic spots. If RNase P also localized to the TAM bodies, its localization should not have been disrupted in the mutant. The specific localization of the TAM body suggested that it might be the daughter cell spindle pole body. To determine whether the TAM body colocalized with the spindle pole body, a wild-type strain was transformed with a CFP-tagged version of the spindle pole component Spc42p (pAH1; ). The TAM body was found to be a distinct focus from that of either of the two spindle pole bodies. This was consistently seen in all cells examined, indicating that the TAM body is not the daughter cell spindle pole (unpublished data). Previous research has demonstrated that mRNA is degraded by the exoribonuclease Xrn1p after the 5′-UTR is cleaved by RNase MRP (; ). An strain accumulates RNase MRP cleavage products that are most likely incompetent for translation, as they have no 5′-cap. Many of the proteins involved in decapping mediated mRNA degradation, including Xrn1p, localize to cytoplasmic foci called P bodies. The decapping activator, Dhh1p, has been shown to display a significant increase in the size and number of P bodies that it localizes to in an (). This too is caused by an accumulation of uncapped mRNAs that are incompetent for translation and are accumulating with the mRNA degradation machinery in P bodies. In an strain examined with the reporter GFP-Pop1p, the same phenomenon occurs. There is a clear increase in the number of TAM bodies (two to five per cell; ) compared with wild type, which has only one spot in most of the cells examined. Localization of the TAM bodies also differs from wild type in that the TAM bodies are present in both the mother and daughter cells. This result suggests that RNase MRP is processing mRNAs in the TAM bodies and that the degradation products are now accumulating in those sites. These results led us to predict that RNase MRP may also localize to P bodies. Because Xrn1p works with RNase MRP to degrade the mRNA, we predicted that it would colocalize if mRNAs were degraded in the TAM body (). Xrn1p has already been shown to be a P body component that degrades mRNAs after decapping. We generated an Xrn1-RedStar2 fusion and introduced it into a wild-type strain with the GFP-Pop1 reporter. As can be seen in , the Xrn1p localized to several discrete foci in cells. In addition, we saw clear colocalization of a single Xrn1p spot with the TAM body. This indicates that the TAM body is a form of P body. We wanted to determine whether the TAM body was a discrete entity or a P body that adds certain degrading activities (RNase MRP) at certain times in the cell cycle. To test this, the same strain was transformed with RFP-Lsm1p (pRP1085), an activator of the Dcp1p–Dcp2p decapping complex known to localize to P bodies (). Interestingly, we observed that GFP-Pop1p and RFP-Lsm1p do not colocalize (unpublished data), suggesting that RNase MRP is not a P body component. To confirm these results, we also examined the localization of GFP-Pop1p and Dcp1p-RFP (pTG003) throughout the cell cycle in wild-type cells. As is evident (), the two proteins do not colocalize at any stage of the cell cycle, indicating that they are distinct cytoplasmic foci. These results indicate that the Xrn1p protein is associated with the TAM body and that the TAM body may represent a specialized form of a P body involved in degrading specific mRNAs. Yet, the TAM bodies are distinct from the P bodies involved in general mRNA decapping and degradation. We have reported that the RNase MRP mutant, (a strong temperature-sensitive point mutation in the gene for the MRP RNA) is synthetically sick in combination with an mutation. mRNA accumulates in an mutant, and in an / double mutant it rises to a level >15-fold higher than wild type (). Because the TAM body appears to be a specialized type of P body, we investigated genetic interactions of the mutation with other genes encoding P body components and genes involved in mRNA degradation. The yeast strain yJA203, in which the chromosomal copy of is replaced with the mutant allele, was mated to strains deleted in genes for , , , , , and . Haploid double mutants were selected and tested for synthetic lethality or sickness as indicated by a lack of growth or compromised growth compared with each of the single mutants. As can be seen in , the mutant displayed synthetic genetic interaction with all of the P body components tested, including , , , , , and . These results suggest that both pathways are potentially interdependent and interconnected. Cytoplasmic exosome components displayed no synthetic interactions with the MRP RNA component, whereas a deletion of the nuclear exosome component was synthetic lethal with the mutation (unpublished data). This result was not unexpected because Rrp6p is involved in 5.8S rRNA processing. Localization of RNase MRP to distinct cytoplasmic foci occurs during the late stages of the cell cycle. To fine tune the appearance and duration of TAM bodies during mitosis and to potentially identify cell cycle components regulating its localization, we examined GFP-Pop1p localization in various FEAR and MEN mutant strains (). These strains arrest at slightly different points late in mitosis. In addition, wild-type strains were arrested using hydroxyurea to provide an S phase arrest and nocodazole to provide a premitotic arrest. Yeast strains with mutations in (TLG208), (TLG285), (TLG206), and (TLG277) were grown at 25°C to midlog phase and shifted to 34°C for 2–3 h. Arrested cells were counted and scored for the presence and localization of a TAM body in the mother or daughter cell. The results are summarized in . As expected, there were no TAM bodies in cells arrested with hydroxyurea. This is consistent with a previous examination of asynchronous cultures. Nocodazole-arrested cells with a nucleus still in the mother did not have TAM bodies, whereas those that had a nucleus in the daughter cell or trapped between the two cells usually did (unpublished data). Because of the difficulty in separating the two phenotypes, quantitation was not done. Because the mother cell–localized nucleus indicates a premitotic cell and the others indicate a failed mitosis with the addition of nocodazole, the results are consistent with the appearance of the TAM bodies immediately or soon after the beginning of mitosis. We postulated that Esp1p might be important for RNase MRP release from the nucleolus. Esp1p is a protease that cleaves the cohesions to initiate chromosome segregation and the FEAR pathway. Esp1p and the FEAR pathway are very important in the initial release of Cdc14p from the nucleolus (). The mutant arrests between metaphase and the onset of anaphase. In this strain, 40% of the arrested cells had a TAM body, 94% of which localized to the daughter cell. This is comparable to what is seen in the wild-type strain (). This result indicates that RNase MRP is released before initiation of the FEAR pathway. Cdc15p is a protein kinase necessary for the MEN pathway. Cdc5p is also a protein kinase that plays a role in the FEAR and MEN pathways. Previously, overexpression of Cdc5p was shown to suppress some RNase MRP mutations, whereas the mutation was synthetically lethal or sick with other mutations (). Yeast strains with the or mutations arrest late in mitosis upon a shift to the nonpermissive temperature. Both and mutants arrested with a high percentage of cells having TAM bodies, similar to what is seen in wild-type telophase cells. In addition, the TAM bodies were restricted to daughter cells. The only cell cycle mutant we tested that differed from wild type was . Cdc14 is a protein phosphatase located at the end of both the FEAR and MEN pathways. Previously, the mutant was shown to be synthetically lethal with a mutation in the RNase MRP protein component Snm1 (). In the mutant, there was a minor decrease in the number of TAM bodies and a small change in the asymmetry, with 83% of the TAM bodies localizing to the daughter cell as compared with 94% in a wild-type strain (). The , , and mutations were also examined for levels of mRNA. The presence of significant amounts of mRNA would indicate that RNase MRP is hampered in TAM body localization or is unable to promote mRNA degradation. As can be seen in , levels of mRNA are elevated in an RNase MRP mutant as has been shown previously (). mRNA levels were found to be elevated in the mutant compared with the and mutants, which arrest at similar points in the cell cycle. This is consistent with it playing a role in the activation of RNase MRP for mRNA degradation and localization. The and mutants displayed low levels of mRNA. To ensure that sufficient levels of nucleolar RNase MRP remain during mitosis, we examined the processing of the 5.8S rRNA (). Movement of RNase MRP to the TAM body may lead to a reduction in processing of the nucleolar substrate. However, no changes in processing of the 5.8S rRNA were seen in any of the cell cycle mutants, even after 2 h at the nonpermissive temperature, indicating that sufficient RNase MRP enzyme remains in the nucleolus to continue rRNA processing. The locasome is a complex of proteins that utilizes the actin cytoskeleton to localize specific mRNAs to the daughter cell (). Asymmetrically localized mRNAs include and mRNAs (). To determine whether TAM body localization is dependent on the locasome, we examined the localization of the GFP-Pop1p reporter in strains with deletion mutations for two of the proteins of the locasome, and . Cultures were grown to midlog phase and examined for the presence and localization of TAM bodies. In each strain, RNase MRP properly localized to the nucleus and nucleolus, as can be seen in the strain in . RNase MRP also localized to the TAM bodies at the correct time in the cell cycle. However, the TAM bodies no longer localized asymmetrically to the daughter. In and strains, the TAM body localized to the daughter cell in only 29.7 and 41.4% of cells, respectively, as compared with 94% in wild type, indicating that localization has become random. Interestingly, a small number of unbudded cells (∼5%) also contained TAM bodies. Because the mRNA is both an RNase MRP substrate and an asymmetrically localized RNA, we examined whether daughter cell localization of RNase MRP was dependent on the presence of the mRNA. RNase MRP may be using the mRNA to carry it out to the daughter cell. However, we saw no changes in daughter cell localization or timing of TAM bodies in a (unpublished data). The RNase MRP mutation was also combined with the mutation to test for genetic interactions. As can be seen in , the double strain grew considerably slower than either of the individual mutations. This genetic interaction indicates that localization of TAM bodies to daughter cells is not essential, but it is important for efficient mRNA degradation, especially when RNase MRP activity is limiting. However, we saw no change in levels of the mRNA in a strain carrying a myo4Δ (unpublished data). RNase MRP is an evolutionarily conserved ribonucleoprotein endoribonuclease that performs several RNA processing events. We recently described a role for this endoribonuclease in degrading the mRNA. A failure to degrade the mRNA leads to persistent mRNA levels and consequently sustained Clb2 protein levels. This keeps the yeast Cdk active and delays exit from the cell cycle (; ). We have demonstrated here that the activity of RNase MRP against mRNAs appears to be regulated during the cell cycle by changes in intracellular localization. Because RNase MRP must be active and localized to the nucleolus throughout the cell cycle to process rRNAs, changes in its localization are a simple and efficient means to regulate the activity of this enzyme against specific substrates. The data indicate that RNase MRP localizes to a discrete spot in the cytoplasm. The earliest signs of this spot are immediately after the initiation of mitosis. The spot often appears to emanate from the nucleus as it pushes into the daughter cell but then clearly becomes a separate entity. The spot can be seen very late into the cell cycle in telophase cells. However, once septation has occurred, cytoplasmic localization is never seen. Occurrence of the discrete spot coincides with the disappearance of mRNA (), suggesting that the TAM body is the site of mRNA degradation. Clb2 protein appears to be degraded in two waves, the first wave being at the initiation of mitosis by Cdc20p and the APC and the second wave coming at the end of mitosis by Cdh1 and the APC (). Degradation of the mRNA during the first wave would be essential for proper regulation of Clb2 protein levels. Late in the cell cycle, RNase MRP loses its concentrated nucleolar localization and localizes homogenously throughout the nucleus. The nucleolus never breaks down in yeast as indicated by Nop1p staining and the continuation in ribosome biogenesis (), so the reason behind the changes in nuclear staining are unclear. It may correspond to an unknown processing event by RNase MRP or P, as we cannot differentiate between RNase P and MRP in the nuclear fluorescence. In yeast, most mRNA is very efficiently exported out of the nucleus and is found in the cytoplasm. Hence, localization of RNase MRP to the cytoplasm is consistent with it degrading the mRNA and other mRNAs in that locale (). Several recent reports have identified similar discrete cytoplasmic spots in several organisms from yeast to humans (; ; ). These spots, called P bodies, contain much of the machinery for mRNA degradation and are believed to be the site of removal of mRNAs. Exponentially growing yeast cells typically contain between 4 and 10 P bodies that are seen throughout the cell cycle. We found that although RNase MRP localizes in a spot with the Xrn1 nuclease, it does not localize with the decapping-enzyme Dcp2 or the decapping-accessory protein Lsm1. Xrn1p has been directly shown to participate with RNase MRP in mRNA degradation. These results indicate that P bodies may be specialized for various functions. Some P bodies that perform traditional decapping and constitutive degradation of mRNA may be present throughout the cell cycle, whereas other P bodies that initiate degradation through mRNA-specific endoribonucleases are present only during certain times or conditions. We propose that one such specialized P body contains the RNase MRP nuclease and is present only in single copy during mitosis. The initial study that identified P bodies in yeast demonstrated that localization with different P body components gave different numbers of P bodies per cell (). The significance of this was not clear at the time but may indicate further specialization of these particles. Likely there are other types of specialized P bodies that may perform other regulated mRNA degradation functions. A temperature-sensitive mutation in the RNase MRP RNA, , is synthetic lethal or synthetic sick with several P body components. The explanation for this phenotype is twofold. First, compromising the regulated MRP-dependent mRNA degradation pathway may make certain messages completely dependent on the constitutive pathway for mRNA degradation. This is clearly the case for the genetic interactions with a deletion of and . double mutant grew extremely slowly, and we were unable to make a strain heterozygous for the mutation and deleted for (unpublished data). The second explanation for the phenotype is if some of these components participate with RNase MRP in degradation of certain mRNAs. This is clearly the case for the Xrn1 exonuclease, which has been shown to degrade RNase MRP cleavage products (). The fact that there is an increase in the number of TAM bodies in an strain and that Xrn1p localizes to the TAM bodies indicates that mRNA degradation occurs there. The synthetic lethality of the double mutant may be caused by the accumulation of mRNAs that cannot be degraded because they have not been deadenylated or are degraded at a much slower rate by an alternate pathway. This may also hold true for the because Pop2p is a regulator of deadenylation (). We examined several cell cycle arrest conditions to determine in detail the timing and control of RNase MRP relocalization. TAM bodies were never seen in cells arrested with hydroxy urea during S phase or with nocodazole before mitosis, consistent with RNase MRP release being a postmitotic event. Localization was also examined in strains with mutations in several cell division cycle regulators that work at the end of mitosis. These include (the yeast separase), and (protein kinases in the FEAR and MEN pathways), and (the protein phosphatase at the end of both the FEAR and MEN pathways). TAM bodies were seen in all of these mutants, indicating that RNase MRP remains relocalized through the end of mitosis. However, the mutant displayed a clear reduction in the percentage of arrested cells with TAM bodies. The activity of the Cdc14 phosphatase is clearly not required for RNase MRP regulation, but it may be required for full localization and activity. This is consistent with elevated levels of the mRNA in an arrested mutant. Cdc14p could be involved in both release and relocalization of RNase MRP. Cdc14p and several other mitotic exit regulators display strong genetic interaction with mutations in RNase MRP components (). Future studies to examine the exact cell cycle regulators that lead to RNase MRP relocalization will be interesting. Loss of proper localization in mutations in the RNase MRP–specific protein Snm1 may indicate that this protein is key in regulating the relocalization pattern. Indeed, Rmp1pp, the other specific protein, may play an important role. Cytoplasmic localization of RNase MRP was reserved to a single spot. Interesting, that spot was exclusively localized to the daughter cell. A mother cell spot was seen in <6% of cells examined and was usually found in the rare cell that had two spots. Identical to the RNase MRP complex, mRNA is asymmetrically localized to daughter cells (). Localization of mRNA is dependent on the locasome (). The locasome is a complex of several proteins that uses the actin cytoskeleton to transport >30 different mRNAs into daughter cells. Our results indicate that this complex is also necessary for localization of RNase MRP to daughter cells. Deletion of the locasome components or leads to a loss of asymmetric RNase MRP localization. Whether RNase MRP localizes to the daughter cell by associating with the locasome or with locasome-associated mRNAs is not clear. We tested to determine whether localization of RNase MRP was dependent of the mRNA and found it was not. However, this does not exclude RNase MRP from traveling with a different RNA. The RNase MRP mutant was found to be synthetically sick with a deletion in the gene, indicating that daughter cell localization of RNase MRP is important for its activity in mRNA degradation. Localization of RNase MRP to the site of its mRNA substrate adds an elegant mechanism of regulating and refining cell cycle control. RNase MRP, like its homologue RNase P, is an ancient enzyme that is thought to be a vestigial leftover from the RNA world (). These enzymes perform ancient processing jobs, including the production of tRNAs, the initiation of DNA replication, and the production of ribosomes. In the world of today, proteins mainly control cell cycle regulation. These proteins often lead to posttranslational modifications of other proteins, producing a plethora of effects from enzymatic activation and inhibition to degradation or relocalization. However, before proteins evolved, cells required a means of cell cycle regulation. Enzymes that degraded or synthesized various RNAs that performed other enzymatic functions would be the simplest mechanism. The identification of RNase MRP functioning in regulating the cell cycle by degrading specific mRNAs adds credence to this hypothesis. Indeed, regulated localization of an RNA enzyme to its site of action or substrate adds an additional level of regulation that could have been performed in the RNA world. Three lines of evidence suggest that TAM bodies are the sites of RNase MRP–directed mRNA degradation. First, the temporal appearance of the TAM bodies coincides with the time of degradation of the mRNA. Second, deletion of the Xrn1p nuclease leads to accumulation of RNase MRP–cleaved mRNA (). The accumulation of RNase MRP products and, hence, Xrn1p substrates is probably what leads to the increase in TAM body numbers. Third, asymmetric localization of RNase MRP is consistent with previous reports of asymmetric localization of the mRNA. There are clearly other important mRNA substrates of RNase MRP (; ). It will be interesting if they are also asymmetrically localized. In our model, the initiation of mitosis produces a signal for the release of RNase MRP from the nucleolus (). Clearly, the Cdc14 phosphatase and Snm1p play some role in the relocalization or activation of RNase MRP for mRNA degradation. Substantial levels of RNase MRP must remain behind in the nucleolus to continue rRNA biogenesis, and this is indicated because neither nucleolar localization of RNase MRP nor RNase MRP–dependent rRNA processing are lost. Once RNase MRP localizes to the cytoplasm, daughter cell localization appears to be mediated by the locasome. It is not clear whether the locasome is specifically carrying RNase MRP into the daughter cell or if it is being transported with one of its substrates or the entire TAM body. Once in the TAM body, the evidence suggests that certain mRNAs, including the mRNA, are degraded. Degradation of these RNAs by RNase MRP and Xrn1p may actually lead to breakdown of the TAM body and relocalization of RNase MRP back into the nucleus and then the nucleolus. A list of yeast strains used in this study is provided in . Strain construction and basic molecular biology techniques were performed as described previously (). The strain used for cloning, DH5α, has the genotype φ80dlacZΔM15 endA1 recA1 hsdR17 (r− m+) supE44 thi-1 λ gyrA96 relA1 Δ(lacIZYA-argF)U169 F. The yeast strain TLG334 () was made by integrating RedStar2, a red fluorescent protein tag, at the chromosomal locus. The RedStar2 tag was amplified off the toolbox plasmid pYM-43 () using the primers OXRN/S2 (5′-TAT ACT ATT AAA GTA ACC TCG AAT ATA CTT CGT TTT TAG TCG TAT GTT CTA ATC GAT GAA TTC GAG CTG G-3′) and OXRN/S3 (5′-AAG TCA CAA AGC AAT GCT GCT GAC CGT GAT AAT AAA AAA GAC GAA TCT ACT CGT ACG CTG CAG GTC GAC-3′). The resulting PCR product was integrated into fresh, competent wild-type cells (DBY2006) using the Li Acetate method (). After the transformation was complete, the cells were resuspended in YPD medium (1% yeast extract, 2% peptone, and 2% dextrose) and incubated at 30°C overnight. The cells were then concentrated, resuspended in TE, pH 7.5, and spread on medium containing 100 μg/ml of nourseothricin. The plasmids used for fluorescence microscopy are listed in . pTD125 and pUN100DsRedNOP1 were gifts from D. Amberg (State University of New York Upstate Medical University, Syracuse, NY) and O. Gadal (Institut Pasteur, Paris, France), respectively. To construct pTG003, pRP1152 [pRS416, ] (a gift from R. Parker, University of Arizona, Tucson, AZ; ) was digested with the restriction enzymes HindIII and SacI to isolate the DNA fragment containing the fusion and cloned into the HindIII and SacI sites of pRS315 (). Subsequent transformations and analyzes were performed as described previously (). To make plasmid pTD125 GFP:POP1, the gene was amplified off the plasmid pTC101 using the primers OTC4 (5′-CTC GTC GAC CAT ATA ATA AGA TTT GTT GCC ACT-3′) and OTC5 (5′-CGC GGA TCC ATG AGC GGGA GTTT GTC TAG AG-3′). The resulting 2.7-kb fragment containing was cloned into the BamHI and SalI sites of pTD125 (). Plasmid pTG105 was constructed by cloning the structural gene into the HindIII and BamHI sites of pTD125. was generated by PCR from the plasmid p33/pMES145RMP1 () using the primer ORMP1/Bam(for) (5′-AT ATA GGA TCC ATG GAT GAG ATG GAT AAT GTG-3′) and the reverse primer ORMP1-2 (5′-AA AAA AGC TTA TCC GAA TAT GCC ATC AAT GGC-3′). For fluorescence microscopy, strains were grown in SCD (2% dextrose, 0.5% ammonium sulfate, 0.17% yeast nitrogen base, 50 μg/ml phenylalanine, 40 μg/ml tryptophan, 30 μg/ml lysine, 20 μg/ml methionine, 40 μg/ml adenine, 20 μg/ml histidine, 30 μg/ml leucine, and 20 μg/ml uracil) lacking leucine and uracil for plasmid selection and supplemented with adenine (20 μg/L) to reduce the fluorescence background that can occur in strains with an mutation. A 3-ml aliquot of SCD (−leucine, −uracil, +adenine) media inoculated with a single colony was grown at 30°C for 20–24 h. A 10-μl aliquot of the overnight culture was passed into fresh media and grown another 20–24 h at 30°C. The second overnight culture was then used to inoculate 20 ml of SCD media and grown to a density of 0.5–1.0 × 10 cells/ml. Live cells with fluorescent-tagged proteins were examined on an Axioskop 2 (Carl Zeiss MicroImaging, Inc.) at 100× magnification using a digital camera (ORCA-ER; Hamamatsu) to capture single-focal plane images. All images were processed using Openlab software (Improvision). Final figures were generated using Photoshop software (Adobe). Wild-type strain TLG205 was used for hydroxyurea arrest experiments. In each experiment, cell cultures were grown to early log phase (0.3–0.7 × 10 cells/ml) in defined SCD media (described in the previous section). Hydroxyurea was added directly to the 20-ml culture for a final concentration of 10 mg/ml. The culture was incubated at 30°C for an additional 3 h, at which time ∼90% of the cells were arrested as medium budded cells. Strains carrying the cell division cycle mutations (TLG206), (TLG208), (TLG285), and (TLG277) were used to arrest cells in late mitosis. Cultures were grown as described in the previous section except at 25°C. When cultures reached a density of 0.5–1.0 × 10 cells/ml, they were shifted to 34°C for 2–4 h. Strains were grown in YPD at 25°C to a density of 0.7–0.8 × 10 cells/ml. Half of the culture was harvested and washed, and the RNA was extracted as previously described (). The other half of the culture was shifted to 37°C for 3 h, after which time RNA was extracted. Equal amounts of whole-cell RNA were separated on a 1% agarose and 2.2 M formaldehyde gel and then transferred to a nylon membrane (). Northern blots were probed using a 1,360-bp HindIII–SpeI fragment of the structural gene radio-labeled with α-[P]dCTP using the Prime-it kit (Stratagene). After a 12-h hybridization at 42°C, the membrane was washed three times at 45°C in 1× SSPE and 0.1% SDS and then three times at 45°C in 0.1% SSPE and 0.1% SDS. The hybridized blot was analyzed on a phosphoimager. Yeast deletion strains were obtained from the Euroscarf gene knockout collection, derived from strain BY4741. These were crossed to the RNase MRP RNA mutant in the strain yJA203, in which the chromosomal gene was replaced with NAT. The RNase MRP mutation in this strain is masked by a plasmid-borne wild-type copy. Diploids were sporulated, and 20 tetrads were dissected for each cross. Three tetrads were selected from each verified interaction, and the level of growth defect was quantified by serial dilution and plating of haploid double-mutant cells alongside haploid single-mutant cells, both in the presence and absence of 5-fluoroorotic acid (5-FOA). Complete absence of growth for the haploid double-mutant cells compared with the haploid single-mutant cells on 5-FOA indicated a synthetically lethal genetic interaction. Reduced growth for the haploid double-mutant cells compared with the haploid single-mutant cells on 5-FOA indicated a growth-defective genetic interaction.

Axons of the central nervous system (CNS) demonstrate no functionally significant regeneration after injury, in contrast to those of the peripheral nervous system, which regenerate vigorously, leading often to complete functional recovery. This lack of regeneration generally results in partial disability or complete paralysis after a CNS injury. However, some adult CNS axons can grow through a peripheral nerve graft (), suggesting that the local glial environment of the adult CNS is a major cause of the lack of regeneration. So far, three major inhibitors—Nogo, myelin-associated glycoprotein (MAG), and oligodendrocyte-myelin glycoprotein (OMgp)—expressed by oligodendrocytes and myelinated fiber tracts have been identified. Interestingly, all these inhibitors were found to bind to the Nogo receptor (NgR) in complex with p75 or TROY, members of the TNF receptor family, suggesting that they have common signaling pathways (). However, some reports suggest that inhibition of these molecules alone is insufficient for regeneration after CNS injury (). MAG knockout mice exhibited little or no enhancement of axonal regeneration in the spinal cord. There seems to be some controversy concerning Nogo knockout mice and NgR-deficient mice. Neither depletion of functional p75 nor administration of a soluble p75-Fc at the lesion site promoted regeneration of the injured spinal cord. These findings prompted us to search for new inhibitors. Repulsive guidance molecule (RGM), which has been reported as the 33-kD mass tectum repellent in chick, induces the collapse of temporal but not nasal growth cones and guides temporal retinal axons in vitro (; ; ). RGM binds to neogenin, identified as a netrin-1 receptor and homologue of DCC (deleted in colorectal cancer), mediating its repulsive activity toward retinal axons (). During chick development, neogenin functions as a dependence receptor, inducing cell death in the absence of RGM (). Three mouse proteins, homologous to chick RGM, termed mRGMa, -b, and -c (; ; ) have been reported. Mouse RGMa is highly homologous (80% identity) to chick RGM. Functional studies in RGMa mutant mice revealed the role of RGMa in controlling cephalic neural tube closure (). We reported that up-regulation of RGMa was observed at the lesioned or damaged site after spinal cord injury (SCI) in rats () and focal cerebral ischemia and traumatic brain injury in humans (). In addition, neogenin and other netrin-1 receptors are constitutively expressed by neurons and glial cells in the adult rat spinal cord (). These findings prompted us to hypothesize that RGMa may play a role in inhibiting axonal regeneration after CNS injury. In this study, we show that RGMa inhibits neurite outgrowth in postnatal cerebellar neurons in vitro. RGMa expression is induced after SCI in rats at the lesion site, in the developing scar tissue, and on the myelinated fiber tracts. Local administration of a neutralizing antibody to RGMa significantly facilitates locomotor improvement and axon regeneration after SCI. We first asked whether RGM contributes to the inhibition of mammalian CNS neurite outgrowth in vitro. Cerebellar granule neurons were used because they express the receptor for RGMa (Fig. S1, available at ). We cultured cerebellar granule neurons from postnatal rats (postnatal days 7–9) on confluent monolayers of either CHO cells expressing rat RGMa (RGMa-CHO cells) or control CHO cells for 24 h and assessed the neurite outgrowth rate (the coculture assay). Neurite outgrowth was significantly inhibited when grown on RGMa-CHO cells (). To explore the signal transduction mechanism involved in the inhibition of neurite outgrowth, we assessed whether the neuronal effects of RGMa are dependent on the small GTPase RhoA or its downstream effector, the Rho-associated serine/threonine kinase (Rho kinase). We cultured the neurons on RGMa-CHO cells in the presence of 10 μM Y27632, a specific inhibitor of Rho kinase (), for 24 h and observed that the inhibitory activity of RGMa was abolished by Y27632 (). To directly assess whether RhoA is involved in the signal transduction of RGMa, the activity of RhoA was determined using the RhoA binding domain of the effector protein, Rhotekin (; ). The assay revealed that within 10 min after the addition of soluble RGMa (see the following paragraph), extracts of the cells contained increased amounts of GTP-RhoA compared with those of the control cells. These results suggest that RGMa inhibits neurite outgrowth by a mechanism dependent on the activation of the RhoA–Rho kinase pathway. Because rat RGMa, as well as chick RGM (), may be a glycosylphosphatidylinositol-linked protein that anchors to the cell surface, we examined whether the soluble form of RGMa shows the same activity. The RGMa-CHO or control CHO cells were treated with phosphatidylinositol-specific PLC (PI-PLC). RGMa was released into the supernatant from the RGMa-CHO cells by the treatment as assessed by immunodetection on Western blots (). We then cultured the neurons with the supernatant for 12 h and measured the neurite length (; the soluble RGMa assay). There was no significant difference in the neurite length between neurons cultured in conditioned media from PI-PLC–treated control CHO cells and those treated with media from nontreated control CHO cells, demonstrating that PI-PLC itself had no effect on neurite outgrowth. Neurite outgrowth was significantly inhibited when the neurons were cultured in conditioned media from PI-PLC–treated RGMa-CHO cells, demonstrating that the soluble form of RGMa inhibited neurite outgrowth. Next, we cultured the neurons on PI-PLC–treated RGMa-CHO cells for 24 h and observed that the inhibitory activity of RGMa was significantly reduced (), suggesting that RGMa was released from the cells by the PI-PLC treatment. The findings that RGMa activates RhoA and inhibits neurite outgrowth prompted us to examine whether RGMa signals via the NgR. To determine whether RGMa binds to the NgR, the neurons were cultured for 24 h on RGMa-CHO cells () or with soluble RGMa () in the presence or absence of 1 μM NEP1-40 (). The inhibitory effect of RGMa was not significantly altered by NEP1-40, suggesting that RGMa inhibits neurite outgrowth by an NgR-independent mechanism. To examine the protein expression of RGMa in the adult CNS, antisera were generated against a synthetic rat RGMa–specific peptide and affinity purified. As shown in Fig. S2 A (available at ), the anti-HA antibody as well as the anti-RGMa antibody recognized a protein with an approximate molecular mass of 33 kD in the lysates from the CHO cells expressing HA-tagged RGMa (RGMa-CHO cells). Identical patterns of immunostaining were observed with the anti-HA and anti-RGMa antibody in the RGMa-CHO cells. The anti-RGMa antibody detected a 33-kD band in the purified CNS myelin () and lysates of the rat spinal cord (). In addition, we confirmed that the anti-RGMa antibody does not cross-react with chondroitin sulfate proteoglycans (CSPGs) or MAG (). The specificity of the antibody was further confirmed by its depleted immunoreactivity with excess antigen peptide (residues 309–322 of rat RGMa) in the injured and noninjured adult spinal cords (Fig. S2 B). We then performed immunohistochemistry to investigate the distribution pattern of RGMa protein in the adult rat spinal cord. RGMa expression was found in both white and gray matter (). Double staining using anti-RGMa and anti–myelin/oligodendrocyte-specific protein (MOSP) antibodies (; ) showed that RGMa was expressed in oligodendrocyte cell bodies and their processes in the white matter (). In addition, RGMa was localized to the somata of neuron-specific β tubulin III protein (Tuj1)–positive neurons in the gray matter but not to the axons of these cells (). Double immunostaining with the anti-RGMa antibody and the antibody against the glial fibrillary acidic protein (GFAP), an astrocyte marker, showed no colocalization, demonstrating that RGMa is not present in astrocytes (). Collectively, RGMa is expressed by neurons and oligodendrocytes, and its expression pattern in the spinal cord is similar to that of Nogo or OMgp (). As our data demonstrate that RGMa inhibits postnatal CNS neurons in vitro () and that RGMa is expressed by neurons and oligodendrocytes in the adult spinal cord (), RGMa may play a role in inhibiting axonal regeneration after CNS injury. To test this hypothesis, we examined RGMa expression after SCI. Western blot analysis revealed an increased RGMa expression in the spinal cord 7 d after injury (). We performed immunohistochemistry on fresh frozen sections obtained at 6 h and 1, 3, and 7 d after dorsal hemisection of the thoracic spinal cord in rats. RGMa expression was induced around the injury site after the surgery and was detected in the lesion epicenter () and the white matter rostral () and caudal (not depicted) to the lesion site. In the epicenter area (), the tissue structure appeared normal 6 h after injury. Subsequently, when degenerative changes began to be observed 1–3 d after injury, immunoreactivity for RGMa was found in the cells around the lesion site and in areas of aberrant extracellular matrix. This extracellular immunoreactivity may be attributable to the degeneration of RGMa-expressing cells or to the secretion of RGMa by RGMa-positive cells. In the white matter adjacent to the epicenter area (), the number of RGMa-immunopositive cells and the intensity of immunoreactivity remained almost unaltered after 6 h. However, 1–3 d after injury, the number of RGMa-positive cells had progressively increased. At 7 d after injury, there was a significant increase in the number of RGMa-immunopositive cells in the epicenter area. We performed a double-labeling experiment 7 d after the injury to characterize the RGMa-expressing cells. Isolectin B4 (IB4)– and RGMa-positive cells were detected in the epicenter area (), suggesting that RGMa is expressed in the microglia/macrophages at the lesion epicenter (). MOSP- and RGMa-positive cells were detected in the white matter adjacent to the epicenter area, demonstrating that RGMa is expressed in the oligodendrocytes in this area. We found that all the MOSP-expressing cells also expressed RGMa, whereas RGMa-positive and MOSP-negative cells were detected particularly around the lesioned site (). The results of the double staining for IB4 and RGMa suggest that an increase of these cells is attributable to an increase in the number of RGMa-expressing microglia/macrophages in the epicenter area. We did not observe GFAP- and RGMa-positive or Tuj1- and RGMa-positive cells in the lesion epicenter or in the adjacent white matter (unpublished data). These results suggest that RGMa with its potent inhibitory activity of neurite outgrowth inhibits axonal regeneration after SCI and that blockage of it offers potential for enhanced recovery after SCI. We therefore investigated whether the anti-RGMa antibody could neutralize the inhibitory activity of RGMa in the coculture assay () and the soluble RGMa assay (). In both assays, the addition of the anti-RGMa antibody but not the control IgG significantly reversed the inhibitory effect of RGMa, demonstrating that the anti-RGMa antibody functions as a neutralizing antibody against RGMa in vitro. As our data demonstrate that RGMa is expressed by the CNS myelin, the anti-RGMa antibody may reverse the inhibitory activity of the CNS myelin on neurite outgrowth. To test this hypothesis, we cultured neurons on poly--lysine (PLL)–coated or PLL + myelin-coated chamber slides for 24 h in the presence or absence of the anti-RGMa antibody (). The inhibitory activity of myelin on neurite outgrowth was partially blocked by the anti-RGMa antibody. We next performed immunodepletion of the CNS myelin with anti-RGMa antibody and cultured neurons on depleted myelin-coated chamber slides for 24 h. Immunodepletion () partially abolished the inhibitory response (). Thus, myelin-derived inhibitors except RGMa contribute significantly to the inhibitory response of CNS myelin. We then assessed whether endogenous RGMa acts as an inhibitor of axon regeneration in the injured CNS. We transected two thirds of the dorsal region of rat spinal cords at the vertebral level Th9/10 (lesions 1.8 mm in depth). This resulted in the transection of the main as well as the lateral corticospinal tracts (CSTs). The neutralizing antibodies to RGMa or the control rabbit IgGs were delivered via osmotic minipumps with catheters placed intrathecally near the thoracic injury site. There was no significant difference between the lesion depths in the control and treated groups (). The locomotor performance of the animals was monitored over a period of 9 wk after injury. The sham-operated rats () achieved full scores according to the Basso-Beattie-Bresnahan locomotor rating scale (BBB; ). All the spinally injured rats became almost completely paraplegic on the first day after the injury () and then gradually displayed partial recovery of locomotor behavior as assessed by the BBB scores. There was no difference in the BBB scores of the anti-RGMa antibody– and control antibody–treated rats up to 4 wk after SCI. In fact, the locomotor performance from 6–9 wk after the surgery was significantly better in rats treated with the anti-RGMa antibody than in those treated with the control antibody. On average, rats treated with the control antibody attained a BBB score of 10.6, whereas those treated with the anti-RGMa antibody achieved a BBB score of 14.9 at 9 wk after the surgery. Thus, the anti-RGMa antibody is effective in treating rats with SCI. The integrity of the dorsal CST in some previously tested rats was assessed by injecting biotin-dextran amine (BDA) into the bilateral sensory-motor cortices. For a group of animals (eight control and six anti-RGMa antibody–treated rats), blocks extending 5 mm rostral and 5 mm caudal to the center of the injury were sectioned in the sagittal plane ( and ). The far rostral as well as the far caudal segments were sectioned in the transverse plane (). For another group (three control and three anti-RGMa antibody–treated rats), the spinal cords were sectioned in the transverse plane (). Samples from the injured rats treated with the anti-RGMa antibody exhibit a completely different pattern of labeling (; and Fig. S5, available at ), as compared with those from the control antibody–treated rats ( and Fig. S4, A–J). Proximal to the lesion, the main CST appears as a tight bundle of fibers, with the traced fibers neither entering nor growing beyond the lesion site in the control rats (). This baseline pattern reflects that axonal regeneration does not normally occur in the CNS. In contrast, numerous ectopic fibers sprouting from the labeled CST were observed both rostral and caudal to the lesion site in rats that were treated with the anti-RGMa antibody (). The longitudinal sections across the lesion site reveal lower retraction of the main CST bundles and a higher number of collateral CST sprouts in rats treated with the neutralizing antibody (; and ). Regenerating fibers with typical irregular meandering growth patterns were frequently observed in the tissue bridges at the level of the lesion in rats treated with anti-RGMa antibody (; and Fig. S5, A and B). These regenerating fibers extended into the lesion scar, the gray matter caudal to the lesion, and along cysts (; and Fig. S3). Notably, many axons traversed the lesion tissue, thereby growing either around cavities or within the lesion epicenter, whereas no BDA-traced CST fibers could be detected in the control rats (). Bundles of axons extended through the lesion tissue and numerous labeled fibers were observed caudal to the lesion sites in the anti-RGMa antibody–treated rats ( and M–R; and Figs. S3 and S5). In only one rat treated with the anti-RGMa antibody, however, did we not detect BDA-positive fiber in the scar tissue or caudal to the lesion site. Serial microscopic images ( and Fig. S4, A–J) and camera lucida drawings of all the BDA-labeled fibers (Fig. S4, K and L) summarize the pattern of CST fiber growth in the rostral-caudal region of the injury. A few undamaged ventral fibers were illustrated in both cases. We then reconstructed the serial sections of the injured spinal cords using all the 70–80 serial longitudinal sections per rat and estimated the number of the labeled fibers. In comparison with the number of labeled fibers observed 4 mm rostral to the lesioned site, >20% of the labeled fibers were seen in the caudal spinal cord of rats treated with the anti-RGMa antibody. However, only a small percentage of the fibers (1% at 1 mm caudal to the injury site) were seen in the control antibody–treated rats (). Most of these fibers follow branched and tortuous courses, instead of a linear trajectory (; Fig. S3, B and C; Fig. S5, B, D, F, and H; and Video 1, available at ). At the caudal-most part of the regenerating fibers, we observed branches sprouting from the main tract ( and Fig. S5 F). We obtained cross-sections of the spinal cord and performed a quantitative analysis of the distribution of the axons to exclude the possibility of axon sparing in the anti-RGMa antibody–treated rats. The total number of labeled fibers 10 mm above the lesion was almost identical between the anti-RGMa and control antibody–treated rats (), indicating that the extent of BDA uptake was the same between the control and anti-RGMa antibody–treated rats. A thorough serial section reconstruction of the lesion site enabled us to conclude that there was no labeled CST axon that extended past the lesion in the normal locations of the dorsal CST in any of the rats (). In rats treated with the anti-RGMa antibody, however, BDA-positive fibers with tortuous appearance were detected in the scar tissue developing at the lesion site (). At 10 mm caudal to the lesion site, the sprouting axons extended through the gray matter to a greater extent than in the white matter in rats treated with the anti-RGMa antibody (). Almost no fibers were seen in the ventral part of the dorsal or dorsolateral column in the anti-RGMa antibody–treated rats as well as the control antibody–treated rats (). The fiber counts in the CST () were not different from those in the negative control (non–BDA-labeled rats; not depicted). There was a little nonspecific staining in the CST, which accounts for the few fiber counts in the CST (). These data strongly suggest that there were no spared CST axons, as axons that are spared and extend into the caudal segments should be present in their normal locations. Multiple fibers with tortuous appearance and unusual branching were seen in the gray matter in injured animals treated with the anti-RGMa antibody (). In only one rat treated with the anti-RGMa antibody, however, did we not detect BDA-positive fiber at 1 mm caudal to the injury site. It should be noted that the labeling of the gray matter caudal to the lesion site might be due to sprouting from undamaged ventral fibers. It is not possible to definitely determine long-tract regeneration. Our findings, however, met many of the criteria for identifying regenerated axons () and raise the possibility that anti-RGMa antibody treatment promotes regrowth of injured CST axons. Finally, we assessed whether the functional improvement is temporally correlated with the anatomical regeneration. We took consecutive sagittal sections of the spinal cords from anti-RGMa antibody–treated animals 3 and 5 wk after injury ( = 3 for each group). shows representative pictures of BDA-labeled CST fibers in the control () and anti-RGMa antibody–treated animals () 5 wk after injury. In anti-RGMa antibody–treated animals, the regenerating fibers extended into the lesion scar but did not extend into the caudal part of the spinal cord (). In the control rats, there was no labeled CST axon that extended past the lesion (). We reconstructed the injured spinal cords and assessed the number of the labeled fibers (). 3 wk after injury, there was no significant difference between the fiber counts at any distance from the lesion site in anti-RGMa antibody– and control IgG–treated rats (). 5 wk after injury, a significant number of BDA-positive fibers (∼25% as compared with the number observed 4 mm rostral to the lesioned site) were detected at the lesion epicenter in anti-RGM antibody–treated rats, whereas almost no fibers were seen in the control rats. These results suggest that functional improvement is temporally correlated with anatomical regeneration/sprouting in animals treated with anti-RGMa. e f a i l u r e o f a x o n r e g e n e r a t i o n i n t h e s p i n a l c o r d i s a t l e a s t p a r t l y a t t r i b u t a b l e t o t h e g r o w t h i n h i b i t o r y p r o p e r t i e s o f t h e C N S w h i t e m a t t e r a n d t h e s c a r t i s s u e d e v e l o p i n g a t t h e l e s i o n s i t e . S e v e r a l o b s e r v a t i o n s i n t h i s s t u d y e s t a b l i s h e d t h e n o t i o n t h a t R G M a i n h i b i t s a x o n r e g e n e r a t i o n a f t e r S C I . R G M a i n h i b i t s n e u r i t e o u t g r o w t h b y a c t i v a t i n g t h e R h o A – R h o k i n a s e s i g n a l i n g p a t h w a y i n v i t r o a n d i s e x p r e s s e d b y o l i g o d e n d r o c y t e s , m y e l i n , a n d n e u r o n s i n t h e a d u l t r a t C N S . A f t e r S C I , R G M a e x p r e s s i o n i s i n d u c e d i n t h e o l i g o d e n d r o c y t e s a n d m i c r o g l i a / m a c r o p h a g e s a r o u n d t h e i n j u r y s i t e . T h e a n t i - R G M a a n t i b o d y , w h i c h b l o c k s t h e i n h i b i t o r y e f f e c t o f R G M a a n d C N S m y e l i n i n v i t r o , e n h a n c e s a x o n a l r e g e n e r a t i o n / s p r o u t i n g a s w e l l a s f u n c t i o n a l r e c o v e r y i n v i v o . Because the previous report () has shown that native chick and functionally active RGM starts with residue 152, we created an HA-RGMa vector in pSecTag2-Hygro (Invitrogen) using the signal peptide of pSecTag2 fused to HA and residues 152–431 of rat RGMa (available from GenBank/EMBL/DDBJ, under accession no. ). FPC-1-MAG-Fc– and MAG-Fc–expressing CHO cells are gifts from M. Endo and K. Mizuno (Tohoku University, Sendai, Japan). Flp-in system (Invitrogen) was used to generate RGMa-expressing cells according to the manufacturer's recommendations. We generated an HA-RGMa fragment containing a signal peptide from pSecTag2 vector using two restriction endonucleases and ligated it into pcDNA5FRT (Invitrogen). This construct (pcDNA5FRT/Igκleader/HA/RGMa) and pOG44 were cotransfected into Flp-in CHO cells, and stable expressing cells were generated after growth in medium containing 500 μg/ml Hygromycin B (Invitrogen) for 2 wk. Expression of HA-RGMa was confirmed by immunodetection using Western blots and immunocytochemistry. We also generated human RGMa–expressing cells and confirmed that human RGMa, either membrane bound or diffusible, potently inhibits the neurite outgrowth of postnatal cerebellar granule neurons. Cerebellar granule cells from rat pups at postnatal days 7–9 were dissociated by trypsinization (0.25% trypsin in PBS for 15 min at 37°C) followed by resuspension in serum-containing medium, trituration, and wash with PBS three times. Cultures were grown in a serum-free DME/F12 medium. For the coculture assay, neurons were plated on confluent monolayers of either RGMa-CHO or control CHO cells in chamber slides (LabTek II; Nunc). For the soluble RGMa assay, confluent monolayers of either RGMa-CHO or control CHO cells were incubated in serum-free DME/F12 medium with or without 2.5 U/ml PI-PLC (Sigma-Aldrich) at 37°C for 3 h. Those treated CHO cells were subjected to the coculture assay. After centrifugation of the culture media at 13,000 for 10 min, the supernatants were collected to remove floating cells. A part of the supernatants was subjected to SDS-PAGE and Western blot analysis to estimate concentration of soluble RGMa with various amounts of the standards (recombinant RGM-A; R&D Systems). Estimated concentration of soluble RGMa was 1.37 (± 0.13 SEM) μg/ml. Neurons were plated in the conditioned media on PLL-coated chamber slides and were incubated for 12 or 24 h. Where indicated, 10 μM of Y27632 (Mitsubishi Pharmaceuticals) was added to the cultures. To inhibit the NgR, 1 μM of NEP(1–40) (Sigma-Aldrich) was added to the culture. For neutralizing antibody assay, anti-RGMa antibody or control rabbit IgG was added to the culture at the concentration of 10 μg/ml. For the neurite outgrowth assay on myelin, we performed myelin preparation from rat brain as described previously (). 3 μg/well of the purified myelin or the depleted myelin were used to coat PLL-coated eight-well chamber slides (). The cells were fixed in 4% (wt/vol) paraformaldehyde and were immunostained with a monoclonal antibody recognizing TuJ1 (1:1,000; Covance). The length of the longest neurite for each β tubulin III–positive neuron was then determined. Cells were lysed in 50 mM Tris, pH 7.5, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 500 mM NaCl, and 10 mM MgCl, with leupeptin and aprotinin, each at 10 μg/ml. Cell lysates were clarified by centrifugation at 13,000 at 4°C for 10 min, and the supernatants were incubated with the 20 μg of GST-Rho binding domain of Rhotekin beads (Upstate Biotechnology) at 4°C for 45 min. The beads were washed four times with washing buffer (50 mM Tris, pH 7.5, containing 1% Triton X-100, 150 mM NaCl, 10 mM MgCl, and 10 μg/ml each of leupeptin and aprotinin). Bound Rho proteins were detected by Western blotting using a monoclonal antibody against RhoA (Santa Cruz Biotechnology, Inc.). To examine the remaining effect of myelin-derived inhibitors except RGMa on neurite outgrowth, we subjected the sample to four rounds of immunodepletion (). The purified myelin was extracted with 20% octylglucoside salt (Dojindo). The extract was centrifuged at 40,000 for 60 min, and the supernatant was incubated with the 20 μg of anti-RGMa antibody overnight at 4°C. 50 μg of protein A beads were added for 3 h with rotation and centrifuged, and the supernatant was subjected to three repeat treatments. The final depleted samples and the protein A beads were subjected to SDS-PAGE and Western blot analysis, and bioassays were performed as described (see Neurite outgrowth assay). CHO cells, cerebellar granule neurons, or adult rat spinal cords were lysed in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, and 0.5% Brij-58 (Sigma-Aldrich), including protease inhibitor cocktail tablets (Roche Diagnostics). The lysates were clarified by centrifugation at 13,000 at 4°C for 10 min, and the supernatants were collected and normalized for protein concentration. Equal amounts of protein were then boiled in sample buffer containing 12% β-mercaptoethanol for 5 min and subjected to SDS-PAGE. The conditioned media, purified myelin, MAG-Fc, and CSPG (Chemicon) were also treated in the same manner except for the lysis buffer treatment. MAG-Fc was purified from stable expressing cells (provided by K. Mizuno). The proteins were transferred onto polyvinylidene difluoride membrane and incubated with 1 μg/ml of a polyclonal anti-RGMa antibody, a monoclonal anti-HA antibody (1:1,000; Sigma-Aldrich), a polyclonal anti-neogenin antibody (1:1,000; Santa Cruz Biotechnology, Inc.), or a polyclonal anti-actin antibody (1:1,000; Santa Cruz Biotechnology, Inc.). For detection, an ECL chemiluminescence system (GE Healthcare) and HRP-conjugated secondary antibodies (1:1,000; Cell Signaling Technology) were used. The domain important for functional activity in chick RGM is the COOH-terminal 150–200 amino acids of the active RGM protein. A synthetic peptide (residues 309–322) was selected as immunogen to generate anti–rat RGMa rabbit antisera. The sequence of the peptide is specific to rat RGMa but has no similarity with RGMb or -c. Antisera were affinity purified and used at 1 μg/ml for immunohistochemistry and immunoblots and at 10 μg/ml for neutralizing antibody assay. To assess the specificity of the anti-RGMa antibody, control and SCI sections were stained in the presence of the rat RGMa–specific peptide (residues 309–322; Fig. S2). An additional control experiment was done by leaving out the primary antibody. These experiments confirmed the specificity of the anti-RGMa antibody. Anesthetized (sodium pentobarbital, 40 mg/kg) female Wistar rats (200–250 g) received a laminectomy at vertebral level T9/10, and the spinal cord was exposed. A number 11 blade was used to cut the dorsal part of the spinal cord at a depth of 1.8 mm. Histologic examination has revealed that these lesions sever all dorsal CST fibers in the dorsal funiculus as well as the lateral CST and extend past the central canal in all animals. For neutralizing antibody assay in animal model, immediately after the spinal cord hemisection, rats were fitted with an osmotic minipump (200 μl solution, 0.5 μl/h, 14-d delivery; Alzet pump model 2002 [Durect Co.]) filled with control rabbit IgG (17 animals, 22.3 μg/kg/day over 2 wk; Sigma-Aldrich) or anti-RGMa antibody (15 animals, 22.3 μg/kg/day over 2 wk). The minipump was placed under the skin on the animal's back, and a silastic tube connected to the outlet of the minipump was placed under the dura at the spinal cord hemisection site with the tip lying immediately rostral to the injury site. The tube was sutured to the spinous process just caudal to the laminectomy to anchor it in place. Afterward, the muscle and skin layers were sutured. The bladder was expressed by manual abdominal pressure at least twice a day until bladder function was restored. Sham-operated rats (five animals) received a laminectomy at vertebral level T9/10, and the spinal cords were exposed. The muscle and skin layers were then sutured. For immunohistochemistry, fresh frozen tissues were obtained from an uninjured spinal cord and from ones at 6 h and 1, 3, and 7 d after injury. After deep anesthesia with diethyl ether, the rats were decapitated and the spinal cords dissected out, embedded in Tissue Tek OCT, and immediately frozen on dry ice at −80°C. Series of parasagittal sections as well as cross-sections were cut at 18 or 50 μm on a cryostat and mounted on APS coating Superfrost-Plus slides (Matsunami). The sections were fixed in 4% (wt/vol) paraformaldehyde for 1 h at room temperature, washed three times with PBS, and blocked in PBS containing 5% goat serum and 0.1% Triton X-100 for 1 h at room temperature. The sections were incubated with primary antibodies overnight at 4°C and washed three times with PBS, followed by incubation with fluorescein-conjugated secondary antibodies (1:1,000; Invitrogen) for 1 h at room temperature. 1 μg/ml of polyclonal anti-RGMa antibody, monoclonal anti-GFAP (1:1,000; Sigma-Aldrich), monoclonal anti-MOSP (1:500; Chemicon), or monoclonal antibody recognizing TuJ1 (1:1,000; Covance) was used as the primary antibody. For double labeling of IB4 from (1:100; Vector Laboratories) and RGMa, the animals were killed by perfusion with PBS followed by 4% paraformaldehyde. The spinal cords were dissected, postfixed, and heated. Samples were examined under a confocal laser-scanning microscope (Carl Zeiss MicroImaging, Inc.) with 4×, 10×, 20×, 40×, and 63× objectives. 1 wk (three control and three anti-RGMa antibody–treated rats), 3 wk (three control and three anti-RGMa antibody–treated rats), or 8 wk (11 control and 9 anti-RGMa antibody–treated rats) after injury, descending CST fibers were labeled with BDA (10% in saline, 3.5 μl per cortex, MW 10,000; Invitrogen) injected under anesthesia at the left and the right motor cortices (coordinates: 2 mm posterior to bregma, 2 mm lateral to bregma, 1.5 mm depth). For each injection, 0.25 μl of BDA was delivered for a period of 30 s via a 15–20-μm inner diameter glass capillary attached to a microliter syringe (ITO). In total, we examined and compared the regenerative responses of 17 control and 15 anti-RGMa antibody–treated rats after SCI. 14 d after BDA injection, the animals were killed by perfusion with PBS followed by 4% paraformaldehyde. The spinal cords were dissected, postfixed overnight in the same fixatives, and cryopreserved in 30% sucrose in PBS. The spinal cord 5 mm rostral and 5 mm caudal to the lesion site (10 mm long) was embedded in Tissue Tek OCT. These blocks were sectioned in the sagittal plane (50 μm) or in the transverse plane, retaining each section. In both cases, transverse sections were also collected from the spinal cord >5 mm rostral and caudal to the injury site (17 control and 15 anti-RGMa antibody–treated rats). Sections were blocked in PBS with 0.5% BSA for 1 h and then incubated for 1 d with Alexa Fluor 488–conjugated streptavidin (1:400; Invitrogen) in PBS with 0.15% BSA. To reconstruct serial parasagittal sections completely, all serial 50-μm-thick sections (∼70–80 sections per animal; eight control and six anti-RGMa antibody–treated rats in , six control and six anti-RGMa antibody–treated rats in ) were evaluated. On each section, the number of intersections of BDA-labeled fibers with a dorsoventral line was counted from 4 mm above to 4 mm below the lesion site. Axon number was calculated as a percentage of the fibers seen 4 mm above the lesion, where the CST was intact. The distance beyond the epicenter of the lesion was scored as positive and otherwise as negative distance. Behavioral recovery was assessed for 9 wk after injury in an open field environment by the BBB. Uninjured and sham-operated rats ( = 5) achieved full scores. Quantification was performed in a blinded manner by two (the first group; eight control and six anti-RGMa antibody–treated rats) or three (the second group; three control and three anti-RGMa antibody–treated rats) observers. The difference of judged counts among the observers was within 1 point on the BBB scoring scale. Fig. S1 demonstrates that cerebellar granule neurons express neogenin. Fig. S2 shows specificity of the anti-RGMa antibody. Fig. S3 shows regenerated CST axons with the anti-RGMa antibody treatment. Fig. S4 displays serial microscopic images of a control IgG–treated rat and reconstruction of the lesioned spinal cords treated with control IgG or the anti-RGMa antibody. Fig. S5 shows injured spinal cords of four animals with anti-RGMa antibody treatment. Video 1 shows regenerated CST axons with the anti-RGMa antibody treatment. Online supplemental material is available at .

The mammalian retina is comprised of both rod and cone photoreceptors (PRs), which initiate the phototransduction cascade upon excitation of their visual pigment by a photon of light. In both PR types, the outer segment (OS) is comprised of stacks of membranous discs in rods and lamellae in cones, which house and compartmentalize the proteins used in the phototransduction cascade. It is commonly thought that the proper development of these organelles is directly linked to normal PR cell function and viability; indeed, mutations in proteins specific to the OS (e.g., the rod visual pigment, rhodopsin) cause a multitude of blinding diseases (). In both PRs, the plasma membrane undergoes further ultrastructural reorganization to form the discs of rod OSs and lamellae of cone OSs (; ). In cones, the membrane lamellae are open and physically contiguous with the plasma membrane, whereas in rods, they become sealed, forming distinct membranous structures (discs) that are separated from the plasma membrane by cytosol. Rod and cone PRs also use redundant and analogous proteins for structural development and phototransduction, and many proteins have a conserved function in both PR cell types (). The precise mechanism of OS morphogenesis is still a matter of active investigation even though the basic features of the process have been known for nearly 40 yr. However, a role for the PR-specific protein Rds (product of the retinal degeneration slow gene) in this process has been suggested based upon its localization to the disc rim, and in vitro data also suggest a fusogenic role for Rds in OS membrane assembly (; ; ; ; ). Rds (also known as peripherin/rds or peripherin-2) is a tetraspanning transmembrane protein that is preferentially expressed in the OSs of rod and cone PRs (; ; ). In the rod-dominated wild-type (WT) mouse retina, the loss of Rds causes a failure of OS generation, a greatly diminished response to light, and a slow degeneration of the PR cell layer (; ; ; ; ). However, these observations are limited by the fact that in the WT mouse retina, the PR population is comprised mostly of rods (>95%), making the study of cones difficult in this animal model. Although Rds is clearly requisite for normal rod OS morphogenesis and function, a similar requirement for Rds in cone PRs has, as of yet, not been established. Furthermore, human mutations in Rds manifest as rod or cone dystrophies with varying severity (; ; ), suggesting this protein has distinct functions in rod and cone PRs. Recently, a knockout of neural retina leucine zipper ( ) has been described in which rod PRs fail to develop and the retina consists entirely of cone PRs (). mouse model as an excellent resource for studying cone PRs (; ; ; ; ). In this study, we took advantage of this model to assess the role of Rds in cone PRs, generating a double knockout mouse that lacked both Nrl and Rds ( / ). We report that in the absence of Rds, cones form atypical OSs consisting of distended membranous structures that do not resemble morphologically normal lamellae. This is in striking contrast to rods, where no OSs form at all in the absence of Rds. retina, where rod function is barely detectable. Finally, our results also suggest that Rds has a role in maintaining interactions between the OS and the specialized extracellular matrix that surrounds cone PRs (the so-called cone matrix sheath [CMS]; ; ; ). To examine the subcellular localization of Rds in cones, we used immunogold cytochemistry and ultrastructural analysis. The majority of Rds immunoreactivity was localized to the rim region of both rod and cone OSs in the WT retina, as previously observed (; ). retina (), establishing the legitimacy of this model for studying Rds function in cones. and mice was verified with PCR genotyping (not depicted) and quantitative RT-PCR (qRT-PCR; ). To determine the impact of the loss of Rds on PR differentiation, we used qRT-PCR to examine the mRNA levels of several genes involved in phototransduction and maintenance of OS structure in the adult retina (). and / mice. retinas (). / retinas relative to the WT retinas and were at comparable levels with those observed in the retinas (; ; ; ). The expression levels of two retinal genes that produce proteins localized to the disc region of PR OSs were also examined. and / retinas (), indicating that these genes were not up-regulated to compensate for the loss of Rds. To evaluate retinal function in vivo, we used electroretinography (ERG) to examine the electrical response of the retina to light stimulation, distinguishing between rod and cone responses by varying the illuminance conditions before delivering a test flash. / mice at postnatal day 30 (P30; ). retina regardless of the light intensity. retinas, a b-wave signal was observed at −0.04 log cd sm and enlarged with increasing flash intensities. / retina, starting at 0.37 log cd sm, which is suggestive of reduced phototransduction sensitivity of nearly 1 log unit. Quantification at 1.89 log cd sm revealed that photopic b-wave amplitudes were undetectable in mice but significantly higher in and / mice as compared with the WT control (). mice results from either the toxic effect of the rapidly apoptosing rod PRs that is detrimental to the surviving cones or loss of the structural support of the rods. / mice show nearly identical a- and b-waves regardless of whether scotopic (dark adapted) or photopic (light adapted) ERGs were recorded, signifying that the ERG responses of this mouse model originate in the cone PRs (Fig. S1, available at ). photopic b-wave had decreased significantly, as previously reported (), and a declining trend in the photopic ERG amplitude of / was observed throughout the time course examined (). At 12 mo of age, the b-wave amplitudes of both strains were not significantly different from those recorded from age-matched WT mice. / is functionally cone dominated, similar to the retina, and capable of significant levels of phototransduction, unlike the retina. / mice compared with that of mice may represent altered phototransduction efficiency and sensitivity because of a structural abnormality resulting from the loss of Rds. / mice (). retinas but absent from and / retinas. The Rds-associated protein Rom-1 displayed an identical pattern of labeling as Rds, which is in support of previous observations that Rds is essential for Rom-1 targeting to the OS (). retinas; however, retinas from and / mice showed only S-opsin immunoreactivity localized in the subretinal space and rosettelike structures and retinal folds. To better define the morphology of the structures labeled by anti–S-opsin, thick sections (25 μM) were examined with confocal microscopy, and rotations of the image stacks were performed (). retinas showed punctate labeling of tightly packed cone OS lamellae that were mostly aligned with the retinal pigment epithelium (RPE), which is typical of OS structure. / retina, a markedly different structure was observed. Instead of normal cone OSs, numerous tubular-shaped structures were located in the subretinal space, representing dysmorphic cone OSs. retinas and were not aligned with the RPE. Upon rotation of the image stack, a latticelike network of these structures was observed, possibly reflecting interactions between these OSs and the RPE microvilli. / mice (). and / retinas (). retinas but was disrupted by whorls and rosettelike structures in the retina and, to a lesser extent, in the / retina. / retinas exhibited an undulating morphology, with numerous retinal folds from the outer nuclear layer toward the inner nuclear layer, which failed to form complete rosettelike structures that were observed typically in retinas of mice. / mice were prone to detachment from the RPE along many regions regardless of the fixation method used (not depicted). The lack of distinct rod OSs was apparent in all panels of except those corresponding to retinas from WT eyes. retina; however, dysmorphic membranous structures lacking the characteristic stacked cone OS lamellae were observed in / retinas. At a higher magnification, these lamellae-less OSs were distended and seemed to conform in shape to the space in which they were located (). Large balloon-shaped contiguous membrane structures also were observed, representing cone OSs that have failed to undergo normal morphogenesis. / retina (), further confirming their identity as cone OSs. / retina, immunohistochemistry was performed to visualize PR-connecting cilia and inner segments (ISs) in the retinas of P30 mice (). The connecting cilium/axoneme and IS appeared intact in both models, as observed by labeling with antibodies to acetylated α-tubulin and to Na/K-ATPase, respectively. Hence, the loss of Rds resulted in gross structural changes in the OS but appeared to have no morphological effect on other PR structures or cellular compartments. We also examined localization of the CMS in relation to cone OSs (). This sheath surrounds the IS and OS of cone PRs to mediate adhesive interactions between the retina and RPE (; ; ). mice, the CMS was labeled by fluorescently conjugated peanut agglutinin (PNA) and appeared within the IS and OS layers. Furthermore, S-opsin immunoreactivity was observed within the OS layer where PNA staining was detected. / retina but was detected solely within the IS layer. In this study, we have demonstrated that the loss of Rds in cone PRs does not affect the differentiation of PRs but causes the formation of morphologically novel distended, membranous OS structures that, nonetheless, are capable of phototransduction. These structures contain S-opsin yet lack the compartmentalization of lamellae, directly implicating Rds in this process during cone OS morphogenesis. Rod PRs lacking Rds fail to form OS structures and possess rhodopsin solely in the tip of the cilium and IS membrane, causing minimal phototransduction activity and subsequent PR degeneration (; ). In contrast, cone PRs lacking Rds form altered OS structures lacking normal lamellar organization but, nonetheless, are capable of phototransduction, albeit with reduced sensitivity. Furthermore, the loss of Rds in cones with concomitant abortive lamellar formation prevents the extracellular matrix from forming around the cone OS, further reinforcing the notion that the OS lamellae are required for proper interactions between cone PRs and their extracellular environment. In these studies, we have used the naturally occurring mutant mouse on a C57BL/6 background, and no photopic ERG signal is detectable using our methods. Previous investigations using mutant mice on a 020/A genetic background revealed a nominal scotopic ERG that would also include the response of surviving rods (). In that study, the ERGs may have been more sensitive, as they were performed by placing a needle electrode into the anterior chamber, whereas our method utilizes a looped platinum electrode placed on the cornea. These differences in genetic background and ERG methodology could explain the variation in results obtained between previous work () and this study. The data presented here support a model of cone OS membrane morphogenesis that predicts OS lamellae rim formation to be a second stage of morphogenesis after evagination of the plasma membrane from the connecting cilium (). A previous study demonstrated ultrastructural localization of Rds in the rim regions in cones opposite to the connecting cilium where the membrane invaginates; however, in the rod PR, the OS plasma membrane is separated from the discs, and Rds localizes to the rim on both sides of the disc (). Several studies have also shown the fusogenic properties of Rds (; ; ), which further implicate its role in disc membrane morphogenesis. Based upon these models, we propose that the loss of Rds in cone PRs causes a morphogenic event in which plasma membrane evagination occurs but invagination fails, resulting in the formation of a dysmorphic OS organelle devoid of lamellae. / retina are consistent with a model of cone OS morphogenesis by which growth of the plasma membrane occurs bidirectionally from the connecting cilium (). Further studies to reveal the subcellular compartmentalization of phototransduction proteins in this novel structure may reveal the exact purpose for the utilization of the disc membrane shape in the normal (WT) OS. This phenotype also demonstrates an inherently different role for Rds in rod versus cone PRs. It appears that cones only require Rds for membrane pinching to form the OS lamellae; however, in rods, Rds may have an additional role in OS development because its absence results in a more severe morphogenic outcome whereby the OS does not form and the rod PRs undergo apoptosis. / retina suggests either a direct involvement of Rds in tethering the CMS to the OS or a more generalized dependence on normal OS structure for PR–matrix adhesional competence. The CMS is required for a variety of cell–matrix and cell–cell interactions, including trophic and metabolic interactions with the adjacent RPE (; ; ). The failed establishment of the CMS around these OSs suggests that disc morphogenesis (or normal OS formation) is somehow integrally linked to CMS association. Future biochemical studies to determine the precise role of Rds in the development and maintenance of this association may provide further insights into the compositional and functional differences between rod- and cone-associated extracellular matrix. These observations also have implications regarding therapeutic strategies for treating human diseases involving Rds mutations that cause diseases specific to cone PRs. It is possible that a complete absence of Rds may be more advantageous than having mutant isoforms of Rds in cone PRs because human mutations in Rds that initially display cone-specific dysfunction could be caused by detrimental protein associations (e.g., aggregation of mutant Rds protein). In this regard, the use of RNA interference methodologies to silence specifically in cone PRs may be a beneficial approach. All mice were bred into and assessed on a C57BL/6 background. All experiments and animal maintenance were approved by the local Institutional Animal Care and Use Committee (Oklahoma City, OK) and conformed to the guidelines on the care and use of animals adopted by the Society for Neuroscience and the Association for Research in Vision and Ophthalmology (Rockville, MD). The Nrl mice were provided by A. Swaroop (University of Michigan Kellogg Eye Center, Ann Arbor, MI). Total RNA was extracted from the retinas of a single mouse using TRIzol reagent (Invitrogen) and DNase treated with RNase-free DNase I (Promega). Reverse transcription was performed using an oligo-dT primer and Superscript III reverse transcriptase (Invitrogen). Primers for all genes were designed to span introns as to avoid amplification from genomic DNA. Primers for spanned exon 2, where a 9-kb genomic insertion of a viral element causes the loss of Rds in the mouse, so amplification only occurred in those samples harboring a WT allele. All primer sequences are available in Table S1 (available at ). qRT-PCR was performed in triplicate on each cDNA sample using a real-time PCR detection system (iCycler; Bio-Rad Laboratories), and ΔcT values were calculated against the neuronal housekeeping gene hypoxanthine phosphoribosyltransferase (). was assigned an arbitrary expression level of 10,000, and relative gene expression values were calculated by the following calculation: relative expression = 10,000/2, where ΔcT = (gene cT – cT). This was repeated with three independent samples for each genotype, and the mean expression value is presented with the SD. Agarose gel electrophoresis and disassociation curve analysis were performed on all PCR products to confirm proper amplification. ERG analyses were performed as previously described (). In brief, after a minimum of 4-h dark adaptation, animals were anesthetized by intramuscular injection of 85 mg/kg ketamine and 14 mg/kg xylazine. For the assessment of scotopic response, a stimulus intensity of 1.89 log cd s m was presented to the dark-adapted dilated eyes in a Ganzfeld (GS-2000; Nicolet). The amplitude of the scotopic a-wave was measured from the prestimulus baseline to the a-wave trough. The amplitude of the b-wave was measured from the trough of the a-wave to the crest of the b-wave. To evaluate photopic response, animals were light adapted for 5 min under a light source of 1.46 log cd m intensity. Afterward, a strobe flash was presented to the dilated eyes in the Ganzfeld with various intensities (−0.99–2.86 log cd s m). The amplitude of the photopic b-wave was measured from the trough of the a-wave to the crest of the b-wave. Significance was determined using one-way analysis of variance and post-hoc tests using Bonferroni's pairwise comparisons (Prism, version 3.02; GraphPad). Tissue fixation and sectioning were performed as previously described (). In brief, eyes from P30 mice were enucleated and fixed in 4% PFA/PBS for 16 h before paraffin embedding. Tissue sections (10-μm thickness) were obtained with a microtome, deparaffinized, rehydrated as described previously (), and blocked in 5% BSA/PBS for 30 min at RT. Slides were briefly washed with PBS and incubated with the primary antibody in 1× BSA/PBS for 2 h at RT followed by a brief wash in PBS and incubation with the secondary antibody in 1× BSA/PBS for 30 min at RT. After a brief washing in PBS, Vectashield with DAPI (Vector Laboratories) was applied, and the slide was coverslipped. Primary antibodies (with dilutions) and sources were as follows: anti–Rds-CT (1:200); anti-Rom1 (1:200); monoclonal anti-rod opsin (Rho 1D4; 1:1,000) provided by R. Molday (University of British Columbia, Vancouver, Canada; ); rabbit anti–mouse S-opsin (1:500), a gift from C. Craft and X. Zhu (Doheny Eye Institute, University of Southern California, Los Angeles, CA); acetylated α-tubulin (1:200) from Sigma-Aldrich; and anti–Na/K-ATPase (1:100) from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA; ). All secondary antibodies (FITC or Cy3 conjugates; Jackson ImmunoResearch Laboratories) were applied at a dilution of 1:1,000 from the original stock. Before incubation with antibodies against acetylated α-tubulin and Na/K-ATPase, antigen retrieval was performed by incubating slides in 10 mM citrate buffer, pH 3.0, for 30 min at 37°C followed by a brief rinsing in PBS. For PNA staining, AlexaFluor488-conjugated PNA (Invitrogen) was applied at a 1:200 dilution during the incubation with secondary antibody. Sections were viewed at RT with a microscope (Axioskop 50; Carl Zeiss MicroImaging, Inc.) in the autoexpose mode using a 40, 63, or 100× objective. Images were captured with a digital camera (Axiocam HR; Carl Zeiss MicroImaging, Inc.) using Axiovision 3.1 software (Carl Zeiss MicroImaging, Inc.). Methods used for tissue collection and processing for plastic-embedment light and electron microscopy and immunohistochemistry were as previously described (). For conventional light and electron microscopy, mice were perfused with 0.1 M sodium phosphate buffer, pH 7.4, containing 2% (vol/vol) PFA and 2% (vol/vol) glutaraldehyde; for plastic-embedment immunohistochemistry, the buffered fixative contained 2% PFA and 0.1% glutaraldehyde. For light microscopy, tissue sections (0.75–1-μm thickness) were viewed and photographed with a photomicroscope (BH-2; Olympus) in the autoexpose mode using a 20 or 60× DplanApo objective, and images were collected with a digital camera system (DXM-1200; Nikon). For electron microscopy, Spur's resin-embedded or (for immunogold) LR White–embedded tissue sections (silver–gold) were viewed with an electron microscope (100EX; JEOL). For immunohistochemistry, primary antibodies (see previous section) were used at a 1:10 dilution; secondary antibodies (AuroProbe 10-nm gold-conjugated goat anti–rabbit IgG; GE Healthcare) were used at a 1:50 dilution. Table S1 presents primers that were designed to generate amplicons of 180–300 bp using Primer3 software (). Fig. S1 shows mice that were dark adapted for a minimum of 4 h or light adapted for at least 5 min, and scotopic ERG analyses were performed as described above. Online supplemental material is available at .

Scramblases are a family of four proteins in humans and mice. They are conserved across species, and at least one member of this family is encoded in organisms ranging from yeast to human beings (). Human Scramblase 1 (PLSCR1) was initially identified as a protein thought to be involved in facilitation of transbilayer movement of phospholipids across the plasma membrane (; ; ). Many cell membranes harbor a Ca-dependent mechanism that can facilitate transbilayer movement of phospholipids between the two leaflets, leading to loss of membrane lipid asymmetry termed scrambling. Scrambling is observed in platelets, erythrocytes, and other cells after elevated intracellular Ca and cell injury by complement and also in cells undergoing phagocytosis and apoptosis (). Attempts to purify and clone proteins active in scrambling led to the identification of human PLSCR1. Scramblases are type 2 single-pass transmembrane proteins (). The four members identified in mice and humans have been named PLSCR1 through -4. All except PLSCR2 contain a proline-rich NH-terminal domain with PXXP and PPXX motifs that can potentially bind to SH3 and WD40 domain–containing proteins. An intracellular EF hand, a Ca binding motif, precedes the transmembrane segment. Mice lacking two of these Scramblases, 1 and 3, were recently generated. Cells derived from PLSCR1-null mice showed alterations in granulocyte production in response to growth factors and antiviral response to interferon (). PLSCR3-null mice and double mutants of PLSCR1 and -3 nulls showed adiposity, dyslipidemia, and insulin resistance (). Neither of the knockout mice shows defects in events associated with lipid scrambling. This suggests either that the other two members of the PLSCR family maintain Scramblase activity or that the presumed activity of this protein in scrambling is incorrect. Hence, the actual in vivo biological function of members of the Scramblase family has yet to be elucidated. The genome encodes two ubiquitously expressed proteins with strong sequence homology to mice and human Scramblases. Using a reverse genetic approach, we have generated flies lacking each of the two Scramblases individually and flies lacking both Scramblases. Our studies using these mutants indicate that Scramblases do not play a deterministic role in scrambling of phospholipids that accompanies developmentally regulated apoptosis or in immune response involving phagocytosis of bacterially infected cells. Further analyses indicate that Scramblases play a modulatory role in the process of neurotransmission at the larval neuromuscular junction (NMJ). The deduced sequences of the mammalian PLSCR family of proteins reveal them to be type 2 plasma membrane proteins with the transmembrane domain at the COOH terminus and a proline- and cysteine-rich acidic domain. The NH-terminal segments contain PXXP and PPXY sequences that are potential SH3 and WW domain binding sites, respectively. The COOH-terminal ends of the proteins are conserved and contain a putative Ca binding motif proximal to the transmembrane domain. There is considerable sequence variation at the NH-terminal segment of Scramblases. The PLSCR2 member of the family lacks the NH-terminal proline- and cysteine-rich sequence. genome analysis identifies two genes with significant homology to mammalian Scramblases (; ). These are CG32056 and CG1893. Protein encoded by CG32056 Scramb 1 shows 37% identity with mouse Scramb 1 (PLSCR1) protein, whereas that of CG1893 or Scramb 2 protein shows 41% identity with Scramb 2 (PLSCR2). The Scramb 1 and 2 proteins of are 56% identical and 69% similar to each other (). Although Scramb 1 includes the proline-rich region NH-terminal domains with PXXP and PPXX motifs that bind SH3 and WD40 region, Scramb 2 protein lacks this stretch of amino acids and in this regard is closer to mammalian Scramb 2 protein. A third gene, CG9804, has been labeled as Scramblase in the annotated genome. However, this protein shows only 25% identity with either mammalian or the Scramblases and seems to have branched out earlier during evolution. Moreover, unlike the other two proteins, this protein has restricted expression, as we fail to see ubiquitous expression by Western analysis (unpublished data). CG10427 stated as a Scramblase in the Flybase was annotated as an uncertain gene and has been eliminated in release 3 of the genome annotation (this aberration resulted from annotation errors of CG32056 gene region). To profile the tissue and developmental expression of Scramblases in , we have generated polyclonal antibodies against both Scramb 1 and 2 proteins and in addition have obtained monoclonal antibodies against Scramb 1 protein. We have also succeeded in expressing a GFP-tagged version of Scramb 2 using the upstream activating sequence (UAS)–Gal4 system (). Western blot analysis of wild-type demonstrates that the two proteins are expressed ubiquitously and throughout development (). Lower levels are observed by Western analysis in early embryos before cellularization. The low abundance in early embryos is attributable to the fact that these are predominantly plasma membrane proteins (see below and Fig. S1, available at ). The embryo begins as a single cell and a syncitium before undergoing cellularization, and the plasma membrane surface grows ∼30-fold during cellularization of the syncytial embryo. Immunofluorescence localization and analysis of transgenic flies (GFP–Scramb 2 fusion) localize the proteins predominantly to the plasma membrane (not depicted). We have colocalized Scramb 1 protein with HRP antigen at the plasma membrane in the larval neurons and the Scramb 2 protein with Notch that localizes to the plasma membrane at the apical margins of the cells of the larval eye-antennal anlagen (Fig. S1). Although Scramb 2 is homogeneously localized at the plasma membrane, Scramb 1 shows a nonhomogenous distribution along the plasma membrane, with areas of intense staining interspersed with areas of homogenously strong staining. The localization of the two proteins suggests a role for them in events involving the plasma membrane. Because the in vivo role of Scramblases in lipid scrambling and other cellular functions remains unresolved, we decided to generate mutant flies lacking the family of Scramblases and then carry out functional analyses. We have used a Western blot–based reverse genetics approach to obtain mutants for both Scramblase genes in (; ). The lack of predictable phenotypes defining the mutant complicates many genetic screens. Therefore, we used a previously documented Western blot–based screening strategy that monitors loss of protein expression (loss of Scramblase expression in our case) on immunoblots (; ). In this protocol, male flies are randomly mutagenized, by feeding them the chemical mutagen, ethyl methane sulfonate (EMS). Mutagenized males are crossed to female balancer flies. Balanced progeny are crossed to flies with a large chromosomal deficiency that uncovers the genomic interval including the gene. Single heads from flies trans-heterozygous for the deficiency and a mutagenized third chromosome are screened for the loss of Scramb 1 antigen by Western blots (). From a screen of 1,950 EMS-mutagenized third chromosomes, 1,900 lines that were viable in trans to were obtained. Western analysis of these lines identified a mutant line expressing no Scramb 1 protein (, bottom). The gene was isolated from this mutant and subjected to sequence analysis. The -null mutant had a T-to-A transition in the first intron of the coding region. This leads to a splicing defect (confirmed by RT-PCR across intron 1) and truncation of the protein after 38 amino acids (+ 3 amino acids of the intron). Thus, the mutant is either a null allele or a severely hypomorphic mutant. Likewise, in the screen, 2,100 lines, viable in trans over were established and screened by Western analysis (, top). One of the lines showed no antigen reactivity against the Scramb 2 antibody (, bottom). Sequencing of the gene from this line showed a T-to-A transition at the first exon–intron junction. In this instance, the splicing defect (confirmed by RT-PCR) leads to termination of the gene after 57 amino acids (an additional eight amino acids from read-through in the intron). Hence, the mutant is either a null or severely hypomorphic mutant. Both mutants were backcrossed three times to (control) to outcross all background mutations. The two mutants are null (or severely hypomorphic) for the respective normal Scramblase proteins. Double mutants of and were generated by meiotic recombination of the individual mutants. The double mutants do not express both proteins as confirmed by Western analysis and also by immunofluorescence at the larval NMJ ( and Fig. S2, available at ).Because the in vivo role of Scramblases in lipid scrambling has not been unequivocally established, we analyzed the mutants in events requiring scrambling of phospholipids, including the exposure of phosphatidylserine (PS) on cell surface in cells undergoing apoptosis and phagocytosis. , , and the double mutants are homozy- gous viable. Programmed cell death is an important feature of development and tissue homeostasis. During development, excess cells and tissues that are no longer useful are removed by apoptosis, and mutants that fail to undergo such normal apoptosis do not survive past embryogenesis (). Because Scramblase mutants are viable and fertile and show no developmental defects, apoptosis during development is clearly not compromised in these mutant flies. As in mammals, apoptotic cell death in is also characterized by PS exposure and clearance of cells by phagocytosis (). In , reaper, hid, and grim genes are necessary for induction of apoptosis. Ectopic overexpression of any of these genes in the fly eye using the eye-specific driver GMR (glass multimer reporter) causes normal photoreceptors to undergo apoptotic cell death, resulting in a severe eye ablation phenotype (). To determine whether Scramblase mutants can modulate this phenotype, we examined eyes from flies overexpressing reaper in each of the three mutant backgrounds. The eye sizes were not significantly different in the three mutant backgrounds compared with control flies (). This indicates that the dynamics of reaper-mediated apoptotic cell death is not altered in Scramblase mutant backgrounds. Because the eye sizes were not different between the control and mutant flies, PS-mediated cell death and phagocytic clearance of apoptotic cells were clearly unaffected in the mutant backgrounds. Thus, Scramblases are not required for either programmed or ectopically triggered apoptotic cell death. Immune defense mechanisms in , like in other metazoan organisms, include physical barriers to infection, innate, and humoral responses (; ; ; ). The immune hemocyte cells, for example, mediate cellular responses such as phagocytosis, encapsulation, and melanization, as well as produce humoral effector proteins. Phagocytic cells recognize a variety of signals on cells primed for phagocytosis (such as apoptotic and bacterially infected cells), including cell surface exposure of PS. We decided to evaluate whether the absence of Scramblases compromised any of these functions, resulting in an observable phenotype. We therefore infected wild-type and mutant cells with a mixture of gram positive and negative microorganisms as described in Materials and methods (; ). Immune competence is measured as viability of flies over time, as shown in . The resistance to an inoculum of mixed microorganisms was comparable between the control and mutant backgrounds. Thus, the immune competence of Scramblase mutants is similar to wild-type flies and not compromised. We conclude that Scramblases do not play a critical role in in vivo events that involve scrambling of phospholipids, such as exposure of PS to the outer leaflet and immunoreactive mechanisms for phagocytic clearance of bacterially infected cells. To demonstrate that PS exposure, a process requiring scrambling of phospholipids, was not compromised in backgrounds where Scramblase protein levels are perturbed, we conducted a series of experiments using Schneider (S2) cells in which we either overexpressed the two Scramblases or depleted them using RNAi. Exposure of cell surface PS during apoptosis has been used as a measure of the ability of cells to scramble phospholipids (; ; ). An early event in apoptosis is exposure of PS on the cell surface, which can be analyzed by binding of fluorescently labeled annexin V that has a high affinity for cell surface PS. The impermeant nucleic acid dye propidium iodide is used in conjunction to exclude cells whose plasma membrane integrity is compromised (necrotic or damaged). To visualize the process of scrambling (i.e., exposure of PS in cells with altered Scramblase expression), we used S2 cells to knock down or overexpress Scramblase and follow PS exposure. Scramb 1 and 2 proteins were knocked down separately or together by RNAi treatment in S2 cells by transfecting with double-stranded RNA (dsRNA) as described in Materials and methods. Western blot analysis of dsRNA-treated cell lysates (+) shows significant depletion of Scramb 1 and 2 compared with untreated lysates (). S2 cells after dsRNA treatment were incubated in a Ca-rich medium containing fluorescently labeled annexin V and propidium iodide. PS exposure was measured as percentage of cells that stained positive for annexin V but negative for propidium iodide (, % scrambling). As seen in (–Act D), there is no difference between control S2 cells and Scramb 1, Scramb 2, and the double (Scramb 1, Scramb 2) knockdown cells. We then tested whether scrambling would be compromised in Scramblase knockdown cells undergoing apoptosis. It has previously been shown that S2 cells undergo apoptosis and expose PS when treated with Actinomycin D (). Using similar conditions, apoptosis was induced in S2 cells by Actinomycin D treatment, and PS exposure was measured as before. There were no apparent differences in the ability of these cells to expose their cell surface PS (, +Act D) as compared with non-RNAi–treated cells undergoing apoptosis. We then stably overexpressed Scramb 1 and 2 proteins under the control of the metallothionein promoter (). depicts Western blots showing dramatic inducible expression of both Scramblases. PS exposure was measured in these cells overexpressing Scramblases. There was no significant difference in the number of cells that were annexin V positive and propidium iodide negative between the control and overexpressors (, % scrambling). The baseline apoptosis was itself slightly higher, probably because of the addition of 0.5 mM copper sulfate to induce protein expression (, −Act D). Next, we induced apoptosis in S2 cells overexpressing Scramb 1 or 2 protein by Actinomycin D treatment and evaluated PS exposure by annexin V binding assays. There was no difference between the vector-transfected control and the overexpressing stable cells (, +Act D). Thus, neither knock down or overexpression of Scramblases resulted in altered exposure of PS in resting cells as well as in cells undergoing apoptosis. The results described above lead us to believe that Scramblases do not play a determining role in scrambling of phospholipids in . We then sought to evaluate the actual in vivo functions of this family of proteins. The Scramblase double-mutant flies appear visibly more active in a vial compared with either of the single mutants or the control flies. Hyperactivity has been associated with neurological alterations in (; ). Because the double mutants were hyperactive and because Scramblases are ubiquitously expressed in the plasma membrane, the NMJ was analyzed in our mutant for defects in synaptic structure and function. larval NMJ is extremely active with intracellular and intercellular signaling involving a plethora of signal-transduction cascades active across plasma membranes at the junction (; ; ; ; ). We rationalized that the structure and function would be sensitive to loss of Scramblase proteins if they had an important role in events involving the plasma membrane. We verified the expression of Scramb 1 and 2 at the larval NMJs by immunofluorescence colocalization with synaptotagmin and HRP (). We also performed coimmunolocalization of Scramb 2 with disc large, cysteine string protein, syntaxin, and Fasciclin II (Fig. S4, available at ). Scramb 1 and 2 proteins are expressed in all segmental nerves and presynaptic and postsynaptic membranes of the larval NMJs (, S3, and S4). We then undertook ultrastructural examination and analysis of the active zones (; ). The analysis revealed a significant increase in the number of synaptic vesicles (SVs) and more vesicles docked at the active one ( and ). The single mutants, on the other hand, did not show significant changes in the ultrastructure of NMJs, except that there was more than one docked vesicle at the active zone of most of the type 1b boutons examined (unpublished data). We reasoned that an increased SV content in double-mutant synapse shown in EM studies should result in an increased loading of FM1-43 at nerve endings after massive vesicle mobilization. FM1-43 binds to membranes and remains trapped in SVs that undergo endocytosis. We analyzed the extent of FM1-43 loading in vesicles cycling through an exo/endo pool of SVs (ECP). Loading of ECP with FM1-43 was achieved by exposing boutons (5 min) to an external solution high in K in the presence of dye (; ). Fluorescence brightness was analyzed 5 min after extensive perfusion with dye-free standard solution. Data analysis revealed that fluorescence brightness in double-mutant boutons was significantly above control (). After ECP loading, we subjected synapses to high-frequency stimulation of the nerve at 10 Hz in a standard external solution containing FM1-43, to also load vesicles cycling through a reserve pool (RP) of SVs. As shown in , this maneuver increases dye load, and fluorescence now distributes over the whole volume of synaptic boutons. Under these conditions also, fluorescence brightness in the double mutant was significantly above control (). Finally, to evaluate the extent of dye loading in SVs cycling through RP, synapses were exposed for a second round to a high K solution void of dye (3 min) to induce massive release of SVs in ECP. The fluorescence that remains after this second exposure to high K labels vesicles that are cycling through the RP (). Data analysis revealed that fluorescence brightness was significantly larger in double-mutant boutons (). Thus, our FM1-43–loading studies are consistent with EM data indicating an excess content of SVs in double-mutant presynaptic terminals. Importantly, as documented in , in all conditions, the extent of FM1-43 loading in the double mutant was rescued to control values by the introduction of a copy of the gene. We then proceeded to investigate ECP exocytosis rate by monitoring the decay of FM1-4 fluorescence brightness after dye loading of vesicles through ECP during low-frequency stimulation of the nerve at 0.5 Hz. Under this condition, transmitter release is maintained essentially by mobilization of ECP (). We found that the time course of decline of fluorescence brightness was similar in double-mutant and control boutons, indicating similar properties of ECP exocytosis (unpublished data). We also investigated the recruitment of SVs from an RP. The RP contains a majority of SVs and has been implicated in use-dependent increase in transmitter release and is mobilized during high-frequency activity at the synapse (; ). 5 min after dye loading of SVs cycling through the RP, RP recruitment rates were assessed by monitoring the decline in fluorescence brightness during high-frequency stimulation at 10 Hz. We found that the rate of RP recruitment was markedly enhanced in double-mutant boutons, as indicated by a significantly faster rate of fluorescence brightness decline (2.4 ± 0.2 min [ = 10] vs. 8.5 ± 0.3 min [ = 10]; ). This result prompted us to investigate whether RP recruitment at rest, in the absence of high-frequency stimulation of the nerve, was also enhanced in the double mutant. Therefore, we evaluated the extent of fluorescence brightness decline at rest, 5 min after SVs cycling through RP were loaded with dye. As shown in (top) at rest, the decline in fluorescence brightness was markedly enhanced in double-mutant boutons (fivefold). Moreover, the extent of fluorescence decline induced by a 5-min stimulation of the nerve at 0.5 Hz in the double mutant was also significantly enhanced relative to control (, bottom). These results reveal that RP recruitment in double-mutant synapses is abnormally enhanced. (D and E) also documents that RP recruitment in double-mutant larvae was rescued to control levels by the introduction of a copy of the gene. We next evaluated the properties of synaptic transmission by recording nerve-evoked postsynaptic currents from larval NMJs (; ; ). Single-mutant synapses displayed transmitter release and plasticity properties that did not differ significantly from control (unpublished data). In contrast, synaptic transmission in the double mutant differed in several aspects from the control. We first observed that the frequency of spontaneous postsynaptic currents was significantly enhanced in the double mutant (). We also found that release probability, quantal content, and nerve-evoked postsynaptic currents were significantly larger in the double mutant (; and ). On the other hand, the amplitude of current and the charge transferred by the release of a single quantum were, within experimental error, the same in control and double mutant (). shows plots of nerve-evoked postsynaptic currents as a function of external Ca concentration in double logarithmic fashion in control, double mutant, and P{}; double mutant. Data analysis revealed a similar Ca cooperativity coefficient for nerve-evoked responses. However, in double-mutant synapses, the curve was shifted toward lower Ca concentrations. We also recorded postsynaptic currents evoked by exposing synapses to a hyperosmotic 0.5 M sucrose solution (). As documented in , neurotransmitter release evoked by application of a 20-s hyperosmotic shock in the vicinity of the synapse was dramatically increased in the double mutant. Importantly, as documented in and , the synaptic defects in the double mutant were rescued by the introduction of a copy of the gene. In further studies, we investigated plasticity properties of double-mutant synapses. Analysis of paired-pulse facilitation revealed that although the extent of facilitation of neurotransmitter release in the double mutant was diminished relative to the control, this difference was not significantly different from the control (). We next evaluated tetanic facilitation and posttetanic potentiation (PTP) by monitoring nerve-evoked currents using a stimulation protocol designed to monitor tetanic facilitation and PTP. Postsynaptic responses were evoked first by stimulating the nerve at 0.5 Hz for 20 s. This was followed by a period of tetanic stimulation at 10 Hz for 50 s. After the tetanic stimulation, the regimen was switched back to 0.5 Hz to monitor PTP. We found that the extent of facilitation of transmitter release elicited by high-frequency stimulation of the nerve did not differ significantly from the control (). On the other hand, we observed that PTP in the double mutant was abnormally prolonged. As shown in , in control synapse, nerve-evoked transmitter release decayed to values before stimulation (, bottom, discontinuous line) in ∼50 s after the cessation of high-frequency stimulation. In contrast, in the double mutant, transmitter release remained significantly potentiated after 100 s, indicating altered short-term synaptic memory. Introduction of a genomic copy of the gene in the double mutant rescued PTP to normal. Also expression of a UAS–Scramb 2 gene driven by elav-GAL4 and neuronal promoter (elav-GAL4; UAS–Scramb 2; double mutant) rescued the content of ECP, RP, mobilization of RP pool of vesicles, postsynaptic current amplitude, quantal content, tetanic facilitation, and posttetanic facilitation to control levels (Fig. S5, available at ). In summary, our electrophysiological inquiries indicate that several aspects of synaptic function, including spontaneous, nerve-evoked release and PTP, are altered in the double mutant. Lipids are asymmetrically distributed between the two leaflets of the plasma membrane. Although the outer leaflet is enriched in choline phospholipids such as phosphatidylcholine and sphingomyelin, the inner leaflet contains anionic phospholipids such as PS and phosphatidylethanolamine. Although these lipids have a tendency to equilibrate along their concentration gradient, specific proteins that catalyze uni- or bidirectional transport of lipids from one leaflet to another maintain the asymmetry. One of the important players in this process is aminophospholipid translocase. Although its molecular identity has not been definitively identified, a P-type ATPase (aminophospholipid translocase) is thought to mediate this function (). However, certain physiological and pathological conditions actively cause collapse of this asymmetry, resulting in scrambling of phospholipids. The most notable situation is in cells undergoing apoptosis wherein cell surface exposure of PS is one of the earliest detectable biochemical events. In fact, the annexin V binding assays for apoptotic cells is based on this exposure of PS. PS exposure has also been proposed as one of the signals that initiate the phagocytic process in cells targeted for phagocytosis. Lipid scrambling is thought to depend on one or more membrane proteins with lipid Scramblase activity. PLSCR1 was thus purified and identified as a putative Scramblase based on in vitro assays of transfer of lipids in bilayer. Subsequently, three other members of this family were discovered in humans and mice. Mice lacking PLSCR1 and -3 have been generated. Phenotypic analysis of the mutants indicated that PLSCR1 mice showed a defective response of hematopoietic precursors to granulocyte colony–stimulating and stem cell factors. PLSCR3 mice show insulin resistance, glucose intolerance, and dyslipidemia. Both knockout mice show no defects in lipid scrambling, suggesting that the presumed activity of this protein is not correct or that other members of the PLSCR family maintain the Scramblase activity. has two definitively identified members of this family, and we have generated mutants in both proteins. Analyses of null (or severely hypomorphic) Scramblase mutants demonstrate that these proteins do not play a critical role in scrambling of phospholipids; instead, they have a modulatory role in the secretory process. We have demonstrated that Scramblases show an overlapping and ubiquitous expression throughout development, and they localize to the plasma membrane in all the tissues examined. The ubiquitous distribution suggests their participation in general housekeeping function involving plasma membrane. We used a Western blot–based genetic screen to obtain null mutants for each of the two Scramblases. We also recombined the two null mutants to generate double mutants of Scramblases. The viability and lack of any developmental defects in the single and double mutants and our studies with reaper expression in mutant animal backgrounds indicate that they do not play a deterministic role in scrambling of phospholipids. This notion is further supported by our data in S2 cells, where neither overexpression nor RNAi-mediated knock down of these proteins has any effect on the ability of resting or apoptotic cells to expose PS. Our studies indicate that loss of both Scramblases results in alteration in both the organization and function of the larval NMJ. The synaptic phenotype of the double mutant includes an increase in the number of vesicles at the synapse (including docked vesicles), a fivefold increase in RP recruitment at rest, a twofold increase in release probability, and an abnormal PTP. The fact that the recruitment of vesicles from RP at rest in the absence of nerve stimulation is dramatically enhanced in the double mutants might explain in part the abnormally elevated number of docked vesicles seen in EM analysis. The increased cycling of vesicles through the RP, as indicated by the almost twofold increase in uptake of FM1-43 (), probably associates with the twofold increase in release probability and evoked release seen in the double-mutant synapse (, A and C; and ). Although increased resting Ca levels can contribute to increased probability of release, we do not believe they are contributing to the double-mutant phenotype. For example, although RP recruitment at rest in the double mutant is enhanced fivefold relative to , the probability of release is only twofold above normal. Because the cooperative factor of Ca for transmitter release is approximately Ca, an increase in Ca at rest in the double mutant would be expected to have a much larger effect on release probability than on RP recruitment. On the other hand, an enhanced recruitment of SVs from RP could contribute to the increased number of SVs available for release observed in the double-mutant presynaptic terminals. An abnormal handling of Ca might underlie the enhanced probability of release found in double-mutant synapses. Indeed, our results indicate a shift to left on the Ca dependence or release, without a change in Ca cooperativity, suggesting that Ca handling might be somehow abnormal in double-mutant synapses. However, paired-pulse facilitation is only slightly reduced in the double mutant synapse, arguing against a grossly abnormal handling of Ca. Also, the rates of ECP exocytosis are known to depend significantly on Ca (). However, contrary to what one would expect if Ca levels were abnormally elevated, we observed that the rates of ECP exocytosis were the same in the control and double-mutant boutons. Thus, our observations would in principle rule out the possibility that grossly elevated Ca levels underlie the abnormal synaptic phenotype observed in double-mutant synapses. In , RP recruitment and transmitter release are fostered by elevations in cAMP levels at synaptic boutons, and RP recruitment is antagonized by PKA inhibitors (). Importantly, in a mutant, with abnormally elevated cAMP levels, RP size is dramatically reduced, probably because of the excess translocation of SVs from RP (). We found that in spite of enhanced RP recruitment, RP size remains vigorous in the double mutant, as documented in . Thus, unlike , enhanced RP recruitment in the double mutant is not accompanied by a reduced RP size. We cannot rule out the possibility that cAMP signaling is altered in double mutant synapses. If so, such alteration should allow for a significant enhancement in RP recruitment without a reduction in RP size. In short, our data seem to rule out in principle a grossly abnormal handling of Ca or a severely altered operation of the cAMP cascade in double-mutant synapses. It is important to note that the effects of cAMP on transmitter release are not limited to fostering RP recruitment. There is evidence that cAMP-gated channels at the presynaptic terminal also play an important role, independent of PKA, in enhancing transmitter release in (). In , RP dynamics has also been associated to the actin cytoskeleton and to the actin binding myosin ATPase complex (; ). Further work should allow us to better understand the role of Scramblases in regulating RP dynamics and transmitter release in . The fact that Scramblases exhibit putative protein binding domains means it is possible that they are able to interact with other proteins and signaling pathways to participate in regulating transmitter secretion and SV trafficking at the presynaptic terminal. We provide the first evidence that Scramblases can play an important role in regulating RP recruitment, transmitter release, and short-term synaptic memory in . We believe that Scramblases have a general role in modulating regulated secretory process. In fact, earlier work done in mammalian Scramblases seems to support this notion, although this conclusion was not drawn in those studies. In one of the earlier studies, Scramblases were found to interact with thrombospondin 1, a secreted protein (). Interferon α, a secreted protein is a known up-regulator of Scramblase (; ). Human Scramblase has been demonstrated to interact with human salivary leukocyte protease inhibitor, a secreted protein (). Activation of Mast cells results in degranulation and secretion of a host of vasoactive amines, proteases, and proinflammatory cytokines. Cultured mast cells RBL-2H3 activated with FcɛRI resulted in up-regulation and phosphorylation of Scramblase protein (). Potentiation of the antiviral action of interferon by Scramblase has also been reported (). PLSCR1 interacts with b-secretase (b-site amyloid precursor protein–cleaving enzyme), which is involved in secretion of the amyloid β-peptide (). PLSCR1 mutants have no defect in scrambling but have deficient granulopoiesis, and PLSCR3 mutant mice display adiposity and dyslipidemia. Finally, B lymphocytes from a patient with Scott syndrome (whose red blood cells show defective scrambling), despite being deficient in Scramblase activity, have normal levels of PLSCR1, and the nucleotide sequence of the corresponding mRNA is identical to controls (). All the aforementioned results and our data from analyzing , , and the double mutants reveal a role for Scramblases in modulating regulated exocytosis and not in the scrambling of phospholipids. A BLAST (basic local alignment search tool) search of the genome database using the mammalian Scramb 2 sequence revealed the existence of two definitive members of Scramblase family of genes. The two Scramblases are Scramblase 67D () and 63B (). genome annotation project has listed these genes as CG1893 and CG32056. CG32056 () is localized to the 67D region on the left arm of third chromosome, and CG1893 gene () is localized to the 63B region on the left arm of third chromosome. A BLAST search against the cDNA database revealed the existence of several cDNA clones from the collection maintained at the time by the Berkeley Drosophila Genome Project and Research Genetics. All clones whose 5′ or 3′ ends showed homology to the mouse Scramblases were obtained from Research Genetics and sequenced. Full-length cDNA clones corresponding to GM13876 for and GH10494 for genes were then used for all subsequent manipulations. The Scramblase cDNAs with appropriate cloning restriction sites were amplified, cloned, and sequenced. For transgenic overexpression in , the appropriate cDNAs were cloned into pUAST vector, and for S2 cell overexpression they were cloned into pRMHa-3 vector (a gift from L. Goldstein, University of California, San Diego, La Jolla, CA). stocks were cultured on standard maize meal agar and maintained at 25°C unless otherwise mentioned. The /TM3, Sb , ; P{w + mC = Actin}-Gal4 stocks were obtained from Bloomington Stock Center. Cyo, 2× GMR Rpr/Sco; MKRS/TM6B flies were obtained from K. White (Massachusetts General Hospital, Boston, MA). males were aged for 3–4 d, starved overnight in empty vials, and subjected to chemical mutagenesis by feeding them on a diet of 25 μM EMS in 3% sucrose as previously described (; ). These flies were then crossed en masse to virgin females with balanced third chromosomes. Single F1 males carrying a mutagenized chromosome over a third chromosome balancer were then crossed to virgin /TM3, , flies in individual vials. The deficiency uncovers the 67D region. From each of these vials, a single progeny that carried the mutagenized chromosome over the deficiency was subjected to Western blot analysis for the loss of respective Scramblase antigen (; ). Vials with lethal mutations in the interval uncovered by the deficiency were set aside to be used in transgenic rescue experiments, if necessary. Single fly heads were homogenized in SDS-PAGE buffer and loaded on gels. The separated proteins were transferred to nitrocellulose and incubated with affinity-purified rabbit polyclonal antibodies raised against the Scramb 1 protein. A single null mutant was isolated in the screen. To outcross all incidental and irrelevant mutations, the null mutant was backcrossed three times to flies and selected by Western analysis. A similar approach was used to isolate the mutant. However, , a deficiency that uncovers the 63B region, was used instead. Again, the null mutant was backcrossed three times to controls to outcross all other mutations and selected by Western analysis. The cDNA was amplified in frame with open reading frame of GFP gene and ligated to generate GFP–Scramb 2 fusion. The gene was then cloned into pUAST vector and injected into embryos to generate transgenic flies expressing GFP–Scramb 2 fusion. pUAST––transgenic flies were generated. These lines were subsequently crossed to actin-Gal4 lines to drive its expression ubiquitously. A similar approach was attempted for gene, but no protein was detected in these flies either by Western analysis or by immunofluorescence. We presumed that the protein was misfolded and degraded. An 11.8-Kb Sal1–Kpn1 fragment of genomic DNA spanning the gene was cloned into pUAST, and transgenic lines were obtained. One of these mapped to the second chromosome and was used in the transgenic rescue experiments. The transgenic expression of Scramb 1 protein was confirmed by Western analysis in the double-mutant background. Rabbit polyclonal and monoclonal antibodies were raised against bacterially expressed, affinity-purified MBP–Scramb 1 protein in which the Scramb 1 protein was further cleaved off from MBP by factor Xa cleavage and purified by electro elution from polyacrylamide gels. Additionally, a COOH-terminal peptide antibody was raised against Scramb 1–derived peptide CFFEKAGNQETDRPGML and affinity purified. Rabbit polyclonal antibodies were raised against Scramb 2 protein expressed using a pET3a expression system and purified by SDS-PAGE and electroelution. For Western analysis, affinity-purified anti-peptide Scramb 1 antibody was used at 1:500 dilution and anti–Scramb 2 antibody was used at 1:2,000–1:4,000. IPP antibodies were a gift from C. Zuker (University of San Diego, San Diego, CA). Anti–Fas II (1:500), anti-synaptotagmin (1:200), anti-cysteine string protein (1:25), anti-synataxin (1:200), and anti-DLG (1:200) monoclonal antibodies were obtained from Iowa Hybridoma Bank. Affinity-purified rabbit anti-synaptotagmin (1:1,000) antibody was a gift from M. Ramaswami and S. Sanyal (University of Arizona, Phoenix, AZ). Anti-HRP (1:500) and secondary antibodies (1:500) were obtained from Jackson ImmunoResearch Laboratories. Anti-CDase (1:3,000) antibodies have been described previously (, ; ). The Notch antibody was a gift from M. Fortini (National Cancer Institute Frederick, Frederick, MD). Embryos and imaginal discs were collected and prepared for immunofluorescent staining as described previously (). Tissues were incubated overnight with the antibodies (1:100–1:500 for anti–Scramb 1 monoclonal and anti–Scramb 2 rabbit polyclonal) at 4°C. These were costained with Alexa 568 or 488–conjugated goat anti–rabbit or goat anti–mouse IgG for 2 h at room temperature, washed, and mounted on Vectashield. For GFP visualization, the wing or eye imaginal discs were isolated from wandering third instar larvae, fixed in 4% formaldehyde, washed, and stained with propidium iodide for nuclear staining. For confocal microscopy, laser-scanning confocal microscopes (models 410 and 510; Carl Zeiss MicroImaging, Inc.) were used in most experiments, using a 63× oil objective, and pictures were obtained at resolutions ranging from 512 × 512 to 2048 × 2048. The pictures were pseudocolored in Photoshop (Adobe) for use in figures. For electron microscopic examinations, wandering third instar larvae grown at 25°C were dissected and fixed using 2% glutaraldehyde and 4% formaldehyde in sodium cacodylate buffer. The tissues were postfixed in 1% osmium tetroxide and dehydrated in graded ethanol and propylene oxide. Specimens were embedded in epoxy resin (Embed-812) and thin sectioned. Transverse sections (90 nm) were stained with uranyl acetate and lead citrate and examined. Mutant Scramblase flies (individual and double mutants) were crossed into GMR-Rpr background to evaluate the effects of these mutations in reaper-induced apoptosis. UAS–Scramb 2 GFP flies were crossed to actin-Gal4 to drive expression ubiquitously. Bacterial infection experiments were done as described previously (; ). , , , and were gifts from D. Kimbrell and R. Schoenfeld (University of California, Davis, Davis, CA). They were individually grown and pelleted. They were subsequently resuspended in LB collectively and pelleted again. The microorganism mixture was then injected into the thorax of individual flies using tungsten needles. A group of 10 individual flies were maintained in each vial, and the vials were replaced with a fresh vial every day. A total of 100 flies were infected for individual samples, and the experiment was repeated three times. and were cloned into pRMHA3 vector, and stable lines were established. The proteins were induced with 0.5 mM copper sulfate, and cells were harvested after 48 h of induction for Western analysis. A 900-bp fragment of and a 600-bp fragment of were PCR amplified with T7 polymerase site as described previously (; ) with minor modifications. RNA was synthesized using the T7 RNA polymerase kit (QIAGEN). The dsRNA were annealed by heating to 75°C for 20 min and slow cooling to room temperature. 1 μg of dsRNA was transfected to S2 cells in a 3-cm dish by Ca phosphate transfection and media changed after 24 h. dsRNA-treated cells were evaluated for protein levels by Western analysis at 24, 48, and 72 h after transfection. For scrambling (apoptotic) index, stable cells induced for protein production for 48 h or cells treated with dsRNA for 48 h were treated with Actinomycin D. They were then incubated in annexin V binding buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, and 2.5 mM CaCl) with annexin V–Alexa 488 propidium iodide. Stained cells were visualized using a confocal microscope (model 510). Approximately 300 cells were counted in three different fields for each experimental sample. Cells that were annexin V positive and propidium iodide negative were then used to calculate the percentage of cells that exposed PS on the surface (scrambling index). Postsynaptic responses evoked by stimulation of the nerve were recorded from segments A2 or A3 of ventral longitudinal muscle 6 or 7 in third instar larvae using a two-electrode voltage clamp as reported previously in a standard working solution containing 128 mM NaCl, 2 mM KCl, 4 mM MgCl, 0.2 mM CaCl, 5 mM Hepes, and 36 mM sucrose, pH 7.3. Osmotic release of neurotransmitter was achieved by application of a hyperosmotic standard solution containing 500 mM sucrose and 50 μM CaCl (; ). The currents were recorded at −80 mV holding potential, and current amplitudes and integrals were analyzed using pClamp software (Axon Instruments, Inc.). Nerves were cut close to the ventral ganglia and sucked into the stimulating pipette. Evoked currents were elicited by stimulation of the nerve at the frequencies indicated using a programmable stimulator (Master-8; MPI). Data acquisition and analysis were done using pClamp software. The quantal content of nerve-evoked responses was estimated by dividing the current integral of individual nerve-evoked currents by the integral of the current evoked by the release of an individual quantum, as described previously (; ). xref #text sup xref #text I n a l l c a s e s , F M 1 - 4 3 u n l o a d i n g w a s m o n i t o r e d i n a s t a n d a r d w o r k i n g s o l u t i o n a t r e s t , i n t h e a b s e n c e o f n e r v e s t i m u l a t i o n o r b y s t i m u l a t i n g t h e n e r v e a t l o w ( 0 . 5 H z ) o r h i g h f r e q u e n c y ( 1 0 H z ) u s i n g a p r o g r a m m a b l e s t i m u l a t o r . Synaptic boutons were imaged using an epi-fluorescence microscope (BX50WI; Olympus) with a 60× objective. Samples were excited for 200–300 ms via a controlled shutter system (Uniblitz; Vincent Associates). Images were captured with a 12-bit cooled charge-coupled device camera (SensiCam; PCO Imaging). Images were acquired and analyzed using Workbench 2.2 software (Axon Imaging). Results are from at least three synaptic boutons from a minimum of six different larvae per experimental condition. Fluorescence analysis was performed after correcting for bleaching and background. Fig. S1 shows colabeling of Scramblase proteins with plasma membrane markers. Fig. S2 shows absence of Scramblase signals in synaptic boutons of the double-mutant larval NMJs. Fig. S3 shows staining of segmental nerves with Scramblases. Fig. S4 shows that Scramb 2 is present in pre and postmembranes of the larval NMJ. Fig. S5 shows rescue of the double-mutant synaptic phenotype by neuronal expression of Scramb 2. Online supplemental material is available at .

The nuclear lamina is part of the nuclear envelope in multicellular eukaryotes opposing the inner nuclear membrane (; ). The major components of the nuclear lamina are type V intermediate filament proteins referred to as lamins (; ) and various lamin-binding proteins (). In mammals, at least one of the B-type lamins, encoded by and , is constitutively expressed and essential for cell viability. The A-type lamins A and C, encoded by , are expressed in most terminally differentiated somatic cells but are absent in early embryos and some undifferentiated cells. Interestingly, lamins A and C are also found in the nucleoplasm as discrete foci, filamentous structures, or are uniformly distributed throughout the nucleus () and have been implicated in chromatin organization (), DNA replication (), RNA Pol II–dependent transcription (), and control of gene expression (). Aside from lamins, numerous lamin-binding proteins are also considered to be important components of the nuclear lamina and of nucleoplasmic lamin structures (; ). Among the best-studied lamin-binding proteins are the LAP2 family members. Of the six alternatively spliced LAP2 isoforms, four—LAP2β, -γ, -δ, and -ɛ—are type II membrane proteins in the inner nuclear membrane (; ) and bind lamin B (; ). LAP2α is structurally and functionally different. It shares only the NH terminus with the other isoforms and contains a unique COOH terminus. LAP2α is localized throughout the nucleus (; ) and has been identified as a specific binding partner of nucleoplasmic A-type lamins (). All LAP2 isoforms as well as the inner nuclear membrane proteins emerin and MAN1 share a common structural motif called the LAP2–emerin–MAN1 (LEM) domain (), which mediates binding to BAF (barrier to autointegration factor), an essential, highly conserved DNA-binding protein in eukaryotes (). Based on these interactions, LEM domain proteins in the nuclear lamina and the nuclear interior have been implicated in chromatin organization (; ). Mutations in the gene have been shown to cause a variety of inherited human diseases (laminopathies) that affect different tissues, including skeletal muscle, heart, adipose tissue, peripheral nerves, and skin, or cause premature aging (; ; ). The molecular basis of these diseases is still unclear. Besides structural defects in lamin complexes, it has also been suggested that disease-causing mutations in may interfere with the proposed gene regulatory functions of lamins (). Among several reported interactions of lamina proteins with transcriptional activators or repressors (), the interaction of A-type lamins () and of nucleoplasmic lamin A/C–LAP2α complexes () with the retinoblastoma (Rb) protein may also regulate gene expression via influencing Rb activity. Rb regulates progression through the cell cycle at the G1→S-phase transition by inhibiting the activity of E2F-type transcription factors in a phosphorylation-dependent manner and by mediating epigenetic changes on the promoter region of E2F/Rb target genes (). It has been shown that Rb is required for the terminal differentiation of many tissues, including adipose and muscle tissue (; ), which are also affected in laminopathies. Based on these data, an intriguing laminopathy disease model has recently been raised, arguing that lamin A/C complexes may cooperate with Rb in controlling cell cycle progression and differentiation of mesenchymal stem cells during tissue regeneration and turnover. Disease-causing mutations in or a lack of lamin A/C may affect the balance between proliferation and differentiation in stem cells, leading to a defect in tissue regeneration (). Intriguingly, mutations in the gene encoding the nucleoplasmic lamin A/C–interacting protein LAP2α has recently been linked to dilated cardiomyopathy in humans, clinically resembling lamin A–linked disease phenotypes (). As LAP2α is only expressed at very low levels in differentiated muscle, it is hard to imagine how mutations in LAP2α can lead to the disease phenotype in fully differentiated heart muscle cells. Therefore, it is tempting to speculate that LAP2α, which has previously been shown to mediate the nuclear retention of Rb (), may also be involved in Rb-mediated control of cell cycle progression and differentiation in muscle precursor cells. Deregulation of this function in patients expressing the disease variant of LAP2α may then lead to impaired tissue turnover. In this study, we investigate whether LAP2α can affect cell cycle progression and differentiation. We found that LAP2α inhibits progression from G1 to S phase and initiates early stages of differentiation in an in vitro adipocyte differentiation culture model. We further show that the cell cycle and differentiation regulatory function of LAP2α requires Rb and involves regulation of the activity of E2F transcription factors. To test the influence of LAP2α expression levels on cell cycle progression, we used two different cell models. First, we generated stable HeLa cell clones expressing a myc-tagged LAP2α under the control of a doxycyclin-dependent promoter, which allowed analysis of the cell cycle phenotypes in a single cell clone expressing different levels of LAP2α. Although HeLa cells are transformed by human papilloma virus E7 oncogene (), they retain a certain level of cell cycle control, as shown by their growth-inhibitory response to serum starvation (see serum starvation data in ). LAP2α-myc HeLa cells grown in the absence of doxycyclin expressed exclusively endogenous LAP2α, which is shown by immunofluorescence microscopy and immunoblot analyses of total cell lysates. The addition of doxycyclin to the culture medium induced the expression of myc-tagged LAP2α in the nucleoplasm (), giving rise to a total LAP2α level twice as high as that in uninduced cells. Cells grown in the absence of doxycyclin proliferated significantly faster than the cells grown permanently in doxycyclin-containing medium (18- vs. 22-h doubling times; ). When doxycyclin was added to cells only upon seeding into culture dishes, they showed intermediate growth rates (∼20-h doubling time; ). The addition of doxycyclin to untransfected cells had no influence on cell growth (). Interestingly, the cell proliferation-inhibitory effect of LAP2α was most prominent in dense cultures. DNA flow cytometry analyses in dense cultures grown in the absence or presence of doxycyclin () showed a slight increase in the relative number of cells in G1 phase (63→70%) and a decrease in S-phase cells (24→19%) upon doxycyclin addition. In addition, we showed previously by BrdU incorporation assays that transient overexpression of LAP2α leads to a reduction in cells in S phase (). Therefore, we concluded that increased LAP2α protein levels negatively affect G1→S-phase transition. To confirm this result in a cell system with an intact cell cycle checkpoint control mechanism, we stably expressed a LAP2α-GFP fusion protein in mouse 3T3 fibroblasts. As shown by immunoblot analyses of cell lysates and by immunofluorescence microscopy, stable 3T3 cell clones expressed roughly equal amounts of ectopic LAP2α-GFP and endogenous LAP2α, both of which localized to the nucleoplasm (). In three independent clones of LAP2α-GFP–expressing cells, we found a subtle but reproducible reduction in their cell proliferation rate compared with that of untransfected or GFP-expressing control cells under optimal culture conditions (10% FCS). DNA flow cytometry also revealed a slight increase in G1-phase cells upon LAP2α-GFP expression (not depicted), indicating an inhibitory role of LAP2α on G1→S-phase progression. However, because both control cells and LAP2α-GFP–expressing cells showed an efficient cell cycle arrest in dense cultures (), the proliferation-inhibitory role of LAP2α may not be detectable as clearly as in dense HeLa cells, which have a compromised, although not completely inhibited, density-mediated growth control. Therefore, we used an alternative approach and forced 3T3 cells into cell cycle arrest by serum starvation (0.5% FCS) for 7 d before we induced cell cycle entry by the addition of complete medium. DNA flow cytometry revealed that LAP2α-GFP–expressing cells showed a significantly delayed reentry into S phase () and a persistent growth reduction within 7 d after induction () when compared with untransfected or GFP-expressing cell clones. We concluded that increased LAP2α protein levels only weakly affected cell proliferation in growing 3T3 cells but interfered with pathways regulating G0/G1 phase to S-phase transition. Having shown that the overexpression of LAP2α negatively affects proliferation and cell cycle entry, we wondered whether reduced levels of LAP2α would have opposing effects. Although stable RNAi-mediated down-regulation of LAP2α was not achieved in 3T3 fibroblasts, stable HeLa cell clones with reduced levels of LAP2α were generated successfully. Immunoblot analyses of total cell lysates showed that unlike an unrelated control RNAi construct (targeting firefly luciferase), stable transfection with a LAP2α-specific short hairpin RNA (shRNA) construct reduced endogenous LAP2α to <10% of the protein level in control cells, whereas the expression of LAP2β remained unchanged (). In addition, immunofluorescence microscopy of LAP2α knockdown versus control cells revealed a significant down-regulation of LAP2α protein. As expected, the proliferation rate of LAP2α knockdown cell clones was increased significantly compared with that of control cells even in optimal growth conditions (, 10% FCS). Interestingly, in low serum conditions (0.5% FCS), control cells exhibited a very low proliferation rate (45-h doubling time), whereas proliferation of LAP2α knockdown cells was similar to that in complete medium (25-h doubling time). DNA flow cytometry confirmed that LAP2α-deficient cells were not as efficiently arrested in G1 phase as control cells in low serum medium (). Altogether, we concluded that LAP2α is involved in cell cycle entry/exit checkpoints and/or in G1→S-phase transition. Next, we wanted to gain more insights into the molecular basis of LAP2α's effect on the cell cycle. Our previous data revealed that LAP2α bound to hypophosphorylated Rb (), which is the predominant Rb form in G1 phase of proliferating cells or in resting cells in G0 phase. Hypophosphorylated Rb blocks transition into S phase by inhibiting the activity of E2F transcription factors that are essential for the expression of S phase–specific genes (). Therefore, we reasoned that overexpressed LAP2α may bind Rb and, in turn, affect its E2F-repressing activity. To analyze the interaction of LAP2α and Rb in more detail, we narrowed down the Rb interaction domain in LAP2α. In vitro–translated and [S]methionine-labeled Rb was overlaid onto transblotted fragments of LAP2α. Full-length protein (aa 1–693) as well as NH-terminally truncated LAP2α fragments missing either the LEM-like domain (aa 81–693) or the LAP2 common domain (including the LEM-like and LEM domain; aa 188–693) strongly interacted with Rb (). Because a COOH-terminal LAP2α fragment (aa 410–693) and a fragment missing the last COOH-terminal 78 residues (aa 1–615) also bound Rb, although less strongly than the other construct, we concluded that the Rb interaction domain of LAP2α is located in LAP2α's unique COOH-terminal domain between aa 415–615, but residues downstream of aa 615 may also contribute to the interaction (). Intriguingly, the Rb interaction domain is located next to the previously identified lamin A/C interaction domain (; ). Thus, it may well be possible that LAP2α, lamin A/C, and Rb form a trimeric complex in cells, as previously suggested based on coimmunoprecipitation of these proteins from lysates of untransformed human dermal fibroblasts (). To test whether this complex can also be formed in transformed HeLa cells, in which at least part of Rb is bound to E6 and E7 oncoproteins, we immunoprecipitated LAP2α complexes from LAP2α-GFP–expressing HeLa cells and analyzed samples by immunoblotting. As shown in , LAP2α/LAP2α-GFP immunoprecipitates obtained with a monoclonal antibody (m) or a polyclonal serum to LAP2α (p) contained Rb and lamin A/C. Control precipitations with mouse IgG (, ctrl) or preimmune serum (, PI) did not pull down any of these proteins. Thus, LAP2α–lamin A/C–Rb complexes also exist in transformed HeLa cells and could potentially be involved in the aforementioned cell cycle regulatory function of LAP2α ( and ). We were unable to detect E6, E7, and E2F proteins in cell lysates and immunoprecipitates most likely as a result of their low concentrations in the samples. Therefore, to test the association of LAP2α with E2F complexes, we performed chromatin immunoprecipitation with monoclonal and polyclonal LAP2α antibodies and tested the association of LAP2α with endogenous E2F-dependent promoters (). Intriguingly, unlike the preimmune serum (, PI), both antibodies pulled down DNA fragments containing E2F-dependent promoter sequences of the thymidine kinase (, TK) genes in HeLa and 3T3 cells. Furthermore, the E2F-dependent promoter sequence of cyclin D1 was efficiently precipitated by monoclonal LAP2α antibody (, m) in 3T3 and HeLa cells and by LAP2α antiserum (, p) in HeLa cells, whereas promoter sequences of the E2F-independent genes histone H4 and E-cadherin were not detected in the immunoprecipitates over background level. In contrast, antibodies to acetylated histone H4 as a positive control pulled down all DNA sequences tested. Thus LAP2α complexes specifically bind to E2F-dependent promoter sequences in vivo. To test whether LAP2α may affect E2F-dependent promoter activity, we tested the effect of overexpressed LAP2α in an E2F-dependent luciferase reporter assay. Proliferating 3T3 cultures exhibited a basal reporter activity, which was induced threefold upon the transient expression of E2F-1. Coexpression of E2F-1 and Rb reduced reporter activity to basal levels as expected because Rb can sequester and inactivate free active E2F-1 (). Importantly, the coexpression of LAP2α with E2F-1 also decreased the reporter activity as effectively as Rb, whereas an NH-terminal LAP2α fragment (LAP2α-N) that did not bind Rb () and did not effect cell cycle progression in HeLa cells () had no significant influence on E2F-dependent activity. Thus, full-length LAP2α efficiently inhibited exogenous E2F-dependent activity, which is consistent with the negative effect of LAP2α on cell proliferation and cell cycle reentry. So far, we have shown that LAP2α interferes with the activity of overexpressed ectopic E2F-1. To confirm this result under more physiological conditions, we wanted to analyze the effect of LAP2α on the activity of endogenous E2F. As expected, endogenous E2F-dependent activity was low in aphidicolin-treated cells, which were arrested at the G1→S-phase transition (shown by DNA flow cytometry; ). Aphidicolin-arrested 3T3 cells also contained predominantly hypophosphorylated Rb that migrates faster on SDS PAGE, whereas proliferating 3T3 cells revealed an Rb double band, reflecting hyper- and hypophosphorylated forms. Thus, unlike transformed HEK cells (shown as a control in ), aphidicolin-treated 3T3 cells were efficiently arrested at the G1–S-phase boundary. Upon the release of cells from the block, they synchronously proceeded through S phase, and E2F-dependent activity was increased 2–6 h after the release, whereas it declined again after 8–10 h when cells were in late S/G2 phase (). We measured the effect of overexpressed LAP2α on endogenous activity at two different cell cycle stages: in late G1 phase (aphidicolin arrested), when E2F-dependent activity is low because of the expression of hypophosphorylated Rb, and in S phase (4 h after release), when E2F-dependent activity is high. In G1 phase, a low basal E2F-dependent activity was further reduced (twofold) upon the expression of exogenous Rb or of LAP2α (). In contrast, in S-phase cells with a higher basal E2F-dependent activity, LAP2α expression was unable to repress E2F-dependent activity, whereas overexpressed Rb still reduced it up to twofold (). These results imply that LAP2α can repress E2F-dependent activity through hypophosphorylated Rb in G1 phase, which is consistent with our previous data that LAP2α binds preferentially to the G1 phase–specific hypophosphorylated Rb (). Overall, our studies strongly support a role of LAP2α in regulating E2F-dependent activity in an Rb-dependent manner. Having shown that LAP2α binds to promoter sequences of endogenous E2F target genes, we wondered whether overexpressed LAP2α also affects the expression of these genes. As E2F target genes are differentially expressed during the cell cycle, we arrested 3T3 control cells and LAP2α-GFP–expressing 3T3 cells in G0 phase by serum starvation. After reentry into the cell cycle by serum addition, we analyzed mRNA levels of three known E2F target genes (cyclin D1, thymidine kinase, and cyclin E) and of one E2F-independent control gene (actin) by semiquantitative RT-PCR. As expected, unlike actin mRNA, the mRNA levels of all three E2F target genes in control 3T3 cells increased between 10 and 14 h after serum addition (), when cells proceeded to the G1–S-phase boundary (for DNA flow cytometry, see ). In LAP2α-GFP–expressing 3T3 cells, the increase in E2F target gene mRNAs was delayed by ∼5 h compared with the control (). These findings are consistent with an inhibitory function of LAP2α repressing endogenous E2F target gene expression. Our findings could be explained if LAP2α regulates cell cycle progression by binding to hypophosphorylated Rb and, as a consequence, increases Rb's repressor function or delays its deactivation during G1→S-phase transition. However, we cannot fully exclude the possibility that LAP2α inhibits E2F activity via an Rb-independent function. To address this question directly, we performed E2F reporter assays in triple knockout (TKO) mouse embryonic fibroblasts (MEFs) in which all three pocket proteins (Rb, p107, and p130) were knocked out (). These cells express normal levels of LAP2α and lamins A/C compared with wild-type fibroblasts (unpublished data). Endogenous E2F activity was not affected by the overexpression of LAP2α in TKO MEFs (). In contrast, the expression of Rb alone or in combination with LAP2α reduced E2F activity to a similar level (approximately twofold), although the coexpression of Rb and LAP2α was slightly more effective than Rb alone. The high level of Rb expression in these transient transfection assays is most likely sufficient to effectively repress E2F activity so that the additional expression of LAP2α may not significantly increase this activity. Overall, we concluded that LAP2α's effect on E2F activity is strictly dependent on the presence of the pocket proteins Rb, p107, or p130. The Rb–E2F pathway is essential for the differentiation of several cell types, including skeletal muscle () and adipocytes (). To test whether LAP2α-mediated regulation of the Rb–E2F pathway can also influence differentiation, we used the 3T3 F442A clone. These cells proliferate in normal culture medium and can be considered as preadipocytes, whereas the addition of insulin and triiodothyronine (T3) to the culture medium induces cell cycle arrest and differentiation into mature adipocyctes within 10 d, as seen in the phase-contrast microscope by a morphological change from spindlelike preadipocytes to rounded adipocytes with numerous fat droplets (). Immunoblot analyses of total cell lysates revealed a switch from the high molecular weight hyperphosphorylated Rb in proliferating cells to the low molecular weight hypophosphorylated form in adipocytes, reflecting cell cycle arrest in G0 phase (). Furthermore, LAP2α appeared as a double band in proliferating cells, representing mitotic hyperphosphorylated and interphase hypophosphorylated LAP2α (), whereas the hypophosphorylated form accumulated in differentiated adipocytes because of the lack of mitotic cells. Finally, the adipocyte-specific differentiation marker peroxisome proliferator–activated receptor γ (PPARγ; ) was expressed exclusively in differentiated cells, and lamins A/C were slightly up-regulated in differentiated cells (). Overall, these data showed that in differentiation medium, preadipocytes enter G0 phase and differentiate to adipocytes in vitro. To test the effect of overexpressed LAP2α on differentiation, we used clones of the 3T3 F442A cells stably expressing LAP2α-GFP, which showed a delayed response in cell cycle entry upon serum starvation and restimulation (). Cell cycle arrest and morphological changes during differentiation as detected by phase-contrast microscopy did not differ significantly in several independent clones of LAP2α-GFP–expressing versus control cells (untransfected or GFP expressing). However, we did detect significant differences by immunoblot analyses of total cell lysates. Hypophosphorylated Rb, LAP2α, and PPARγ, all of which are early markers of adipogenesis, clearly accumulated in proliferating LAP2α-GFP preadipocytes compared with control cells, suggesting that the overexpression of LAP2α initiated early events in the differentiation pathway, accompanying a partial cell cycle arrest. Upon insulin/T3 stimulation of LAP2α-GFP cells, hypophosphorylated Rb and LAP2α accumulated similarly to the control cells. In contrast, the expression level of PPARγ did not reach the levels observed in the control (), indicating that differentiation was incomplete. To test this possibility, we stained for differentiation-specific lipid droplets using oil red. As shown in , LAP2α-GFP–expressing cells had significantly fewer oil red–positive droplets than control cells. Thus, although the overexpression of LAP2α-GFP initiated differentiation events in the absence of insulin and T3, further differentiation did not occur even in the presence of insulin/T3. In this study, we analyzed the effects of LAP2α on cell cycle progression and differentiation and showed that decreased LAP2α levels negatively affected the growth arrest response to serum starvation, whereas increased levels delayed the transition from G0 to S phase. We further showed that this function of LAP2α is mediated by pocket proteins. What are the molecular mechanisms of the Rb-mediated cell cycle control by LAP2α complexes? Rb regulates the activity of E2F transcription factors, which control the expression of cell cycle regulatory genes (; ). In noncycling cells, Rb is hypophosphorylated and is active as a transcriptional repressor. Hypophosphorylated Rb is bound to E2F transcription factors, blocks its transactivation domain, inhibits the recruitment of preinitiation complexes, and represses genes directly by recruiting chromatin-modifying proteins, including histone deacetylase HDAC1 (), methyl transferase Suv39h, and heterochromatin protein HP1 (; ). Through sequential phosphorylation of Rb by cyclin D and E–dependent kinases (), Rb is released from E2F, causing its activation as a transcription factor. Based on our previous findings (; ; ) and results reported here, we postulate that nucleoplasmic LAP2α–lamin A/C complexes bind hypophosphorylated Rb, thereby delaying its deactivation and maintaining E2F in a repressed state. The molecular mechanisms as to how LAP2α–lamin A/C complexes exert their regulatory function on Rb are unknown. However, several possibilities can be envisaged: first, LAP2α may regulate Rb phosphorylation. LAP2α preferentially bound to Rb phosphorylated on serine 780, which is the first serine to be targeted at the G1→S-phase transition (). Recruitment of Rb-S780 to LAP2α–lamin A/C may delay the phosphorylation of downstream sites. Alternatively, LAP2α–lamin A/C complexes may favor the dephosphorylation of Rb, leading to its deactivation. In line with the latter hypothesis, have recently shown that in knockout fibroblasts, TGFβ-mediated cell cycle arrest is impaired through a less efficient phosphatase PP2A-mediated dephosphorylation of Rb. Second, LAP2α–lamin A/C may recruit regulatory proteins to the Rb complex. Third, LAP2α–lamin A/C may recruit Rb– E2F complexes to subnuclear compartments or chromatin regions. Our findings that LAP2α complexes associate with E2F-dependent promoter regions imply that the cell cycle regulatory activity of LAP2α acts directly on the promoter of E2F target genes. fibroblasts by proteasomal degradation (), the interaction of Rb with LAP2α–lamin A/C complexes may stabilize the proteins. However, this would imply that proteasomal degradation also contributes to the deactivation of Rb repressor activity at the G1→S-phase transition, although no direct evidence for this has been reported so far. Furthermore, we did not see any changes in Rb protein level upon the overexpression or knockdown of LAP2α. Finally, it should be mentioned that another LAP2 isoform, LAPß, was also found to down-regulate E2F activity, although the mechanisms are most likely different from that of LAP2α, involving the binding of LAP2β to the protein germ cell less () and to histone deacetylase 3 (). What is the physiological relevance of LAP2α–lamin A/C complexes in cell cycle control? In cycling cells, these proteins may provide additional checkpoint mechanisms, preventing premature entry into S phase. However, we consider it more likely that LAP2α–lamin A/C complexes are involved in controlling cell cycle exit during the differentiation of adult stem cells. Rb has been found to be essential for the differentiation of various tissues, including skeletal muscle and adipocytes (; ). Inactivation of the gene in proliferating myoblasts and in differentiated muscle fibers in mice revealed that Rb is required for the initiation of myogenic differentiation but not for the maintenance of the terminal differentiation state (). Also in adipocyte differentiation, Rb facilitates the initiation of differentiation by inducing cell cycle arrest (). Thus, Rb seems to be involved in maintaining a balance between proliferation and differentiation in adult stem cells (). Adult stem cells do not proliferate but enter the cell cycle upon specific signals and self propagate to maintain a stable population of stem cells in the tissue. At the same time, they differentiate to regenerate damaged tissue or maintain a basic turnover of differentiated cells (). Our data suggest that LAP2α–lamin A/C complexes are essential cofactors of Rb in maintaining the balance between proliferation and differentiation in adult stem cells. Similar to Rb, LAP2α's critical role would be in transit amplifier cells rather than in terminally differentiated cells. Consistent with this hypothesis, the highest levels of LAP2α expression during the differentiation of skeletal muscle satellite cells is just before entry into a postmitotic state (). Moreover, by artificially maintaining high levels of LAP2α expression in preadipocytes, early events in differentiation were enhanced but terminal differentiation was inhibited (). Therefore, we propose that the LAP2α–lamin A/C complex is critical for Rb-mediated entry of adult stem cells into a postmitotic state but is not required for terminal differentiation. During the last 10 yr, geneticists have identified mutations in the human gene that give rise to pathological phenotypes in muscle, adipose, bone, neuronal, and skin tissue or cause premature ageing (; ). Mice lacking lamin A/C or expressing mutated lamin A/C show similar phenotypes (; ; ). The molecular basis of these diseases is unclear. It may involve structural defects in mutated lamins, leading to nuclear fragility, or deregulated functions of lamins in gene expression (). As LAP2α is involved in targeting a subfraction of lamin A/C to the nucleoplasm (, ), it may be relevant for the nucleoplasmic functions of lamin A/C. If nucleoplasmic lamins together with LAP2α control Rb in cell cycle progression and differentiation, disease-causing mutations in lamin A/C may cause differentiation defects. In support of this model, the expression of a disease-linked lamin A mutant in C2C12 myoblasts () inhibited the in vitro differentiation into myotubes, and the expression of wild-type and mutated lamin A inhibited the differentiation of 3T3 cells into adipocytes (). We propose that disease-causing mutations in A-type lamins perturb the balance between proliferation and differentiation in adult stem cells, leading to less efficient tissue regeneration. This model would help to explain the late onset of the disease during childhood or adulthood, an age when tissue regeneration becomes relevant, as well as progressive muscle weakness, which may be caused indirectly by a defect in regeneration rather than by direct degeneration (). Intriguingly, mutations in the gene encoding LAP2α have recently been linked to laminopathy-like cardiomyopathy (). Disease-causing LAP2α mutants bind less efficiently to lamin A/C, supporting the hypothesis that nucleoplasmic LAP2α–lamin A/C complexes fulfill important functions. 3T3-F442A and HeLa cells were cultured at 37°C and 8% CO in high glucose DME medium supplemented with 10% FCS, 50 U/ml penicillin, 50 μg/ml streptomycin, 0.2 mM glutamine (Invitrogen), 0.8 μg/ml biotin, and 0.4 μg/ml pantothenic acid (Sigma-Aldrich) for 3T3-F442A cells. Confluent 3T3-F442A cells were differentiated with 17 nM insulin and 2 nM T3 (Sigma-Aldrich) for 7–10 d. Lipid drops were stained for 1 h with 0.3% Oil Red O (Sigma-Aldrich) after fixation in 3.7% formaldehyde (Merck) in PBS for 2 min. For DNA flow cytometry, 5 × 10 cells in 1 ml PBS were mixed with 4 ml 85% ethanol at −20°C. Cell pellets were briefly treated with 0.05% aqueous pepsin, stained with 2 μg/ml DAPI and 15 μg/ml sulforhodamine (Sigma-Aldrich) in 100 mM Tris-HCl, pH 8.0, 2 mM MgCl, and 0.1% Triton X-100, and analyzed with a flow cytometer (PAS II; Partec). For cell proliferation assays, 5 × 10 cells were seeded onto 100-mm dishes (Nunc), and cell numbers were determined within 7 d using a coulter counter Z2 system (Scharfe System). For serum starvation, cells were grown in medium with 0.5% FCS for 5 d. LAP2α-myc HeLa Tet-on cells () were grown in medium with 200 μg/ml hygromycin and induced with 2 μg/ml doxycyclin (Invitrogen). pTD35 encoding LAP2α-GFP was generated by cloning LAP2α cDNA via NheI–XhoI from pTD15 into a modified pEGFP-N3 vector (CLONTECH Laboratories, Inc.) as described previously () and was used for stable transfection of 3T3 cells with LipofectAMINE 2000 (Invitrogen). Stable clones were selected in medium with 700 μg/ml geneticin. For reporter assays, cells were cotransfected with 2 μg luciferase reporter p3xE2F-Luc (provided by H. Rotheneder, Medical University of Vienna, Vienna, Austria; ) and 2 μg β-galactosidase control reporter plasmid pAD-CMV1-βgal () together with various combinations of the following plasmids (1 μg each): pCI-neo-HA-E2F1 (), pCMV-Rb (provided by M. Busslinger, Research Institute of Molecular Pathology, Vienna, Austria; ), pSV61 (full-length LAP2α in pcDNA3.1), pSV62 (LAP2 common in pcDNA3.1), and pcDNA3.1 (Invitrogen). Luciferase activity was measured with the lumat system (Berthold) and normalized for β-galactosidase activity. Relative luciferase activities were calculated as the light units relative to the reporter plasmid controls. Data were obtained from at least three independent experiments, and mean values and standard errors were calculated. pSHAG-1 vector containing a human U6 promoter fragment (−265 to 1) was derived from pTOPO-ENTR/D (Invitrogen) as described previously (; ). Human LAP2α-specific oligonucleotides encoding a shRNA were designed as described (): 5′-ACTGAT-CAATTCTCTTCTGGAAGCTTGCAGAAGAGAATTGATCAGTCTTTTTT-3′ and 5′-GATCAAAAAAGACTGATCAATTCTCTTCTGCAAGCTTCCAGA- AGAGAATTGATCAGTCG-3′. Annealed oligonucleotides were cloned into pSHAG-1 via BseRI–BamHI. As a control, we used pSHAG-FF1 encoding a shRNA targeting firefly luciferase (gift of Greg Hannon, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY; ). Using the Gateway system (Invitrogen), shRNA expression cassettes in pSHAG-1 were shuttled into pTRACER-EF/Bsd plasmids in which a Gateway-compatible conversion cassette had been introduced in a reverse orientation via EcoRV. The resulting RNAi vectors (pJG118 targeting LAP2α and pJG121 control vector targeting luciferase) were transfected into HeLa cells, and stable clones were selected with 20 μg/ml blasticidin (Invitrogen). Chromatin immunoprecipitation was performed as described previously () with a few modifications. Cross-linking with formaldehyde was performed for 15 min, and the soluble chromatin fraction was diluted 1:10 in 16.7 mM Tris-HCl, pH 8.0, 167 mM NaCl, 1.2 mM EDTA, 1.1% Triton X-100, and 0.01% SDS and was precipitated with monoclonal antibody 15 or antiserum 245 to LAP2α or rabbit preimmune serum as a control. Chromatin–antibody complexes were isolated with 30 μl of protein A/G–Sepharose beads and solubilized in 2% SDS, 0.1 M NaHCO, 10 mM DTT, and 250 mM NaCl overnight at 65°C. Samples were analyzed by immunoblotting, or DNA was extracted with phenol-chloroform, precipitated with ethanol, and dissolved in water. The immunoprecipitated DNA was analyzed for cyclin D1, thymidine kinase, E-cadherin, and histone H4 promoter sequences by PCR using a thermocycler (PTC-200; MJ Research) and PuReTaq Ready-To-Go PCR beads (GE Healthcare). The linear range for each primer pair was determined empirically using different amounts of genomic DNA. PCR products were resolved on 2% agarose gels. Primers used are as follows: 5′-GACCCTGGCCAGGATAAAC-3′ and 5′-AGACGAGCCCTAAGCTCTC-3′ for mouse cyclin D1; 5′-GACCGACTGGTCAAGGTAGG-3′ and 5′-GTTTCATTCCGGCGCACAGG-3′ for human cyclin D1; 5′-AGACCCCGCACCTGAATCTG-3′ and 5′-TTCACGTAGCTGAGAGGTGG-3′ for mouse and 5′-CGATCAGCCACGTCCATC-3′ and 5′-CGCCGACCGCTTTAAACC-3′ for human thymidine kinase; 5′-GACACCGCATGAAAAGAATAGCTG-3′ and 5′-CTTTCCCAAGGCCTTTACCACC 3′ for mouse histone H4; and 5′-AACTCCAGGCTAGAGGGTCA-3′ and 5′-GGGCTGGAGTCTGAACTGA-3′ for human E-cadherin. For RT-PCR, poly(A) mRNA was extracted from cells with an mRNA Isolation Kit and reverse transcribed using the First Strand cDNA Synthesis Kit (both from Roche). Aliquots of the resulting products were used as templates for specific PCR amplifications using PuReTaq Ready-To-Go PCR beads (GE Healthcare). The conditions for PCR reactions were optimized for each primer pair. All cDNAs were normalized according to actin expression levels. Primers used are as follows: 5′-AGTGCGTGCAGAAGGAGATT-3′ and 5′-CACAACTTCTCGGCAGTCAA-3′ for cyclin D1; 5′-ATGAGCTACATCAATCTGCCC-3′ and 5′-TTCCGATCATGTGTGGAGAA-3′ for thymidine kinase; 5′-CAAAGCCCAAGCAAAGAAAG-3′ and 5′-CCACTGTCTTTGGAGGCAAT-3′ for cyclin E; and 5′-ATCTGGCACCACACCTTCTAC-3′ and 5′-CAGCCAGGTCCAGACGCAGG-3′ for actin. PCR reactions were quantified with QuantiScan software (Biosoft). Immunofluorescence microscopy and immunoblotting were performed as described previously (; ). For microscopy, samples were mounted in mowiol and analyzed using a microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) equipped with a confocal laser-scanning unit (LSM510 META; Carl Zeiss MicroImaging, Inc.) and α–plan-Fluar 100× NA 1.45 oil (a = 0.11 mm) and plan-Apochromat 63× NA 1.40 oil differential interference contrast MC27 (a = 0.19 mm) objectives (Carl Zeiss MicroImaging, Inc.). Phase-contrast images were taken on a microscope (Axiovert 40C; Carl Zeiss MicroImaging, Inc.) equipped with a digital camera (PowerShot G5; Canon) and plan-Neofluar 10× NA 0.30 (a = 5.6 mm) and Achromat 20× NA 0.4 objectives. Images were prepared with Adobe Photoshop software. Total protein lysates of cells for immunoblotting were prepared by dissolving cells of one 10-cm dish in 500 μl of 2× SDS-PAGE sample buffer. For detection of blotted proteins, the Super Signal ECL system (Pierce Chemical Co.) was used. The preparation of recombinant LAP2α and blot overlay assays was described previously (). Primary antibodies used were hybridoma supernatants of antibodies to LAP2 and LAP2α (), rabbit antiserum to LAP2α (ImmuQuest; ), monoclonal myc 1-9E10.2 antibody (American Type Culture Collection), monoclonal antilamins A/C antibody 1E4 (a gift of F. McKeon, Harvard Medical School, Boston, MA; ), monoclonal anti-Rb G3-245 (BD Biosciences), monoclonal anti-PPARγ (Santa Cruz Biotechnology, Inc.), polyclonal anti-actin (Sigma-Aldrich), and anti-acetylhistone H4 (Upstate Biotechnology). Secondary antibodies were goat anti–mouse IgG and goat anti–rabbit IgG conjugated to Texas red (Jackson ImmunoResearch Laboratories), goat anti–rabbit IgG conjugated to AlexaFluor488 (Invitrogen), and goat anti–mouse IgG and goat anti–rabbit IgG conjugated to peroxidase (Jackson ImmunoResearch Laboratories).

During development, precisely regulated signaling pathways instruct cells to adopt particular fates. Wnt signaling mediates many developmental decisions (). Highlighting its importance, the misregulation of Wnt signaling causes improper fate specification, tumor formation, and early lethality (). Although proper Wnt signaling is essential, the mechanisms that control ligand distribution and signaling levels are not fully understood. One process proposed to affect Wnt signaling is intracellular transport (; for review see ). In endocytosis, membrane proteins are recruited to small plasma membrane invaginations. These forming endocytic vesicles are cleaved from membranes via the function of dynamin (), a protein encoded by () in . These vesicles then undergo Rab5-mediated fusion with the early endosome (; ). There, internalized proteins are sorted and redistributed within the cell. Proteins slated for degradation are sorted by hepatocyte growth factor–regulated tyrosine kinase substrate (Hrs) into the inner vesicles of the multivesicular body (MVB; ). When MVBs fuse with lysosomes, these internalized proteins are degraded. Work in other signaling pathways has suggested that by regulating the level and distribution of ligand, endocytosis can affect the induction of signaling (; for review see ). Indeed, studies examining the relationship between endocytosis and Wingless (Wg) signaling suggest effects on Wg levels and spread. In the wing, loss of dynamin eliminates extracellular Wg (Wg(ex)) in 50% of samples (), suggesting that dynamin may mediate Wg secretion. However, other studies indicate that dynamin is not involved in forming secretory vesicles from the Golgi (; ; ). Thus, the effect of dynamin on Wg production remains unclear. After Wg is secreted, it must travel to reach target cells. The role of endocytosis in Wg spread is heavily debated, as Wg may spread by either diffusion or intracellular transport. Supporting extracellular spread, dynamin-mediated internalization is not required for Wg spread in the wing (). The efficiency of diffusion has been questioned, however, because Wg interacts with proteins in the extracellular matrix (). Alternatively, Wg may spread through vesicle intermediates. GFP-tagged Wg can be internalized and recycled to the cell surface in embryonic cells (). Visualization of membrane phospholipids also suggests that Wg may spread via vesicular structures in the wing (). Given that evidence supporting both extracellular and intracellular transport exists, the extent to which endocytosis affects Wg spread is controversial. Thus, although previous studies suggest that endocytosis may regulate Wg levels and spread (for review see ), many questions remain. Aside from affecting ligand levels and distribution, endocytosis may also regulate signal transduction (; ). Determining whether endocytosis directly affects Wg signal transduction has been complicated, however, by difficulties in distinguishing effects on protein levels from signaling levels. In mutant embryos, Armadillo (Arm) staining is reduced but not eliminated (). The presence of Arm indicates that Wg signaling can occur at the cell surface; however, it is unclear whether the reduction is caused by altered Wg spread or impaired signal transduction. Although this is consistent with the facilitation of Wg signaling by dynamin, it has been suggested that signaling is negatively regulated by Rab5 (), raising doubt as to the necessity of endocytosis in Wg signaling. Given these contradictory results, the effect of endocytosis on Wg signaling is unclear. In this study, we use genetic tools to alter vesicle transport and study the effect on Wg production, transport, degradation, and signaling. In cells treated with Wg media, the knockdown of dynamin or Rab5 reduces Wg reporter activity, suggesting that internalization and endosomal transport facilitate signaling. In the wing, we find that dynamin mediates Wg transcription but is not required for Wg secretion or spread. Independent of altered Wg protein levels, endocytosis appears to regulate Wg signaling. Although impaired internalization and endosomal fusion increase Wg levels, signaling is reduced. Conversely, increased endosomal transport and obstructed transport from the endosome enhance Wg signaling. This correlates with the presence of endosomal accumulations of Wg, Arrow (Arr), and Dishevelled (Dsh). Thus, our data suggest that trafficking to the endosome facilitates Wg signaling possibly through the formation of an endosomal protein complex. To determine endocytic effects on Wg signaling, a cell-based Wg assay was used. S2R+ cells were transfected with a cytomegalovirus (CMV)-driven Renilla luciferase (RL) transfection control and Super8XTOPFlash (TOPFlash), a Wg reporter driving the expression of firefly luciferase. In response to Wg, the TOPFlash/RL ratio increases, serving as a quantitative measure of Wg signaling. Additionally, cells were transfected with double-stranded RNA (dsRNA) to determine the effect of particular genes on signaling. Knockdown of Arm, a mediator of Wg signaling, profoundly reduces TOPFlash/RL (; and Table S2, available at ). Conversely, knockdown of casein kinase 1a (ck1a), a negative regulator of signaling, strongly increases TOPFlash/RL (; and Table S2). These signaling levels are consistent with prior cell culture (; ) and in vivo analyses (for review see ; ). We next transfected cells with dsRNA against the coding region. Wg media was added 7 d after transfection to induce signaling. Luciferase levels and protein knockdown were assessed on day 8. In stimulated cells with reduced dynamin, TOPFlash/RL decreased by 79% ( and Table S2 A), indicating that dynamin promotes Wg signaling. Similarly, the effects of endosomal transport were evaluated by transfection with dsRNA against the Rab5 coding region (R5; ). These cells showed a 93% decrease in luciferase ratio ( and Table S2 B). This is surprising because a recent study argues that R5 transfection increases Wg signaling (). To understand this discrepancy, we first reduced Rab5 using a dsRNA against the highly specific 3′ untranslated region (R5). Similar to R5, R5-treated cells show an 82% decrease in TOPFlash/RL ( and Table S2 B). Second, we transfected cells with a Wg DNA construct, as performed by , in lieu of adding Wg media. We initially examined cells 4 d after transfection as performed by . Although Rab5 was still present at this time point (), a 38% reduction in TOPFlash/RL was observed ( and Table S2 C). At 8 d after transfection, we observe strong knockdown and an 82% reduction in luciferase ratio, which is consistent with our results ( and Table S2 D). Finally, we examined TOPFlash/RL ratios upon transfection with different RL control vectors ( and Table S2 D). Although cells transfected with the polIII-RL transfection control used by show a 68% increase in luciferase ratio, transfection with tk-RL and s-188-cc-RL show 34 and 25% reductions in luciferase ratio, respectively. These varied TOPFlash/RL ratios indicate that transfection control vectors can produce different RL levels that dramatically impact the quantification of Wg signaling. However, given that R5 transfections with three out of four RL vectors show reduced luciferase ratios, our data suggest that impaired Rab5-mediated endosomal fusion hinders Wg signaling. To determine the relevance of our cell culture data, we studied the effects of endocytosis on signaling in the wing. Wg forms a morphogen gradient in the larval wing that regulates proliferation and cell fate specification (; ). Wg is secreted at the dorsal–ventral (DV) boundary of the wing disc and is detected at high levels spanning approximately three cell widths (; ; ). Spots of Wg are also present in the wing pouch, decreasing with distance from the DV boundary. As a morphogen, Wg can induce different target genes depending on signaling levels (). High levels of signaling induce Senseless (Sens) in cells bordering the DV boundary (; ). Low levels of signaling are sufficient to induce Distal-less (Dll) broadly across the wing pouch (; ; ). Both Sens and Dll function in wing margin bristle development (; ). Formation of a normal-sized wing is also dependent on Wg signaling, as mutants lack wings (). Thus, by examining the expression of Wg targets and adult wing morphology, we can assess Wg signaling levels. To study the effect of internalization on signaling, we expressed dominant-negative () to impair internalization from the cell surface (). Because it has been suggested that dynamin mediates Wg secretion (), we used two Gal4 drivers to analyze Wg distribution and signaling. induces expression at and near the DV boundary (; ), thereby permitting analysis of Wg transcription and secretion. Conversely, () induces expression throughout the wing pouch except for cells at the DV boundary (). Because this does not include Wg-expressing cells, allows analysis of Wg spread and degradation independent of Wg production. By combining data from these drivers, we can study changes in Wg production, spread, degradation, and signaling. When was overexpressed at the DV boundary, Wg distribution is narrow compared with controls (), which is indicative of altered Wg transcription or secretion. In situs show less Wg RNA (), indicating that dynamin facilitates Wg transcription likely through its regulation of Notch signaling (; ; ). However, as shown in , discs exhibit elevated Wg(ex) levels compared with controls. Thus, our data suggest that when dynamin function is blocked, Wg transcription is reduced, but Wg is secreted and accumulates extracellularly. To investigate the effect of dynamin on Wg spread, we expressed using . These discs show a dramatically widened Wg distribution compared with controls (). Consistent with impaired internalization, this protein can be detected extracellularly (). Because Wg expression is similar to controls (), the enhanced Wg(ex) likely results from reduced Wg degradation when function is inhibited. Notably, the normal Wg expression also indicates that Wg can spread from the DV boundary in a dynamin-independent manner. Thus, dynamin regulates Wg levels through transcription and degradation but does not appear to be required for Wg secretion or spread. To determine whether dynamin affects signaling, Wg target gene expression was examined. Although both and overexpression of show enhanced levels of Wg(ex), we find that Sens expression is nearly absent (), indicating that dynamin is required to achieve high signaling levels. Furthermore, Dll levels in cells are decreased compared with cells outside the wing pouch that do not express (). Dll expression is similarly reduced in temperature-sensitive () mutant clones at the restrictive temperature (not depicted). The progressively weaker effects of dynamin on Sens and Dll are consistent with our understanding of the Wg morphogen gradient and indicate that impaired internalization reduces but does not eliminate Wg signaling. Notably, the reduced protein expression is unlikely to be the result of cell death, as little to no TUNEL-positive columnar cells are observed in the -, and discs studied (Fig. S1, B and C; available at ). Additionally, the differential decrease in Sens and Dll suggests that these reductions do not arise from cell death. Further supporting reduced Wg signaling, wings show a loss of margin tissue that resembles the mutant phenotype (; ; ; ). adult wings are small with altered morphology ( AA), exhibiting bristle loss consistent with decreased Sens expression and Wg signaling (; ). Thus, consistent with our cell culture data, these data indicate that impaired dynamin function reduces Wg signaling even when significantly more Wg(ex) is present. This effect is more obvious for Wg targets requiring high signaling levels, suggesting that Wg(ex) can induce only low signaling levels in the absence of dynamin-mediated internalization. After dynamin-mediated internalization, endocytic vesicles undergo Rab5-mediated fusion with the endosome (; ). As our cell culture data suggest that the loss of Rab5 reduces Wg signaling, we determined whether endosomal transport affects signaling in vivo by expressing dominant-negative Rab5 (Rab5, also called Rab5), a constitutively GDP-bound form that inhibits endosomal fusion (; ). In discs, Wg staining is more punctate but otherwise similar to controls (). Despite this, Sens expression near the DV boundary is eliminated (), indicating that high levels of signaling are blocked by impaired endosomal transport. Dll expression is also much reduced compared with levels outside of the wing pouch (). The stronger effect on Sens than Dll is similar to , further indicating that high Wg signaling levels cannot be reached when endocytosis is blocked. Evaluating cell death, we find TUNEL-positive columnar cells upon the expression of (Fig. S1 D). However, this cell death likely causes only minor changes in protein expression as indicated by the large number of Wg-expressing cells present (). Additionally, overexpression of results in the loss of wing tissue similar to the loss of Wg (; ; ), further suggesting that endosomal transport significantly affects Wg signaling. Similarly, we have analyzed the expression of . As shown in , Wg distribution is significantly expanded, which was caused, in part, by increased Wg transcription (). Despite high Wg levels, Sens expression is absent, and Dll expression is markedly reduced (). Again, the differential effects on Sens and Dll expression are consistent with impairment of the Wg signaling gradient. expression of causes almost a complete loss of wing tissue (), as documented for mutants (). Notably, cooverexpression of Arm (), a mediator of Wg signaling, partially restores bristles and wing size (unpublished data). This indicates that although some TUNEL-positive cells are observed in columnar cells expressing Rab5 (Fig. S1 D), cell death does not account for the observed phenotypes. Together, these data suggest that early endosomal transport facilitates Wg signaling in vivo. Although Rab5 is constitutively inactive, wild-type Rab5 (Rab5) is subject to the regulation of cellular factors (). To examine the effect of Rab5, we induced clones of the expression of . As shown in (A–C), Rab5 overexpression causes no change in Wg, Sens, or Dll expression. Interestingly, it has been reported that expression of Rab5 decreases Sens levels (). However, we find that the overexpression of Rab5 by either or causes no change in Wg, Sens, or Dll expression (; and not depicted). Furthermore, the adult wings are indistinguishable from controls ( and not depicted). These data indicate that Rab5 overexpression does not change Wg signaling. Given the different results obtained from Rab5 and Rab5, our data also suggest that endocytic regulators can control Wg signaling by altering endosomal transport. Consistent with this, the overexpression of dominant active Rab5 (Rab5, also called Rab5) causes an enhancement in signaling that was not observed with either Rab5 or Rab5. In these discs, Sens expression is detected more than two cell diameters from the DV boundary, and Dll expression is enhanced in the Gal4 expression domain (Fig. S2, D–I; available at ). Consistent with increased Sens expression and Wg signaling (; ), the adult wings develop ectopic bristles (Fig. S2, K and L). These data suggest that although Rab5 expression reduces Wg signaling, the promotion of endosomal transport by Rab5 enhances signaling. These data further indicate that in addition to proteins like dynamin and Rab5 that directly mediate vesicle trafficking, regulators of endocytic proteins can modulate Wg signaling. Upon internalization to the endosome, proteins slated for degradation are sorted into MVBs via the function of Hrs (; ). We analyzed wings with altered Hrs function to determine whether trafficking from endosomes to MVBs affects signaling. In mutant clones, Wg distribution is slightly expanded, with much of the protein localized in large puncta (; see ). Wg(ex) staining fails to detect these accumulations (not depicted). Although most Wg is located intracellularly in mutants, Sens is sometimes more broadly expressed within mutant clones than in the internal control (). Similarly, some mutant clones show enhanced Dll levels (). These changes in expression are most evident in large clones induced early in development. As Hrs is a very stable protein (), small clones induced later show minor or no changes in Wg target gene expression, probably as a result of Hrs protein perdurance. Thus, although less Wg(ex) is present than in controls, the impairment of endosome to MVB transport augments Wg signaling. Our data strongly suggest that internalization and protein localization to the early endosome play a critical role in Wg signaling. We next examined the effect of enhanced MVB transport. The overexpression of Hrs by and facilitates trafficking through MVBs as demonstrated by enlarged LAMP-positive lysosomes (not depicted). When Hrs is overexpressed at the DV boundary, Wg distribution is disrupted (). However, when Hrs is broadly expressed, Wg distribution is slightly widened (). Despite differences in Wg levels, both genotypes show a reduction in Sens and Dll (). Consistent with reduced signaling, wings have a loss of margin tissue (), and wings are reduced in size with fewer bristles near the wing margin (not depicted). Further supporting an effect on Wg signaling, the cooverexpression of Wg with Hrs largely suppresses the loss of margin tissue (). In canonical Wg signaling, Wg associates with Frizzled receptors and Arr coreceptors to phosphorylate the cytoplasmic protein Dsh and activate signaling (; ; ). We find that the coexpression of Frizzled () and myc-tagged Dsh () are each capable of suppressing the phenotype. These data indicate that Hrs phenotypes arise specifically from changes in Wg signaling rather than other factors (Fig. S1 G). Together, these data suggest that although internalization and early endosomal transport facilitate Wg signaling, progression to the MVB negatively regulates signaling. Our data indicate that localization to early endosomes enhances Wg signaling. This is similar to receptor tyrosine kinase signaling, where the formation of endosomal signaling complexes is proposed to facilitate signaling (; ). To determine whether Wg signaling occurs in a similar manner, we first studied Wg localization. We find that Wg partially colocalizes with the early and late endosome marker FYVE-GFP (not depicted; ) and the late endosomal protein Rab7-GFP (; ). Additionally, electron microscopy was performed on wing discs expressing HRP-tagged Wg protein (). Based on compartment morphology, HRP activity is localized to small vesicles, endosomes, MVBs, and lysosomes (). Thus, Wg is internalized and trafficked through intracellular compartments. We further examined the localization of Arr, the homologue of LRP5/6, and Dsh, which are two proteins that are necessary for Wg signaling (; ; ). In controls, small puncta of HA-tagged Arr (ArrHA) sometimes colocalize with Wg (). ArrHA and Wg also occasionally colocalize with Dsh, which is present at low levels in the cytoplasm as well as in small puncta ( and not depicted). In mutants, ArrHA and Wg often localize to large puncta that colocalize with the endosome/lysosome marker Benchwarmer (also called Spinster; , C and E; ; ). Intracellular ArrHA and Wg often also colocalize with accumulations of Dsh (, D and F; and not depicted). Thus, in mutants, enhanced Wg signaling correlates with greater colocalized Wg, Arr, and Dsh on endosomes than in wild-type cells. Our analysis has revealed the surprising finding that intracellular transport affects the efficiency of Wg signaling. In cell culture, knockdown of dynamin, a protein essential for clathrin-mediated internalization, reduces the TOPFlash/RL ratio, which is suggestive of decreased Wg signaling. Similarly, Rab5 knockdown causes reduced TOPFlash/RL ratios under most conditions, suggesting that internalization and endosomal transport are important for Wg signaling. Interestingly, transfection with polIII-RL, a control vector used in a recent screen for modifiers of Wg signaling (), produces conflicting results for Rab5 compared with other RL controls, indicating that cell culture–based Wg signaling assays are very sensitive to experimental conditions. Thus, although our cell culture results indicate an endocytic regulation of Wg signaling, in vivo validation is critically important. In the wing, we found further evidence that Wg signaling levels are highly dependent on intracellular transport. When endocytosis is altered, ligand levels and signaling levels are uncoupled such that high Wg levels do not necessarily enhance signaling. Therefore, we have limited usage of the term morphogen gradient, which could refer to either ligand or signaling levels. We instead describe Wg distribution and signaling readouts. When internalization is inhibited in a domain that does not affect Wg production, we find high levels of Wg(ex), likely as a result of reduced degradation. However, Wg target gene expression is diminished, indicating that impaired internalization decreases Wg signaling in vivo as well as in cell culture. When early endosomal transport is impaired, Sens and Dll expression are also reduced despite abundant Wg levels. In both cases, markers of high signaling levels are especially affected, indicating that intracellular signaling is important to achieve robust Wg signaling levels. The differential decrease also argues that changes in Sens and Dll expression are not merely the result of cell death or global changes in transcription (). Further supporting this, we find the normal expression of other genes in the wing pouch (unpublished data). Additionally, when endosomal transport is enhanced or when transport from the endosome is impaired, Wg signaling is increased. These data suggest that protein localization to the endosome facilitates Wg signaling. Conversely, increased transport to MVBs decreases the expression of Wg readouts. This causes an adult wing phenotype that can be suppressed by Wg signaling components. Thus, we propose that in addition to low levels of cell surface signaling, intracellular Wg signaling is critical for proper signaling levels (). Because endocytosis is tightly regulated, intracellular Wg signaling may allow for the rapid modulation of signaling levels. For example, endosomal transport can be regulated merely by changing the GDP/GTP state of Rab5. Our work indicates that impaired endosomal transport by GDP-bound Rab5 reduces Wg signaling, whereas enhanced endosomal fusion by GTP-bound Rab5 increases signaling. Because the GDP/GTP-binding state of Rab5 is controlled posttranslationally by GTPase-activating proteins and guanine nucleotide exchange factors, endocytic regulation likely allows more of a rapid adjustment of signaling than regulatory mechanisms requiring transcription and translation. Furthermore, because endocytic rates vary between cell types, this regulation may allow signaling to be adjusted in particular parts of the body or cells of a tissue. Thus, regulated endocytosis allows for precise temporal and spatial control of Wg signaling. Endocytosis is hypothesized to regulate signaling through several mechanisms. For example, lysosomal degradation of internalized active receptor tyrosine kinases serves to attenuate signaling (; ). However, our data suggest that Wg signaling is enhanced by endocytosis. One theory by which intracellular transport facilitates signaling is that the internalization of ligand–receptor complexes promotes interactions with other signaling members recruited to or already present on endosomes. In MAPK signaling, ERK1 receptors form protein complexes with endosomal MP1 and p14 (), leading to greater activation of signaling. Similarly, TGFβ signaling may be enhanced by receptor internalization to endosomes where the Smad2 anchor protein SARA is enriched (). Although our work and that of others suggests that Wg undergoes receptor-mediated internalization in the wing (; ), these data alone cannot explain the enhanced Wg signaling observed. However, not only are Wg and Arr colocalized in large endosomal accumulations in mutants, but they also colocalize with the cytoplasmic signaling component Dsh. The colocalization of Wg, Arr, and Dsh correlates with the increased expression of Wg readouts. These data suggest that internalization and endosomal transport may promote Wg signaling by facilitating associations between the Wg–receptor complex and downstream signaling components like Dsh. Interestingly, Dsh is reportedly present on intracellular vesicles, and mutations that impair vesicular localization do disrupt canonical Wg signaling (for review see ). Axin, a protein that inhibits Wg signaling by down-regulating Arm levels (), has also been shown to colocalize with Dsh on intracellular vesicles (). Upon Wg signaling, Axin relocalizes from intracellular puncta to the plasma membrane (). This correlates with Arm stabilization and increased Wg signaling. Because Axin associates with Dsh and the cytoplasmic tail of Arr (for review see ), we propose that internalized Wg forms an endosomal signaling complex that may relocalize Axin, thereby stabilizing Arm and facilitating signaling. S2R+ cells express all of the signaling components necessary to respond to exogenously added Wg (), making them well suited to study Wg signaling. S2R+ cells (a gift from P. Beachy, Johns Hopkins University School of Medicine, Baltimore, MD) were maintained in Schneider's Media (Invitrogen) with 10% heat-inactivated FBS (JRH Biosciences). For protein knockdown, dsRNAs were synthesized using the MEGAscript RNAi kit (Ambion) from PCR products containing the T7 promoter (taatacgactcactataggg). Primer pairs are shown in Table S1 (available at ).Several transfection protocols were tested in this study. Amounts for six-well plate transfections are shown as follows: 0.2 μg dsRNA, 0.2 μg Super8XTOPFlash or Super8XFOPFlash (), and 2 ng pRL-CMV (Promega) in 100 μL were sequentially combined with 3.2 μL Effectene Enhancer (QIAGEN), 10 μL Effectene (QIAGEN), and 10 S2R+ cells in 1.6 mL of growth media. Knockdown was assessed by Western blotting at multiple time points. Strong knockdown of dynamin was observed after 8 d. To assess the effect of s on Wg signaling, 1 mL of media containing or lacking Wg protein (see next section) was added 7 d after transfection. 1 d later, the cells were lysed to reconfirm protein knockdown and to assess luciferase levels using the Dual-Luciferase Reporter Assay System (Promega). For Rab5, however, only limited knockdown was observed using this protocol even after 8 d. To test the effect of Rab5, an alternative protocol was used. 2.5 μg Super8XTOPFlash or Super8XFOPFlash and 25 ng pRL-CMV in 1.275 mL were sequentially combined with 20 μL Effectene Enhancer, 12.5 μL Effectene, 2.5 μg dsRNA, and 2 × 10 S2R+ cells in 2.5 mL of growth media. Knockdown was assessed by Western blotting at multiple time points. Strong knockdown of Rab5 was observed after 8 d. To assess the effect of Rab5 on Wg signaling, Wg-conditioned media was added, and cells were lysed as described in protocol 1. To induce Wg signaling using Wg DNA rather than Wg-conditioned media, the following protocols were used: 1.25 μg Super8XTOPFlash or Super8XFOPFlash, 1.25 μg pMK33-Wg (a gift from N. Perrimon, Harvard Medical School, Boston, MA) or empty vector, and 12.5 ng pRL-CMV in 1.275 mL were sequentially combined with 20 μL Effectene Enhancer, 12.5 μL Effectene, 2.5 μg dsRNA, and 2 × 10 S2R+ cells in 2.5 mL of growth media. 4 or 8 d later, the cells were lysed to assess protein knockdown by Western blotting and luciferase levels. To test the effect of polIII-RL () and s-188-cc-RL (), the following protocol was used: 0.625 μg Super8XTOPFlash or Super8XFOPFlash, 1.25 μg pMK33-Wg or empty vector, and 0.625 μg polIII-RL or s-188-cc-RL in 1.275 mL were sequentially combined with 20 μL Effectene Enhancer, 12.5 μL Effectene, 2.5 μg dsRNA, and 2 × 10 S2R+ cells in 2.5 mL of growth media. 8 d later, the cells were lysed to assess protein knockdown by Western blotting and luciferase levels. To test the effect of tk-RL (Promega), the following protocol was used: 1.13 μg Super8XTOPFlash or Super8XFOPFlash, 1.25 μg pMK33-Wg or empty vector, and 0.13 μg tk-RL in 1.275 mL were sequentially combined with 20 μL Effectene Enhancer, 12.5 μL Effectene, 2.5 μg dsRNA, and 2 × 10 S2R+ cells in 2.5 mL of growth media. 8 d later, the cells were lysed to assess protein knockdown by Western blotting and luciferase levels. All luciferase results are presented as the mean Super8XTOPflash/RL or Super8XFOPFlash/RL and SEM of multiple independent trials relative to the EGFP control (Table S2). Significance was based on a two-tailed test. To obtain media containing and lacking Wg protein, S2 Tub-Wg cells (Drosophila Genomics Resource Center) and S2 cells were grown in M3 Media (Sigma-Aldrich) with 1 g/L of yeast extract, 2.5 g/L bactopeptone, and 10% heat-inactivated FBS. 125 μg/ml hygromycin (Sigma-Aldrich) was added to the S2 Tub-Wg media. Cells were pelleted by centrifugation. Media was used immediately or stored at −80°C. The presence of Wg protein was confirmed by Western blotting. Cells were washed with PBS and lysed in 1× Passive Lysis Buffer (Dual-Luciferase Assay; Promega) or radioimmunoprecipitation assay lysis buffer (0.150 M NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, and 0.05 M Tris, pH 8) supplemented with protease inhibitor cocktail (Complete). Proteins were quantified by Bradford assay. Blots were probed as described previously () using the following antibodies: mouse antidynamin (1:2,000; BD Biosciences), mouse anti-actin (1:5,000; MP Biomedicals), mouse anti-Arm (1:2,500; ), mouse anti-Wg 4D4 (1:2,000; ), and rabbit anti-Rab5 (1:500; ). Secondary goat HRP-conjugated anti–mouse and anti–rabbit antibodies were used at 1:2,500 (Jackson ImmunoResearch Laboratories), and bands were visualized by Western lightning chemiluminescence plus reagent (PerkinElmer). Blots were developed in a processor (M35A X-OMAT; Kodak), scanned with a scanner (ScanMaker 8700; Microtek) and the accompanying ScanWizard Pro software (Microtek), and processed for brightness using Photoshop software (Adobe). Crosses were maintained at 21°C unless otherwise stated. Wing discs were equal in size to controls and morphologically normal unless otherwise stated. Representative wings of eclosed flies are shown. Wings were either mounted in Permount (Fisher Scientific) or just placed on a slide and visualized with a stereomicroscope (MZ16; Leica) fitted with a planApo 1× objective and a camera (Microfire; Optronics). Wing pictures were captured using Image-Pro Plus (MediaCybernetics) and In-Focus (Meyer). For curled wings, images were processed by extended focus in Image-Pro Plus. Images were recolored, adjusted for brightness, and painted to remove excess wings in Photoshop (Adobe). Expression patterns of () and () were determined by crossing to and staining resultant larvae for β-galactosidase. Patterns did not alter with the cooverexpression of (). To inhibit dynamin function, the Gal4 drivers were crossed to (). Our analysis of expressed by was performed on discs with relatively normal morphology, as changes in gross morphology were observed in some discs. mutant clones were generated by crossing females to males and heat shocking the progeny for 1 h at 38°C 12–36 h after egg laying. Larvae were raised at 18°C and shifted to 35°C for 7 h immediately before dissection. Female larvae were processed as in conventional antibody staining (see next section) except that dissection and fixation were performed at the restrictive temperature to maintain a blockade in endocytosis. To affect early endosomal fusion, the Gal4 drivers were crossed to (), (a gift from M. Gonzalez-Gaitan, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany), and (). Our analyses of and were performed on wing discs with relatively normal morphology, as changes in gross morphology were observed in many discs. () flies were dissected from pupal cases to examine wing morphology. Wild-type Rab5 overexpression was also analyzed by crossing to < and heat shocking progeny for 5–15 min at 38°C during early larval development. To generate mitotic clones, or males were crossed to females. Progeny were heat shocked at 38°C for 1 h during early first instar development. Because maternally deposited Hrs is very stable, the phenotypes described in this study may not be evident in small clones induced late in development. The overexpression of Hrs was studied using , and (a gift from H. Krämer, University of Texas Southwestern Medical Center at Dallas, Dallas, TX). Genetic interactions were examined using (), (a gift from K. Bhat, Emory University School of Medicine, Atlanta, GA), and (). Wg signaling components were localized using the following stocks: (), (), (), , , and (). For conventional antibody staining, wandering third instar larvae were dissected in PBS, fixed in 4% formaldehyde in PBS, and incubated in primary antibody overnight. The following primary antibodies were used: mouse anti-Wg 4D4 (1:10; ), rabbit anti–β-galactosidase (1:1,000; Cappel), guinea pig anti-Sens (1:1,000; ), mouse anti-Dll (1:500; a gift from G. Boekhoff-Falk, University of Wisconsin, Madison, WI), rabbit anti-Dll (1:100; ), mouse anti-HA (1:100; Covance), guinea pig anti-Spinster/Benchwarmer (1:100; ), and rat anti-Dsh CB (1:1,000; ). Samples were later incubated in fluorescent conjugated secondary antibodies (1:300; Invitrogen and Jackson Immunochemicals). Samples were mounted in Vectashield mounting medium (Vector Laboratories) and were imaged using a confocal microscope (LSM 510; Carl Zeiss MicroImaging, Inc.) and accompanying software. Additional details of image acquisition and processing are shown in Table S3 (available at ). Control and experimental samples of each figure were taken at identical confocal settings. Single confocal sections of representative samples are shown unless otherwise stated. Extracellular protein staining was performed as described previously () using tubulin as a negative control. TUNEL labeling was performed as described previously () except that larvae were dissected in PBS and fixed in 4% formaldehyde in PBS. The TMR red In Situ Cell Death Detection Kit (Roche) was used. Changes in the columnar cell layer were evaluated. As a positive control, larvae underwent a 1-h heat shock at 38°C 1 d before TUNEL staining (Fig. S1 H). In situ hybridization was performed as described previously () and mounted in 50% glycerol in PBS. Images were acquired with an imaging system (Imager.Z1; Carl Zeiss MicroImaging, Inc.) fitted with a 63× NA 1.4 plan-Apochromat lens and a camera (Axiocam MRm; Carl Zeiss MicroImaging, Inc.) using Axiovision software (Carl Zeiss MicroImaging, Inc.). Images were recolored using Photoshop (Adobe). To determine the extent of wing notching, the intact wing perimeter of each wing was measured using ImageJ software (National Institutes of Health) and divided by the respective total estimated wing perimeter. For each genotype, the mean and SEM were calculated. Significance was based on a two-tailed test. To quantify the extent of protein colocalization in wing imaginal disc stainings, the number of colocalized pixels in a fixed area near the center of the wing pouch was measured using LabelVoxel and TissueStatistics functions of Amira (Indeed-Visual Concepts GmbH). Relative results are presented. and larvae were dissected in PBS and incubated in 0.5 g/L 3,3′-DAB (Sigma-Aldrich) + 0.003% HO to visualize HRP. Samples were fixed in 2% PFA, 75 mM lysine, 10 mM NaIO, 37 mM phosphate buffer, pH 7.4, and postfixed in 3% OsO. Samples were dehydrated and embedded. 55-nm thin sections were stained in 4% uranyl acetate and then in 2.2% lead nitrate and 3.5% sodium citrate. Images were acquired with an electron microscope (JEM-1010; JEOL) fitted with a digital camera (2k; Gatan). No HRP-positive structures were detected apically in the C5-Gal4–negative control, indicating that the staining is specific for expressed Wg HRP. The adult wing phenotype is consistent with increased Wg signaling, indicating that the fusion protein is functional. Fig. S1 shows our analysis of cell death in wing discs with altered endocytosis by TUNEL. Fig. S2 shows the effect of the enhancement of Rab5-mediated endosomal fusion on Wg signaling. Table S1 describes the specific sequences of dsRNA used for knockdown in our cell culture assays. Table S2 provides quantitative data from the cell culture Wg signaling assay, including negative controls. Table S3 describes additional methods for image acquisition and processing. Online supplemental material is available at .

Though the proper distribution of lipids among organelles is critical for numerous cellular functions, exactly how lipids are sorted and moved among intracellular membranes remains poorly understood. Some lipid sorting likely occurs during vesicular trafficking (; ). For example, sphingomyelin and cholesterol are partially excluded from COPI-coated vesicles, a process which may contribute to the enrichment of these lipids in the late Golgi network and the plasma membrane (PM; ). In addition, many classes of lipids are transferred between organelles by nonvesicular pathways. In most cases, the molecular mechanisms and regulation of these pathways have not been characterized. It has been known for some time that sterols, such as cholesterol in mammals and ergosterol in yeast, can move between cellular compartments via nonvesicular pathways whose functions are poorly understood. Treating mammalian cells with brefeldin A, which inhibits protein secretion by disassembling the Golgi complex (; ; ), only slightly slows the delivery of newly synthesized cholesterol from the ER to the PM (; ; ). Similar results were obtained in yeast using mutants with conditional defects in the proteins required for vesicular trafficking (; ). We have found that the movement of exogenous sterols from the PM to the ER sterol in yeast is also not blocked in mutants (). Nonvesicular transport pathways move sterols between other cellular compartments as well (; ; ; ; ). The proteins required for these nonvesicular transport pathways have not been identified. In mammals, nonvesicular cholesterol transport could be facilitated by lipid transfer proteins known to bind sterols, including some StART domain–containing proteins or SCP2 homologues (). However, the yeast lacks homologues of these proteins. We wondered if another class of lipid-binding proteins, the oxysterol-binding protein (OSBP)–related proteins (ORPs), might facilitate nonvesicular sterol transport in yeast. OSBP, which is the founding member of the ORP family, was identified as a cytosolic receptor for oxysterols (). In mammals, these compounds down-regulate cholesterol biosynthesis and uptake and have been implicated in several other cellular processes (). Oxysterols may also modulate sterol biosynthesis in (). The cloning of OBSP () was followed by the identification of a large family of ORPs, including 16 in humans and 7 in (). All of these proteins contain an OSBP-related domain, which binds oxysterols and other lipids. Most also contain one or more other domains, including pleckstrin homologue (PH) domains (which bind phosphoinositides [PIPs]), the ER-targeting motif FFAT, ankyrin repeats, and transmembrane domains; these domains serve to localize these proteins to various cellular compartments (, ; ; ). In addition, the localization of ORPs is not static; OBSP alters its intracellular location in response to oxysterols (). Some ORPs may function as lipid sensors that integrate lipid metabolism with other cellular processes. It has recently been shown that OSBP acts as a sensor that regulates two phosphatases in a cholesterol-dependent manner (). In addition, two yeast ORPs (Osh6p and Osh7p) interact with Vps4p, which is a member of the AAA ATPase family, and could regulate its function in response to cellular sterol levels (). It has also been suggested that ORPs are lipid transfer proteins (; ; ). We have recently solved the structure of the yeast ORP Osh4p/Kes1p and found that it looks remarkably like other lipid transfer proteins; it binds a single sterol molecule in a hydrophobic binding pocket covered by a “lid” domain (). A role for some yeast ORPs (called Osh proteins) in sterol transport is also supported by studies on mutants missing one or more of these proteins; intracellular sterol distribution is severely altered in mutants missing all of these proteins (; ). We investigated the role of these proteins in the nonvesicular transport of exogenous sterols from the PM to the ER in yeast. We measured the rate of PM to ER sterol transport in mutants lacking Osh proteins. These studies exploited the ER localization of the enzymes (known as acyl-coenzyme A:cholesterol acyltransferases [ACATs]) that convert free sterols to steryl esters (); the esterification of exogenous radiolabeled sterol indicates that the sterol has been moved to the ER. Because does not take up exogenous sterol during aerobic growth, we use strains that have an altered allele of a transcription factor () that permits aerobic sterol uptake (). These strains take up and use several sterols, including cholesterol and ergosterol. Both of these sterols support the growth of yeast mutants that cannot make sterols. Because the Osh proteins likely have overlapping functions (), we first measured PM to ER sterol transport in a mutant lacking all of these proteins. Yeast requires any one of the seven Osh proteins for viability (). Therefore, we used a strain ( ) that has a temperature-sensitive allele of one of the genes () and deletions of the other six genes (). At permissive temperatures, this strain esterifies exogenous cholesterol at less than half the rate of a strain that has all the Osh proteins ( ; ). strain esterifies exogenous cholesterol more than seven times more slowly than the strain, despite the fact that the two strains take up similar amounts of free cholesterol (). We also examined the esterification rate of exogenous ergosterol, which we previously showed is transported to the ER much more slowly than exogenous cholesterol (). cells; at 37°C, H-ergosteryl esters were not detectable (). To facilitate comparisons of the strains in different conditions, we have calculated their relative rates of esterification of exogenous sterols (). It should be noted that we have previously shown that the fraction of exogenous sterol that becomes esterified over time is not affected by the total amount taken up; therefore, comparisons between strains that take up slightly different amounts of sterol are valid (). at elevated temperature suggests that other pathways may also move cholesterol to the ER inefficiently. cells take up and esterify cholesterol more slowly at low temperatures (compare to and ), perhaps because steryl ester synthease activity decreases or membrane fluidity changes. Collectively, these findings show that efficient esterification of exogenous sterols requires Osh4p and likely other Osh proteins, particularly at elevated temperatures. , except that it has a wild-type allele. This strain does not have a conditional growth defect. esterifies exogenous cholesterol at a rate similar to (). strain, but still significantly slower than a strain that has all the Osh proteins (). cells also esterify exogenous ergosterol significantly more slowly than cells. (). Therefore, Osh proteins other than Osh4p are required for most of the esterification of exogenous sterol, though Osh4p makes some contribution. The slow esterification of exogenous sterols in the strains suggests that Osh4p and other Osh proteins are required for the efficient transfer of exogenous cholesterol from the PM to the ER. We performed several control experiments to rule out other explanations of the slow esterification. First, we found that this difference is not explained by low ACAT activity in the cells (). cells, this difference is not enough to explain the dramatically slower esterification of exogenous cholesterol by cells. cells, which is a possibility because intracellular sterol distribution is altered in cells (). and strains using the method of Lange and Steck, which makes use of the fact that ACAT only has access to sterols in the ER (). Whole-cell lysates are reacted to completion with an excess of C-oleoyl-CoA; the amount of C-steryl oleate formed indicates the quantity of sterol in the ER. cells (). Thus, the slow esterification of exogenous cholesterol by cells is not caused by elevated amounts of endogenous sterol in the ER. cells. To address this possibility, we performed pulse-chase experiments. and strains were grown at a nonpermissive temperature for 30 min with radiolabeled cholesterol and then chased with a large excess of unlabeled cholesterol for 40 min. strain then the amount of radiolabeled cholesteryl ester should have decreased after the chase. cells. cells because they have impaired sterol transport to the ER. Because numerous mutations that severely inhibit vesicular transport do not block the movement of exogenous cholesterol to the ER (), we conclude that the Osh proteins perform overlapping functions necessary for nonvesicular transport of exogenous cholesterol to the ER. Our findings suggest that, in addition to Osh4p, other Osh proteins are required for the efficient transfer of exogenous cholesterol from the PM to the ER. To estimate the role of these proteins, we determined the rate at which exogenous cholesterol is esterified in mutants missing each of the Osh proteins individually (). The relative rate of esterification slowed somewhat in cells missing either Osh3p or Osh5p (), but was not significantly altered in mutants lacking the other Osh proteins. These defects were additive, i.e., a mutant lacking both Osh3p and Osh5p esterified cholesterol more slowly than cells missing just one of these proteins (). These findings suggest that Osh3p and Osh5p, in addition to Osh4p, may move sterols from the PM to the ER. cells (compare and ), other Osh proteins may also facilitate PM to ER sterol transfer. Because the Osh proteins are needed for PM to ER sterol transport in vivo, we wanted to determine if one of the Osh proteins could extract sterols from donor membranes and transfer them to acceptor membranes in vitro. Osh4p was chosen because it is one of the best characterized of the Osh proteins and because it binds to phosphatidylinositol bisphosphate (PIP; ), which is a lipid enriched in the PM. Consistent with a role for Osh4p in sterol transport, purified Osh4p-extracted cholesterol from liposomes (). We also examined its ability to extract other lipids and found it could also extract PIP (1-arachidonyl-2-steroyl-PIP), though approximately five times less efficiently than cholesterol (). Osh4p could not extract any of the other lipids we tested, including the following: dipalmitoyl phosphatiylcholine (PC), phosphatidylserine (PS), a ceramide (oleoyl phytosphingosine), dioleoyl glycerol (DAG), and triolein (TAG; unpublished data). Thus, Osh4p extracts cholesterol, but not most other lipids, from membranes, suggesting Osh4p directly moves sterols. We next determined the ability of Osh4p to transfer sterols between membranes. Purified Osh4p moved cholesterol between liposomes in vitro in a time- () and concentration-dependent () manner. It also transported ergosterol, which is the primary sterol in yeast (). Thus, Osh4p can function as a sterol transport protein for both cholesterol and ergosterol. Osh4p transferred most other lipids poorly, though it could mediate some transfer of PIP and PS (). Consistent with a role for Osh4p in sterol transport, it moved even less of most other lipids from liposomes containing 10% cholesterol than from those that lacked cholesterol (). In contrast, as discussed in the section Sterol transport by Osh4p is regulated by PIPs, liposomes including 10% PS in donor membranes stimulated cholesterol transfer by Osh4p (). We conclude that Osh4p likely preferentially transports sterols between membranes, but can also move PIP and PS. If sterol transfer is one of the functions of Osh4p, then mutations in Osh4p that render it nonfunctional by complementation analysis might disrupt the ability of the protein to transport sterols. We studied proteins with mutations in the following three regions of Osh4p: the binding tunnel (Y97F and L111D), the deletion of the conserved 29–amino acid lid domain (Δ29) that covers the binding tunnel, and the conserved charged residues near the mouth of the cholesterol-binding tunnel might be needed for interaction of Osh4p with bilayers (K109A, K168A, K336A, HH143, and 144AA; ). With the exception of K168A, all of these mutations render Osh4p nonfunctional by complementation analysis (). All but one of the mutations decreases the ability of Osh4p to transport cholesterol () and substantially reduce its ability to extract cholesterol (). Therefore, we conclude that lipid transfer is likely one of the physiological functions of Osh4p. We found that PIP stimulates sterol transfer by Osh4p. Because cholesterol transport was still stimulated by this PIP in all of the Osh4p mutants (, compare red and blue bars), none of the residues we altered is required for the PIP stimulation of sterol transport by Osh4p. Because Osh4p can also transfer some PIP and PS between membranes, we wondered if Osh4p mutants that were unable to transfer cholesterol efficiently also transferred these lipids poorly. PIP and PS transport were not affected by Y97F, but were reduced by L111D (). Because Y97 interacts with the hydroxyl group of sterols (), it is perhaps not surprising that altering this residue does not affect the ability of Osh4p to move other lipids. We speculate that the acyl moieties of PIP and PS bind in place of sterol in the hydrophobic tunnel and that the charged headgroups of these lipids are accommodated by an open conformation of the lid (). Because Osh4p has been shown to bind PIPs (), we wondered if they might regulate sterol transfer by Osh4p. It has previously been shown that the PIP-binding domain of a ceramide transfer protein is needed for it to function in vivo (). To determine if PIPs affect sterol transport by Osh4p in vitro, donor and acceptor liposomes were prepared either with or without 0.5 mol% PIP because Osh4p binds this PIP (). We found that PIP stimulates cholesterol transport by Osh4p when it is on either donor or acceptor vesicles (). These effects were additive because transfer was fastest when both donor and acceptor vesicles contained PIP. To further characterize the affect of PIP on Osh4p, we measured the rate of cholesterol extraction from liposomes by Osh4p and found it was stimulated ∼10-fold when the liposomes contained 0.5 mol% PIP (). We determined the ability of other lipids to stimulate sterol transfer by Osh4p (). No enhancement of transport was found when the donor membrane contained 0.5 mol% of other acidic phospholipids such as phosphatidic acid or PS, suggesting that the enhancement of transport with PIP is specific for this PIP. A smaller enhancement of transport was also found when donor membranes contained 10 mol% PS. The stimulatory affects of 10 mol% PS and 0.5 mol% PIP were additive. In addition, we found that only PIP, and, to a lesser extent PIP, were able to increase the rate of cholesterol transport by Osh4p (). It should be noted, however, that lacks detectable levels of PIP. Therefore, we conclude that PIP is the primary PIP species able to stimulate sterol transport by Osh4p in cells. Sterol transfer is also slightly stimulated by PS. Because PIP is enriched in the PM and is also found in other compartments including Golgi membranes, it likely stimulates sterol transfer at these compartments. PIP may stimulate sterol transfer by Osh4p by increasing the affinity of Osh4p for PIP-containing membranes, probably by binding to PIP headgroups on the bilayer surface. Although low levels of PIP stimulate transport, we wondered if high levels of PIP might inhibit transfer by effectively trapping Osh4p on membranes. Consistent with this, we found that, with a 1:1 ratio of donor to acceptor vesicles, increasing the amount of PIP in donor vesicles by 10-fold slowed transport (; 5.0% PIP2 in donor). In addition, even low levels of PIP in acceptor vesicles can inhibit transfer when the ratio of donor to acceptor vesicles is 1:10 (, column 3). Collectively, these results suggest that PIP modulates sterol transfer by Osh4p, likely by affecting its affinity for membranes. Our findings suggest that PIPs regulate sterol transport by Osh4p and possibly other Osh proteins, some of which also bind PIPs (; ; ). To confirm the requirement of PIPs for efficient PM to ER sterol transport in cells, we used mutants with conditional defects in PIP and PIP synthesis. Stt4p and Pik1p are essential PI-4 kinases in yeast. Mutants with conditional defects in these proteins have reduced levels of PIP and PIP at elevated temperature (). strain (). The differences in the total amount of sterol taken up by the strains does not affect the analysis because we have previously shown that the fraction of exogenous sterol that becomes esterified over time is not affected by the total amount taken up (). To look more directly at the requirement of PIP, we used a strain, which has a conditional defect in PIP 5 kinase. At elevated temperature, mutants have reduced levels of PIP, but normal amounts of other PIPs (; ). We found that cells depleted of PIP had defects in sterol transfer to the ER (), suggesting that this PIP is particularly important for rapid sterol transport. We ruled out that these differences are caused by differences in ACAT levels in these strains (). Therefore, normal levels of PIP and PIP are required for efficient transfer of exogenous cholesterol to the ER. PIP depletion likely disrupts nonvesicular PM to ER sterol transport because mutations that dramatically slow vesicular trafficking do not affect the movement of exogenous sterols to the ER (). In addition, because yeast lacks putative nonvesicular sterol transporters other than the Osh proteins (i.e., it does not have StART or SCP2 homologues), PIP depletion may affect sterol transport by Osh4p and other PIP-responsive Osh proteins. Because Osh4p can transfer PS between liposomes in vitro, we wondered if some of the Osh proteins might transport PS in cells. PS is synthesized in the ER from cytidine diphosphate-diacylglycerol and serine and can subsequently be converted to PE by Psd (). Yeast has two Psds, one in the Golgi complex (Psd2p) and the other in mitochondria (Psd1p; ; ). Because these enzymes are not in the ER, the conversion of PS to PE requires the transfer of PS to the Golgi complex or mitochondria. Nonvesicular transport pathways mediate this transfer (). The PE generated by the Psds can be returned to the ER and converted to PC. Though some of the components of this pathway have been identified (), the lipid transfer proteins that move PS have not been determined. If the Osh proteins were required for nonvesicular PS transfer in vivo, then the conversion of newly synthesized PS to PE might be expected to slow in cells lacking Osh proteins. When wild-type cells were labeled with H-serine for 30 min, ∼50% of the PS synthesized was converted into PE and PC (). To confirm that the radiolabeled PE and PC were generated from H-PS, we found that a strain missing the two Psds ( ) makes small amounts of radiolabeled PE or PC (). The conversion of PS to PE and PC was not blocked in the strains (). It is not clear why these strains convert a greater proportion of newly synthesized PS to PE and PC than a wild-type strain, but lipid metabolism may be altered in these strains. The finding that the conversion of PS to PE is not blocked in strains suggests that Osh proteins may not be needed for the nonvesicular transfer of PS to the Psds. However, because Psd2p is located in the Golgi complex, newly synthesized PS might still reach Psd2p by vesicular transport. strain. Sec18p is the yeast homologue of -ethylmaleimide-sensitive factor and is essential for most vesicular trafficking (; ). cells were still able to convert newly synthesized PS to PE and PC after half an hour at a nonpermissive temperature (), which are conditions that result in a severe block in vesicular transport (). It is also possible that Osh proteins might be needed to move PS to only one of the Psds. Because several Osh proteins have been localized to the Golgi complex, but none to the mitochondria, we thought it more likely that they would transfer PS to Psd2p than Psd1p. To investigate this possibility, we determined the rate of conversion of PS to PE and PC in cells missing Psd1p. Cells lacking Psd1p still convert PS to PE and PC, though slightly less efficiently than a wild-type strain (). strain. At nonpermissive temperatures, this strain is still able to convert PS to PE and PC (). Collectively, these results suggest that the Osh proteins are not required for the nonvesicular transfer of PS in the ER to Psds in the Golgi complex or mitochondria. The transport of exogenous sterols from the PM to the ER in yeast is not blocked in numerous mutants that have severe defects in vesicular trafficking (). We demonstrate that this transport slows significantly in cells lacking Osh proteins. Consistent with a direct role for these proteins in sterol transport, we find that a representative member of this family, Osh4p, transfers sterols between liposomes in vitro. This is the first demonstration that ORPs are lipid transfer proteins. We propose that nonvesicular transport of endogenous sterols is one of the functions of the ORPs in yeast, and probably in higher eukaryotes as well. It seems likely that some Osh proteins have other functions. They may transport other classes of lipids in addition to sterols. We found that Osh4p also transfers PIP and PS between liposomes, though less efficiently than cholesterol and ergosterol. However, Osh proteins may not transport significant amounts of PS in vivo because they are not required for nonvesicular PS transport to mitochondria or the Golgi complex. It is also likely that some of the Osh proteins function as lipid sensors that regulate other proteins, a function that has recently been demonstrated for OSBP in mammals (). PM to ER sterol transport is likely mediated by several Osh proteins because it is unaffected or only moderately slower in cells missing any one of them. Our findings suggest that, in addition to Osh4p, Osh3p and Osh5p may also transfer sterols from the PM to ER in vivo. A role for Osh3p and Osh5p in PM to ER sterol transport is consistent with their largely cytoplasmic localization and the finding that the Osh3p PH domain (without the OSBP-related domain) localizes to the PM (; ). Because cells missing Osh3p and Osh5p still move exogenous sterols to the ER more rapidly than cells, other Osh proteins, including Osh4p, probably also transfer PM sterols to the ER in vivo. We have previously shown that efficient PM to ER transfer requires either of two ATP-binding cassette transporters in the PM (). These transporters are also needed for efficient uptake of exogenous sterols by yeast (). How the ATP-binding cassette transporters affect the rate of PM to ER sterol transport is not known. They might directly facilitate the movement of sterols out of the PM, perhaps to Osh proteins. Alternatively, they might affect the composition or distribution of lipids in the PM in such a way that Osh proteins can more easily extract sterols and move them to internal compartments. Why yeast has such a large number of Osh proteins remains an open question. Several Osh proteins localize to various compartments throughout the cell (; ; ; ). Therefore, some Osh proteins may transport sterols or other lipids primarily to or from specific organelles. Genetic analysis has revealed that all seven Osh proteins have a single overlapping essential function and suggests that loss of the Osh proteins causes a defect in the recycling of ergosterol back to the PM (; ). Thus, while individual Osh proteins may move lipids primarily to or from particular organelles, they may all have the ability to transfer sterols from endosomes back to the PM. Alternatively, the Osh proteins could have a shared essential function as lipid sensors. Osh4p is partially localized to the Golgi complex and has been genetically implicated in Golgi function (; ). Sec14p is an essential PI-PC transfer protein that is required for proper Golgi function. However, cells lacking Osh4p do not require Sec14p for viability. It has been proposed that Sec14p acts as a lipid-sensor that regulates a functionally important pool of DAG in the Golgi complex by modulating enzymes that consume or generate DAG (; ). What role Osh4p plays in this process is not clear. Like Sec14p, it could act as a lipid sensor at the Golgi complex. Alternatively, Osh4p could affect the lipid composition and function of the Golgi by moving sterols, PIPs, or some other lipid to or from this organelle. We have also found that PIPs stimulate sterol transfer by Osh4p. PIP stimulation is probably a wide-spread property of ORPs. Indeed, we find that PIP-depleted cells have defects in PM to ER sterol transport. Some large mammalian and yeast ORPs interact with PIPs via PH domains (; ). The fact that Osh4p and some other Osh proteins lack these domains () but bind PIPs (; ) suggests that PIPs may interact with ORPs via multiple mechanisms. Osh4p likely binds to PIP headgroups using some part of its external surface. Because altering the conserved charged residues near the mouth of the sterol-binding tunnel did not affect the ability of PIP to stimulate transport, these residues are probably not required for PIP binding. Bankaitis and coworkers showed that an Osh4p with mutations in three basic residues fails to bind PIP (). When we altered these residues, the resulting protein was largely insoluble (unpublished data) and we were not able to obtain enough protein to assess sterol transport by this mutant. Our findings suggest that PIPs stimulate sterol transport by Osh4p, and likely other ORPs, by increasing the affinity of Osh4p for PIP-containing membranes. Because high levels of PIPs can also inhibit sterol transfer by Osh4p, changes in PIP levels, perhaps in response to cell-signaling events, could regulate sterol transfer by ORPs. In addition, because different PIP species are enriched in various cellular compartments, PIP stimulation of ORPs likely serves to regulate the movement and distribution of sterols (and possibly other lipids) among cellular compartments by ORPs. The identification of ORPs as PIP-regulated nonvesicular lipid transfer proteins in yeast provides a good model system to help us begin to understand the role of these proteins in maintaining intracellular lipid distribution. The strains used in this study are listed in . To allow these strains to take up exogenous cholesterol during aerobic growth, their alleles were replaced with by homologous recombination, as follows. A 757-bp fragment of the allele (from nucleotide 1985 to the end of the gene) was cloned into YIPlac211 (). The resulting plasmid was used as a template for a PCR reaction with primers UPC2-A2 and UPC-C (). The product of this reaction was used to transform cells, and the transformants were selected on medium that lacked uracil. The presence of the allele was confirmed, as previously described (). To introduce into cells, a 901-bp fragment of sec18-1 containing the mutation in this allele was cloned into YIPlac211. The resulting plasmid was used as template for a PCR reaction with the primers SEC18-C and SEC18-B1 (). The product of this reaction was used to transform cells, and the transformants were selected on medium that lacked uracil. To delete in cells, a PCR reaction using primers PSD1∷URA-f and PSD1∷URA-r () using the plasmid pRS416 () as a template. The resulting product was used to transform cells, and the transformants were selected on medium that lacked uracil. was confirmed by PCR. in WPY817, this strain was plated on medium containing 5-fluoroortic acid () to select cells unable to grow without uracil. Lipids were purchased from Avanti Polar Lipids, Inc. Radiolabeled lipids were purchased from American Radiolabeled Chemicals, Inc., except for H-ergosterol, which was isolated as previously described (), and H-PS, which was made as follows: yeast cells were labeled with H-acetic acid, and phospholipids were extracted and separated as previously described () using an Agilent 1100 series HPLC. The peak containing H-PS (detected at 203 nm) was identified by comparison to known standards and collected. The species of purchased radiolabeled lipids used were as follows: 1-arachidonyl-2-steroyl-PIP; PC; oleoyl phytosphingosine; dioleoyl glycerol; and triolein. Purified Osh4p was obtained as previously described (). The uptake and esterification of C-cholesterol was determined as previously described (). In brief, C-cholesterol in Tween80/ethanol (1:1) was added at the indicated concentration to the medium of growing cells so that the final Tween80 concentration was 0.5%. For pulse-chase experiments, cells were grown and 23°C, shifted to 37°C for 30 min, and 2.0 μM C-cholesterol added to the medium. After 30 min, 250 μM of unlabeled cholesterol was added to the medium and cells were grown for an additional 40 min. Samples were removed at the indicated times and lipids were extracted and analyzed as previously described (). ACAT activity of the strains was measured as previously described (). The relative amount of sterol in the ER was determined basically as previously described (). Cells from rapidly growing cultures were pelleted, washed once with ice-cold HO, resuspended in 20 mM Tris, pH 7.5, 100 mM NaCl, 1 mM DTT, and lysed by homogenization with glass beads in a Bead-Beater 8 (BioSpec Products, Inc.). The lysate was centrifuged for 5 min at 500 to remove debris and unlysed cells. For the assay, 200 μg lysate was mixed with 25 μM C-oleoyl-CoA in a total volume of 100 μL of 20 mM Tris, pH 7.5, and 100 mM NaCl. The reaction was incubated at 37°C for 10 min and stopped by the addition of CHCl and MeOH. Lipids were extracted and quantitated as previously described (). Liposomes were prepared by mixing dioleoyl phosphatidyl choline (DOPC) and C-cholesterol at a molar ratio of 99:1. The lipids were dried under a stream of nitrogen and resuspended in 20 mM Tris-HCl, pH 7.4, and 100 mM NaCl (TS) at a final concentration of 1 mM. The membranes were allowed to swell at room temperature for at least 1 h and vigorously vortexed, and then liposomes with a mean diameter of 1 μm were made using a mini-extruder from Avanti Polar Lipids, Inc. For the extraction assay, purified Osh4p was incubated with 50 μL of membranes at 30°C for the indicated times, after which the membranes were kept at 4°C and pelleted by centrifugation in a rotor (model TLA100; Beckman Coulter) at 70,000 rpm (191,000 ) for 1 h. The amount of C-cholesterol in the supernatant was determined by scintillation counting and was taken to be cholesterol that had been extracted from the liposomes. The amount of radiolabel in the supernatant in control reactions without Osh4p (usually ∼2% of the total) was subtracted from the amount measured with Osh4p. Lipid transport was assayed as previously described (), with the following modifications. Unless otherwise indicated, the lipid composition of the donor membranes was DOPC/egg PE/lactosyl-PE/cholesterol/PIP (59.5:20:10:10:0.5). Acceptor liposomes were prepared by mixing DOPC and egg PE at a ratio of 80:20, unless otherwise indicated. The lipids were dried under a stream of nitrogen, resuspended in 20 mM Hepes, pH 7.4, 100 mM NaCl, and 1 mM EDTA (HES) at either 1 or 10 mM (acceptor vesicles) and allowed to swell at room temperature for at least 1 h; they were then vigorously vortexed. Acceptor vesicles to be used for 1:10 donor to acceptor lipid transport assays were prepared by sonication with a Branson tip sonicator. Before use, the liposome solution was cleared by centrifugation at 15,000g at 4°C for 10 min. All donor vesicles and acceptor vesicles used in 1:1 donor to acceptor transport assays were prepared by extrusion using a filter with 0.2 μm pore size. All vesicles were used within 24 h of preparation. For the transport assay, 50 μL of donor membranes and acceptor membranes were mixed together with Osh4p in a total volume of 110 μL. After incubation at 30°C, 30 μg of agglutinin (Vector Laboratories) was added, and the reactions were placed on ice for 15 min, with occasional mixing. The donor membranes were then pelleted by centrifugation at 15,000 for 5 min at 4°C. The amount of radiolabel in the supernatant was determined by scintillation counting and taken to be lipid that had been transferred to acceptor membranes. The amount of radiolabel in the supernatant in control reactions without Osh4p or with Osh4p and incubated on ice was subtracted from the amount measured with Osh4p. 30-ml cultures were grown in synthetic complete medium with 1 mM ethanolamine at 25°C. They were shifted to 37°C, grown for 25 min and then 1 mM S-adenosyl methionine, 2 mM methionine, and 10 μg/ml cerulenin (from a 5 mg/ml stock in DMSO) were added to the medium. Control experiments showed that these compounds were needed to prevent radiolabel from being incorporated into PE and PC in cells missing Psd1p and Psd2p. The cultures were grown for 5 min, 50 μCi H-serine (American Radiolabeled Chemicals, Inc.) was added, and the cultures were grown for 30 min more. They were placed at 4°C and washed once with ice-cold HO. Lipids were extracted () and phospholipids were separated as previously described (). The peaks containing PS, PE, and PC were collected and counted in a scintillation counter.

The activation of natural killer (NK) cells, which is mediated by the recognition of their ligands by activation receptors, triggers a complex, highly regulated response leading to the death of a target cell. Cytolytic responses require reorganization of the actin cytoskeleton for receptor translocation at the cell surface to facilitate conjugate formation, initiation, the continuation of signaling, and effective killing (; ). This dynamic and highly complex process is regulated by a variety of actin-binding proteins, such as cofilin, profilin, and Wiskott-Aldrich syndrome protein (WASp), as well as Scar family proteins, thymosins, capping proteins, and the Arp2/3 complex (; ). The latter is crucial for actin nucleation and formation of new actin filament branches (). The Arp2/3 complex, which is formed from seven subunits (; ), is closely regulated by WASp family proteins (). WASp binds to the Arp2/3 complex (), increasing its affinity for ATP (). WASp activity, in turn, is tightly controlled by a variety of adaptor and regulatory proteins (), including the WASp-interacting protein (WIP; ). NK cells are very useful for studying the regulation of cytotoxic cell activity because of the expression of the killer cell immunoglobulin-like receptors (KIR; ) that are capable of inhibition of NK cell cytotoxicity (). A characteristic feature of inhibitory KIR molecules is a long cytoplasmic tail (76–95 amino acids) containing one or two immunoreceptor tyrosine-based inhibition motifs (ITIMs; ). Upon tyrosine phosphorylation, inhibitory KIR ITIMs serve as docking sites for the protein tyrosine phosphatases SHP-1 and -2 (; ), which can dephosphorylate kinases and adaptor proteins that are involved in early events of signal transduction (), leading to inhibition of NK cell activity. The number of proteins that are dephosphorylated is presently unknown. In this study, we describe a multiprotein complex comprised of WIP, WASp, actin, and myosin IIA that is formed during NK cell activation, but is not formed in the presence of KIR2DL1 inhibitory signaling. The identification of this multimolecular protein complex that is regulated by activating and inhibitory signaling provides insight into understanding the cytoskeletal rearrangements and mechanisms essential for NK cytolytic activity. First, YTS/KIR2DL1/FLAG-WIP cells were activated by mixing with the appropriate target cells. Immunoprecipitation with anti-FLAG mAb revealed a substantial number of proteins that coimmunoprecipitated with FLAG-WIP (). Specific components of this complex (marked by asterisks in ) were identified by tandem mass spectrometry (MS). The presence of WIP and WASp in the complex was confirmed, and the additional presence of actin, myosin IIA heavy chain, and two myosin light chains was revealed (). An example is shown for myosin IIA identification in . The presence of these proteins was subsequently verified by immunoblotting (). Because WIP has been shown to interact with WASp (), and actin has been demonstrated to bind both WIP () and WASp (), the presence of WASp and actin in the complex was not surprising. However, the inducible recruitment of actin and the constitutive presence of WASp in the complex are noteworthy. The presence of both heavy and light chains of myosin IIA (, A [left] and C) indicates their probable involvement in cytotoxic NK cell activity. An interaction between WIP and a myosin in mammalian cells has not been previously reported. Treatment of cells with latrunculin A and B, which bind to monomeric actin and prevent its polymerization (), eliminated actin binding to WIP in a dose-dependent manner (), as well as myosin IIA binding (unpublished data); therefore, the complex could not be formed. However, either stabilization of actin filaments by jasplakinolide or treatment of cells with swinholide A, which has filament-severing ability (), both resulted in increased actin binding to WIP (). Thus, the actin found in the complex appears to be F-actin. The WIP–WASp interaction was stable under all conditions. In resting NK cells, FLAG-WIP and the majority of WASp were eluted with the partition coefficient K = 0.43 (), corresponding to a mass of ∼350–400 kD on the calibration curve. WIP and WASp migrated on SDS-PAGE as proteins of 60 and 62 kD, respectively. The presence of WIP and WASp in the relatively high mass fraction may reflect either their interaction with other proteins in the cell or oligomerization. Importantly, neither myosin IIA nor actin was found in the fractions containing WIP and WASp, demonstrating that the complex was not formed in resting cells. However, the activation of NK cells resulted in the shift of the majority of WIP and WASp to a fraction, with the partition coefficient K = 0.3, corresponding to a mass of ∼1,200–1,500 kD (). Moreover, the change of WIP–WASp position was accompanied by the simultaneous appearance of myosin IIA and actin in the same fraction, indicating that a multiprotein complex of ∼1,300 kD was formed after NK cell activation. The majority of actin in both resting and activated cells, likely representing the monomeric actin pool, was eluted with K = 0.61 (), which is the same K as that of the molecular mass standard ovalbumin (43 kD), thus validating the measurements. To evaluate the function of inhibitory receptor KIR2DL1 in YTS cells conjugated with physiological target cells, phosphorylation of the receptor was investigated. First, phosphorylation of KIR2DL1 that had previously been transduced into YTS cells () after antibody cross-linking was determined over time. Maximum KIR2DL1 phosphorylation in YTS cells was found at 3 min (), in agreement with a previous study () showing maximum phosphorylation of KIR at 3.5 min after ligand binding on the surface of insect cells. Induction of KIR2DL1 phosphorylation was specific because irrelevant antibody did not induce KIR phosphorylation. Importantly, a signaling-deficient KIR2DL1*250 mutant lacking both ITIMs () was not phosphorylated. Interestingly, KIR2DL1 migrated on the gel as a doublet after induction, likely because of differential phosphorylation of the two ITIMs in KIR2DL1 because at time 0 unphosphorylated KIR2DL1 migrated as a single band. Similarly, KIR2DL1*250 migrated as a single nonphosphorylated band (). Next, the effect of inhibitory signaling in YTS/KIR2DL1/FLAG-WIP cells was examined using 721.221/Cw6 cells as the target, as HLA-Cw6 is a ligand for KIR2DL1 (). Complex formation was strongly diminished in the presence of KIR2DL1-mediated signaling (, A [right] and C). Specifically, actin was recruited to the complex after 3 min of NK cell activation, with maximum phosphorylation at 10 min, and its amount decreased with time. In contrast, inhibitory signaling largely prevented the appearance of actin in the complex. Association of myosin IIA with the WIP–WASp complex increased substantially by 10 min after NK cell activation and correlated with the appearance of myosin regulatory light chain and actin. An increased amount of myosin IIA was observed, even at 60 min. In contrast, NK cell inhibition only resulted in a slight increase in the amount of myosin IIA and its association with the WIP-based complex was quickly disrupted. Thus, KIR2DL1 inhibitory signaling effected the interaction of myosin IIA with the complex, similarly to actin. Interestingly, WASp that is present in FLAG-WIP immunoprecipitates was not affected by inhibitory signaling. To test if the observed effects were KIR2DL1 mediated, the YTS/KIR2DL1*250 cell line that is deficient in KIR signaling was used. The binding of KIR2DL1*250 to its ligand had no effect on complex formation (), and the resulting complex was identical to that found in activated YTS/KIR2DL1 cells. Thus, KIR2DL1*250 was unable to influence complex formation, demonstrating that the KIR inhibitory receptor signal affects complex assembly and/or function of the multiprotein complex created during NK cell activation. Formation of the complex and its alteration by inhibitory signaling raised the question of whether or not the complex is a part of the immune synapse. Therefore, we examined the localization of the complex in the cell after activation or inhibition. In resting YTS/KIR2DL1/FLAG-WIP cells, nearly all FLAG-WIP colocalized with WASp, but not all WASp colocalized with FLAG-WIP, likely reflecting a pool of WASp either bound to endogenous WIP or not associated with WIP (). Importantly, both proteins were dispersed throughout the cyto-plasm, whereas myosin IIA (as well as F-actin; Fig. S1, available at ) was located at the cell periphery. The activation of NK cells by mixing with target cells resulted in the relocation of FLAG-WIP and WASp and the accumulation of myosin IIA at the cell–cell contact site (), where F-actin was also localized (Fig. S1; ). FLAG-WIP was polarized toward the contact site in 76% of conjugates of NK cells with susceptible target cells, and 16% of it was found at the synapse. Similarly, WASp was polarized in 64% of cytolytic conjugates, with 22% found at the cell–cell interface. The remainder of the WIP–WASp complex polarized to a region adjacent to the synapse. Myosin IIA accumulated at the contact site in 60% of conjugates () and colocalized with FLAG-WIP (). F-actin also accumulated at the cell–cell interface in 85% of cytolytic conjugates (Fig. S1), in agreement with a previous study (). Strikingly, in conjugates formed between NK cells and nonsusceptible 721.221/Cw6 cells (that initiated inhibitory signaling in YTS/KIR2DL1/FLAG-WIP cells) neither FLAG-WIP, WASp, myosin IIA, nor F-actin polarized toward the cell–cell contact site ( and Fig. S1), and the distribution of those proteins remained as it was in resting cells (). Thus, in response to NK cell activation, WIP, WASp, and myosin IIA moved to the cell–cell contact site with F-actin, whereas KIR2DL1 inhibitory signaling prevented that relocalization. The constitutive presence of a WIP–WASp complex, which is unaffected by inhibitory signaling (), raised the question of whether recruitment of actin and myosin IIA to the complex is mediated by WIP or by WASp. To test this, the YTS/KIR2DL1 cells were transfected with a WIP COOH- terminal deletion mutant, FLAG-WIPΔ460–503, which excluded the region required for WASp binding (amino acid residues 461–485 of WIP; ) and should prevent WASp from binding to WIP. Analysis of anti-FLAG immunoprecipitates from activated YTS/KIR2DL1/FLAG-WIPΔ460–503 cells revealed that WASp was not pulled down by the mutant FLAG-WIPΔ460–503 protein. However, this mutant still allowed recruitment of actin and myosin IIA to the complex (). Recruitment of actin, but not myosin IIA heavy chain, to the complex was decreased in activated YTS/KIR2DL1/FLAG-WIPΔ460–503 cells, suggesting that WASp may contribute to actin recruitment or deletion of the COOH-terminal part of WIP and may cause changes in WIP conformation affecting actin binding. Nevertheless, actin and myosin IIA interaction with the complex appears to be independent of WASp and to require functional WIP. To assess possible mechanisms governing the control of generation of the multiprotein complex, phosphorylation of the complex components was examined. Because WIP, but not WASp, seemed to play an essential role in complex formation (), WIP phosphorylation in response to NK cell signaling was investigated, with particular focus on whether the phosphorylation of WIP could affect interaction of the components of the complex with WIP. WIP has been shown to undergo phosphorylation in T cells in response to CD3ζ chain cross-linking (). Because WIP has been demonstrated to inhibit WASp activity in vitro (), it has been proposed that WIP phosphorylation allows WASp to dissociate form WIP, releasing WASp from WIP inhibition (). To evaluate this paradigm in NK cells the YTS/KIR2DL1/FLAG-WIP cells were labeled with [P]orthophosphate and WIP phosphorylation status was analyzed after the activation and inhibition of NK cells with appropriate target cells. After the activation or inhibition of NK cells, anti-FLAG immunoprecipitation demonstrated two distinct bands of phosphoproteins, which were identified as WIP and WASp by immunoblotting (). WIP was phosphorylated even at time 0, suggesting that WIP is constitutively phosphorylated in NK cells. WIP phosphorylation quickly increased upon NK cell activation to twofold after 3 min, reached a maximum of 2.7-fold at 10 min, and decreased to the basal level at 60 min (). On the contrary, inhibition of NK cell activity by KIR2DL1 signaling resulted in no increase of WIP phosphorylation, indicating that KIR2DL1 inhibitory signaling influenced WIP phosphorylation. Surprisingly, phosphorylation of WASp pulled down by FLAG-WIP did not change significantly during activation or inhibition of NK cells (unpublished data), but it is likely that small WASp phosphorylation changes could have been undetected because of its constitutive phosphorylation at S483 and S484 (). Most interestingly, changes in WIP phosphorylation states did not affect WASp binding, as defined by coimmunoprecipitation studies, indicating that the function of WIP phosphorylation during NK cell activation may be different than the regulation of WIP–WASp association. WIP was indirectly shown to be phosphorylated by PKCθ in T cells (). Because WIP phosphorylation was not required for regulation of the WIP–WASp interaction, the possible function of WIP phosphorylation and the identification of the kinase responsible for WIP phosphorylation in NK cells were examined next. Analysis of the WIP amino acid sequence revealed several consensus sites for PKC- and casein kinase II–mediated phosphorylation. Consequently, the effects of PKC and CKII inhibitors were studied. To test if PKC family kinases are involved, a general inhibitor of PKC activity, bisindolylmaleimide I, was used, as well as PKCα,β and PKCθ pseudosubstrates, which specifically inhibit activity of α, β, and θ isoenzymes, respectively. Pretreatment with bisindolylmaleimide I, followed by activation of YTS/KIR2DL1/FLAG-WIP cells, resulted in severely decreased WIP phosphorylation compared with control, indicating that a PKC family kinase is involved in WIP phosphorylation (). Pretreatment of cells with PKCα,β pseudosubstrate () or with Gö6983, which inhibits PKC α, β, γ, δ, and ζ (80 nM; unpublished data), did not decrease WIP phosphorylation after NK cell activation, suggesting that these isoenzymes were not involved in WIP phosphorylation. Similarly, treatment with specific CKII inhibitor did not affect the WIP phosphorylation state, indicating that casein kinase II is not involved in WIP phosphorylation in vivo. However, pretreatment with the pseudosubstrate inhibitor specific for PKCθ prevented WIP phosphorylation after NK cell stimulation (), demonstrating that PKCθ is responsible for phosphorylation of WIP in NK cells. Because WIP phosphorylation was affected by activation and inhibitory signaling (), and the changes in WIP phosphorylation levels over time appeared to be related to recruitment of actin and myosin IIA to the complex, the relationship between the WIP phosphorylation state and the formation of the multiprotein complex was investigated. In addition to PKC inhibitors, the YTS/KIR2DL1/FLAG-WIP cells were treated with protein tyrosine kinase inhibitors. Myosin inhibitors were also used to assess the effect of myosin activity inhibition on complex formation. Pretreatment of cells before activation with an Src family kinase inhibitor, PP2, did not affect WIP phosphorylation or the recruitment of actin and myosin IIA to the complex, indicating that complex formation is not regulated by Src family kinases (). A decrease of actin and myosin IIA binding was seen after treatment with 100 μM of the general protein tyrosine kinase inhibitor genistein (unpublished data), but the observed effects were most likely nonspecific because of the inhibition of all tyrosine kinases and an impairment of NK cell functions. Inhibition of myosin IIA heavy chain or regulatory light chain activity with blebbistatin or ML-7, respectively, before NK cell activation did not disrupt complex formation. However, inhibition of PKC activity abolished both actin and myosin IIA, but not WASp, recruitment. The disruption of complex formation was caused only by specific inhibition of the θ and not the α, β (), γ, δ, or ζ (unpublished data) isoforms. Hence, PKCθ activity is crucial for WIP phosphorylation. The results suggest that actin and myosin IIA recruitment and proper complex formation during NK cell activation correlate with WIP phosphorylation and are likely dependent on it. Changes in WIP phosphorylation during NK cell activation that correlate with complex formation, combined with the observation that the WIP–WASp interaction is independent of the WIP phosphorylation state (), implicate WIP in functions other than WASp regulation. To assess WIP functionality in NK cells, RNAi was used to decrease WIP expression in NK cells (). Introduction of siRNA to YTS/KIR2DL1 cells caused a marked reduction of WIP expression and almost complete inhibition of NK cell cytotoxicity (). Comparatively, RNAi in YTS/KIR2DL1 cells resulted in only a moderate decrease in cytotoxicity. Importantly, RNAi eliminated the formation of the multiprotein complex. Gel filtration analysis of WASp, actin, and myosin IIA from RNAi-treated cells revealed that these proteins were eluted with the same partition coefficients in both resting and activated cells, indicating that the complex was not assembled in these cells (). Actin rearrangements are vital for cell conjugation and immune synapse formation in NK and T cells. Formation of filamentous actin rings at the immune synapse allows for firm coupling of the lymphocyte with its target cell and ensures localized delivery of lytic granules (; ). Cytoskeleton reorganization is required for sustained signaling in T cells (), and inhibition of actin polymerization by cytochalasin D prevents NK cell cytotoxic activity (). Thus, affecting the actin cytoskeleton itself or key regulatory proteins involved in its polymerization may present a rapid mechanism for inhibiting cytotoxic activity. We show that activation of NK cells by their target cells results in formation of a large (∼1.3 mD) multiprotein complex comprised of WIP, WASp, actin, and myosin IIA. These proteins polarize to the cell–cell contact site after NK cell activation ( and Fig. S1). Assembly and localization of this complex is affected by KIR inhibitory signaling (, A and C; and ), indicating the functional significance of the complex for NK cell activity. The essential role of WIP in the formation of this complex and in cytotoxicity may be dependent on its phosphorylation by PKCθ (– ). WIP, which was originally identified as a protein interacting with WASp (), was later shown to bind both monomeric and filamentous actin, to retard WASp-mediated actin polymerization in vitro (), and to have a variety of functions, many related to cell signaling and motility (, ; ; ; ). WIP immunoprecipitates from activated YTS cells contain WASp, actin, and myosin IIA (, A and C; and ). The presence of actin and WASp in WIP preparations is in agreement with previous studies (; ). Actin association with WIP is independent of WASp, as indicated by the fact that a WIP mutant unable to interact with WASp interacts with actin (). However, this association is dependent on WIP, as the disruption of WIP expression inhibits multiprotein complex formation (). Actin binds to the NH-terminal part of WIP (), whereas WASp interacts with the WIP COOH-terminal sequences (; ), explaining the observed result. However, the decreased amount of actin in WIPΔ460–503 immunoprecipitates suggests that WASp may contribute to actin recruitment, likely by binding additional actin monomers (). The presence of a class II myosin as the part of the multiprotein complex ( and ) has not been described previously. A WIP homologue from yeast, verprolin, has been demonstrated to bind class I myosin (; ). Although mammalian class I myosin function is still poorly understood, recent studies provide support for the role of another member of the myosin family, myosin IIA, in leukocyte activity (; ; ; ). WIP, as well as WIPΔ460–503, immunoprecipitates from YTS cells contain myosin IIA, indicating an interaction between WIP and myosin IIA. This interaction could be direct, through actin, or through some other protein. WIP and WASp from nonstimulated cells segregate as an ∼350–400 kD protein complex. Purified WASp and its homologue N-WASp were previously shown to behave as dimers or multimers (; ). WASp dimer, together with bound WIP, would create a heterotetramer structure with a theoretical mass of ∼250–260 kD. Additionally, WASp is known to bind a plethora of adaptor and regulatory proteins (), which could explain why WIP and WASp are parts of an ∼400 kD complex. Importantly, neither myosin IIA nor actin is observed in the fractions containing WIP and WASp, demonstrating that the complex is not formed in nonactivated NK cells (). Activation of YTS cells results in the shift of a substantial portion of WIP and WASp to the high molecular mass fraction, accompanied by the occurrence of myosin IIA and actin in the same fraction (), indicating that the multiprotein complex is formed in response to NK cell stimulation. The main protein contributing to the mass of ∼1.3 mD is myosin II because class II myosins are composed of two heavy (171–244 kD) and four light (16–25 kD) chains, forming a heterohexamer (). Activation of YTS cells by target cells causes association of actin and myosin IIA with WIP and WASp. The complex is formed after 3 min of activation, with the maximum actin and myosin IIA recruitment at 10 min (). The short time that the myosin IIA regulatory chain is present suggests that myosin IIA is fully functional only for the short period required to fulfill its task. However, the prolonged association of myosin IIA heavy chain with the complex () may favor a quick response to new stimuli, as short-term interaction would generate a delay required for full myosin IIA heterohexamer assembly. Because the role of myosin IIA in the complex is not clear yet, one speculation would be that myosin IIA moves along the actin cytoskeleton and transports WIP and WASp as “cargo” to places of dynamic actin assembly, i.e., the immune synapse. F-actin is a part of the complex (). In cytolytic conjugates, WIP and WASp polarize to the cell–cell contact site where myosin IIA and F-actin also accumulate ( and Fig. S1). After translocation to the cell–cell interface, WASp could stimulate the Arp2/3 complex to create new actin branches (), whereas WIP could stabilize newly formed actin filaments () to allow more firm cell–cell coupling and provide a rigid scaffold for the immune synapse. Signaling-deficient KIR2DL1*250 provides support for a specific role of KIR2DL1 in the inhibition of multiprotein complex formation. Contrary to wild-type KIR2DL1, recognition of HLA-Cw6 on the surface of target cell by KIR2DL1*250 has no effect on complex formation, indicating that the complex is formed despite the recognition of MHC class I ligand (). Thus, KIR2DL1-mediated signaling can affect proteins involved in actin cytoskeleton reorganization. Inhibitory signaling may affect the complex directly by dephosphorylating one or more of its components, or indirectly by targeting regulatory molecules (e.g., adaptors and kinases) upstream of WIP and WASp. Phosphorylation of KIR ITIMs results in the recruitment of SHP-1 and -2 phosphatases (), which may dephosphorylate Vav-1, thus preventing Cdc42 activation. At the least, a “trapping” mutant of SHP-1 binds Vav-1 (). Because Cdc42 is required for WASp activation and myosin IIA activity (), Vav-1 dephosphorylation by SHP-1 could affect complex functionality. The possibility that SHP-1 and -2 dephosphorylate several of the regulatory components, as well as proteins of the multiprotein complex, needs to be carefully examined. A specific inhibitor for PKCθ effectively prevents WIP phosphorylation, showing that WIP is phosphorylated by PKCθ in NK cells (). This result, directly demonstrating WIP phosphorylation in vivo, supports previous indications of PKCθ involvement in WIP phosphorylation in T cells using WIP-specific peptide antibodies (). Interestingly, a low level of WIP phosphorylation was observed before NK cell stimulation (). WIP could be constitutively phosphorylated at one site and undergo regulated phosphorylation at another site in response to cell stimulation. Ser488 of WIP was identified as a WIP phosphorylation site using mutagenesis (), and this residue may be a site of regulated WIP phosphorylation. Further direct identification of WIP phosphorylation sites is an important next step. Interestingly, no correlation between WIP phosphorylation level and WASp association is observed. WASp is associated with WIP even in resting cells ( and ; unpublished data), suggesting constitutive interaction between WIP and WASp (in agreement with ). WIP phosphorylation has been proposed to dissociate the WIP–WASp interaction in T cells (). However, an increase of WIP phosphorylation in response to NK cell activation is not accompanied by a change in the amount of WASp bound to WIP. Inhibition of WIP phosphorylation also does not affect the WIP–WASp interaction (). Although WIP phosphorylation does not regulate interaction of WIP and WASp in NK cells, the inhibition of PKCθ activity, but not Src family kinases, and the consequent abrogation of WIP phosphorylation, prevents the recruitment of actin and myosin IIA to the complex. Additionally, inhibition of myosin II activity by blebbistatin and ML-7 does not influence interaction between any of the four proteins, indicating that myosin IIA activity is not required for the assembly (). Ligation of KIR2DL1 by HLA-Cw6 results in the suppression of WIP phosphorylation and correlates with prevention of a multiprotein complex formation ( and ). Disruption of WIP expression by RNAi leads to a nearly complete loss of cytotoxicity, supporting the unique role of WIP in NK cell activity. In contrast, RNAi results in incomplete loss of cytotoxicity (), similar to that observed in WAS patients (). Because actin cytoskeleton rearrangements are essential for lymphocyte activation (), the observed defects in NK cell cytotoxic activity are, at least in part, likely caused by faulty actin polymerization. Alternatively, other proteins may be able to substitute for WASp function in activation of the Arp2/3 complex (; ). Depletion of WIP causes more profound effects. WIP regulates WASp activity (), is required for cortical actin network organization (), is essential for multiprotein complex formation, and appears to play a central role in NK cell cytotoxicity. Polyclonal anti-KIR2DL1 antibodies were generated using full-length KIR2DL1 protein. Other antibodies used include the following: anti-KIR mAb (EB6; Beckman Coulter), anti-WASp (US Biological), anti-WIP (a gift from R. Geha and N. Ramesh, Children's Hospital Boston, Boston, MA), anti-ERK1/2 (Upstate Biotechnology) and MOPC-21 (isotype control), anti-phosphotyrosine (PT66), anti–myosin IIA heavy chain, anti-FLAG (M2), anti-myosin regulatory light chain (MY-21), anti-actin, anti–rabbit IgG (RG-96), and anti–mouse IgG (all from Sigma-Aldrich). Alexa Fluor–conjugated reagents were obtained from Invitrogen. The FLAG-pMX vector was generated using pMX-IRES-GFP plasmid (a gift from T. Kitamura, University of Tokyo, Tokyo, Japan). Two DNA oligos encoding the FLAG peptide sequence were inserted into pMX-IRES-GFP vector to generate FLAG-pMX-IRES-GFP plasmid. The full-length WIP and WIPΔ460–503 cDNA were amplified from pcDNA3-WIP vector (a gift from N. Ramesh). Both, WIP and WIPΔ460–503 constructs were subsequently inserted into the FLAG-pMX-IRES-GFP vector. All constructs were verified by DNA sequencing analysis. Vector-based RNAi, targeting the GGAGGTTTCCTGTGCCTTCT sequence of WIP and the GGGAACAGGAGCTGTACTCAC sequence of WASp, was used to generate WIP and WASp knockdown cell lines. Two DNA oligos designed to produce WIP or WASp siRNA hairpins were inserted into pBS-U6-RNAi-dual-CMV-GFP vector (a gift from X. Liu, Harvard University, Cambridge, MA; ; ). The construct was verified by DNA sequencing and introduced into YTS/KIR2DL1 cells using Nucleofector I (Amaxa Biosystems). The GFP-positive cells were sorted using a MoFlo high performance cell sorter (DakoCytomation). WIP or WASp expression levels in each stable cell clone were examined by Western Blot. The clones with a substantial decrease of WIP or WASp expression levels were selected and sorted. At least six clones were analyzed and one representative clone was used for studies of WIP or WASp knockdown effects. Cells were maintained in RPMI 1640 medium supplemented with 10% FCS, -glutamine, and 1.6 mg/ml genectin (Invitrogen) or 2 μg/ml puromycin (Sigma-Aldrich). FLAG-WIP and FLAG-WIPΔ460–503 constructs were transfected into the retrovirus-packaging Plat-E 293 Eco cell line (). Retroviruses were used to infect tumor NK YTS/KIR2DL1 or YTS/KIR2DL1*250 cell lines, as previously described (). GFP-positive transfected cells were sorted using a MoFlo high performance cell sorter. For antibody-mediated KIR2DL1 cross-linking, 4 × 10 YTS/KIR2DL1 cells were incubated with EB6-coated protein G magnetic beads (2.8 μm; Dynal) for 3, 10, or 30 min at 37°C in PBS supplemented with 2% FCS. For negative controls, 4 × 10 YTS/KIR2DL1 cells were incubated with MOPC21-coated protein G magnetic beads and 4 × 10 YTS or YTS/KIR2DL1*250 cells were incubated with EB6-coated protein G magnetic beads for 3 min at 37°C in PBS/2% FCS. After stimulation, cells were mixed with an equal volume of ice-cold 2× lysis buffer and lysed for 30 min at 4°C. Magnetic beads were isolated from the total cell lysates using a magnet (Dynal) and washed extensively with ice-cold lysis buffer (1% NP-40, 50 mM TRIS/HCl, pH 7.4, 150 mM NaCl, and 1 mM NaVO). For cell mixing, 2 × 10 cells, either YTS/KIR2DL1/FLAG-WIP, YTS/KIR2DL1-FLAG-WIPΔ460–503, or YTS/KIR2DL1*250/FLAG-WIP, were mixed with 2 × 10 721.221 or 721.221/Cw6 cells in complete RPMI medium. Mixed cells were immediately transferred to a 37°C water bath and incubated for 3, 10, 30, or 60 min, followed by brief centrifugation and lysis in ice-cold lysis buffer for 30 min at 4°C. For time 0, cells were mixed and immediately lysed. Cleared cell lysates were incubated with anti-FLAG mAb coupled to protein G Sepharose beads (Sigma-Aldrich). Sepharose beads were pelleted by centrifugation (5,000 RPM for 2 min at 4°C) and washed extensively with ice-cold lysis buffer. 2 × 10 YTS/KIR2DL1/FLAG-WIP cells were labeled for 4 h at 37°C with [P]orthophosphate (0.2 mCi/ml, 285.6 Ci/mg; Perkin Elmer) in phosphate-free RPMI 1640 media (CHEMICON International, Inc.) supplemented with 10% phosphate-free FCS. Labeled cells were mixed with either 721.221 or 721.221/Cw6 target cells and incubated for 3, 10, 30, or 60 min at 37°C, lysed, and immunoprecipitated with anti-FLAG antibody. Radiolabeled proteins were visualized by autoradiography. Incorporated radioactivity was quantified using SigmaGel software v1.0 (Jandel Scientific), normalized to phosphorylation, observed at time 0, and expressed as arbitrary units. In experiments involving kinase inhibitors, the following inhibitors were used: 100 nM bisindolylmaleimide I, 100 μM PKCα,β pseudosubstrate inhibitor, 100 μM PKCθ pseudosubstrate inhibitor, 2 μM casein kinase II inhibitor, 1 μM PP2, 100 μM blebbistatin, and 5 μM ML-7 (all from Calbiochem). Labeled cells were treated with kinase inhibitors or 0.1% DMSO (control) for 30 min at 37°C, followed by mixing with 721.221 target cells for 10 min at 37°C, cell lysis, and immunoprecipitation with anti-FLAG antibody. Immunoprecipitated proteins were visualized by autoradiography and subsequently blotted with the indicated antibodies. YTS/KIR2DL1 cell cytotoxicity was evaluated by Cr release assay as previously described (). Gel filtration was performed on a Superose 6 column (106.5 ml; GE Healthcare) using a BioLogic Workstation (Bio-Rad Laboratories). Cleared cell lysates from 2 × 10 YTS/KIR2DL1/FLAG-WIP or YTS/KIR2DL1/dsRNA cells, either resting or stimulated with 721.221 (1:1 ratio), were applied to the Superose 6 column. Samples were eluted at 4°C with TBS and 1.25-ml fractions were collected. Proteins from each fraction were precipitated with 10% TCA and the neighboring fractions (e.g., 1 + 2, 3 + 4, etc.) were pooled to reduce the number of samples for analysis. Precipitated proteins, which were transferred to PVDF membrane, were immunoblotted with anti–myosin IIA, anti-FLAG, anti-WASp, or anti-actin antibodies. Calibration curves were prepared by measuring the partition coefficients (K) of molecular mass markers (GE Healthcare) and plotting them against log molecular mass. K values for proteins of interest were calculated based on elution volumes and were fitted to the calibration curve. All partition coefficients were calculated using the following equation:where = elution volume for the protein, = void volume of column, and = total volume of column. YTS/KIR2DL1/FLAG-WIP cells were stimulated, lysed, and cleared, and cell lysates were immunoprecipitated with anti-FLAG mAbs and resolved on a 4–12% NuPage gel. Gels were stained with a Colloidal blue staining kit (Invitrogen), and individual protein bands were subsequently excised and digested with 12.5 ng/μl trypsin (Promega). Resultant protein samples were analyzed using a Finnigan LTQ linear ion trap mass spectrometer (Thermo Electron Corp). Ion trap spectra were obtained and MS data was analyzed using the Sequest Cluster with BioWorks software, which was supplied by the manufacturer. Certain peptides were further positively identified by MS peptide sequencing, which was facilitated through the use of the ProFound peptide mapping software (The Rockefeller University). MS experiments and analysis were performed in the proteomics facility of the Children's Hospital of Philadelphia. YTS/KIR2DL1/FLAG-WIP cells were conjugated to either 721.221 or 721.221/Cw6 cells at a 1:1 ratio for 10 min at 37°C in complete RPMI medium, followed by adherence to poly--lysine–coated slides (Sigma-Aldrich) for 15 min at 37°C. Next, the adherent cells were washed with PBS, fixed, and permeabilized with Cytofix/Cytoperm buffer (BD Biosciences) supplemented with 0.1% Triton X-100. After blocking with 1% BSA, cells were stained with anti-WASp mAbs followed by Alexa Fluor 647–conjugated goat anti–mouse, anti–myosin IIA followed by Alexa Fluor 405–conjugated goat anti–rabbit, and Cy3-cojnugated anti-FLAG mAb. Antibodies were used in the range of 1–20 μg/ml. F-actin was stained using Alexa Fluor 647–conjugated phalloidin. Cell conjugates were visualized by a laser-scanning confocal microscope (LSM510 Axiovert 100M; Carl Zeiss MicroImaging, Inc.). NK cells were identified by GFP fluorescence and positive anti-FLAG staining. The percentage values were determined by evaluation of 150–200 conjugates in randomly selected fields, in 3–4 separate experiments. The images were obtained using 63× Plan-Apochromat objective and LSM510 software v. 3.2 (both Carl Zeiss MicroImaging, Inc.). Because of extensive photobleaching, the γ value of blue channel was increased from 1.0 to 1.4. Fig. S1 shows the localization of F-actin and FLAG-WIP in YTS/KIR2DL1/FLAG-WIP cells after the formation of cytolytic and noncytolytic conjugates with target cells. Online supplemental material is available at .

Cytotoxic T lymphocytes (CTLs) form a major part of the body's defense against viral infection and cancer by seeking out and killing infected or tumorigenic cells. CTLs kill their targets either by ligation of death receptors on the target cell or by granule exocytosis (for review see ). During granule-induced cell death, perforin and granule-specific serine proteases (granzymes) are delivered to the target cell, where they induce cell death by pathways that have yet to be completely defined (; ). Perforin is a key molecule in this process, as CTLs from perforin mice are inefficient in killing target cells in vitro, and perforin mice have a decreased ability to clear virus and diminished tumor surveillance manifested as increased susceptibility to spontaneous B cell lymphoma (; ; ; ; ). The primary role attributed to perforin is to ensure the correct trafficking of granzymes into the target cell (; ); however, perforin has also been shown to induce direct lysis of the target cell (; ; ; ; ; ). Granzyme A has trypsinlike proteolytic specificity that induces cell death by cleaving a variety of substrates, including lamins, DNase within the SET complex, and PHAPII (for review see ). Granzyme B induces apoptotic cell death by cleaving the proapoptotic Bcl-2 family member Bid (; ; ) and the regulatory prodomain of caspase-3 (; ). Although somewhat less efficient than CTLs isolated from wild-type mice, CTLs from mice deficient in granzymes A and B (granzyme AB) maintain their ability to kill many target cells by granule exocytosis (; ; ; ), and, in contrast to perforin mice, granzyme AB mice develop normally, do not develop spontaneous malignancy, and clear many viruses normally. Granzymes C, K, and M have also been shown to induce cell death when delivered by perforin in vitro (; ; ). Therefore, death induced by CTLs from granzyme AB mice may be caused by perforin-mediated delivery of other death-inducing granzymes. CTLs of mice deficient in granzyme B have been reported to have reduced expression of granzyme C as a result of inadvertent silencing of its gene (), and granzyme M is preferentially expressed by NK cells (). Consistent with these observations, it has understandably been proposed that cell death induced by granzyme AB CTLs occurs through direct perforin lysis (). The question of how CTLs kill their target cells in the absence of granzymes A and B is likely to be physiologically important in light of recent findings that individual activated lymphocytes within a population may express different combinations of perforin and granzymes or even no granzymes at all (). Indeed, in this study, up to 10% of perforin-positive lymphocytes did not express either granzyme A or B at day 3 after IL-2 activation. In addition, it has been proposed that the mechanism of target cell death (apoptosis or lysis) can subsequently influence the quality and quantity of the inflammatory response by modulating antigen processing and presentation (for review see ). The mechanism of death induced by granzyme AB CTLs has been studied at a population level (, ); however, using these assays, it has not been possible to distinguish apoptosis from necrosis, lysis, or any alternate death mechanisms. Analysis of morphology remains one of the most decisive ways by which to distinguish the various forms of cell death. Time-lapse microscopy has previously been used to investigate the subcellular organization of polarized CTLs () and to measure the kinetics of cell death induced by cytotoxic drugs (,). Therefore, we have used time-lapse microscopy to study in real time the morphological and molecular parameters of single cells undergoing granule-induced death after attack by alloreactive CTLs. We found that wild-type CTLs induced apoptosis, but the absence of granzyme A and B resulted in a modified form of cell death that was distinct from both apoptosis or perforin-mediated lysis. These data show that death induced by CTLs from granzyme AB mice cannot simply be attributed to perforin lysis; rather, death is likely to be initiated by other granule components. Furthermore, the unexpected resistance of the granzyme AB animals to a variety of viral pathogens cannot simply be attributed to perforin. To characterize cell death induced by CTLs by time-lapse microscopy, it was first necessary to characterize cell death induced by purified perforin alone or perforin together with granzyme B. During apoptosis induced by cytotoxic drugs, cells initially undergo a period of intense ruffling of the plasma membrane known as blebbing, and phosphatidylserine (PS) becomes externalized from the inner leaflet of the plasma membrane to the outer leaflet (). Such a cell would normally be rapidly phagocytosed in vivo; however, under typical cell culture conditions, the integrity of the plasma membrane is eventually lost, resulting in the rupture of the cell. PS externalization can be detected by binding of annexin V, and the loss of plasma membrane integrity can later be detected by the passive uptake of propidium iodide (PI; ,). This process is in contrast to necrosis, which involves rapid cell swelling and rupture of the plasma membrane in the absence of rounding and blebbing and results in simultaneous binding of annexin V and PI uptake. Cells undergoing early apoptosis stain with annexin V but exclude PI (annexin VPI), whereas necrotic cells or cells that have undergone late apoptotic changes stain with annexin V and PI. Granzyme B has been shown to induce apoptosis in a perforin-dependent manner (), and purified perforin is believed to induce cell lysis (; ). However, little is known about the kinetics and morphological features of death induced by these proteins. To investigate the kinetics of granzyme B–induced cell death, we used real-time imaging to follow annexin V binding and PI uptake in MS9II cells treated with 10 nM perforin (a concentration that does not cause cell death when applied to cells alone) and 25 nM granzyme B. Under these conditions, the majority of cells (∼70%) died within 4 h, and all showed typical signs of apoptosis (example provided in ). Individual cells showed the initial morphological features of apoptosis (rounding/blebbing) in an asynchronous manner between 0.5 and 2 h and invariably followed a distinct sequence of events: rounding and blebbing (seen at 50 min in , c); PS exposure as detected by annexin V binding (at 1 h 20 min; , d); and plasma membrane rupture as detected by PI uptake (2 h 20 min; , f). By calculating the relative intensity of each fluorochrome as a function of time, we determined when individual cells became positive for annexin V staining and permeable to PI (). In the example shown, the individual frames shown in correspond to the arrows (a–g) shown in . By studying multiple target cells in the population, we determined that the mean time from cell rounding/blebbing until it became annexin V positive was ∼20 min, and the mean duration from annexin V binding to the loss of plasma membrane integrity (PI staining) was ∼70 min (). Therefore, the mean total time for cell death, defined as commencing with the onset of morphological changes (rounding/blebbing) and ending with PI staining, was ∼90 min. Inevitably, this estimate did not include the time taken to achieve Bid cleavage, cytochrome c release, or caspase activation, as these parameters are known to occur extremely rapidly and precede rounding (; ). We also used the same methodology to investigate the characteristics of perforin-induced cell lysis by assaying annexin V binding and PI uptake in cells treated with sufficient purified perforin to induce direct lysis (80 nM) in the absence of granzyme B (). After perforin addition, the plasma membrane developed protrusions that rapidly expanded, and the cells simultaneously became permeable to PI and stained positive with annexin V (). Staining with each fluorochrome typically commenced 2–3 min after perforin addition and reached maximal levels by 10 min (). Interestingly, when the concentration of perforin was progressively reduced so that only 10–50% of the cells died, the changes in cell morphology and staining did not vary significantly (not depicted). In particular, the apoptotic changes seen in the presence of granzyme B were never observed; independent of the dosage used, purified perforin induced only cell lysis. Using the aforementioned findings, we next turned to the question of how intact CTLs induce target cell death in the presence or absence of granzymes A and the B cluster (granzyme AB). We raised alloreactive CD8 mouse CTLs (H-2 anti–H-2) that killed MS9II fibroblasts (H-2) exclusively in a perforin-dependent manner (). CTLs from wild-type mice efficiently killed MS9II in a 4-h assay as indicated by release of preloaded Cr (). The CTLs of granzyme AB mice were somewhat less potent than wild-type CTLs over the 4-h time frame (), but these CTLs maintained significant killing ability. CTLs derived from perforin mice did not induce significant cell death (), and cytotoxicity was completely blocked by chelating calcium with EGTA (not depicted), which is consistent with a perforin-dependent cell death mechanism. Furthermore, the CTLs of gld mice that lack functional FasL were as efficient as wild type (not depicted), confirming that cell death in this assay was completely dependent on the granule pathway. Using RT-PCR, we showed that CTLs from granzyme AB mice expressed comparable levels of perforin and granzyme K mRNA to CTLs isolated from wild-type mice but, as expected, were completely deficient in message for both granzymes A and B (). Consistent with a previous study (), CTLs from granzyme AB mice expressed mRNA for granzyme C; however, this was ∼10-fold less than that of wild-type mice. Therefore, the CTLs of granzyme AB mice were capable of synthesizing both granzyme C and K as well as perforin. To characterize the morphology of death, we coincubated CTLs from wild-type mice with MS9II cells at effector/target ratios (2:1) that easily permitted us to identify CTLs (motile and small) from MS9II cells (large, flat, and adherent). We then followed the death by time-lapse microscopy (example presented in ). We found that MS9II cells underwent changes that were virtually identical to those of cells killed by purified perforin and granzyme B. After conjugation, the target cells rapidly rounded and detached from the culture dish. The plasma membrane then became ruffled (blebbed), and, after a further period of time, the cell began to swell. A clear increase in annexin V binding was observed during the blebbing phase of death, but PI uptake was only observed when the cell had become swollen. The sequence of events (rounding, blebbing, annexin V binding, and PI uptake) was invariant for all of the cells examined. Using the same methodology described in Time-lapse microscopy to determine the relative fluorescence intensity of annexin V and PI, we could determine the specific timeline of events for MS9II cells killed by wild-type CTLs. For example, in the cell depicted in , conjugation of the CTL was observed at 20 min, cell rounding was observed by 30 min, and annexin V binding was detectable by ∼1 h. The cell remained round and the plasma membrane remained intact until 1 h 45 min, after which the cell swelled and its outline became less distinct. PI uptake was detected at 1 h 50 min, which is coincident with cell swelling (). Pooled analysis of single cells ( = 33) revealed that the interval between rounding and annexin V binding was ∼30 min, whereas the duration between annexin V binding and PI uptake was ∼55 min (). Thus, the overall time of death (rounding to PI) was ∼90 min. Similar to death induced by wild-type effector cells, CTLs from granzyme AB mice induced rapid rounding and blebbing of the MS9II target cells followed eventually by plasma membrane swelling and rupture (example presented in ). In contrast to death induced by wild-type CTLs, the onset of annexin V staining and PI uptake were virtually simultaneous during death induced by granzyme AB CTLs (). In the example presented in , interaction between the CTL and target cell was first observed at 1 h 42 min; however, annexin V binding and PI uptake were not observed until 2 h 48 min (). When several target cells ( = 19) were analyzed in the same way, the overall time from cell rounding to annexin V binding was ∼90 min, but, in contrast to wild-type CTLs, the duration from annexin V binding to PI uptake was <5 min. Using cell rounding as a common reference point, it was possible to compare the time course of CTL-induced cell death with apoptosis induced by granzyme B that was delivered by sublytic concentrations of purified perforin (). The execution phase (duration from rounding to PI uptake) of CTL-induced death (87.9 ± 3.6 min for wild-type CTLs and 87.4 ± 3.2 min for granzyme AB CTLs) was very similar to that of purified perforin/granzyme B (94.6 ± 3.2 min). Therefore, in our model system, we found that regardless of whether death was induced by purified granzyme B/perforin, wild-type CTLs, or granzyme-deficient CTLs, the duration of death was ∼1.5 h in each case. During death induced by wild-type CTLs or granzyme B/perforin, the target cells became annexin V positive ∼30 min after rounding, and the cells remained annexin V positive but excluded PI for the following ∼60 min. During death induced by granzyme AB CTLs, annexin V binding was not evident until ∼90 min and was almost concomitant with PI uptake. The major discrepancy between wild-type CTLs and granzyme AB CTLs was that annexin positivity was markedly delayed during death induced by granzyme AB CTLs so that the time during which the cells were annexin VPI was reduced virtually to zero. Because not all CTLs that came in contact with targets killed that target and some targets were contacted by numerous CTLs (Fig. S2, available at ), it was not possible to get an accurate measure of the initiation phase of apoptosis (i.e., duration from conjugation to rounding). Although overall the morphology of cell death in response to granzyme AB CTLs was strikingly similar to apoptosis, simultaneous staining of annexin V and PI was reminiscent of perforin lysis. Therefore, we considered that individual target cells might be attacked, in turn, by several CTLs (Fig. S2) delivering consecutive sublytic doses of perforin so that their cumulative effects would eventually kill the cell. To model this, we added multiple sublytic doses (5 nM) of perforin at 20-min intervals (). We found that at this concentration, a single addition of perforin induced death of <5% of cells after 4 h (). However, the same concentration plus granzyme B induced apoptosis in >80% of cells (). Two or three sequential doses of perforin every 20 min did not result in an increase in cell death, but four sequential doses induced death in ∼30% of cells. As three sequential doses of perforin did not induce more death than one single dose, the death observed after four doses represented a cumulative effect similar to what may occur if perforin was delivered by several CTLs. In contrast to death induced by granzyme AB CTLs, the sequential addition of sublytic perforin did not result in rounding and blebbing (). Rather, this death involved formation of the very large membrane protrusions similar to that induced by single lytic doses of perforin (). These data indicated that death induced by granzyme AB CTLs does not reflect multiple hits of sublytic perforin and more likely reflects the action of granule proteins delivered by perforin. Much of our current knowledge of granule-mediated cell death by intact CTLs is derived from population-based studies that cannot easily distinguish between various death mechanisms (; ; ). Therefore, relatively little is known about the nature of cell death induced by intact CTLs at a single cell level. Furthermore, the ability of CTLs from mice deficient in key granzymes A and B to kill target cells through the granule pathway has been documented in several experimental systems (; ; , ; ). However, it has not been possible to determine how the target cells were killed. It has recently been shown that individual CTLs in wild-type mice may contain perforin but neither granzyme A or B (). The killing observed by the granzyme AB CTLs may, therefore, be caused by a specific subset of cells that are commonly present in populations of CTLs isolated from wild-type mice. As the nature of death (apoptosis, necrosis, lysis, or another alternate programmed cell death) can influence how the dying cell is presented to the immune system (), the mode of death induced by granzyme AB CTLs is likely to play a bona fide role in the immune response of wild-type animals to pathogens or tumor cells. It has been proposed that the default pathway adopted by CTLs in the absence of most of the proapoptotic granzymes is perforin-induced lysis (). In contrast, granzymes C, K, and M have been shown to induce the death of target cells when delivered by perforin. Therefore, death induced in the absence of granzymes A and B may be caused by various other death-inducing granzymes. We found that death induced by lytic concentrations of perforin was virtually immediate (plasma membrane rupture within 2–3 min and complete by ∼10 min) and involved the progressive swelling of plasma membrane protrusions followed within a few minutes by the loss of plasma membrane integrity. Cell rounding and detachment from the culture flask were notably absent. This is consistent with a recent study that also showed that perforin induced large “airbaglike” protrusions of the plasma membrane followed by plasma membrane rupture (). These protrusions were reported to be reversible when perforin was delivered at sublytic concentrations, probably as a result of plasma membrane repair by lysosomes (). In our model, we never observed large plasma membrane protrusions when we used sublytic concentrations of perforin alone or in combination with granzyme B. This may reflect cell type differences or may suggest that the formation of plasma membrane protrusions is not required for granzyme uptake. In contrast to perforin lysis, MS9II cells killed by granzyme B added with sublytic concentrations of perforin or by wild-type or granzyme AB CTLs underwent more gradual death (taking ∼90 min), which was characterized by rapid rounding followed by a discrete period of plasma membrane blebbing before the eventual loss of plasma membrane integrity. Our thorough kinetic analysis of cell death at a single cell level has revealed that the morphology of death induced by either wild-type CTLs or CTLs from granzyme AB mice was quite clearly distinct from cell lysis induced by purified perforin. The rapidity of cell death induced by CTLs or perforin/granzyme B probably reflects the essential need to kill virus-infected cells as efficiently as possible to limit viral replication and spread. But strikingly, although the duration of target cell death induced by intact CTLs was ∼1.5 h in the presence or absence of granzyme A and the B cluster, the timing of specific molecular events differed somewhat. During apoptosis induced by granzyme B/perforin or the CTLs of wild-type mice, PS exposure occurred shortly after the target cell rounded up, and the cell maintained plasma membrane integrity over the following 45–60 min. In contrast, PS exposure was not evident shortly after rounding during death induced by granzyme AB CTLs, and binding of annexin V to PS was only evident virtually at the same time as it lost plasma membrane integrity. Under our experimental conditions, it is not clear whether PS ever becomes exposed on the outer leaflet of the plasma membrane or whether the dying cell becomes permeable to annexin V. Regardless of the nature of annexin V binding, these data indicate that granzyme B and wild-type CTLs induce classic apoptosis, whereas granzyme AB mice induce a novel form of programmed cell death that shares similar features with but is distinct from apoptosis. These findings are consistent with our previous study that showed that caspase-dependent DNA fragmentation, a classic hallmark of apoptotic cell death, was detected in cells killed by wild-type CTLs but was not detected in cells killed by granzyme AB CTLs (). The physiological consequences of killing a target cell by apoptosis or by nonapoptotic cell death are unclear. It has been proposed that cells undergoing apoptosis are packaged and cleared by phagocytosis before plasma membrane rupture, avoiding the leakage of potentially toxic cytoplasmic material into the cellular milieu and reducing bystander cell damage. It has also been suggested that apoptotic cells may be presented to antigen-presenting cells in a different way than necrotic cells and that this may modulate the resultant immune response (for review see ). PS exposure is an important event in the phagocytosis of apoptotic cells, as macrophages deficient in PS receptor bind but do not ingest dying cells. As cells killed by granzyme AB CTLs do not expose PS before plasma membrane rupture, it is possible that these cells would not be efficiently engulfed by macrophages and, thus, may be perceived by the immune system as having undergone necrosis. If the target cell under attack contains potentially viable virions, the outcome for the host might differ significantly. First, if viral particles were released extracellularly, the opportunity for spread to neighboring cells or for systemic infection is heightened. Second, the efficiency of viral uptake into antigen-presenting cells might well influence subsequent antigen presentation, chemotaxis, and the quality of the local inflammatory reaction. On the basis of our findings, we can speculate that one reason a virus might express a specific inhibitor of intrinsic apoptosis (e.g., IAPs) or granzyme B (e.g., the L4 100-kD protein of adenovirus) might be to influence the mechanism of cell death, thereby reducing the likelihood of phagocytosis and encouraging viral spread rather than significantly prolonging the life of the infected cell. It must be noted that it has not been determined whether cells killed by granzyme AB CTLs can be recognized and cleared by phagocytic cells through a PS-independent mechanism. It is intriguing to speculate on the molecular mechanisms responsible for the cell death induced by granzyme AB CTLs. Several granzymes have been identified in mice in addition to granzymes A and B, including C, D, E, F, G, J, K, M, and N, of which granzyme C, K, and M have been shown to induce cell death when delivered by perforin (; ; ; ). Granzyme M expression is restricted to natural killer cells (; ). As our data is generated from CD8 T cells, granzyme M is unlikely to be responsible for the death in our model. Granzyme AB mice were generated by crossing granzyme A mice with granzyme B mice (). As disrupting the granzyme B gene also resulted in the reduction in expression of multiple genes, including granzymes C, D, E, F, and G (), we considered the possibility that granzyme C might be excluded as a cause of cell death. However, we found significant levels of granzyme C mRNA expressed in our alloreactive CTLs (10,000 times higher than that detected in naive T cells from wild-type mice). So, although granzyme C expression may be reduced in granzyme AB mice, it cannot be excluded as causing the death of MS9II target cells in our model. We further showed that message for granzyme K is expressed at comparable levels in CTLs from granzyme AB and wild-type mice. Therefore, the death induced by CTLs from granzyme AB mice may be caused by either granzyme C, K, or other granule proteins for which a death-inducing ability has not been described. The distinct specificity, alternate regulation, cell-specific expression profile, and species-to-species variation of granzymes suggest that each of the granzymes may play a very specific role in an immune response while collaborating to thwart most types of viral infection or cells undergoing oncogenic transformation. Understanding the mechanism by which granzyme AB CTLs kill their targets will be important in developing our knowledge of how CTL-induced target cell killing helps prevent infection and may unmask the key granule constituents involved in maintaining the relatively healthy immune status of granzyme AB mice compared with perforin-deficient mice. Cr (as sodium dichromate) was purchased from GE Healthcare. Annexin V–FLUOS was obtained from Roche. Cell culture reagents were purchased from Invitrogen, and PI and all other chemicals were obtained from Sigma-Aldrich. Mouse perforin expressed in baculovirus-infected insect cells was purified essentially according to . Human granzyme B was expressed and purified from as described previously (). The adherent mouse fibroblast cell line MS9II (derived from the C3H strain H-2) was obtained from W. Sly (Washington University, St. Louis, MO) and cultured at 37°C in a humidified CO incubator in RPMI medium supplemented with 2 mM glutamine, 10% FBS, and 1 mM sodium pyruvate. C3H/J (H-2) mice and C57BL/6 (H-2) mice were maintained in specific pathogen-free conditions at The Walter and Eliza Hall Institute (Melbourne, Australia). Mice deficient in granzyme A and the granzyme B cluster (granzyme AB; ; ; ; ) were obtained from M. Simon (Max-Planck-Institut für Immunbiologie, Freiburg, Germany; ) and maintained at the Peter MacCallum Cancer Centre (Melbourne, Australia). The mice were generated by backcrossing granzyme B cluster–deficient mice (generated on strain 129 embryonic stem cells; H-2; ) for six generations with C57BL/6 mice and for a further two generations with granzyme A–deficient mice (generated from C57BL/6 embryonic stem cells; ). Thus, the granzyme AB mice were backcrossed a total of eight generations to C57BL/6. Allogeneic CTLs were generated in a one-way mixed lymphocyte reaction in which splenocytes isolated from C57BL/6 or granzyme AB mice were cocultured for 7 d in RPMI containing 10% FBS, 2 mM glutamine, 50 μM β-mercaptoethanol, 100 μM nonessential amino acids, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM sodiun pyruvate, and 50 U/ml rIL-2 with a similar number of lethally irradiated stimulator splenocytes from C3H/J mice (H-2). In some experiments, the activated lymphocytes were restimulated with C3H/J splenocytes and cultured for a further 3–4 d. Phenotypic analysis by FACS demonstrated that the responder cells were >90% CD3CD8 T cells. The activation markers CD25, CD44, and CD69 did not vary significantly between the responder cell populations. 10 target cells were incubated with 75 μCi Cr in 100 μl of culture medium for 1 h at 37°C. Cells were washed three times in culture medium to remove the unincorporated Cr and resuspended at 2 × 10 cells/ml. CD8 CTLs were incubated with the target cells at the ratios indicated, and the supernatant was harvested using a Skatron supernatant collection system (Molecular Devices). Cr released into the supernatant was detected using an automatic γ counter (Wallac Wizard 1470; PerkinElmer). In each case, the spontaneous release of radiolabel over the time of the assay was no higher than 10% of the total incorporated radioactivity. The percentage of Cr release was calculated by the following equation: [(cpm of Cr released from sample − cpm of Cr released from untreated cells/cpm of Cr released from cells treated with 1 M HCl − cpm released from untreated cells) × 100]. MS9II cells were plated in 96-well culture plates and incubated overnight at 37°C in a humidified CO incubator. The plates were transferred to a temperature-controlled stage (Prior Proscan; GT Vision) maintained at 37°C on a microscope (IX-81; Olympus). PI was added to the cultures at 50 ng/ml, and annexin V–FLUOS was added at 2 μg/ml. Cells were exposed to an equal number of activated CD8 CTLs and viewed using a 20× LCplanFL NA 0.40 lens (Olympus) for the times indicated. Images were captured at specified intervals using a CCD camera (model ORCA-ER; Hamamatsu) controlled by MetaMorph software (Universal Imaging Corp.). As the fluorescence intensities of annexin V–FLUOS and PI were significantly different, we used MetaMorph and Excel software (Microsoft) to plot the fluorescence reading for each frame relative to the maximum for that fluorochrome over the course of the experiment after subtraction of the background fluorescence reading from each. The maximum fluorescence plotted for each fluorophore was 1.0, and baseline fluorescence was zero. Splenic lymphocytes from a 7-d mixed lymphocyte reaction (>95% CD8 T cells) were pelleted and resuspended in 1 ml TRIzol reagent (Invitrogen). RNA was extracted with 200 μl chloroform and precipitated with isopropanol. The RNA pellet was washed with 70% ethanol, resuspended in 40 μL RNA Storage Solution buffer (Ambion), and incubated at 60°C for 5 min. cDNA was synthesized using an Omniscript kit (QIAGEN) according to the manufacturer's instructions. In brief, cDNA mix (0.2 U/μL Omniscript reverse transcriptase [QIAGEN], 1× cDNA buffer, 0.5 mM each deoxynucleotide triphosphate, 1 μM oligo-dT [Promega], and 0.5 U/μL RNAsin [Invitrogen]) was added to 100 ng RNA and heated at 37°C for 1 h. cDNA encoding granzymes A, B, C, K, and perforin were quantified using real-time analysis, which was performed using an Assays-on-Demand TaqMan Gene Expression Assay kit (Applied Biosystems), and singleplex PCR was performed using TaqMan (FAM dye labeled) minor groove-binder probes. cDNA encoding the mitochondrial ribosomal protein L32 was quantified as an internal standard to provide a reference point for comparing mRNA levels for granzyme with perforin in naive and activated lymphocytes. Real-time PCR was performed using a thermocycler (ABI PRISM 7700; Applied Biosystems) with denaturation at 95°C for 2 min followed by 40 cycles of denaturation at 95°C for 10 s and synthesis for 30 s at 60°C. Video 1 is a time-lapse video of the death of an MS9II cell induced by 10 nM perforin and 25 nM granzyme B as described in . Video 2 is a time-lapse video of the death of an MS9II cell induced by 80 nM perforin as described in . Video 3 is a time-lapse video of the death of an MS9II cell induced by a wild-type CTL as described in . Video 4 is a time-lapse video of the death of an MS9II cell induced by a granzyme AB CTL as described in . Fig. S1 is a montage of a target cell killed by granzyme AB CTLs as described in , and Fig. S2 is a montage of a target cell killed by granzyme AB CTLs in which the target cell depicted is touched by several lymphocytes before rounding. Online supplemental material is available at .

Localized translation of mRNAs is important for numerous biological processes in eukaryotes, with a particularly significant role in the control of early development (for reviews see ; ). For example, the anterior–posterior and dorsal–ventral axes in are defined by the proper localization and translation of , , and maternal mRNAs to three different positions within the oocyte (for review see ). Similarly, in , the establishment of the germ cell lineage is dependent on the partitioning of maternal mRNAs in P-granules into the posterior cell during the first embryonic cell division (; ). Localized translation also plays a role in somatic cells. For example, the spindle pole is regulated by the translation of a cyclin mRNA localized to this structure (). Similarly, the control of local translation in neurons influences growth cone guidance and synaptic plasticity (for review see ). There are many processes involved in localized translation of mRNAs. mRNAs must accumulate at specific subcellular sites either by active transport or by diffusion and then entrapment. The mRNA as it is being transported, as well as the corresponding unlocalized mRNA, must be translationally repressed. Finally, the translation of the localized mRNA must be activated. These processes are illustrated by the regulation of β-actin mRNA by zipcode binding protein 1 (ZBP1), which binds specific sequences in the 3′ untranslated region (UTR) of β-actin mRNA and is required for its localization to sites of actin polymerization at the cell periphery (). ZBP1 binds β-actin mRNA in the nucleus and represses its translation while the mRNA is transported. Once β-actin mRNA reaches the cell periphery, ZBP1 is phosphorylated by Src, which reduces binding of ZBP1 to the mRNA, thus allowing for translation (). Translation repression complexes associated with mRNAs, including those that are localized, are often found in large aggregates referred to as RNP particles or granules. The aggregation of RNP complexes may facilitate the transport of specific mRNAs to discrete regions of the cell and possibly help to establish or maintain the repressed state. Recent results have identified a conserved set of proteins that are involved in both translation repression and packaging of repressed mRNAs into RNA granules in both somatic cells and early development. A variety of eukaryotes, including yeast, , and mammals, concentrate at least some untranslated mRNAs in cytoplasmic RNP granules referred to as processing bodies (P-bodies) but also called GW- or Dcp-bodies (; ; ; ; ; ). P-bodies can be thought of as a purgatory for mRNA because transcripts within these complexes can either return to translation or be subject to degradation by mRNA decapping and 5′ exonuclease digestion (; ; ). P-bodies may be involved in a wide variety of translation repression events. For example, proteins within P-bodies are required for general translation control in yeast (), and translation repression by microRNAs appears to involve the accumulation of target mRNAs in P-bodies (; ; ). The dual role of P-bodies in translation repression and mRNA degradation leads to these complexes containing both translation repressor proteins and enzymes for mRNA degradation, such as the decapping enzyme. The same proteins that function in translation control and P-body formation in somatic cells appear to underlie the function of maternal mRNA storage granules. For example, the yeast Dhh1p and its homologue in mammalian cells RCK/p54, which are members of the DEAD-box family of RNA helicases, function both in translation repression and P-body formation in yeast and mammals (; ). Importantly, homologues of Dhh1p are required for proper translation control of maternal mRNAs and are found in maternal storage granules in a variety of organisms. For example, the homologue Me31b is found within maternal mRNA granules and is required for the proper translational control of the mRNA (). The homologue CGH-1 colocalizes with P-granules, the maternal mRNP granules that are localized to the germline in (). Xp54, the homologue, is present in stored RNP granules (), and tethering Xp54 to specific mRNAs represses their expression (). P-bodies and maternal mRNA storage granules share other components, as the decapping enzyme can also be found in P-granules in (; ). Similarly, the homologue of yeast Pat1p, a key component of yeast P-bodies, is a component of translationally repressed maternal mRNPs in oocytes (; ; ). These results suggest that maternal mRNA storage granules and P-bodies are both structurally and functionally related, although one anticipates that maternal mRNA storage granules will have additional mechanisms to allow spatial and temporal control of mRNAs in a more precise manner. In a current set of papers, members of the Scd6 family of proteins have been identified as a new conserved component of the translation repression complex found in P-bodies and maternal mRNA storage granules. This protein family was first identified through computational analyses as a new class of Lsm proteins (; ). The family is named after the member Sdc6p () and contains several subclasses of closely related proteins (). Each Scd6 family member contains two RNA-associated motifs: the Lsm, or like-Sm domain, at the NH terminus, and clusters of RGG motifs, which would be predicted to form an RGG box RNA binding domain. Like-Sm domains are ancient protein domains that are found in eubacteria, archaea, and eukaryotes (for reviews see ; ). Like-Sm domains interact with each other to form six- or seven-membered ring structures that can bind RNA. Both the Lsm region and a protein fragment containing the RGG region of the Scd6 orthologue CAR-1 bind poly(U) in vitro, verifying that this class of Lsm proteins is capable of directly binding RNA (). Scd6 family members also contain the newly recognized FDF motif of unknown function, although it is shared with the Edc3 family of Lsm proteins (; ), which are also components of P-bodies in yeast and mammals (; ). Scd6 orthologues have been found to associate with RNP particles containing translationally repressed mRNAs in numerous organisms. In oocytes, Trailer hitch colocalizes with RNP particles containing Me31b that have previously been shown to contain translationally repressed mRNAs (; ). In developing oocytes and embryos in , CAR-1 localizes to two types of RNP granules, both of which contain CGH-1, P-granules, and smaller cytoplasmic foci, which may be analogous to P-bodies in yeast and humans (; ; ). Scd6p localizes with Dhh1p in P-bodies in yeast (Johnson, N., personal communication). In addition, the mammalian homologue Rap55 localizes to P-bodies in mouse and human cells (). The localization of Scd6 proteins with these RNP particles suggests that they function in the control of mRNA translation and/or degradation. This idea is supported by the observation that Scd6 proteins physically associate with other proteins involved in translation repression. Trailer hitch and CAR-1 affinity purify with Me31b or CGH-1, respectively, as well as with other proteins associated with translation repression (; ; ). Some of these interactions are dependent on RNA, suggesting that the proteins may be linked together on the same RNA molecules. In other cases, the proteins may directly interact with each other because their copurification is resistant to RNase treatment. The observations that mRNAs copurify with Trailer hitch RNP complexes () and that CAR-1 can bind RNA () suggest that Scd6 proteins may bind specific mRNAs and control their translation (see last section). An unanticipated connection is that these Lsm proteins appear to have functional and physical links with the ER. This connection was first suggested by the observation that overexpression of the homologue Scd6p suppresses a deficiency in clathrin (), suggesting that Scd6p could affect the flux of proteins to or from the membrane. Subsequently, the homologue was found to be required for the efficient secretion of Gurken, a member of the TGF-α family, and Yolkless, the vitellogenin receptor, in oocytes (). Both Gurken and Yolkless proteins accumulated in large foci within the oocyte, although some of the protein was still secreted, suggesting that has a general role in the secretory process. An important step in secretion is the exit of proteins from the ER to the Golgi. The COPII complex is required for ER-to-Golgi trafficking and is located at discrete sites in the ER associated with ER exit (for reviews see ; ). In mutants, a component of the COPII complex, Sar1, is mislocalized from small discrete foci to abnormally large patches in nurse cells and the oocyte (). The mislocalization of Sar1 protein indicates that is required for normal ER exit site distribution and morphology. The defect in ER exit sites would lead to defects in the secretion of proteins like Gurken and Yolkless. The Scd6 orthologue CAR-1 has also been found to be required for ER dynamics in embryos (). Normally, the organization of the ER undergoes changes in conjunction with the cell cycle (). In embryos, the ER is in a dispersed state during interphase. The ER changes to a more ordered, reticulated state in mitosis. During mitosis, the ER also associates strongly with the mitotic spindle, both at the poles and in the area between the poles, referred to as the midzone. As the cleavage furrow forms, the reticulated state rapidly disassembles back to the dispersed state. In embryos that have been depleted of CAR-1 by RNAi treatment, the ER is found in large patches and thick strands both at interphase and during mitosis (). During mitosis, the association of the ER with the spindle, particularly with the midzone region, is reduced compared with wild-type. Thus, depletion of CAR-1 significantly disrupts ER organization. Whether the defects in ER organization in are due to a disruption of ER exit site function or the abnormal distribution and morphology of ER exit sites in are due to a more global disruption of ER organization remains to be determined. However, it is clear that this family of Lsm proteins is required directly or indirectly for normal ER organization. CAR-1 and Trailer hitch may affect ER organization directly, given that both proteins can localize at or near to the ER (; ). In embryos, the majority of small RNP granules containing CAR-1 overlap with ER structures in interphase. During mitosis, CAR-1 concentrates on the spindle region, including the midzone, as mitosis progresses, suggesting that it may associate with the ER at the spindle; however, this needs to be confirmed by colocalization studies. In , Trailer hitch colocalizes with the ER in nurse cells and oocytes (). Although the colocalization in oocytes is difficult to interpret because the oocyte is so densely packed with ER, it is likely to be relevant, given that there is a defect in exit from the ER in both cell types. The physical connection between RNP granules containing CAR-1 and Trailer hitch and the ER has led to the suggestion that these proteins regulate mRNAs that are associated with the ER (; ). Depletion or mutation of CAR-1 or results in a diverse set of phenotypes, some of which may be a consequence of the disruption of ER organization. Mutations in cause defects in dorsal–ventral patterning in , so that mutant eggs lack or have reduced dorsal appendages (). The ventralization of mutant eggs is likely due to the lack of secretion of Gurken, as it is a major determinant of dorsal fate. A major phenotype of CAR-1 in may also result from the disruption of the ER. Depletion or mutation of CAR-1 causes dramatic failure in cytokinesis in early embryos (; ; ). When CAR-1 is defective, the cleavage furrow begins to form normally during anaphase, but then it regresses and membrane fails to accumulate at the cleavage furrow. In wild-type embryos during anaphase, interzonal or midzone microtubule bundles form between the two separating masses of chromosomes associated with the spindles. These interzonal microtubule bundles are thought to signal the proper positioning of the cleavage furrow () Interzonal microtubule bundles are completely absent in CAR-1–depleted embryos (; ). propose that the lack of microtubules in the midzone is a result of the reduced accumulation of ER with the spindle in CAR-1 RNAi embryos. The close association of the ER with the spindle could allow the ER to stabilize the microtubules by regulating the local calcium concentration. In addition, the normal distribution and function of exit points in the spindle-associated ER may be necessary for the elaboration of new membrane and membrane proteins required for completion of cytokinesis. Therefore, one simple model is that the failure of cytokinesis results from the disruption of the ER organization caused by the lack of functional CAR-1. Alternatively, because CAR-1 associates with the spindle and defects in midzone spindle formation can affect ER organization at the midzone (), it is formally possible that CAR-1 affects spindle function, which then leads to the defect in ER organization at the spindle; however, this model does not explain why depletion of CAR-1 has a global effect on ER organization. CAR-1 defects in also give rise to other phenotypes. CAR-1 got its name from its effects on cytokinesis and apoptosis as well at its association with RNA. In , about half of all oocytes undergo apoptosis just before individual oocytes form from a syncytium in the gonad, apparently to provide cytoplasmic components to the oocytes that do form (). Depletion of CAR-1 increases physiological germ cell apoptosis and impairs oogenesis (). It is not clear whether these alterations in physiological apoptosis and oogenesis are related to defects in ER function. However, it is interesting that depletion of CAR-1 causes extracellular cytoplasmic “spheres” to accumulate that contain high levels of the secreted yolk receptor protein, RME-2 (). It is tempting to speculate that these “spheres” are produced by some perturbation in the secretory process. Alternatively, CAR-1 may regulate a variety of mRNAs that affect cellular processes other than ER function. xref #text An unresolved issue is the precise function of these Scd6 proteins. Three lines of arguments lead to the hypothesis that the Scd6 family of Lsm proteins plays some role in the regulation of mRNA translation and/or degradation. First, CAR-1, Trailer hitch, Rap55, and Scd6p can all be found in mRNA granules containing untranslated mRNAs and coimmunopurify with other proteins involved in translational control. Second, depletion of the DEAD box RNA helicase, CGH-1, causes defects in cytokinesis and apoptosis similar to those seen with CAR-1 depletion, supporting the idea that CAR-1's function is related to mRNA (; ). Third, these proteins can bind RNA and coimmunoprecipitate mRNAs. In this view, phenotypes of strains lacking these proteins would be due to misregulation of specific mRNAs. For example, Scd6 proteins could regulate particular mRNAs that control the spatial organization of the ER. Localized translation would allow the concentrated expression of the encoded proteins, which would then influence the formation of functional subdomains of the ER. Consistent with this idea, the mRNAs for Sar1 and Sec13, components of the COPII complex, which is associated with ER exit sites, are present in immunoprecipitates of Trailer hitch (). Concentrated localized expression of Sar1 mRNA on the ER would be an effective mechanism for controlling the size and distribution of ER exit sites, as recruitment of Sar1 is the first step in assembly of COPII complexes on the ER membrane. The idea that localized mRNA translation may be involved in determining ER organization is novel and raises the possibility that this is an important mechanism for controlling ER dynamics through the cell cycle and during development. It should be pointed out that there is no direct evidence that members of this family of Lsm proteins act as translational repressors. Because targeted translation of mRNAs involves the reactivation of the repressed mRNA, CAR-1 and Trailer hitch could act as translational activators or as both activators and repressors. This latter possibility is suggested by the fact that cytoplasmic polyadenylation element binding protein, an mRNA binding protein involved in local control of translation in a variety of settings, functions first to repress the translation of mRNAs and later to activate them (for review see ). Finally, one has to consider the possibility that these proteins may work independently of translational control. This is suggested by the observation that the Scd6 orthologue Rap55 is associated with RNP particles that are thought to control mitotic spindle assembly by a mechanism proposed to be independent of new translation (). This finding suggests that RNA or proteins within the particle may serve a direct structural or regulatory role in spindle assembly. Thus, Scd6 proteins could theoretically influence RNP function without modulating translation. Future work identifying the mRNAs bound by this class of Lsm proteins and how their function is affected by these proteins will be important for understanding the specific role of this protein family and for determining the roles that control of mRNA localization, translation, and decay play in eukaryotic cells.

The activation of NF-κB can be mediated, at least partially, by interaction motifs present in the prodomains of specific caspases (; ; ). Indeed, several studies have shown that caspases-1, -2, -8, and -10 use similar mechanisms to activate NF-κB, but caspase-9, -11, and -12 do not activate this transcription factor (; , ). The prodomains of caspases-2, -8, and -10 were shown to be capable of recruiting the NF-κB signaling molecules such as the E3 ubiquitin ligase activity containing factor, tumor necrosis factor-receptor associated factor 2 (TRAF2), the kinase receptor interacting protein 1 (RIP1), and NEMO/IKKγ, eventually leading to increased kinase activity of the IKK complex and NF-κB activation (; ; ) (). Caspase-2 complex formation, initiated by PIDD, has been associated with DNA damage repair (). The death domain–containing serine/threonine kinase RIP1 is also recruited in the caspase-2 complex (), and has been shown to be essential for DNA damage–induced NF-κB activation (). Therefore, the caspase-2 complex, also called PIDDosome (), may act as an integrator or molecular switch between inflammatory/anti-apoptotic and apoptotic signaling pathways (). When DNA damage occurs, the PIDDosome complex may emit signals that activate NF-κB, hence allowing the cell to initiate cell survival and DNA repair pathways. Whether caspase-2 has a physiological role in this pathway remains to be established. However, massive DNA damage may also generate too much proteolytically active caspase-2, leading to the continuous activation of apoptotic pathways. Similarly, caspase-8 forms a complex with FLASH (FLICE-associated huge protein) and TRAF2 to initiate the NF-κB activation pathway during TNF signaling in fibroblast and epithelial cell lines (). Although in these cells caspase-8 enzymatic activity is apparently not required for TNF-induced NF-κB activation, antisense-mediated caspase-8 depletion resulted in inhibition of TNF-induced NF-κB activation (), suggesting that caspase-8 can have signaling activities independent of its enzymatic activity. In addition, showed that RNAi-mediated knock-down of caspase-8 abrogates NF-κB activation induced by Apo2L/TRAIL (a member of the TNF family), which would occur when caspase-8, FADD, TRAF-2, RIP-1, and NEMO are recruited to a secondary cytosolic protein complex. The authors propose that caspase-8 enzymatic activity is implicated in Apo2L/TRAIL-dependent NF-κB activation because addition of the pan-caspase inhibitor zVAD-fmk blocks this NF-κB activation. As NF-κB activation upon TRAIL receptor stimulation is observed rather late (hours), as compared with TNF-induced NF-κB activation (minutes), one cannot exclude that the identified signaling complex is formed in response to general apoptotic stress, independently from the TRAIL receptor–bound DISC complex. In contrast to caspases-2, -8, and -10, the caspase-1 CARD domain is able to recruit the RIP2 kinase, which is involved in TLR- and TcR-mediated NF-κB activation () (). Experiments using a RIP2 dominant-negative mutant have indicated that RIP2 mediates caspase-1 CARD-induced NF-κB activation (). Caspase-1 can be recruited to different types of inflammasomes that are initiated upon recognition of intracellular bacteria by NLRs (). However, the in vivo relevance of the activation of NF-κB by caspase prodomains remains to be proven. The initial observation that caspase inhibitors efficiently abrogate the proliferation of primary human T cells in vitro and block the production of IL-2 in CD3/CD28-stimulated Jurkat cells suggested a role for caspases beyond the apoptotic context (). Several reports have now provided genetic evidence for the role of caspase-8 in the proliferation of immune cells (; ; ; ). Patients with inactivating mutations in caspase-8 suffer from impaired proliferation of T, B, and NK cells (). Interestingly, mice in which caspase-8 was conditionally deleted in T cells suffered from similar defects (). Peripheral T cells from these mice were unable to proliferate after T cell receptor activation. IL-2 production after CD3/CD28 stimulation decreased significantly in caspase-8–deficient T cells of both human and murine origin (; ). Recently, the circle was closed by demonstrating the essential role of caspase-8 in T cell receptor (TcR)–induced NF-κB activation (), confirming its function upstream of IL-2 production and human T cell proliferation. Moreover, through reconstitution experiments in caspase-8–deficient primary human T cells, enzymatically active caspase-8, although unprocessed, was shown to be required for TcR-mediated NF-κB activation (). Through labeling of the active center it was estimated that only 10–15% of the total caspase-8 content became enzymatically active after TcR stimulation. In contrast, TNF-induced NF-κB activation in T cells does not require caspase-8 enzymatic activity (), similar to observations in fibroblasts and epithelial cells, as discussed above (). However, did not observe a difference in TcR-stimulated NF-κB activation in caspase-8–deficient mouse T cells. But these authors looked at NF-κB activation at a rather late time point (6 h), which may explain this apparent contradiction concerning the involvement of caspase-8. In this respect, it has been shown previously that TRAF2 deficiency only delays the kinetics of NF-κB activation in response to TNF, which normally occurs within minutes after treatment (). At a later time point (90 min), the extent of activated NF-κB in TRAF2-deficient cells was equivalent to that in control cells. Finally, T, B, and NK cells of patients with caspase-8 deficiency do not show NF-κB activation after stimulation of antigen receptors, Fc receptors, or TLR-4 (), confirming the essential role of caspase-8 in NF-κB signaling. What is the mechanism of caspase-8 activation? It was shown that TcR stimulation triggered the recruitment of FADD and caspase-8 to the CARMA-Bcl10-MALT1 (CBM) complex (). Absence of caspase-8 abrogated IKK recruitment and activation by the CBM complex. Although it is clear that enzymatically active caspase-8 is required, the identity of the substrates that need to be processed to initiate the NF-κB signaling cascade remains to be determined. MALT1, also called paracaspase, harbors a caspase-like domain and a death domain, and is required for TcR-induced NF-κB activation (). In addition to the CBM complex (), TRAF2, TRAF6, and RIP2 were also shown to be essential for TcR-induced NF-κB activation (; ). Next to caspase-8, TLR-4 signaling shares several other molecules required for NF-κB activation with the TcR pathway, such as Bcl10 (), RIP1 (), RIP2 (), and TRAF6 (). In contrast to B cells of patients harboring homozygous caspase-8 mutations (), the deletion of caspase-8 in murine B cells does not lead to impaired NF-κB activation upon TLR-4 stimulation (). Nevertheless, caspase-8–deficient B cells fail to proliferate in response to dsRNA or LPS, ligands for TLR-3 and TLR-4, respectively. It remains to be determined whether the observed discrepancy in the response of human and murine caspase-8–deficient B cells is due to a fundamental difference in the signaling pathways involved, to the different nature of caspase-8 deficiency, or to the different experimental designs used to analyze NF-κB activation. Why would activation of the apoptotic initiator caspase-8 in one cellular context lead to initiation of apoptosis, whereas in another cellular context it activates NF-κB and induces proliferation? This may, at least in part, depend on the level of caspase-8 activation. In proliferating cells, caspase-8 apparently remains unprocessed and would therefore be less active or at least only be active in the complex (). In FasL-induced apoptosis, caspase-8 is processed and strongly activated (). One way to modulate the level of caspase-8 activation is by the expression of c-FLIP, a structural homologue of caspase-8 lacking caspase activity (; ). High concentrations of c-FLIP prevent full processing and release of active caspase-8 from the DISC, thus blocking the induction of cell death. Mice conditionally lacking c-FLIP in T lymphocytes display severe defects in the development and proliferation of mature T cells (; ), possibly due to enhanced apoptosis sensitivity. However, TcR-induced activation of NF-κB, a caspase-8–dependent phenomenon as discussed above, in cFLIP-deficient thymocytes appears largely intact (), whereas overexpression of cFLIP in Jurkat T cells increases NF-κB activation upon TcR ligation (). Therefore, the role of FLIP proteins in NF-κB activation remains unclear. Indeed, conflicting data were reported in the literature concerning the role of caspase-8 and cFLIP in TcR-dependent NF-κB activation (). The different levels of cFLIP expression in the different transgenic lines and the various extents of TcR triggering used may explain the apparent discrepancies observed in the literature. Indeed, it was shown that dependent on the anti-CD3 concentration used one can observe increased or reduced T cell proliferation in FLIP transgenic mice (). Taking together the data presented in , we can formulate the following conclusions. Deficiency of either FADD, caspase-8, or cFLIP results in a common phenotype, namely reduced proliferative response upon TcR stimulation, demonstrating that these proteins act in a common pathway starting from the TcR. Although it has been extensively documented that NF-κB is required for T cell proliferation and activation (), the data discussed above suggest that TcR-induced proliferation does not solely depend on NF-κB activation. Viral FLIPs are proteins that resemble the prodomain of caspase-8 and inhibit the recruitment of caspase-8 to the DISC (). Human herpesvirus 8 (HHV-8) vFLIP can also activate the NF-κB pathway, a property not shared by other vFLIPs (). This observation reminds us of the similar situation at the level of CARD-containing caspases—some are capable of inducing NF-κB activation, whereas others are not (, ). Transgenic overexpression of HHV-8 vFLIP in the lymphoid compartment leads to constitutive NF-κB activation and increased incidence of lymphomas, but has no significant effect on Fas-induced apoptosis or the development and maturation of lymphocytes. Therefore, one can fairly conclude that FADD, caspase-8, and cFLIP are molecularly linked, at least in normal T cell development and functioning. The exact role of these molecules in the NF-κB activation pathway remains to be determined. In this respect, it would be interesting to analyze the NF-κB signaling pathways in transgenic mice overexpressing the caspase-8 inhibitor CrmA (cytokine response modifier A) (). What really determines the final decision between FADD/caspase-8–mediated apoptosis or FADD/ caspase-8–mediated proliferation is not clear today, but availability and specific proteolytsis of particular substrates will certainly contribute. Recently, a possible involvement of executioner caspases in NF-κB activation was suggested. Although cleavage of poly(ADP-ribose)polymerase-1 (PARP-1) at the DEVD site has been used as a canonical hallmark of caspase-3/-7 activation and apoptosis, the in vivo significance of this cleavage was largely unknown. Initially, it was thought that PARP-1 cleavage in apoptotic cells represented a way to inactivate the DNA repair capacity. Recent genetic evidence suggests a novel role for caspase-generated fragments of PARP-1 in inflammatory responses. Indeed, LPS-stimulated macrophages from knock-in mice expressing caspase-resistant PARP-1 display impaired NF-κB–mediated gene activation despite normal binding of NF-κB to DNA (). As discussed above, caspase-8 is required for nuclear translocation of NF-κB (), whereas executioner caspases apparently act at the NF-κB transactivation level following DNA binding through a PARP-1–mediated mechanism () (). Indeed, it had previously been shown that PARP-1 can interact with both subunits of NF-κB (p50 and p65) and with the transcriptional coactivator p300 (, ). Whether the enzymatic activity of PARP-1 is required for the enhanced transcriptional activity of NF-κB is a matter of dispute. Some reports show that enzymatic activity is not required for coactivation of NF-κB by p300 in TNF- or LPS-stimulated primary fibroblasts or macrophages (; ). However, other studies using PARP-1 inhibitors do indicate a role for PARP-1 enzymatic activity in the enhancement of NF-κB transcriptional activity (; ). The strong phenotype of the caspase-resistant PARP-1 knock-in mice, such as their resistance to endotoxic shock and to intestinal and renal ischemia-reperfusion, further support an in vivo contribution of executioner caspases to inflammatory responses through PARP-1 cleavage. Hence, caspase-mediated PARP-1 cleavage under nonapoptotic conditions could contribute to the level of NF-κB transcriptional activity. The possible role of executioner caspases in the activation of NF-κB may at least partially explain the multitude of nonapoptotic functions that have recently been attributed to this caspase subfamily. For example, caspase-3 was reported to be active in the nuclei of dividing cells in the proliferative regions of the rat forebrain (). Furthermore, caspase-dependent cleavage of p27 occurs in proliferating lymphoid cells (), and B cells from caspase-3–deficient mice can hyperproliferate (). The latter finding indicates that caspase-3 may function as a negative regulator of B cell cycling. In this respect, caspase-3 was shown to be required for the cleavage of the Cdk inhibitor p21 in B cells (). Similar studies in conditional knock-out mice may reveal additional roles for caspases in cell cycle regulation. sub #text

Both kinetochore microtubules (MTs [kMTs]) and nonkMTs in mitotic and meiotic bipolar spindles of higher eukaryotes exhibit poleward translocation or flux (). Most kMTs normally extend the full length of the kinetochore fiber from their plus end attachment sites at kinetochores to minus end anchorage sites at spindle poles (). In animal cells, the flux of kMTs is coupled to minus end depolymerization at spindle poles. This poleward flux of kMTs can account for 20–100% of chromosome to pole movement depending on cell type (). The remaining poleward movement is produced by kinetochore “Pacman” motility that is coupled to kMT depolymerization at the kinetochore. The molecular mechanisms that generate kMT poleward flux are still poorly understood. Several studies have reported that Eg5 (kinesin 5) is responsible for the sliding component of flux for both nonkMTs and kMTs (; ; ). This plus end–directed kinesin cross-links antiparallel MTs and slides them toward their minus ends. Because the plus ends of nonkMTs overlap with each other and with kMTs in the central region of a bipolar spindle, Eg5 is an ideal candidate for the role of flux driver. Forces could be applied to kMTs by interaction with Eg5 or through lateral cross-links to adjacent fluxing nonkMTs to the same pole (; ; ; ). On the basis of these studies, proposed a mechanistic model in which sliding forces generated by Eg5 drive poleward MT flux and activate MT minus end depolymerization at poles. A salient feature of this model is that pole-associated MT depolymerases (e.g., kinesin 13) sense sliding forces to regulate the depolymerization rate and spindle length. In agreement with this model, the inhibition of KLP10A (kinesin 13 in ) did not affect the sliding component of flux but blocked depolymerization at the poles, resulting in highly elongated metaphase spindles (). In egg extract spindles, perturbation of the normal localization of Kif2a (kinesin 13) by the disruption of dynein/dynactin blocks MT minus end disassembly at poles, but antiparallel MT sliding continues (). Here, we test whether Eg5 is the dominant mechanism of kMT poleward flux in mammalian PtK1 cells using specific inhibitors of Eg5. We assay flux in monopolar spindles that lack antiparallel MTs and test two polar complex proteins for their possible role in poleward flux. An important aspect of our studies is the use of quantitative fluorescent speckle microscopy (FSM [qFSM]) and fluorescence photoactivation techniques combined with two-color spinning disk confocal imaging to obtain much more accurate measurements for kMT poleward flux than achieved in previous studies on the roles of kinesin 5 and 13 for all spindle MTs (Miaymoto et al., 2004; ; ; ). Kinetochores in mammalian cultured cells exhibit directional instability (), although the character of movement is usually somewhat different for individual bioriented chromosomes at the spindle equator. Those chromosomes that are positioned near the spindle axis oscillate regularly between phases of poleward and antipoleward movement. In contrast, chromosomes aligned at the periphery of the metaphase plate show little, if any, oscillation (; ). We found by kymograph analysis that flux rates of kMTs were not significantly different for kinetochore fibers attached to oscillating and stationary chromosomes (; Video 1, and supplemental material, available at ). Although we consistently detected new speckles entering the kinetochore fiber as kMTs polymerized during antipoleward movement and detected the disappearance of speckles near the kinetochore as kMTs depolymerized during poleward movement (, arrows), the rates of flux remained constant (). The mean tension for stationary kinetochores as measured by the distance between sister kinetochores (2.6 ± 0.2 μm; = 30 time points for each of seven kinetochores) was typically higher than the mean tension for oscillating sisters (2.36 ± 0.52 μm; = 30 time points for each of six kinetochores; centromere rest length was 1.3 ± 0.25 μm in cells lacking MTs). Thus, the higher tensions developed at kinetochores as they stretched their centromeres during poleward movement for oscillating kinetochores had no detectable effect on the rate of kMT poleward flux in PtK1 cells, unlike what has been proposed in meiotic egg extract spindles and crane fly spermatocytes (; ). To improve the accuracy and statistical significance of our kMT flux measurements, we used qFSM (). qFSM involves computer-vision methods to automatically detect and track speckles along kinetochore fibers (; supplemental material; and Fig. S1, A–C; available at ; Yang, G., A. Matov, and G. Danuser. 2005. Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Computer Vision and Pattern Recognition. 9–17). This method yielded a mean kMT poleward flux velocity in bipolar spindles of 0.65 ± 0.08 μm/min ( = 1,428 tracks; supplemental material). The variation of speckle velocities within a spindle was larger (SD = 0.26 μm/min; ) than between different spindles (supplemental material). This indicates that MT flux involves significant heterogeneity that is possibly associated with spatially and temporally varying contributions of multiple force-generation mechanisms. However, the flux pattern did not exhibit a clear spatial organization (), precluding the identification of the various sources of flux from heterogeneity analysis alone. Kinetochore fibers are composed of ∼25 kMTs and 25 nonkMTs in PtK1 cells (). The half-life of kMTs has been estimated to be ∼5 min at metaphase, whereas the half-life of nonkMTs is much shorter (∼0.2 min) because of the high dynamic instability of nonkMT plus ends and their rapid growth (11 μm/min) and shortening (∼20 μm/min) velocities (; ). The width of a kinetochore fiber (<0.4 μm) is not much larger than the resolution limit of the light microscope (0.25 μm). As a result, fluorescent speckles seen at a specific site in images of kinetochore fibers contain fluorophores from adjacent nonkMTs that produce fluctuation in speckle intensity that could potentially confuse automated tracking (supplemental material). Hence, we sought to verify flux rates by an independent method. Tracking the movement of fluorescent marks on kMTs produced by photoactivation is less sensitive to signal instability caused by nonkMT turnover. Fluorescence from nonkMTs rapidly disappears after photoactivation (supplemental material and Video 2, available at ), leaving fluorescent marks on the more stable kMTs in higher contrast (). We applied fluorescence photomarking methods to PtK1 cells expressing photoactivatable (PA) GFP (PA-GFP) fused to α-tubulin. PA-GFP–tubulin within spindle fibers was photoactivated by a 0.1-s pulse from a 408-nm laser focused by a cylindrical lens into a pseudo-Gaussian line profile with a 1.4-μm width at half-maximum intensity. The persistent marks on kinetochore fibers in metaphase cells were seen to move poleward at a constant velocity with a mean of 0.62 ± 0.26 μm/min (Video 2), which is similar to our qFSM measurements and earlier studies (; ). As reported previously (; ; ), the width of the mark along the fibers appeared relatively constant until it reached the ends of the kinetochore fibers at the pole, where it shortened at the velocity of poleward flux (Video 2). We simulated mark movements along the 5–7-μm length of kinetochore fibers in metaphase spindles using the velocity distributions obtained by qFSM (). These simulations produced movements of photoactivated marks that closely approximated the experimental data and indicated that the heterogeneity of speckle velocities does not detectably widen the mark over its short travel to the pole (supplemental material and Video 3). To determine whether kMT poleward flux in PtK1 cells depends on the activity of kinesin 5, we used several specific small-molecule inhibitors. Monastrol (), HR22C16-E1 (), and s-trityl--cysteine () were used at concentrations reported to substantially inhibit flux in egg extract spindles (100–200 μM monastrol) or induce monopolar spindles in mammalian cultured cells (10 μM HR22C16-E1 and 2 μM s-trityl--cysteine). Both qFSM and photoactivation approaches revealed that the inhibition of Eg5 in bipolar metaphase spindles did not abrogate kMT poleward flux, although the rate decreased by ∼25% (to 0.5 ± 0.27 μm/min; and see , supplemental material, and Videos 4 and 5; available at ). Spindle length and the character of chromosome movement also remained normal upon kinesin 5 inhibition (mean sister kinetochore distance = 2.16 ± 0.23 μm; either kinetochores with 20–30 time points each; P = 0.2, which is not significantly different from the control). Thus, Eg5 is responsible for only a minor contribution to kMT poleward flux during metaphase in PtK1 cells. This differs substantially from reports of kMT poleward flux and spindle length regulation in S2 cells () and MT flux in meiosis II egg extract spindles (; ) in which Eg5 has a dominant role. Although Eg5 is thought to be the molecule responsible for antiparallel MT sliding, it is conceivable that other molecular motors can contribute to this mechanism. Therefore, we analyzed kMT poleward flux in monopolar spindles that do not have overlapping MTs of opposite orientation. The addition of kinesin 5 inhibitors before nuclear envelope breakdown results in the formation of monopolar spindles with unseparated centrosomes (). Plus ends of MTs in the monopolar spindles are uniformly oriented away from the pole, as evident from time-lapse recordings of PtK1 cells expressing the GFP form of the +TIP protein EB1 ( and Video 7, available at ; ). Chromosomes attach to monopolar spindles via either one (monotelic) or both sister kinetochores (syntelic) and oscillate toward and away from the pole in a normal fashion (Video 6; ). The rate of kMT poleward flux in monopole spindles measured by qFSM was 0.56 ± 0.1 μm/min ( = 7 cells; 792 tracks; ). We also measured flux in monopoles using photoactivation of a bar of fluorescence across a monopolar spindle in between the chromosomes and the pole (). Much as in bipolar spindles, photoactivated marks disappeared quickly for nonkMT fibers and persisted for fibers containing kMTs. These persistent marks moved poleward at a constant velocity () until they reached the ends of their kinetochore fibers. Then, the width of the mark decreased at the rate of flux (, Fig. S1 E, and Video 8, available at ). Photoactivation measurements yielded a velocity of poleward flux of 0.44 ± 0.26 μm/min ( = 13 cells; 17 tracks; see and supplemental material). Thus, when Eg5 is inhibited, the velocity of flux is similar between monopolar and bipolar spindles. This, in turn, implies that most of kMT poleward flux in mammalian spindles is not based on the sliding of antiparallel MTs. Recently, reported that MT poleward flux in bipolar mitotic spindles is inhibited by siRNA depletion of the depolymerase Kif2a in human U2OS cells. These cells were also depleted of another depolymerase, mitotic centromere-associated kinesin, which alone did not block flux. In this study, they were unable to resolve kinetochore fibers. In PtK1 cells, we found that Kif2a was concentrated at the polar minus ends of kinetochore fibers from late prometaphase through anaphase, and it also localized to centromeres throughout mitosis and to centrosomes during interphase and mitosis (Fig. S2, available at ). Microinjection of antibodies to either human or Kif2a into prophase or prometaphase cells had no effect on the progression of chromosome alignment or segregation as assayed by live cell imaging (unpublished data). Because siRNA for Kif2a is currently not possible in PtK1, we attempted to alter Kif2a localization at spindle poles by microinjecting a recombinant protein fragment (p150_CC1) of the NH-terminal coil-coiled domain of the p150 subunit of dynactin. P150_CC1 disrupts dynein/dynactin, induces spindle lengthening, and has been shown to greatly diminish Kif2a localization to spindle poles in egg extract spindles (). In PtK1 cells, p150_CC1 microinjection altered spindle morphology, causing centrosomes to become disconnected from the ends of the spindle and minus ends of kinetochore fibers to become less focused ( and Fig. S2). Nevertheless, poleward flux continued in both bipolar (0.70 ± 0.08 μm/min; Video 9) and monopolar spindles, and Kif2a remained most concentrated at the minus ends of the kinetochore fibers (). In addition, most sister chromosomes continued to oscillate as assessed by themovement of dark bars (absence of fluorescence) between growing and shortening sister kinetochore fiber pairs (Video 9). These results indicate that Kif2a is an integral part of a complex of proteins that concentrates at the minus ends of individual, fluxing kinetochore fibers even when the minus ends of the fibers are not attached to the centrosome and are not focused together. Our results reveal that neither Eg5 nor antiparallel MTs are required for the majority of kMT poleward flux in PtK1 cells. The velocity of flux in PtK1 cells is slow relative to flux in embryo and egg extract spindles, and there is little () to no () flux in monopolar spindles in these systems. These results suggest that mechanisms responsible for flux contribute differently in different systems (e.g., mammals vs. lower animals). Although polar ejection forces on chromosome arms push nonkMTs poleward, they pull kinetochores and their kMTs away from the pole (). As a result, polar ejection forces cannot be the source of force that produce a constant rate of kMT poleward flux in monopolar spindles, particularly during kinetochore poleward movement when kinetochores pull strongly on their kMTs in a direction opposite to flux. Instead, our data indicate that in PtK1 cells, the dominant mechanism of kMT poleward flux is a poleward pulling-in mechanism coupled to minus end depolymerization (; ; ; ; ; ). This mechanism may be similar in principle to the kinetochore Pacman mechanism. However, the mechanism of dynamic attachment and pulling force generation at a depolymerizing end is not well understood. It might involve sleeve mechanisms as envisioned by , rings of DASH/DAM proteins like those discovered in budding yeast, anchorage proteins like NuMA, and pulling forces generated by inside-out peeling of tubulin protofilaments at depolymerizing ends promoted by the ATPase activity of depolymerases like Kif2a (). Our p150_CC1 data suggest that a kind of self-feeding mechanism is working in which both the depolymerase and pulling activity are in the same complex that remains at minus ends of K fibers and does not require anchorage to the centrosome or focused pole. Flux and depolymerization rates may depend critically on load, as proposed by . A plus end–directed kinesin motor attached to the complex of proteins at the minus ends of kinetochore fibers may be critical for driving poleward flux () as well as reducing the load on the depolymerization mechanism (“thresher mechanism”). Sensitivity of the minus end depolymerization mechanism to load would also explain how kinesin 5–driven sliding forces in bipolar PtK1 spindles produce a 25% increase in the poleward flux rate of kMTs. Finally, we found little change in the length of bipolar spindles when Eg5 was inhibited, a result typical for mitotic mammalian spindles in late prometaphase or metaphase (). Unlike meiosis II spindles in egg extracts, other factors such as astral or polar ejection forces may play a much more significant role in determining metaphase spindle length than Eg5 sliding forces. PtK1 cells were maintained in Ham's F-12 medium (Sigma-Aldrich) and were complemented with 10% FBS, antibiotics, and antimycotic (). A PtK1 cell line stably expressing PA-GFP fused to α-tubulin was created by selecting cells transfected with a vector that was made by removing α-tubulin with BamHI and XhoI from the GFP-tubulin vector (CLONTECH Laboratories, Inc.) and ligating it into pPA-GFP-C1 (gift of G. Patterson and J. Lippincott-Schwartz, National Institute of Child Health and Human Development, Bethesda, MD; ). PA-GFP–tubulin PtK1 cells were selected with 1 mg/ml geneticin and maintained in Ham's F-12 as PtK1 cells except they were sometimes complemented with 15% FBS to promote the growth and health of cells. A PtK1 cell line stably expressing GFP-EB1 was a gift of J. Tirnauer (University of Connecticut Health Center, Farmington, CT) and was maintained in Ham's F-12 with 0.25 mg/ml geneticin. For live cell imaging, all cells were incubated in Leibovitz's L-15 (Invitrogen) complemented with 10% FBS, antibiotics, antimycotic, 7 mM Hepes, 4.5 g/L glucose, and 0.45 U/ml oxyrase (Oxyrase). We used two-color fluorescence live cell imaging. PtK1 cells were coinjected with low levels of X-rhodamine–labeled tubulin to fluorescently speckle MTs () and AlexaFluor488-labeled antibodies to CENP-F protein (gift of D. Cleveland [University of California, San Diego, San Diego, CA] and T. Tafari [National Institute of Environmental Health Sciences, Research Triangle Park, NC]) to label kinetochores and spindle poles (). Microinjection was performed at room temperature to prevent tubulin polymerization inside the microneedle. For dynein/dynactin inhibition experiments, 0.2 μg/ml p150_CC1 recombinant protein (needle concentration; gift of J. Gaetz and T. Kapoor, The Rockefeller University, New York, NY) was coinjected with rhodamine-labeled tubulin (to designate which cells were injected for the immunofluorescence of Kif2a or for speckle labeling of MTs in live cell imaging experiments to measure flux rates). Some cells were injected with tubulin alone and fixed and stained for Kif2a (human antibody to Kif2a was a gift of D. Compton, Dartmouth Medical School, Hanover, NH; antibody to Kif2a was a gift of R. Ohi, Harvard Medical School, Boston, MA) to ensure that rhodamine-tubulin injection was not altering Kif2a localization. Tubulin and antibody or p150_CC1 were diluted into tubulin injection buffer (50 mM K-glutamate and 0.5 mM MgCl, pH 7.0, with KOH) and coinjected into prophase and/or early prometaphase cells. Digital images were collected with a cooled CCD camera (Orca ER; Hamamatsu) coupled to a Yokogawa spinning disk confocal unit (CSU10; PerkinElmer), which was attached to an inverted microscope (TE300; Nikon []) with a 100× 1.4 NA plan-Apochromatic differential interference contrast objective. Stage temperature was maintained at ∼35°C using an air curtain incubator (ASI 400; Nevtek). Near simultaneous fluorescence images were acquired at 488 and 568 nm at a single focal plane every 5–20 s. For photoactivation, a 408-nm laser (56ICS323; Melles Griot) was focused by a cylindrical lens into a pseudo-Gaussian line profile with a 1.4-μm width at half-maximum intensity. Images were collected before and after a 0.1-s photoactivation through a 100× 1.4 NA plan-Apochromatic phase objective to track chromosome and spindle location (phase contrast) and the photoactivated bar (488-nm fluorescence and 1-s exposure). Each frame of the FSM image sequence was aligned to the previous frame in the sequence using a correlation-maximization algorithm (). Speckle detection was then performed on the aligned images as previously described (). Speckle tracking was formulated as a modified optimal bipartite graph linear assignment problem and was solved using the Jonker-Volgenant algorithm (Yang, G., A. Matov, and G. Danuser. 2005. Proceedings of the Institute of Electrical and Electronics Engineers International Conference on Computer Vision and Pattern Recognition. 9–17). The flux rate for each track was defined as its mean velocity. When computing flux rates in a specific region, only those tracks that were contained completely within that region were considered. Software for image alignment, speckle detection, speckle tracking, and statistical analysis of tracking results was developed using MATLAB (Mathworks) and C++. For photoactivation image analysis, the distance between the photoactivated mark and the closest pole was tracked and measured over the time course of the video using MetaMorph software (Molecular Devices). A graph of distance versus time was plotted, and the slope of the best-fit line was determined to be the flux rate for that mark. Online supplemental material includes methods for fluorescent speckle detection and tracking, a simulation of photoactivation, and additional analysis of flux velocity distributions in monastrol-treated cells. Fig. S1 shows an example of speckle tracking (A–C) and a cumulative velocity histogram comparison (D). Fig. S1 E is a kymograph corresponding to the monopolar spindle in . Fig. S2 shows the localization of Kif2a and the effects of p150_CC1 injection in PtK1 cells. Table 1 contains a summary of all poleward flux measurements. Video 1 corresponds to . Video 2 shows a photoactivated mark of tubulin fluorescence moving to the pole in a control spindle. Video 3 shows four photoactivation simulations that are described in the supplemental material. Videos 4 and 5 correspond to , and Videos 6–8 correspond to . Video 9 is a time lapse of p150_CC1 plus a rhodamine-tubulin–injected PtK1 cell (as in ). Online supplemental material is available at .

Replication checkpoint signaling initiates when the DNA structures that form at stalled replication forks activate the ataxia-telangiectasia and Rad-3–related (ATR) protein kinase (for review see ). ATR then phosphorylates and activates the Chk1 kinase. The pathway leading from activated ATR to Chk1 is complex and involves numerous intermediaries. In fission yeast, activation of Chk1 by the ATR homologue Rad3 requires, amongst other factors, Rad9, Crb2, and Cut5. Cut5 plays a central role in transducing the checkpoint signal from activated Rad3 to Chk1, as it forms complexes with both Crb2 and Rad9 (for review see ). In metazoans, Chk1 activation also requires ATR, Rad9, and the Cut5 homologue TopBP1/Mus101. Despite this conservation, there is an important difference between the fission yeast and metazoan Chk1 activation pathways. In metazoans, the Claspin protein plays an essential role in Chk1 activation (), whereas in fission yeast the Claspin homologue Mrc1 is not involved (). Likewise, in fission yeast, Crb2 is essential for Chk1 activation (), whereas in metazoans the Crb2 homologue 53BP1 is not known to be involved. Because Crb2/53BP1 is not required for Chk1 activation in metazoans, it is unclear what role, if any, TopBP1/Mus101 plays in promoting activation of Chk1 by ATR. Cut5/TopBP1 is unique amongst Chk1 activators in that it is also essential for DNA replication (). This is an important feature of Cut5/TopBP1, as replication is required to generate the DNA structures that activate ATR and initiate the checkpoint response (). In fission yeast, temperature-sensitive alleles of Cut5 have allowed a separation of the replication and checkpoint functions of the protein (); however, it is not yet known if the replication function can be uncoupled from checkpoint function for TopBP1. One recent study of the TopBP1 homologue Xmus101 (also known as Xcut5) showed that Xmus101 is required to recruit both ATR and DNA polymerase α (pol α) to chromatin during a checkpoint response (). Pol α is required to generate the DNA structures that activate ATR (), and it is therefore possible that the role of Xmus101 in Chk1 activation is limited to this early phase of the process. Alternatively, Xmus101 could also have a later function, during relay of the checkpoint signal from activated ATR to Chk1. The differences in the mechanism of this signal relay between fission yeast and metazoans preclude clear predictions about what role, if any, Xmus101 might play in promoting phosphorylation of Chk1 by ATR. To address this important issue, we have used conditions in egg extracts that bypass the requirement for pol α and for generating DNA structures in activating Chk1. We report on the role that Xmus101 plays in Chk1 activation under these bypass conditions. Our results demonstrate that Xmus101 has a late checkpoint function, to promote phosphorylation of Chk1 by activated ATR. To study Chk1 activation, we used the AT70 system in egg extracts (). In this system, two short oligonucleotides, A70 and T70, are annealed to one another and then added to extracts. Addition of the duplex, but not either single oligonucleotide alone, triggers robust Chk1 phosphorylation. Chk1 activation in this system is dependent on ATR, Claspin, and other checkpoint proteins (, ; ). To monitor checkpoint activation, we used a previously established assay based on a fragment of the Chk1 protein, Chk1ΔKD, that is phosphorylated in an ATR- and Claspin-dependent manner in egg extracts (; ). Phosphorylation of Chk1ΔKD results in an easily detectable mobility shift on SDS-PAGE gels. An example of the AT70 checkpoint system is shown in . Egg extracts were supplemented with Chk1ΔKD and either the single A70 oligonucleotide, the AT70 duplex, or no DNA at all. After a 100-min incubation, samples were taken and probed by immunoblotting for Chk1ΔKD. A Chk1ΔKD mobility shift was observed in the samples containing AT70 (, lane 3) but not in the sample containing A70 (lane 2) or no DNA (lane 1). This demonstrates that AT70 specifically triggers Chk1ΔKD phosphorylation. Control experiments, detailed in the supplemental text (available at ), showed conclusively that Chk1ΔKD is a reliable surrogate for the endogenous Chk1 in these assays (Fig. S1). Previous work has shown that when sperm chromatin and the DNA replication inhibitor aphidicolin are used to activate the checkpoint in egg extracts, both the MCM complex and pol α are required for Chk1 activation (). Both of these proteins promote replication (for review see ). The MCM complex is a hexameric assembly that unwinds DNA during replication, whereas pol α synthesizes the primers that initiate replication. To determine whether these proteins are required for Chk1 activation in the AT70 system, we removed MCM and pol α by immunodepletion. shows that all detectable MCM5 protein was removed from the extract, and shows that same was true of the p70 subunit of pol α. Despite the loss of MCM and pol α in these extracts, addition of AT70 still activated Chk1, as shown by Chk1ΔKD mobility shift (). MCM and pol α are thus dispensable for Chk1 activation in the AT70 system. We conclude that the AT70 system bypasses the requirement for replication proteins in Chk1 activation. If the role of Xmus101 in Chk1 activation was restricted to generating the checkpoint-activating DNA structure, then it, like MCM and pol α, should be dispensable for Chk1 activation in the AT70 system. To address this, Xmus101 was removed from extract by immunodepletion () and the samples were assayed for Chk1 activation after addition of AT70. As shown in , and in contrast to MCM and pol α, removal of Xmus101 prevented Chk1ΔKD phosphorylation. Importantly, Chk1 activation was restored in Xmus101-depleted extract after supplementation of the extract with recombinant Xmus101 that had been produced in rabbit reticulocyte lysates (). Supplementation of Xmus101-depleted extract with unprogrammed reticulocyte lysates did not rescue Chk1ΔKD phosphorylation (unpublished data). We conclude that Xmus101 is essential for Chk1 activation in the AT70 system. This result thus distinguishes Xmus101 from the replication fork components MCM and pol α and demonstrates that the role of Xmus101 in Chk1 activation extends beyond generating the checkpoint-activating DNA structure. We considered the possibility that our anti-Xmus101 antibody, HU142, might inhibit Xmus101 function when added to extract. To test this, we added purified HU142 antibody directly to extract along with AT70. Addition of HU142 prevented AT70-induced phosphorylation of Chk1ΔKD, whereas nonspecific IgG did not (). This result demonstrates that HU142, which recognizes the COOH-terminal 333 amino acids of Xmus101 (), blocks AT70-mediated checkpoint signaling. One explanation for this is that binding of HU142 to the COOH-terminal 333 amino acids of Xmus101 prevents an interaction between this region and a factor that is required for checkpoint signaling. If so, then we might expect overexpression of the isolated 333 amino acid domain to also inhibit Chk1 activation through sequestration of this presumptive factor away from the full-length endogenous Xmus101. To test this, we titrated a recombinant protein consisting of GST fused to the 333 COOH-terminal amino acids of Xmus101 (GST-CT333; ) into extracts and then added AT70 to activate Chk1. As shown in , addition of GST-CT333, but not GST alone, inhibited Chk1ΔKD phosphorylation in a dose-dependent manner. We conclude that when the function of the extreme COOH terminus of endogenous Xmus101 is antagonized, through either binding of HU142 or overexpression of the isolated domain, then Chk1 activation is blocked. The results in and show that loss of Xmus101 blocks AT70-mediated Chk1 activation and that two independent inhibitors of Xmus101, a neutralizing antibody and a dominant-negative fragment, do the same. We next used these inhibitors to define the Xmus101-dependent step in Chk1 activation. In , Chk1 activation is (minimally) a three-step process (). Upon ATR activation, the first identifiable step is phosphorylation of the Claspin Chk1 binding domain (CKBD; ). This phosphorylation is ATR dependent and thus also serves as an indicator of ATR activation. After Claspin CKBD phosphorylation, the next step (step 2) is assembly of a Claspin–Chk1 complex. Claspin–Chk1 complex assembly is dependent on step 1 and is required for step 3 in the process, direct phosphorylation of Chk1 by ATR (). To assay for completion of step 1, CKBD phosphorylation, we used a fragment of Claspin corresponding to the CKBD and assessed its phosphorylation by mobility shift on SDS-PAGE according to published procedures (). The GST-Claspin CKBD did not undergo a mobility shift in extracts that lacked DNA, nor did it shift in extracts containing just the A70 oligonucleotide; however, we did detect a shift in extracts containing AT70 (, lanes 1–3). Interestingly, the GST-Claspin CKBD mobility shift occurred normally in extracts containing both AT70 and HU142 (, lane 4). This result shows that although HU142 prevents AT70-mediated Chk1ΔKD phosphorylation, it has no effect on CKBD phosphorylation. We next determined whether step 2, assembly of a Claspin–Chk1 complex, occurs in extracts containing Xmus101 inhibitors. For this, we used a previously published Claspin Chk1 binding assay (; ). A full-length GST- and His-tagged Chk1 protein (Xchk1-GH) was coupled to Ni–nitrilotriacetic acid (NTA) agarose beads, and the beads were then added to extracts. After a 100-min incubation, the beads were recovered, washed, and probed for the presence of Claspin by immunoblotting. A Claspin–Chk1 complex could be detected specifically in extracts that had activated the checkpoint through addition of AT70 (, lane 3) and was not detected in extracts lacking DNA or containing A70 (, lanes 1 and 2). Importantly, the Claspin–Chk1 complex readily formed in extracts containing both AT70 and HU142 (, lane 4). These data demonstrate that HU142 inhibits the ATR-dependent phosphorylation of Chk1 but not assembly of a Claspin–Chk1 complex. To confirm the results obtained with HU142, we repeated the experiments with an independent inhibitor of Xmus101 function, the recombinant GST-CT333 protein. Despite the ability of GST-CT333 to prevent phosphorylation of Chk1 (), it had no effect on either mobility shift of GST-Claspin CKBD or formation of a Claspin–Chk1 complex (). These results are fully consistent with those obtained with HU142 and demonstrate that although the COOH terminus of Xmus101 is required for Chk1 phosphorylation, it is not required for the ATR-dependent assembly of a Claspin–Chk1 complex. We conclude that one position of Xmus101 in the checkpoint activation pathway is after assembly of the Claspin–Chk1 complex and before phosphorylation of Chk1 by ATR. The results presented thus far show that Xmus101 plays a direct role in AT70-mediated Chk1 activation. To see if this is also true when checkpoint activation occurs under more physiological conditions, we examined the role of Xmus101 in Chk1 activation by replication-blocked sperm chromatin templates. Under these conditions, Xmus101 is required for replication fork assembly; thus, depletion of Xmus101 would indirectly affect Chk1 activation by virtue of a failure to generate DNA structures. To get around this, we asked if we could separate the replication and checkpoint functions of Xmus101 mutationally. We constructed a deletion mutant of Xmus101 named Mini Xmus101 or Mini (). Mini corresponds to the first 759 amino acids of the protein and thus lacks the COOH-terminal 333 amino acids that have been implicated in Chk1 activation. Endogenous Xmus101 was immunodepleted, and the depleted extracts were supplemented with rabbit reticulocyte lysates that had been programmed for in vitro transcription/translation reactions using constructs encoding either full-length Xmus101 or Mini. shows that the full-length Xmus101 and Mini were produced to an equivalent extent in the rabbit reticulocyte lysates. The reconstituted extracts were supplemented with sperm chromatin and α-[P]dATP, and replication of the sperm chromatin was assessed. Supplementation of depleted extract with either full-length Xmus101 or Mini rescued the replication defect completely, whereas add-back of unprogrammed reticulocyte lysates did not promote replication (). We conclude that all of the replication functions of Xmus101 are contained within Mini. Because Mini was competent for DNA replication, we asked whether it was also sufficient for Chk1 activation. To examine this, we added sperm chromatin and the DNA polymerase inhibitor aphidicolin to extracts that had been immunodepleted of Xmus101 and supplemented with either full-length Xmus101 or Mini. Chk1 activation was then monitored by immunoblotting of the extracts with antibodies specifically recognizing Chk1 that had been phosphorylated on serine 344. Serine 344–phosphorylated Chk1 represents the activated form (). As shown in , depletion of Xmus101 prevented Chk1 activation, and this was restored by add-back of full-length Xmus101. Importantly, add-back of Mini did not rescue Chk1 activation (). These data show that Mini is replication competent and Chk1 activation deficient. Based on the ability of Mini to separate the replication and Chk1 functions of Xmus101, we conclude that Xmus101 functions directly in Chk1 activation, independent of its role in replication fork assembly. In a final experiment, we sought to connect the function of Xmus101 in Chk1 activation to the checkpoint response that prevents mitosis when replication is blocked. For this, we used cycling egg extracts, which when treated with aphidicolin are prevented from entering mitosis in a checkpoint-dependent manner (). Mitosis was assessed by examining sperm nuclei for nuclear envelope breakdown, as described previously (). Addition of aphidicolin delayed mitosis, as expected (). Importantly, this delay was reversed by addition of the Xmus101 inhibitor HU142 (). HU142 was as effective as the known checkpoint inhibitor caffeine in releasing the checkpoint-mediated arrest. Addition of HU142 to interphase extracts had no adverse effect on DNA replication (). We conclude that Xmus101 is required for both Chk1 activation and for the checkpoint-mediated delay in entrance into mitosis when replication is blocked. Together with previous work (), the results presented here demonstrate that Xmus101 has at least two functions during a checkpoint response: it acts early to recruit ATR and pol α to damaged DNA and it functions later to promote phosphorylation of Claspin bound Chk1 by activated ATR. The challenge for future studies will be to determine the exact mechanism whereby Xmus101 performs this late function. egg extract preparation and DNA replication analysis were performed as described previously (). For the experiment shown in , cycling extracts were used, and they were prepared as described previously (). For immunodepletion of Xmus101, affinity-purified HU142 antibody was cross-linked to protein A–Sepharose beads (GE Healthcare) using the dimethylpimelimidate direct-coupling method, as described previously (). Immunodepletion was then performed as described previously (). Immunodepletion of pol α has been described (), and immunodepletion of MCM5 was performed in an analogous manner. A70 and AT70 oligonucleotides were added to 50 ng/μl of extract, and the phosphatase inhibitor tautomycin was included, at 3 μM, as previously described (). For Chk1ΔKD shift assays, a bacterially expressed recombinant Chk1ΔKD was added, to 400 nM, to extracts and then visualized on immunoblots using a T7 monoclonal antibody (Novagen). GST-Claspin CKBD shift assays were performed by adding bacterially expressed GST-Claspin CKBD (final concentration: 50 ng/μl) to extract followed by visualization with anti-GST antibody (GE Healthcare; ). The Claspin–Chk1 interaction assay was performed exactly as described previously (). Xmus101 BRCT 1–8 (FL) corresponds to full-length Xmus101 (nucleotides 1–4542) subcloned into pCS2 + MT. Mini corresponds to Xmus101 nucleotides 1–2277 subcloned into pCS2 + MT. For transcription/translation in vitro, a TNT SP6 Quick Master Mix kit (Promega) was used in all cases. Chk1ΔKD has been described (). GST-Claspin CKBD was produced by subcloning Claspin nucleotides 2609–2779 into pGEX-4T-1. Xchk1-GH has been described previously (). GST-CT333 was produced by subcloning Xmus101 nucleotides 3540–4542 into pGEX-4T-1. Chk1ΔKD was expressed as a His-tagged fusion protein in and purified over nickel NTA agarose according to standard procedures. Recombinant Xchk1-GH was produced via infection of Sf9 insect cells. Purification of Xchk1-GH over a nickel NTA agarose column was performed as described previously (). GST, GST-Claspin CKBD, and GST-CT333 were expressed in and purified over glutathione agarose according to standard procedures. The anti-Xmus101 HU142 antibody and its affinity purification have been described (). Antibodies against the p70 subunit of pol α have been described (). Antibodies against Claspin were produced by W. Dunphy (California Institute of Technology, Pasadena, CA). Antibodies against Chk1 and serine 344–phosphorylated Chk1 were obtained from Santa Cruz Biotechnology, Inc., and Cell Signaling Technology, respectively. Antibodies against MCM5 were obtained from Bethyl Laboratories. Figure S1 shows that Chk1ΔKD is a reliable surrogate for endogenous Chk1 in the AT70 checkpoint system. Online supplemental material is available at .

Excitation–contraction coupling in skeletal muscle depends on motor neuron–induced cell depolarization and the subsequent interaction between the dihydropyridine receptor (DHPR) and the ryanodine receptor (RYR), resulting in the release of Ca from the terminal cisternae of the sarcoplasmic reticulum (SR). Although much has been done in this field, studies of the quantitative aspects and kinetics of the concentration of free Ca in the SR lumen ([Ca]) have been marred by technical challenges. Most of the available data come from biochemical studies on isolated fractions (), x-ray microanalysis studies on rapidly frozen samples (), or extrapolations measuring the rise of cytosolic [Ca] ([Ca]; ). Recently, direct monitoring of [Ca] made use of the fluorescent dyes fluo-5N () or mag-indo-1 () in isolated frog muscle fibers. These approaches still suffer from major drawbacks; the subcellular localization of the dyes is not SR specific, they are difficult to apply to live animals, and, thus far, no [Ca] kinetics during excitation–contraction coupling with high temporal resolution have been determined. Cameleon Ca sensors potentially overcome most of these problems. First, as they are genetically encoded, they can be selectively targeted to subcellular compartments. Second, their ratiometric nature ensures that changes in probe quantity and movement artifacts are inherently corrected (). Third, they can be introduced into intact tissues and organisms by standard techniques (). Finally, the recent molecular engineering of cameleons have functionally silenced the two central domains (i.e., CaM and the M13 peptide), rendering these probes virtually inert as cellular signaling molecules while maintaining their Ca-sensing properties (). Using an SR-targeted cameleon and two-photon confocal microscopy in live mouse, we have addressed two unsolved issues in muscle physiology: direct quantitative measurement of the kinetics and amplitude of [Ca] transients during single twitches and tetanic stimulation, and the effect of β-adrenergic stimulation on SR Ca handling. It is known that the force of contraction can be enhanced by β-receptor agonists in both heart and skeletal muscle (). In cardiac muscle, it involves PKA-dependent phosphorylation of troponin I (), DHPR (), phospholamban (), and RYR II (). In skeletal muscle, the mechanism is less studied, but, as in the heart, it seems to rely on PKA-dependent phosphorylation of different targets, such as DHPR () and RYR I (). Regarding RYR I in particular, it is still a matter of discussion whether phosphorylation of the channel is physiologically relevant (; ). We demonstrate not only that a massive decrease of [Ca] occurs during tetanic stimulation in vivo, but also that a substantial drop is elicited even during single muscle twitches. Using Epac1–cAMP sensor (), we show the first dynamic measurement of [cAMP] in a live animal and provide direct evidence that during β-adrenergic force potentiation the [Ca] at rest, as well as the SR Ca efflux and reuptake, are markedly increased. Tibialis anterior (TA) muscles were transfected in vivo with cDNA encoding YC6.2ER or D1ER, which was targeted to the SR, as previously described (). As shown in for D1ER, the probe exhibited the typical striation pattern for SR. This pattern was always observed for D1ER, whereas YC6.2ER showed a more diffuse staining when strongly overexpressed. Data obtained with YC6.2ER was similar to that with D1ER; given the precise localization pattern of D1ER, however, only data with this probe is included in our study. depicts confocal images of longitudinal slices of muscles transfected with D1ER (, green) and immunostained against sarcoendoplasmic reticulum Ca-ATPase (SERCA) type 1 (, red). The overlay shows the colocalization of the two proteins (). This colocalization, along with the functional data revealing a drop in [Ca] upon stimulation (see the following section), shows that D1ER is correctly targeted when expressed in live mouse skeletal muscle. To quantify the drop of [Ca] upon contraction, we studied the probe response at a stimulation frequency of 1 Hz. The background-subtracted YFP/CFP ratio images were scanned along the fiber length in 10-ms windows; in parallel, fiber deflection was measured as an indicator of muscle contraction (). The fibers exhibited a twitch profile of ∼100 ms in duration, which is typical of fast-twitch fibers and is as expected for this muscle. The ratio changes followed almost exactly the kinetics of fiber deflection, with the drop in ratio clearly preceding muscle contraction. However, the nadir in [Ca], as measured with D1ER, is reached ∼30 ms after the start of contraction, which is similar to the time course of [Ca] rises that was previously measured (), but definitively slower than the time-to-peak level in [Ca] as measured with a fast Ca indicator (). This discrepancy may be attributable to a variety of factors. First, a relatively slow dissociation of Ca and conformational change of the cameleon may lead to a delay between the actual drop in [Ca] and the decrease in fluorescence resonance energy transfer. We do not consider this possibility very likely because, according to fast kinetics measurements in vitro (), the dissociation rate constant (k) of D1ER is 256 s and, thus, the indicator should be able to monitor a faster drop if it occurred; indeed, we calculated an even higher apparent dissociation constant ( ) in HeLa cells than the one previously determined in vitro (Fig. S1, available at ), but the k could not be measured in situ. Second, the signal reflects the mean of the whole SR, as it is localized not only in the terminal cisternae but also, as suggested by the colocalization of the probe with SERCA 1 (), in the longitudinal SR; in the latter, the changes in [Ca] may be smaller and delayed. Third, the kinetics shown may reflect the actual situation, indicating that a decrease of [Ca] may also continue during the decay phase of [Ca]. The maximal drops in absolute YFP/CFP ratio from the baseline differed considerably between different analyzed fibers, ranging from ∼3–25% of the initial value. The calibration of ratio signals into absolute [Ca] was derived from in situ measurements in culture cells (Fig. S1). Based on the known properties of D1ER and assuming a 200-μM for Ca within the SR lumen (which is higher than that calculated in vitro; Fig. S1), the basal [Ca] was 308 ± 30 μM and exhibited a drop during single twitches of 53 ± 6 μM (for both, mean ± SEM; = 18 fibers). We then tested the effect of partial inhibition of the SERCA, using an appropriate dose of cyclopiazonic acid (CPA) to verify the importance of the SERCA on the kinetics of [Ca] changes during stimulation. Intramuscular injection of CPA (100 μl of 100 μM CPA) caused a time-dependent drop of YFP/CFP ratio that typically stabilized ∼30–40 min after application of the drug at a value of 1.87 ± 0.17 (mean ± SEM; = 3 fibers), corresponding to ∼134 μM. Under these conditions, SERCA inhibition was only partial; we could repetitively induce fiber contraction (), whereas, as shown in skinned rat fibers, a muscle fiber may contract only few times after complete block of the SERCA (). Indeed, nerve stimulation of fibers treated with this CPA concentration revealed an amplification of contraction intensity (), albeit the resting [Ca] was significantly lower than in controls and the maximal level of [Ca] decrease was reduced (, top). In parallel experiments, we confirmed that the peak level in [Ca], as measured with cameleon YC2, is increased by this CPA treatment (unpublished data). Fiber contractions upon a single nerve impulse in the presence of CPA () were also stronger than in control conditions; similar results have been obtained in another study (). This cannot be easily explained if all Ca release from the SR was caused by electromechanical coupling, i.e., if it occurred only during the action potential. In this case, one would expect the contraction and rise in [Ca] to be the same, or smaller, in CPA-treated muscle compared with controls. Several possibilities, which are not mutually exclusive, may be suggested: a reduced cytoplasmic Ca buffering in the presence of CPA, caused by a slight increase in [Ca] (with partial saturation of [Ca] buffers); a small capacitative Ca entry might be activated under these experimental conditions (); however, we found no significant increase in basal [Ca] in the presence of CPA (unpublished data); a reduced contribution of the Ca-binding activity of the SERCA, given that SERCA bound to CPA are in a low Ca affinity conformation (); and that the massive release caused by electromechanical coupling was followed by a small, but prolonged, release of Ca (apparently, as monitored with D1ER), which is possibly a consequence of Ca-induced Ca release. A slow release could be partially counteracted by the SERCA activity and, thus, be unmasked in fibers treated with CPA. The role and existence of Ca-induced Ca release in fast skeletal muscle is surely of minor relevance (, ; ), but to our knowledge it has never been completely excluded. Our experimental setup enabled us to address an unanswered question of the utmost biological relevance: the role of the SR in β-adrenergic force potentiation. First, we directly studied the effect of isoproterenol, which is a β−adrenergic agonist, on [cAMP] by transfecting TA muscles with Epac1–cAMP sensor (). The subcellular distribution of fluorescence was homogeneous (cytoplasmic) within the fibers (). We found a highly reproducible rise in [cAMP] upon injection of isoproterenol, as shown by the increase of the CFP/YFP ratio to 1.33 ± 0.03 (mean ± SEM; = 8 fibers; data normalized to control; ). The rise was already maximal 10 min after application of the drug and decreased slightly thereafter (). Similar results were obtained with another fluorescent PKA-based cAMP probe () and upon injection of high doses of forskolin (250 μM); given the probe's ∼2-μM cAMP (), the rise in [cAMP] that is attributable to β-adrenergic stimulation is likely in the micromolar range. As shown in (bottom), the application of isoproterenol also led to a larger fiber deflection, which is indicative of the expected force potentiation. This increase in single-twitch amplitude was paralleled by four interesting phenomena. First, the basal-normalized YFP/CFP ratio increased from 1.00 ± 0.02% to 1.06 ± 0.01% (mean ± SEM; = 3 fibers) after injection of isoproterenol (, top). According to the calibration procedure described in the supplemental material, this would account for a rise in [Ca] from ∼278 to ∼311 μM. Second, the decrease of YFP/CFP ratio during single twitches was enhanced in the presence of isoproterenol without interim stimulation (, top), accounting for drops in [Ca] of ∼65 and ∼99 μM in the absence and presence of isoproterenol, respectively; this effect was even stronger upon 50 Hz stimulation, where the corresponding changes were calculated to be ∼83 and ∼153 μM. Third, the kinetics of the Ca-release/reuptake cycle during single twitches in the presence of isoproterenol had the same duration as in controls (160 ms). Thus, the rise in [cAMP], presumably via PKA-dependent phosphorylation, increases Ca release from the SR, but does not change the kinetics of release/reuptake. Finally, although the drop in [Ca] was higher in the presence of isoproterenol, the time constants from peak to half-maximal recovery were virtually identical in the absence (35 ms) and presence (36 ms) of the adrenergic stimulus. Thus, during the same time period more Ca is taken up by the SR in fibers stimulated by isoproterenol compared with controls. In summary, the data presented show that stimulation of β-adrenergic receptors in skeletal muscles leads to a major and prolonged rise of [cAMP] accompanied by an increase of basal [Ca] and a drop in [Ca] during contraction. The increased fiber contraction is clearly dependent on the larger Ca release because previous work has shown that, unlike in cardiac fibers, cAMP-dependent PKA activation does not modify the Ca sensitivity of the contractile apparatus (; ). In particular, only myosin-binding protein C and troponin I of cardiac muscle may be phosphorylated by PKA (). Regarding the effect of isoproterenol on Ca release, different, nonmutually exclusive explanations may be considered. First, a higher basal [Ca] might account for a larger release of Ca during contraction. However, no correlation between basal [Ca] and the drop during contraction was observed; neither by ourselves (Fig. S2, B and C, available at ) nor in a previous study (). Second, the increased Ca efflux could be attributable to direct cAMP/PKA-dependent activation of RYR I; this possibility has been previously proposed (; ), but no convincing evidence has been provided. Third, and most likely, it could be caused by PKA-dependent phosphorylation of DHPR that increases its coupling to RYR I. Strong evidence supporting this mechanism has been obtained in myoblasts (). The effect of β-adrenergic stimulation on Ca accumulation was unexpected for fast skeletal muscle, which is known to be devoid of phospholamban (). Evidence supporting a role for cAMP in SERCA stimulation has been described in the skinned fibers of cat (), but this finding has been neglected for almost 30 yr. A simple, conservative explanation may be an indirect effect of cAMP/PKA on SERCA activity, mediated by PKA activation of glycogen metabolism (); both pyruvate () and local ATP generation by SR-bound glycolytic enzymes appear particularly important for the activity of SERCA (). Increased availability of glucose (and, thus, ATP) may explain the higher Ca loading of the SR upon β-adrenergic stimulation. Transfection experiments used the following constructs in pcDNA3 (Invitrogen): YC2, D1ER, and Epac1–cAMP sensor were gifts from R.Y. Tsien (University of California, San Diego, San Diego, CA) and M.J. Lohse (University of Würzburg, Würzburg, Germany). The A23187, ascorbic acid, CPA, digitonin, histamine, isoproterenol, and thapsigargin (all from Sigma-Aldrich) used were of the highest available grade. All injected drugs and cDNAs were diluted in sterile physiological solution (0.9% NaCl). SERCA 1 antibody (clone VE121G9) was purchased from Affinity BioReagents and Alexa Fluor 568 goat anti–mouse IgG was purchased from Invitrogen. All experiments used C57BL/10 mice (aged 6–12 mo; Charles River Italia). Animal handling was approved by the local authority for veterinary services, in accordance with Italian law. For anesthesia, Rompun (Bayer) and Zoletil 100 (Laboratoires Virbac) were injected i.p. Transfection was carried out using an electroporation-based method of the TA muscle, as previously described (). For immunostaining, muscles were first stretched, fixed overnight at 4°C in 4% paraformaldehyde in 0.1 M phosphate buffer (PB), and dehydrated in 10% sucrose/PB; they were then snap frozen in liquid nitrogen–cooled isopentane. The muscle was embedded in Jung tissue-freezing medium and 10-μm-thick cryosections were prepared using a cryostat (model CM1850; Leica). After drying, the sections were quenched with 50 mM NHCl/PB and permeabilized with 0.1% Triton X-100/PB. Sections were then rinsed with PB, blocked with 0.2% gelatin/PB and incubated with anti-SERCA 1 antibody (1:500, diluted in 0.2% gelatin/PB) overnight at 4°C. Sections were blocked again with 0.2% gelatin/PB and incubated with the secondary antibody (1:250, diluted in 0.2% gelatin/PB) for 2 h. Finally, the preparations were rinsed in 0.2% gelatin/PB and PB and mounted in Mowiol. Confocal images of immunostained sections were obtained with a microscope (Radiance 2100MP; BioRad Laboratories) equipped with an Ar (488-nm line) laser and a HeNe laser (543-nm line), a 60×/1.4 NA objective (Nikon), a 560DCLPXR dichroic mirror, HQ515/30 and E570LP emission filters (Chroma Technology Corp.) for the detection of cameleon, and Alexa Fluor 568 fluorescence signals. Images were taken at 1,024 × 1,024 pixel resolution and 50 lines s scan speed. This procedure was performed as previously described (), with minor modifications. Unless otherwise indicated, video microscopy was performed at 256 × 256 pixels with a 1.95-Hz acquisition frequency. For [Ca] studies, images were acquired at 1,024 × 1,024 pixels at 166 lines s. Drug injection (1 mM CPA, 100 μM CPA, or 10 μM isoproterenol + 1 mM ascorbic acid) was performed locally using a 30-gauge needle (Artsana) with a typical volume of 100 μl. To allow the muscle to recover from the injection-induced swelling, microscopic observation was interrupted for at least 5 min. Datasets were analyzed essentially as previously described (), using ImageJ software (National Institutes of Health). Fibers were oriented in the microscopic field with the longitudinal axis perpendicular to the line scan direction, matching the timeline to the long fiber axis. A mask containing the whole observed part of the fiber was created from the corresponding median-filtered YFP video by intensity thresholding. Next, both CFP and YFP videos were mean filtered (1 pixel kernel) to suppress hot pixel noise, background intensity was subtracted when necessary, and the mask was applied. For the data shown in Fig. S1, the mean CFP and YFP values were determined inside the mask region; subsequent calculations and graphs used Excel 2002 (Microsoft). Ratio plots were normalized using the last 10 images preceding the first contraction as reference values. , were obtained as follows. A YFP/CFP ratio video was created by applying the custom-made ImageJ plug-in Ratio Plus. Ratio values were then determined along the fiber length in regions of interest (ROIs) with a width of 5 pixels, equaling 10-ms time frames. is the mean ratio value at point a. In , is equal to , whereas in and it is equal to of the corresponding controls. In parallel, the relative fiber deflection was determined using the last frame before the first contraction as a blank. , individual twitches were synchronized by setting the first ROI with deflection as the start of contraction. For the timelines shown in , subsequent frames were treated as continuous in time; this assumption is valid because the time-lapse between the last scan line of an image and the first one of the following frame was of the same order as the jump from one line to the next in the same frame, and because contractions that initiated in one frame and continued in a following frame, appeared smooth in procedure after recomposition, as executed in . Experiments showing the functionality and correction for movement artifacts of D1ER as well as a detailed description of D1ER in situ calibration are available online. Online supplemental material is available at .

To avoid deleterious consequences of DNA damage, eukaryotic cells activate a signaling network that coordinates rapid detection of the DNA lesions with temporary delay of cell cycle progression and activation of repair machinery (). One important aspect that determines the effectiveness of these genome surveillance pathways is a carefully orchestrated redistribution of their components to nuclear regions containing the damaged DNA (). The cytological manifestation of nuclear rearrangements in response to ionizing radiation (IR) and/or radiomimetic drugs is the formation of the so-called IR-induced foci (IRIF; ). IRIF are dynamic, microscopically discernible structures containing thousands of copies of proteins involved in various aspects of double-strand break (DSB) metabolism. As such, IRIF are widely used as a convenient marker of DSB location. Apart from various repair-associated DNA transactions (), proteins associated with IRIF also participate in restructuring of large segments of chromatin in the vicinity of the DNA lesions (), thereby increasing the accessibility of damaged DNA to the repair factors (). In addition, sustained protein assembly in the DSB-flanking chromatin seems to be required to preserve the integrity of the epigenetic information encrypted in these regions (; ). Despite the general consensus that IRIF formation signifies an important step in cellular protection against the deleterious effects of DSB-generating insults, the question of how the genome surveillance pathways actually benefit from the increased local concentration of their regulators remains poorly understood. For instance, recent results unmasked an unexpected level of complexity by showing that Chk2, the integral component of the genome surveillance machinery, interacts with DSBs only transiently, without forming cytologically discernible foci. The existence of such a “cryptic” mode of Chk2–DSB interaction indicates that some enzymatic transactions associated with the DNA damage response do not strictly require massive protein concentration at the DSB sites. In addition, relatively little attention has been paid to whether all damage-induced foci, once formed, are structurally similar. Furthermore, it is unclear to which extent the protein assemblies at the DSB sites vary during the cell cycle progression. We performed a systematic survey of protein redistribution after a defined DSB-generating insult and under standardized experimental conditions. Our intention was to subclassify the key components of the DSB network (sensors, signaling components, mediators, repair factors, and checkpoint effectors) according to the residence sites after DNA damage and to determine whether and how the protein assemblies at the DSB sites fluctuate during the cell cycle. To obtain standardized experimental conditions, we generated DSBs by laser microirradiation (). The key advantage of the microlaser approach is its ability to target defined nuclear volumes and generate a similar amount of DNA damage both in different areas in the same nucleus and in distinct nuclei within the microirradiated cell population (; ). Although several laboratories have used this approach to study various aspects of the DNA damage response (for review see ), we initiated this study by revisiting a key parameter that must be considered when interpreting redistribution of proteins after genotoxic insults: the type and amount of the DNA lesions. It has been shown that, like IR, the laser-induced DNA damage produces single- and double-strand DNA breaks and base modifications (). However, the relative distribution and the density of these chromosomal alterations are hugely influenced by the laser type and energy output and by the type of photosensitizers. Our choice of a laser line within the UV-A spectrum (λ = 337 nm) and sensitization of cells with halogenated thymidine analogues has been dictated by the need to generate experimental conditions where we can still benefit from targeting the damage to the defined nuclear volumes, yet induce protein redistribution that both qualitatively and quantitatively resembles that of the “classical” IRIF. To rigorously test this approach, we took advantage of the fact that a subset of DSBs in mammalian cells is repaired by homologous recombination () and that these lesions could be identified by detection of the single-stranded DNA (ssDNA) intermediates coated by the replication protein A (RPA; the RPA detection method used here recognizes selectively long stretches of ssDNA that result from enzymatic DSB resection). Because these RPA-coated ssDNA regions manifest as distinct, microscopically discernible foci that could be counted, we used this approach to quantify the amount of homologous recombination–repaired DSBs generated by the microlaser and IR, respectively. We exposed cells to the microlaser and/or increasing doses (0–10 Gy) of IR, detected the ssDNA foci by immunostaining with an antibody to RPA, and recorded a series of 3D images to ensure detection of all RPA foci in the entire nuclear volumes. Examination of these images revealed that both laser and IR produced RPA foci of a similar size (, left) and, importantly, that the total amount generated by the microlaser (under settings used for all experiments in this study) was very similar to that generated by 3 Gy of IR (, right). Because the biological effect of DSBs is a combination of their absolute amount and their relative density in a given nuclear volume, we set out to also assess the local impact of both laser microirradiation and IR. We first determined the size of the region spanning the entire nuclear volume exposed to the microlaser (, top left) and then placed the same region over the maximum nuclear diameter of cells exposed to increasing doses of IR (, right; note that these images are also 3D projections). Subsequent counting of the RPA foci revealed that the DSB density in the microlaser tracks was similar to that produced in comparable nuclear volumes of cells exposed to 10 Gy of IR (). This approach allowed us to locate the microlaser response to a dose range of between 3 and 10 Gy of IR. To further refine the estimate of the damage extent generated by the microlaser, we directly assessed its impact on p53 phosphorylation on Ser15. This DSB-induced, ataxia telangiectasia–mutated kinase (ATM)–mediated phosphorylation event was chosen because of the unique capacity of p53 to become targeted by activated ATM without the sustained focal accumulation of either of these proteins at the DSB sites (). As a result, cells treated with a DSB-generating insult respond by a homogeneous increase of phosphorylated p53 throughout the entire nucleus (, left), thereby providing a sensitive surrogate for the extent of DNA damage induced by diverse DSB-generating stimuli. Quantitative measurement of the immunofluorescent signal associated with Ser15 phosphospecific antibody revealed that the amount of DNA damage delivered by the microlaser was comparable to that generated by IR in a dose range of between 4 and 6 Gy (, right). First, a group of proteins assembled within the entire regions of modified chromatin that surrounds the DNA breaks and spans up to a megabase distance from the initial DSB lesion (). The diagnostic cytological manifestation of this pattern was a complete colocalization with the γ-H2AX–decorated chromatin and could be best illustrated on the example of the DSB-induced redistribution of Mdc1 and 53BP1, both in microlaser-generated DSB tracks () and in IRIF induced by moderate doses of IR (Fig. S1 A, available at ). Interaction of both proteins with the DSB-modified chromatin was also reproduced in primary human fibroblasts (Fig. S2) and in living cells expressing GFP-Mdc1 and GFP-53BP1 (see ). The observed spreading of these so-called checkpoint mediators throughout the entire DSB-flanking “microenvironment” is consistent with the recent findings describing the ability of Mdc1 and 53BP1 to interact with posttranslationally modified histones (; ; ; ). Other proteins in this category include the ATM kinase and the components of the Mre11–Rad50–Nbs1 nuclease complex (MRN; ; see for Nbs1). Interaction of all these proteins with the DSB-flanking chromatin makes sense in light of the recent discoveries. We have previously shown () that the retention of MRN at the DSB sites required direct binding of its Nbs1 component to Mdc1, the latter being the main recognition module of γ-H2AX within the DSB-flanking chromosomal microenvironment (). In addition, it has been reported that the recruitment of activated ATM to the DSB sites is mediated via the COOH-terminal region of Nbs1 (; ). Thus, the ability of Nbs1 to bind Mdc1 on one hand and Nbs1's potential to recruit ATM on the other provide a basis for a large-scale concentration of these factors around the DSB-containing chromosomal lesions. An important denominator shared by all proteins in this category is their ability to assemble at the DSB-flanking chromatin throughout most of the cell cycle. We have consistently seen that all of the ∼200 cells microirradiated in each experiment responded by a robust accumulation of these proteins in the microirradiated tracks (unpublished data). Such a uniform response did not indicate cell cycle–dependent interactions. More specifically, Mdc1 (; see ) and other proteins described in this section (unpublished data) readily accumulated in the microirradiated nuclear regions in cells with various intensities of cyclin A or B1 (cells in S and G2), as well as in cells lacking a detectable amount of both cyclins (cells in G1). Together, these data suggest that the assembly of proteins at the DSB-flanking chromatin can occur throughout the interphase. Another group of proteins assembled in much smaller areas (“microfoci”) located to the center of the microirradiated tracks and surrounded by relatively vast regions of γ-H2AX–modified chromatin (). This spatial pattern could be resolved on the standard IRIF level (Fig. S1 A), it could be observed in primary human fibroblasts (Fig. S2), and it is discernible in living cells expressing GFP-tagged proteins (see ). We also verified that the cytological appearance of these microfoci was not affected by various fixation protocols, the relative affinity of antibodies, and/or the image acquisition conditions (Fig. S1 B; unpublished data). As the proteins in this category typically include nuclear factors involved in DNA repair by homologous recombination (Rad51, Rad52, BRCA2, and FANCD2), these microfoci likely represent either individual DSBs or closely neighboring DSBs assembled in repair centers (). The same compartments are occupied by the ATR kinase and the protein machinery (ATR interacting protein [ATRIP] and RPA) required for ATR's assembly at the DSB sites (). Finally, Rad17 and -9, both components of the proliferating cell nuclear antigen–like sliding clamp loaded on DNA after DNA damage, also show similar localization patterns (). Two prominent features discriminate these DSB-induced microcompartments from the chromatin-mediated interactions. First, accumulation of all proteins in this category is restricted to areas of ssDNA formed after resection of the initial DSB lesions (). This is illustrated by a close overlap of FANCD2 with ssDNA (), the latter structure being revealed by immunostaining of BrdU without previous DNA denaturation (; see Materials and methods). Second, not all microirradiated cells were able to generate detectable stretches of ssDNA (and the corresponding focal protein accumulation), despite the fact that all such cells responded by an equally robust phosphorylation of H2AX () and/or Mdc1 assembly (). Indeed, coimmunostaining experiments revealed that the interaction of FANCD2, Rad51, and other proteins from this category with the ssDNA microcompartments could only be detected in cells that expressed cyclin B1 and/or A ( and ). These findings are consistent with the recent study showing that the DSB resection is cell cycle dependent () and suggest that, unlike the chromatin-mediated interactions, accumulation of proteins in the ssDNA microcompartments is temporally restricted to S and G2 phases. The latter conclusion is also consistent with a report showing that the recruitment of Rad51 to the laser-generated DSBs and/or IRIF requires postreplicative DNA () and with experiments in yeast that failed to detect Rad51 in IRIF during the G1 phase (). On the other hand, we note that a recent study reported accumulation of Rad51 in laser-damaged nuclei also during G1 (). However, these authors used a relatively high laser energy in unsensitized cells, resulting in visible morphological destruction of the microirradiated nuclear areas, a condition that could have overpowered the physiological restrictions for Rad51 to attempt DSB repair in the context of prereplicative chromatin (see also the following sections). The components of the MRN complex share the striking feature to interact with both the DSB-flanking chromatin () and the microfoci described in the previous section (). The ability of Nbs1 to interact with the latter compartments could be revealed by its ability to form microfoci even in cells with reduced levels of H2AX, that is, after disruption of the key step in forming the DSB-induced chromatin microenvironment (). These data extend our earlier observation that a very similar pattern of Nbs1 redistribution was observed in cells where the integrity of the DSB-flanking chromatin was disrupted by the down-regulation of the γ-H2AX–binding protein Mdc1 (). To determine whether these microfoci represent the ssDNA compartments, we microirradiated Mdc1-deficient cells and coimmunostained Nbs1 with antibodies to RPA and BrdU (the latter without previous DNA denaturation). Indeed, both of these ssDNA markers showed tight colocalization with a fraction of Nbs1 that assembled in the microirradiated tracks under these conditions (). Consistent with the cell cycle–dependent formation of the ssDNA compartments, the retention of Nbs1 in these microfoci was much more pronounced in cells capable of resecting the primary lesions and generating cytologically discernible stretches of ssDNA (, bottom). Thus, in addition to its ability to assemble in the large DSB-flanking chromosomal regions, Nbs1 can interact with the ssDNA microcompartments in a chromatin-independent manner. This is consistent with a study detecting an extraction-resistant pool of Mre11 in the center of γ-H2AX–coated chromosomal domains () and with the recently reported causative role of MRN in the formation of the ssDNA microcompartments (). Furthermore, the ability of Nbs1 to assemble in the ssDNA microcompartments can explain the H2AX-independent accumulation of the MRN components in early stages of the DSB response (). The only other protein capable of simultaneous interaction with both ssDNA compartments and the DSB-flanking chromatin is BRCA1. After laser microirradiation, BRCA1 clearly spreads throughout the entire chromatin regions marked by γ-H2AX and/or by retention of typical chromatin binding proteins, such as 53BP1 (, A and B, top). Like in all other chromatin-specific interactions described in the previous sections, this pattern of BRCA1 redistribution could be detected throughout the interphase, although the amount of BRCA1 retained in the microirradiated G1 cells is less pronounced because of the lower abundance of the total BRCA1 protein in this cell cycle stage (). Importantly, although disruption of this compartment by Mdc1 down-regulation triggered dissociation of BRCA1 from the DSB-flanking chromatin, it did not impair a productive assembly of BRCA1 at the subchromatin microfoci (, A and B, bottom). Additional evidence for the chromatin-independent role of BRCA1 comes from the observation that reducing BRCA1 levels by siRNA precluded assembly of important repair factors (BRCA2 and Rad51; both are known to function downstream of BRCA1) at the ssDNA microcompartments (Fig. S3 D, available at ). Collectively, these results add an important spatial dimension to the emerging functional interplay between Mdc1 and BRCA1 (; ) by showing that it is the chromatin bound (but not the ssDNA associated) fraction of BRCA1 whose retention at the DSB sites is controlled by Mdc1. It is important to emphasize that the complex interaction pattern of Nbs1 and BRCA1 described in this section is quite unique, likely reflecting the central position of the MRN complex and BRCA1 in DSB recognition and signaling (; ). For instance, uncoupling of the typical chromatin binding proteins 53BP1 (Fig. S3 A) and Mdc1 (Fig. S3 B) from their respective histone residues was sufficient to completely abrogate their ability to accumulate at the DSB sites, despite the fact that, under the same experimental conditions, the ssDNA compartments were clearly formed (, B and C; and ). Furthermore, although down-regulation of Mdc1 impaired stable interaction of 53BP1 with the DSB-flanking chromatin (; ), it did not prevent productive assembly of ATRIP and Rad51, the typical components of the ssDNA compartments (Fig. S3 C). Thus, the majority of the DSB regulators studied so far tend to interact rather exclusively with either ssDNA or the DSB-flanking chromatin, and the retention of proteins in these respective compartments seems to be regulated by mutually independent mechanisms. Interestingly, one study () reported that cells deficient in H2AX (a chromatin component) are not able to retain ATR (the ssDNA component) at stalled replication forks. Thus, the rather strict spatial subcompartmentalization of checkpoint regulators described here might be specific for DBSs, likely because of the complexity of the DNA and chromatin rearrangements required for efficient repair of these serious chromosomal lesions. Some proteins intimately involved in the genome surveillance network do not visibly accumulate at the damage sites, a striking phenomenon that we have previously described for the Chk2 kinase (). The new addition to this spatial category provided here is Chk1 (, top; see for GFP-Chk1 in living cells). The inability of Chk1 to form cytologically discernible foci was not caused by the lack of its activation after laser microirradiation because the microirradiated cells readily induced phosphorylation of Chk1 on Ser317 (, bottom), one of the ATR target sites whose phosphorylation accompanies activation of Chk1 (). An important aspect of this spatial pattern is that phosphorylated Chk1 () and Chk2 () do not concentrate around the DNA damage sites but rapidly spread to the entire nucleus. Together with the evidence that Chk1 phosphorylation by ATR requires physical interaction of these two components directly at the sites of damaged DNA and/or stalled replication forks (; ), these results indicate that Chk1 (like Chk2) facilitates signal transduction between focal DNA lesions and relatively immobile effector structures (replication origins, stalled replication forks, and gene promoters) elsewhere in the undamaged parts of the nucleus. The distinction between Chk1 and -2 in this respect is that although Chk2 could be activated throughout the interphase (), Chk1 function is temporally limited because of the fact that the ssDNA formation and ATR signaling is restricted to S/G2 phases of the cell cycle (). Several other proteins known to function on various levels of the DSB-induced signaling did not readily accumulate at the DSB sites (). The inability of DNA-PK, Ku70, Smc1, and Smc3 to form cytologically discernible foci was not restricted to a single time point (we failed to detect increased local accumulation of any of these proteins between 5 min and 8 h after DNA damage); neither was it influenced by the cell type, DSB insult, fixation, and/or imaging conditions (, , S1, and S2). However, we note that the DNA damage–induced redistribution of some of these factors had been studied before and produced conflicting results. We therefore set out to critically reexamine some of these cases and explain the discrepancies. On one hand, our inability to detect cytologically detectable accumulation of DNA-PK and its regulatory subunit Ku70 both in locally microirradiated cells ( and A) and after global exposure to the commonly used dose range of IR (Figs. S1 A and S2) is consistent with the results obtained by , who generated DSBs by irradiating cells with charged ion beams. On the other hand, another group has reported local accumulation of DNA-PK in nuclei exposed to Nd:YAG laser (). To find the reason for these differences, we tried to modify our experimental conditions and test whether any of those would be able to locally increase the concentration of DNA-PK/Ku70 to the levels detectable by light microscopy. Neither variation of the laser dosage in sensitized cells nor the extension of the assay period from several seconds to many hours produced such results, despite generating clear and robust DSB response, including induction of γ-H2AX and recruitment of the DNA-PK–related ATM and ATR kinases (see and for examples). Only a substantial increase of the laser output in nonsensitized cells was able to induce weak but detectable accumulation of DNA-PK/Ku70 at the microirradiated tracks (, left). However, such treatment also produced massive local damage of the overall nuclear structure (never seen after microirradiating the sensitized cells by moderate laser doses and/or after IR), manifest, for instance, by a decrease of DNA staining within the microirradiated areas (, inset). Similar signs of general nuclear disruption (areas with markedly altered optical density) were observed also in the previous study describing local accumulation of DNA-PK/Ku70 (). Interestingly, the same group applied these assay conditions to demonstrate local accumulation of Smc1, the structural component of the multiprotein cohesin complex (). Also in this case, we were able to reproduce these results but, again, only after exposing the unsensitized cells to very high laser energy that was accompanied by local disruption of the overall nuclear structure (, right). Neither IR (up to 10 Gy; Fig. S1) nor moderate laser microirradiation in sensitized cells () showed any signs of cytologically discernible accumulation of Smc1. Similar results (no accumulation after moderate doses of IR and/or laser) were obtained with GFP-Smc3, the heterodimerizing partner of Smc1 ( and Fig. S4, available at ). Hence, despite both DNA-PK/Ku70 and Smc1 being integral components of various facets of DBS response (; ; ; ; ; ), local accumulation of these proteins to a degree that could be resolved by light microscopy seems to require enormous local concentration of DSBs. Indeed, when we tested the conditions compatible with Ku70 and/or Smc1 recruitment to DSBs by the same assay described in , we observed that the nuclear areas exposed to the high laser energy were unable to resolve clear RPA foci. Instead, these regions showed signs of a uniform and strong RPA accumulation, indicating extremely high density of DSBs (). At the same time, the relative abundance of the typical chromatin binding protein, such as 53BP1, was reduced in these regions compared with moderate laser doses in sensitized cells (, right), indicating that the high laser energy output not only generates a massive DNA damage by itself but also triggers local destruction of histones and/or the other chromatin-associated protein complexes. Such extreme density (and complexity) of chromosomal damage likely saturates the cellular capacity to repair the lesions, thereby generating conditions that may stabilize (or aggregate) the templates for the assembly of proteins that specifically interact with DNA ends (). Based on these results and considerations, we propose that the assembly of the DNA-PK/Ku70 holoenzyme and loading of the Smc1/3-containing cohesin complex is spatially restricted to unprocessed and/or only partially processed DNA breaks and that, within a range of physiologically relevant doses of DSB-generating insults, these microcompartments are beyond the resolution of light microscopy and do not manifest as cytologically discernible nuclear foci. Importantly, the latter conclusion does not contradict a recent report describing IRIF containing the ATM-phosphorylated Smc1 (). Indeed, we were able to reproduce these results by detecting local phosphorylation of Smc1 on serine 957 (one of the key ATM target sites) after moderate doses of laser microirradiation that did not cause any discernible accumulation of the total Smc1 protein (, compare the two bottom panels). Thus, it appears that in addition to cohesin, which assembles at the DSBs de novo (; ), and the amount of which (at least in mammalian cells) is likely limited below the levels that could be resolved by light microscopy, there must be a sizable pool of Smc1 throughout the undamaged chromatin as a result of the physiological processes accompanying DNA replication. Indeed, loading of the vertebrate cohesin on chromatin is triggered by formation of the prereplication complexes, an event that occurs early during the cell cycle (telophase in cycling cells and G1 after stimulation from quiescence; ). The spatial pattern of the Smc1 response to DNA damage reported here (readily detectable local phosphorylation without discernible increase of the total protein concentration) indicates that after DSB generation, this prereplication complex–loaded fraction of cohesin remains bound to chromatin and becomes locally accessible to the ATM-mediated phosphorylation. In this regard, it is important to note that phosphorylated Smc1 spreads throughout the entire regions of γ-H2AX–decorated chromatin (, bottom). This is consistent with our findings that all three components required for the DSB-induced Smc1 phosphorylation—ATM, Nbs1, and BRCA1 (; )—avidly interact with this DSB-generated nuclear subcompartment (, , and ). Some integral components of the DSB-induced genome surveillance network did not accumulate at the damaged areas under any of the conditions explored in this study, including the high-energy laser illumination described in the preceding section (; ). The key proteins in this category, p53 and Cdc25A, share their functional position within the DSB network by serving as the key effectors of the DNA damage–induced genome surveillance pathways (). Their lack of direct physical engagement with DSBs indicates that the efficient DSB-induced gene expression (p53) and cell cycle arrest (Cdc25A and p53) requires a specific signaling component capable of rapid and efficient connection of these effector molecules with the focal DNA lesions. The spatial properties of activated Chk1, Chk2 (), and ATM () render these kinases the most plausible candidates for a “messenger” function. Our study provides evidence that after DSB-generating insults, a mammalian nucleus undergoes a complex compartmentalization reflected by distinct patterns of protein redistribution. The essence of our results is summarized in . For the sake of clarity, the key implications of how the residence sites of the studied proteins help us better understand their roles in the DSB response were systematically discussed while describing the individual spatial categories in the preceding sections. We would like to complement these specific conclusions with more general and conceptual ramifications of the reported results. In particular, we would like to emphasize that among the diverse modes of protein redistribution after a DSB-generating insult, only the proteins assembled in the DSB-flanking chromatin regions and the ssDNA microcompartments could be readily detected as intranuclear foci. An important implication of this finding is that within the range of physiologically relevant doses of DNA damage, the IRIF formation (and/or the protein assembly at the laser-damaged nuclear tracks) cannot serve as the only criterion for the direct involvement of a given protein in the DSB response. This conclusion is supported by adding several new members to the expanding family of proteins that, although functionally distinct, share the capability of a productive interaction with DSBs without a massive increase in their local concentration. Thus, in addition to signaling kinases whose interaction with the DSB sites is too transient to manifest as foci (Chk1 and -2), other proteins in this category assemble at DSB intermediates whose size is below the resolution of light microscopy (DNA-PK/Ku70). In addition, Smc1 is likely just one example of a larger group of proteins that stably interact with chromatin even in undamaged nuclei and yet become engaged in the DSB signaling and/or repair after local modification by enzymes recruited to the sites of DNA damage. In conclusion, we hope that the results reported here will provide the necessary framework to add more proteins to the emerging “spatial map” of the DSB-induced genome surveillance network. If coordinated in terms of experimental conditions, subclassification of proteins according to their residence sites before and after DNA damage may help validate, predict, or even exclude their roles in the increasingly complex DSB response. One example illustrating the potential usefulness of the spatial dimension in approaching some lingering questions in the field is the redistribution of the key apical kinases induced by DSBs. Most notably, ATM, ATR, and DNA-PK each occupy distinct nuclear subcompartments (DSB-flanking chromatin, ssDNA microcompartments, and unprocessed DSB ends, respectively). Because deficiency of the respective kinases is accompanied by distinct phenotypes (), it is clear that these kinases, despite sharing several downstream substrates, have limited capability to substitute each other. Although the regulatory network determining the exact function of each of these kinases is very complex, their relocation to distinct compartments after DNA damage can, at least partly, explain their overlapping versus nonoverlapping potential. The U2OS cell line and the BJ fibroblasts were seeded onto glass coverslips (Menzel) and grown in DME supplemented with 10% fetal bovine serum and standard antibiotics. The U2OS-derived cell lines stably expressing GFP-Mdc1, -53BP1, -ATR, and -Chk1 were described previously (; ; ; ). The YFP-Rad52 construct was generated by subcloning the Rad52 cDNA (a gift from R. Kanaar, Erasmus Medical Centre, Rotterdam, Netherlands) to the pEYFP-N expression plasmid (CLONTECH Laboratories, Inc.). Generation of the GFP-Smc3 plasmid is described in detail in Fig. S4. The U2OS cell lines stably expressing YFP-Rad52 and GFP-Smc3 were generated by cotransfecting the respective expression plasmids together with the pBabe-puro plasmid containing the puromycin resistance cassette. Upon selection with 1 μg/ml puromycin (Sigma-Aldrich) for 10 d, resistant clones were tested for the expression and functionality of the GFP/YFP-tagged proteins (see Fig. S4 for characterization of the GFP-Smc3–expressing cells). Laser microirradiation to generate DSBs in defined nuclear volumes was performed essentially as described previously (; ; ). In brief, the culture medium was supplied with 10 μM BrdU (Sigma-Aldrich) for 24 h to sensitize the cells to DSB generation by UV-A laser (λ = 337 nm). Before laser treatment, the coverslips were transferred to a phenol red–free CO-independent medium (Invitrogen). After microirradiation of ∼200 cells (a procedure lasting in total for <10 min), the coverslips were incubated for 1 h in the incubator before fixation. IR was delivered by an x-ray generator (HF160 [Pantak]; 150 kV; 15 mA; dose rate: 2.18 Gy/min) as previously described (). siRNAs against H2AX and Mdc1 were described previously (). The control siRNA was against HSP70B (). Cells were transfected with the siRNA duplexes with oligofectamine (Invitrogen) according to the manufacturer's instructions and incubated for 96 h before further treatment. For generation of stable Mdc1 knockdown, the oligonucleotides 5′-gatccccgtctcccagaagacagtgattcaagagatcactgtcttctgggagacttttt and 5′-agctaaaaagtctcccagaagacagtgatctcttgaatcactgtcttctgggagacggg were annealed and ligated into the pSUPER plasmid digested with HindIII and BglII. For generation of stable 53BP1 knockdown, the oligonucleotides 5′-gatccccgaacgaggagacggtaatattcaagagatattaccgtctcctcgttcttttt and 5′-agctaaaaagaacgaggagacggtaatatctcttgaatattaccgtctcctcgttcggg were annealed and processed as above. The resulting constructs were transfected into U2OS cells and selected with Puromycin as described previously (). Rabbit polyclonal antibodies against the following targets were used: Smc1 (Abcam), Smc3 (Abcam), Smc1-S957P (Novus Biologicals), p53-S15P (Santa Cruz Biotechnology, Inc.), TRF2 (Santa Cruz Biotechnology, Inc.), ATRIP (a gift from R. Abraham, The Burnham Institute, La Jolla, CA), Rad17 (Santa Cruz Biotechnology, Inc.), Rad9 (Santa Cruz Biotechnology, Inc.), Rad51 (Santa Cruz Biotechnology, Inc.), FANCD2 (Novus Biologicals), Nbs1 (Novus Biologicals), ATM (Abcam), 53BP1 (Santa Cruz Biotechnology, Inc.), γ-H2AX (Upstate Biotechnology), cyclin A (Santa Cruz Biotechnology, Inc.), and Chk1-S317P (Cell Signaling). The following mouse monoclonal antibodies were used: DNA-PK (Lab Vision), Ku70 (Lab Vision), Chk1 (DCS-310; Abcam), RPA p32 (Lab Vision), BRCA2 (Calbiochem), Mre11 (GeneTex), BRCA1 (Santa Cruz Biotechnology, Inc.), γ-H2AX (Upstate Biotechnology), and cyclin B1 (Santa Cruz Biotechnology, Inc.). Sheep anti-Mdc1 antibody was a gift from S. Jackson (Wellcome Trust/Cancer Research UK Gurdon Institute, Cambridge, UK) and M. Stucki (University of Zürich, Zürich, Switzerland). Anti-BrdU mouse monoclonal antibody (RPN20AB) to detect ssDNA was obtained from GE Healthcare and was applied without any preceding DNA denaturation or nuclease treatment (). Other immunostaining steps were identical to those described in the following section. Cells were fixed for 15 min in 4% formaldehyde and permeabilized in 0.2% Triton X-100 for 5 min. To facilitate discrimination of the chromatin-associated versus ssDNA bound pools of Nbs1 (), cells were preextracted for 5 min at 4°C with a buffer containing 25 mM Hepes, pH 7.5, 50 mM NaCl, 1 mM EDTA, 3 mM MgCl, 300 mM sucrose, and 0.5% Triton X-100 as described previously (). Coverslips were incubated with primary antibodies for 1 h followed by secondary antibodies coupled to Alexa 488, 568, or 647 (Invitrogen) for 30 min. Where indicated, the DNA stain ToPro3 (Invitrogen) was added to the last washing solution. Coverslips were mounted onto glass slides (Menzel) with DAPI-containing mounting medium (Vector Laboratories) and subject to two- or three-color confocal microscopy on an LSM-510 (Carl Zeiss MicroImaging, Inc.) mounted on an Axiovert 100M (Carl Zeiss MicroImaging, Inc.) equipped with Plan-Neofluar 40×/1.3 oil-immersion objective, as previously described (). For quantitative assessment of the DNA damage–induced p53 phosphorylation, masks were manually drawn around the individual nuclei, and the mean fluorescence associated with antibody to phosphorylated serine 15 of p53 (S15-P) subtracted for the background fluorescence was determined using the ImageJ software (NIH). The obtained values were exported to Prism4 (GraphPad) software for further data processing. Fig. S1 shows the spatial patterns of DSB-induced protein redistribution resolved on the level of the IRIF and provides evidence that the DSB-induced redistribution of proteins to distinct nuclear subcompartments was assayed under unsaturated conditions for image acquisition. Fig. S2 shows the main spatial patterns of DSB-induced protein localization in primary cells (the BJ strain of human diploid fibroblasts). Fig. S3 provides evidence for autonomous protein interactions with distinct DSB-generated nuclear subcompartments. Fig. S4 describes generation and characterization of the GFP-Smc3 cohesin subunit. Online supplemental material is available at .

The 100–200 H/ACA RNPs of each mammalian cell affect several basic functions, such as protein synthesis, gene expression, and chromosome stability. They do so by catalyzing site-specific pseudouridylation of ribosomal and spliceosomal small nuclear RNAs, by processing ribosomal RNAs, and by stabilizing telomerase RNA. In addition to C/D RNPs, H/ACA RNPs form one of the two major classes of small nucleolar RNPs (snoRNPs) functioning in nucleoli and Cajal bodies. They consist of four evolutionarily conserved core proteins and one H/ACA RNA (; ; ; ; ). The four core proteins—the pseudouridylase NAP57 (dyskerin; Cbf5p in yeast), GAR1, NHP2, and NOP10—are capable of forming a protein-only complex (). Specifically, NOP10 and GAR1 bind independently to NAP57, whereas docking of NHP2 onto NAP57 is mediated by NOP10, thereby forming the core trimer NAP57–NOP10–NHP2 (noncovalent interactions are indicated with an en-dash). These interactions are conserved in yeast and archaeal H/ACA RNPs (; ; ). Stable expression of H/ACA RNAs requires incorporation into RNPs, in particular, association with the core trimer, as genetic depletion of each of its proteins causes a specific loss of H/ACA RNAs in yeast (; ; ; ). Therefore, assembly of H/ACA RNPs may follow the simple but unusual pathway of protein-only particle formation followed by association with an H/ACA RNA. In this manner, functional H/ACA RNPs can be assembled in archaea (; ) but not in eukaryotes, where additional factors may be required (). Indeed, two proteins involved in H/ACA RNP biogenesis in yeast, the nuclear assembly factor Naf1p and Shq1p, have been identified based on their interaction with Cbf5p (yeast NAP57) and Nhp2p (; ; ). Although neither Naf1p nor Shq1p is part of mature H/ACA RNPs, their genetic depletion leads to a specific loss of H/ACA RNAs. While this study was in progress, chromatin immunoprecipitation (ChIP) analyses in yeast showed cotranscriptional recruitment of Naf1p and Cbf5p to H/ACA RNA genes (; ). Although Naf1p and Shq1p have putative homologues in mammals, insects, nematodes, and plants, none have been detected in the sequenced genomes of archaea, suggesting a eukaryotic-specific function (unpublished data; ). In this study we demonstrate how human Naf1p (NAF1) is involved in H/ACA RNP assembly. We show this using a unique combination of biochemical and single-cell labeling approaches. Because most mammalian snoRNAs are processed from introns (), we designed cell lines that expressed the H/ACA snoRNA E3 from the intron of an inducible pre-mRNA reporter construct stably integrated into their genomes. In this manner, we were able to monitor and control the transcription of a specific H/ACA RNA in a single cell and to delineate the stepwise assembly of H/ACA RNPs in the cell. Because yeast Naf1p interacted with some of the H/ACA core proteins, we tested whether NAF1 did as well and, if so, with which ones (; ; ; ; ). For this purpose, we used an in vitro–translation/immunoprecipitation assay that we had previously used to dissect the molecular architecture of mammalian H/ACA RNPs (). Specifically, NAF1 was in vitro translated in the presence of various combinations of H/ACA core proteins (, I lanes). Subsequently, one of the proteins was immunoprecipitated and the precipitates analyzed for coprecipitating proteins (, IP lanes). Analysis of the autoradiograms showed that NAF1 efficiently coprecipitated with NAP57 alone (, lane 4) and in the presence of the core trimer (lane 6) but not in the absence of NAP57 (lane 2). Neither NAF1 nor NAP57 alone or together precipitated with HA-NHP2 (, lanes 4, 6, and 8). However, addition of NOP10, which mediates the interaction between NHP2 and NAP57, caused NAP57 (, lane 10) and associated NAF1 to precipitate with HA-NHP2 (lane 12). Although NAP57 efficiently coprecipitated with HA-NOP10 (, lane 4), NAF1 did not (lane 2). However, when NHP2 was included with HA-NOP10, some NAF1 did coprecipitate (, lane 6), suggesting that the association of NHP2 with NOP10 exposed a minor NAF1 interaction domain. As in the case of HA-NHP2, further addition of NAP57 to form the core trimer led to the efficient coprecipitation of NAF1 (, lane 8). In contrast, the box C/D core protein fibrillarin, added as a negative control, was not precipitated in any combination (). Collectively, these findings demonstrated that, similar to its yeast counterpart (; ; ), human NAF1 specifically interacted with NAP57 alone or in the context of the core trimer (). Even in the context of the full complement of core proteins, NAF1 efficiently precipitated with NAP57 (, lane 2). Moreover, when the amount of NAF1 was increased in the input (, compare lanes 1 and 3), the amount of GAR1 relative to that of NHP2 and NOP10 was reduced in the precipitate (, compare lanes 2 and 4). This result indicated that NAF1 competed with GAR1 for NAP57 binding. Indeed, increasing the amount of NAF1 in the input while keeping the amount of the H/ACA core proteins the same (, lanes 1–3) resulted in a concomitant reduction of GAR1 relative to NHP2 and NOP10 in NAP57 precipitates (lanes 4–6). NAF1 may bind more tightly than GAR1 to NAP57 because in this system GAR1 expression was insufficient for converse competition and because of the previously observed inconsistent NAP57 association of GAR1 (). Given the presence of a homologous domain in NAF1 and GAR1 (), it is plausible that this domain of each protein binds to the same site in NAP57. Collectively, NAF1 and GAR1 bind NAP57 in the context of the core trimer, but their binding sites overlap (or are identical), precluding simultaneous interaction (). This competition between NAF1 and GAR1 suggests a sequential interaction with NAP57. Although these biochemical analyses clearly showed that NAF1 interacted with NAP57, it was imperative to ascertain that this interaction occurred in the biological context of the cell nucleus. Toward this end, we developed a novel nuclear tethering assay. For this purpose, NAF1 was fused to the lac repressor, LacI, which binds to lac operator sequences that were stably integrated at a single chromosomal locus of the human U2OS osteosarcoma cell line (2-6-3; ). Transient transfection of NAF1-LacI into the 2-6-3 U2OS cells resulted in the tethering and concentration of NAF1-LacI at one site in the nucleus, easily detectable by indirect immunofluorescence with anti-NAF1 or -LacI antibodies (, panel 1, arrows). Using double immunofluorescence, we tested what endogenous proteins NAF1-LacI recruited to its tethering site. NAP57 accumulated above its nucleolar levels at the sites of NAF1-LacI concentration, indicating that NAF1 also bound to NAP57 in the nucleus (, panel 2). Similarly, NHP2 was observed at the tethering site, apparently recruited to NAP57 via NOP10 (). Despite its ability to interact with NAP57 (), GAR1 was not concentrated at the nuclear site of the NAP57–NAF1-LacI complex (), consistent with our finding that GAR1 and NAF1 binding to NAP57 was mutually exclusive (). All observed interactions were due to the NAF1 moiety of the LacI fusion protein, as LacI alone failed to recruit any of the proteins tested (unpublished data). These results supported our findings from the in vitro–translation/immunoprecipitation approach (). Although the absence of GAR1 suggested that NAF1-LacI failed to recruit intact H/ACA RNPs, we confirmed this by probing for the presence of four H/ACA RNAs—E3, ACA8, ACA 18, and ACA25. By using simultaneous FISH detection of the four snoRNAs, no or only negligible amounts of H/ACA RNAs were detected at the sites of NAF1-LacI concentration (). Moreover, Nopp140, which binds mature H/ACA RNPs (; ; ), failed to be recruited by NAF1-LacI (unpublished data). Therefore, NAF1 in the nucleus interacted with NAP57 alone and/or in the context of the core trimer NAP57–NOP10–NHP2 but not with mature H/ACA RNPs. Although NAF1-LacI bound to the lac operator sequences in the nucleus, excessive expression led to its accumulation in the cytoplasm (, panel 1). Therefore, to investigate the localization of endogenous NAF1, we raised polyclonal antibodies against a synthetic NAF1 peptide. On Western blots of whole cell extracts, affinity-purified NAF1 antibodies recognized a single protein band that migrated at the same position as in vitro–translated NAF1 (compare and , lane 1). By indirect immunofluorescence, they yielded a nucleoplasmic staining pattern excluding nucleoli (the sites of mature snoRNPs, which were double labeled with antibodies to the C/D snoRNP protein fibrillarin; ). Furthermore, NAF1 was not enriched in Cajal bodies, another site of mature snoRNPs, identified by coilin immunofluorescence (). Given the nuclear and cytoplasmic localization of the NAF1-LacI fusion protein, we tested whether NAF1 shuttled between the nucleus and cytoplasm using a heterokaryon approach (). For this purpose, we fused human HeLa and mouse NIH3T3 cells (, outlined cells), whose nuclei can be distinguished by a difference in DNA staining (, panel 2, asterisks and arrows). Because the NAF1 antibodies barely recognized the mouse protein while brightly staining the HeLa nuclei (, panel 1), we could test whether NAF1 equilibrated between the nuclei of different origin in fused cells. To inhibit de novo protein synthesis, the fusions were performed in the presence of cycloheximide. Although NAF1 barely redistributed between the different nuclei right after fusion, it fully equilibrated 4 h later (, compare panel 1 in H and I). Specifically, reduced staining in the HeLa nuclei and increased staining in the 3T3 nuclei of the fused cells indicated equilibration of HeLa NAF1 between the nuclei of different origin, i.e., shuttling of NAF1. So far, our data have indicated that NAF1, although absent from mature H/ACA RNPs, bound NAP57 and the core trimer. Based on the shuttling of NAF1, such an interaction could occur in the cytoplasm followed by nuclear translocation for association with H/ACA RNAs. We reasoned that the most likely sites for nuclear RNP assembly were the sites of H/ACA RNA transcription. To test our hypothesis, we developed a unique cell system. Our four requirements for this system were the inducible transcription of a specific H/ACA RNA, the visualization of its transcription site, the processing of the H/ACA RNA from an intron, and the generation of a functional mRNA. Specifically, two cell lines were generated (based on the U2OS cell line) that had tandem arrays of the following constructs stably integrated at a single locus in their genome. Both constructs consisted of the human β-globin gene with three exons, two introns, and a short 5′ untranslated region (UTR; ; ; ). Exon 3 was truncated and fused in-frame to CFP containing the peroxisomal targeting tripeptide Ser-Lys-Leu (SKL) at its COOH terminus (); thus, fluorescent peroxisomes resulted from the transcription of a functional mRNA. The 3′ UTR encompassed 18 MS2 stem loop repeats, which provided tight binding sites for the MS2 coat protein when transcribed (; ), thereby allowing the detection of the transcription locus with a MS2-GFP fusion protein. Most important, the rat H/ACA snoRNA E3 was placed in the second intron of the β-globin gene of the first construct (). Expression of the gene was induced by doxycycline through the transactivator rtTA (reverse tetracycline-responsive transcriptional activator), which was stably integrated in these U2OS cells (). This cell line was termed E3. A control cell line that contained the same construct but without the E3 snoRNA was named E3-minus. After doxycycline induction of the transgene, proper expression and processing of the RNAs in our cell lines was confirmed by fluorescence microscopy and RNase protection assays. First, using FISH with DNA probes for exon 1 and 2 of the β-globin construct, the exons were detected as a single spot in the nucleoplasm and in mature β-globin–CFP mRNPs in the cytoplasm (, panel 1). Hybridization to the same nucleoplasmic spot with a probe for intron 1 confirmed its identity as a transcription site (, panel 2). Alternatively and for potential live imaging studies, the transcription sites and cytoplasmic mRNPs were detected by transiently transfected MS2-GFP (, panel 3). The presence of the mRNA and its 3′ UTR in the cytoplasm suggested that it was full-length and that the β-globin–CFP pre-mRNA was correctly spliced and exported. This was corroborated by the visualization of its translation product, β-globin–CFP, in peroxisomes where it localized by virtue of its COOH-terminal SKL targeting sequence (, panel 4). Importantly, in the absence of doxycycline and transgene transcription, no transcription sites, mRNPs, or β-globin–CFP were detected (unpublished data). The presence of a single transcription site per nucleus suggested that tandem arrays of our constructs integrated at a single site in the genome of these U2OS cells (; ). Indeed, FISH on metaphase spreads corroborated a single insertion site (unpublished data). In E3 cells, the snoRNA produced from the transgene was detected at the transcription site by FISH with two probes complementary to the rat E3 snoRNA (, panels 1 and 2, arrows and insets). Additionally, E3 accumulated in nucleoli in a granular pattern typical for snoRNPs, suggesting its proper incorporation into H/ACA RNPs (, panel 1). Induced cells were routinely identified by the β-globin–CFP protein product of the transgene in peroxisomes (, panel 3). Because the endogenous human E3 snoRNA differed in 13 nucleotides from the induced rat E3, hybridization to endogenous E3 was weaker (, panel 1, asterisk). In the E3-minus control cell line, no E3 was observed at the transcription site (, panels 1 and 2, arrows and insets) and no enhanced fluorescence occurred in nucleoli of induced versus uninduced cells (, panel 1, asterisk). The transgene, however, was properly expressed as attested by the β-globin–CFP signal in peroxisomes (, panel 3). As a final quality test of our stable cell lines, we investigated the length of the induced snoRNA and the regulation of gene expression using RNase protection assays. Specifically, a radiolabeled RNA probe was synthesized that spanned the intronic snoRNA and exon 2 of β-globin (). When incubated with total RNA from the cell lines, fragments from this probe hybridizing to the stable target RNAs were protected from digestion by RNase A and T1 and migrated at their expected positions on a sequencing gel (, lanes 2–5). However, hybridization to yeast tRNA yielded no protected fragments (, lane 1). After induction by doxycycline, prominent bands of full-length E3 snoRNA and exon 2 were protected by total RNA from E3 cells (, lane 3) but only exon 2 by RNA from E3-minus cells (lane 5). The generation of full-length E3 snoRNA indicated its proper processing and incorporation into H/ACA RNPs. Full-length exon 2 represented correctly spliced globin-CFP mRNA. Transcription of the transgene was tightly regulated because no E3 and exon 2 were detected in the absence of doxycycline (, lanes 2 and 4). Endogenous E3 snoRNA was present independent of induction of the transgene (, lanes 2–5). Note that only smaller fragments of the probe were protected by partial hybridization to endogenous human E3 snoRNA because of slight sequence differences to the rat E3 of the transgene. As a control, a separate probe was included that protected a 100-nt fragment of the unrelated RNA of the signal recognition particle (SRP). Equal amounts of SRP RNA were protected by total RNA from induced and uninduced cells but not by tRNA (). These data indicated that our transgene was correctly induced, transcribed, and spliced and that the rat E3 snoRNA was properly processed and assembled into stable H/ACA RNPs. To test which proteins associated with the nascent H/ACA RNA at its site of transcription, we transfected GFP-tagged constructs into the E3 cell line and induced expression of the transgene. Localization at the transcription site was monitored by double fluorescence with cotransfected MS2-RFP or by FISH with the intron 1 probe (, panel 2). All core proteins localized to nucleoli in their characteristic granular pattern, suggesting their incorporation into H/ACA RNPs (, panel 1). In addition, the proteins of the core trimer, NAP57-, NHP2-, and NOP10-GFP, were identified at the site of H/ACA RNA transcription, indicating an early association of the core trimer with the nascent RNA (, insets). However, the fourth core protein, GAR1-GFP, was consistently absent from H/ACA RNA transcription sites (, arrow and insets). The putative assembly factor, NAF1-GFP, concentrated at the transcription sites but was excluded from nucleoli (). Overexpressed NAF1-GFP accumulated in the cytoplasm like NAF1-LacI ( and , panel 1). These observations suggested an early involvement of NAF1 in H/ACA RNP biogenesis, followed by GAR1 replacement during or after release of the maturing particles from the site of transcription. Indeed, the absence from transcription sites of GFP-Nopp140, a chaperone associated with functional H/ACA RNPs in nucleoli (), supported the lack of mature particles at the site of transcription (). Additionally, the C/D snoRNP protein GFP-fibrillarin, although localizing properly to nucleoli, was absent from the transcription site (unpublished data). Importantly, none of the proteins was observed at the transcription site of E3-minus cells, confirming that all colocalization was due to the nascent H/ACA RNA. The results with the transfected constructs were corroborated with endogenous proteins by indirect immunofluorescence and quantitated. In the same way as the transiently transfected constructs and in addition to their known location in nucleoli and nucleoplasm, endogenous NAP57, NHP2, and NAF1, but not GAR1, were concentrated at the E3 transcription site (). For quantitation, the presence of each protein, transfected and endogenous, at an E3 transcription site was scored. The identifications of each factor at the site of transcription from at least 100 cells are given as a percentage (). These data documented the presence of NAP57, NOP10, NHP2, and NAF1 at the E3 transcription site but a lack of association of GAR1. The <100% coincidence of the localizing factors with the transcription site was likely caused by variable stages of H/ACA RNA transcription and the fact that uncertain identifications were discounted. Nevertheless, the quantitation supported our conclusions and proved the images in and to be truly representative. To confirm the recruitment of NAP57 and NAF1 to the snoRNA transcription site by an independent method, we used ChIP on our induced cell lines transiently transfected with GFP-tagged constructs. Using PCR, we tested for the presence of three DNA fragments in the inputs and ChIPs: the promoter region, the rat snoRNA E3, and the CFP-SKL of the transgene. The presence of the correct fragment in the inputs was confirmed in all cases (, odd lanes). Immunoprecipitation with anti-GFP antibodies detected an association of the positive control, the core histone H2b, with all fragments (, lane 2). Although GAR1 and the control C/D snoRNP protein fibrillarin were not associated with any of the three fragments (, lanes 4 and 6), NAP57 and NAF1 precipitated the fragment corresponding to snoRNA E3 but not the other two (lanes 8 and 10). These results indicated that NAP57 and NAF1 were recruited specifically during transcription of H/ACA snoRNA E3 but not before or after. Altogether, the ChIP data supported our single-cell fluorescence localizations. For the following three reasons, we also investigated whether Cajal bodies coincided with H/ACA RNA transcription sites. First, Cajal bodies have been implicated as sites of H/ACA RNP–mediated pseudouridylation (). Second, the transcription of other small nuclear RNAs, e.g., U2, had been physically associated with Cajal bodies (; ). Third, we wanted to verify that the observed H/ACA RNA transcription sites were unique entities and distinct from Cajal bodies. This was to be expected because GAR1 and Nopp140, both verified Cajal body constituents (; ), were not present at transcription sites (; and ) and because only 27% of our U2OS cells contained Cajal bodies. Indeed, most Cajal bodies, identified by their marker protein coilin, were distinct from the snoRNA E3 transcription site (). Nevertheless, in a small number of instances (18 out of 245 cells with both Cajal bodies and transcription sites), an association of the two was observed, suggesting that, as in the case of U2 snRNA, there may be some relationship between active snoRNA genes and Cajal bodies. Having established that both NAP57 and NAF1 were recruited to the site of H/ACA RNA transcription, we used a siRNA approach to investigate how the appearance of the two proteins related to each other. For this purpose, E3 cells were transiently transfected with NAP57 siRNA for 72 h and the knockdown confirmed by indirect immunofluorescence. NAP57 was knocked down in the nucleoli of >50% of cells, whereas the levels of fibrillarin in the same nucleoli remained unaffected (, compare panels 1 and 2). Similarly, knockdown of fibrillarin left the amount and distribution of NAP57 unaffected (, compare panels 1 and 2). As judged by comparison of fluorescence intensities of nucleoli in silenced versus unsilenced cells (, arrows), NAP57 and fibrillarin were knocked down by ∼60 and ∼80%, respectively. To test the impact of NAP57 and fibrillarin knockdown on the expression of the transgene, E3 cells were transfected with siRNAs and induced with doxycycline before total RNA isolation and RNase protection analysis. In this manner, fibrillarin knockdown had no impact on the expression and stability of the E3 snoRNA and the globin-CFP mRNA (, lane 1). However, NAP57 knockdown abolished production of stable E3 snoRNA but left globin-CFP mRNA induction unaffected (, lane 2). The impact of NAP57 knockdown on endogenous E3 snoRNA was less severe, reflecting the increased half-life of the snoRNA once incorporated into mature RNPs. In separate experiments, knockdown of NAF1 also caused a specific loss of induced E3 snoRNA but not of exon 2 (, lane 3), even though NAF1 was only reduced by ∼50% (). These data indicated that reduced levels of NAP57 and NAF1 abolished packaging of newly synthesized H/ACA RNA into stable RNPs. Because both NAP57 and NAF1 proved essential for H/ACA RNP biogenesis, we tested whether their recruitment to nascent snoRNA was interdependent. For this purpose, NAF1-GFP was transfected into E3 cells knocked down for NAP57 but induced for transgene expression. Surprisingly, in NAP57-silenced cells, NAF1-GFP was never detected at E3 snoRNA transcription sites identified by FISH with the intron 1 probe (). However, in control fibrillarin knocked down cells, NAF1-GFP was observed at snoRNA transcription sites (unpublished data). Therefore, recruitment of NAF1 to nascent H/ACA RNAs was NAP57 dependent. Finally, we tested whether NAP57 was also dependent on NAF1. Because of the partial knockdown, individual cells reduced in NAF1 were difficult to unambiguously identify by indirect immunofluorescence. Therefore, Western blotting was used to compare the amount of the two proteins in whole cell lysates (). After NAF1 knockdown, NAP57 levels were reduced compared with those of tubulin and of control cells knocked down for fibrillarin (, compare lanes 1 and 2). Thus, NAP57 levels were dependent on NAF1. The capability to detect and image discrete regions in the cell nucleus at which specific complexes assemble, in conjunction with detailed biochemical analysis of these same processes, is a powerful tool for the unraveling of intricate interactions taking place during the generation of molecular complexes. Following this approach, we dissected the spatiotemporal assembly of human H/ACA RNPs using newly established cell lines that allow induction and visualization of H/ACA RNA transcription. Our data support the following three-step model of human H/ACA RNP biogenesis (). First, NAF1 stabilizes newly synthesized NAP57 before integration into the RNP. NOP10 and NHP2 may associate with NAP57 at this point to form the core trimer because the NAF1–NAP57 interaction is unimpeded by the presence of NOP10 and NHP2. Second, the NAF1–NAP57–(NOP10–NHP2) complex is recruited to the site of H/ACA RNA transcription. Third, GAR1 replaces NAF1 in the maturing particle either at the transcription site triggering its release or during transit through the nucleoplasm to the site of function in nucleoli and Cajal bodies. Therefore, assembly of these simple five-component RNPs displays a surprising spatiotemporal complexity that serves as paradigm for the assembly of larger RNPs. For example, ribosome assembly requires an unexpected number of nonribosomal proteins (). A general concept that emerges from our studies is that newly synthesized, free NAP57 is always paired with NAF1. Specifically, the two proteins are interdependent (). The need for such a stabilizing activity is supported by the fact that mammalian NAP57, when expressed in bacteria or insect cells, precipitates and is insoluble (unpublished data). The shuttling of the normally nuclear NAF1 is consistent with its association with NAP57 already in the cytoplasm followed by their coimport (). If free NAP57 and NAF1 are normally associated, then not only may NAF1 recruitment to the H/ACA RNA transcription site be dependent on NAP57 but also vice versa. For example, NAF1 may target NAP57 to nascent H/ACA snoRNAs by binding to the phosphorylated COOH-terminal domain (CTD) of RNA polymerase II and remain there through the interaction of NAP57 with the snoRNA (). Such a mechanism is supported by findings with yeast Naf1p, which binds to the CTD (; ; ). Two recent reports using ChIP analysis showed that Naf1p and Cbf5p (yeast NAP57) are also recruited cotranscriptionally in yeast; however, it remains to be resolved whether phosphorylated CTD is required or not (; ). Regardless, our dissection of the role of human NAF1 in H/ACA RNP biogenesis suggests that it is the orthologue of yeast Naf1p and that it functions as a NAP57 chaperone. The last step of H/ACA RNP biogenesis, the replacement of NAF1 by GAR1, may occur in Cajal bodies, possibly aided by the survival of motor neurons (SMN) protein. The following three observations support such a scenario. First, SMN, all snoRNP core proteins, and most of their RNAs tested have been detected in Cajal bodies (; ). Second, SMN, which assembles the Sm proteins onto spliceosomal U snRNAs, interacts with GAR1, raising the possibility of its involvement also in H/ACA RNP biogenesis (; ). Third, fibrillarin, the glycine and arginine domain–containing protein of C/D snoRNPs, also interacts with SMN and also is involved in a late RNP assembly step (; ; ; ). Therefore, and in analogy to the C/D snoRNP U3 (), H/ACA RNP biogenesis may be added to the many functions of Cajal bodies. The occurrence of the final step of H/ACA RNP maturation in Cajal bodies is further attractive because it would preclude the presence of functional H/ACA RNPs in the nucleoplasm, where they might endanger nontarget RNAs by pseudouridylation. In conclusion, we have developed a novel cell system to visualize the site of transcription of an intronic H/ACA RNA and to define factors recruited to the nascent H/ACA RNA after induction of its transcription. These stable cell lines were based on a previously established system to monitor gene expression in living cells (; ) and allowed us to dissect the biogenesis of H/ACA RNA beyond what was possible with ChIP analysis (; ). We have defined some roles of the H/ACA-specific factor NAF1 in RNP assembly. Nevertheless, additional factors may be involved, as in the case of C/D RNP biogenesis, where the export factors CRM1 (chromosome region maintenance 1) and PHAX (phosphorylated adaptor for RNA export) are important (). Our cell lines will now allow testing if these and other factors also play a role in H/ACA RNP biogenesis. The human NAF1 clone was obtained from American Type Culture Collection (clone 3447276). For transient transfections, the following plasmids were used: NAF1-GFP (pNK16; NAF1 fused to monomeric GFP in monomeric GFP–C1; ); RFP-NAF1 (pNK31; NAF1 fused to monomeric RFP in monomeric RFP–C1; ); NAF1 (pNK40; in pcDNA3; Invitrogen); NAP57-GFP, NOP10-GFP, NHP2-GFP, and GAR1-GFP (); GFP-Nopp140 (pTM95; ); GFP-fibrillarin (); MS2-GFP (pNK35; the NLS was removed from and photo-activatable GFP replaced by GFP in MS2–photo-activatable GFP; ); MS2-RFP (pNK36; same as for MS2-GFP); and LacI-NAF1 (pNK41; LacI from pAFS51 [] via pJRC70 [a gift from J. Chubb, University of Dundee, Dundee, UK] cloned into pNK40). As templates for in vitro transcription/translation, the following constructs were used: NAP57, NOP10, NHP2, GAR1, HA-NOP10, and HA-NHP2 as described previously (); NAF1 (pSR29; NAF1 under T7 promoter in pBSII SK+; Stratagene); and HA-NAF1 (pSR36; NAF1 with a single HA tag in pTM93; ). The pTet-globin-snoE3-CFP-18MS2-2 (for E3 cells) and pTet-globin-CFP-18MS2-2 (for E3-minus cells) were generated by sequentially inserting the following, mostly PCR-amplified fragments, into pCMV-globin (): CFP with a COOH-terminal SKL tripeptide from pECFP-N1 (CLONTECH Laboratories, Inc.); 18 MS2 repeats from pSL-24X (; six repeats were lost because of recombination in ); Tet promoter from pTRE-2 (CLONTECH Laboratories, Inc.) replacing the CMV promoter; and, after excision of the neomycin-resistance gene, rat snoRNA E3 from pTM105 (). As a probe for RNase protection assays, an 883-nt fragment encompassing part of intron 1 through intron 2 after the snoRNA insertion site () was cloned into PCR4-TOPO vector (Invitrogen), yielding pSR41. As a probe for SRP RNA, nucleotides 119–221 of the RNA were amplified from plasmid p7swt2 () and cloned into pBSII KS+ to yield pSR27. The probe for U19 was as published (). For indirect immunofluorescence, cells were fixed and permeabilized as we described previously () and mounted in p-phenylenediamine–containing medium (). The following primary antibodies were used: polyclonal anti-NAP57 serum (RU8 at 1:200) raised in rabbits (Covance Research Products) against the synthetic peptide KSLPEEDVAEIQHAE (GenScript Corporation) of human NAP57 with an additional NH-terminal cysteine as described previously (); anti-NAF1 IgGs (CRX6 at 1:100) raised in rabbits against the synthetic peptide CISSLPPVLSDGDDDLQVEKENKN of human NAF1 as described for the NAP57 antiserum and affinity purified over a peptide column (); anti-GAR1 IgG (p16 at 1.5 μg/ml; ); anti-NHP2 IgG (p14 at 3 μg/ml; ); anti-fibrillarin IgG (D77 at 1 μg/ml; ); anti-coilin ascites fluid (5P10 at 1:1,000; ); and anti-LacI IgG (9A5 at 1 μg/ml; Upstate Biologicals). The following fluorescently labeled secondary antibodies were used at a 1:200 dilution: goat anti–rabbit IgGs (rhodamine; Chemicon); goat anti–mouse IgGs (fluorescein; Boehringer); and goat anti–rabbit F(ab′) fragments (Cy5; Jackson ImmunoResearch Laboratories). Imaging was performed at room temperature using a 60×/1.4 NA planapo objective on an inverted microscope (1X 81; Olympus) containing automatic excitation and emission filter wheels connected to an air-cooled charge-coupled device camera (Sensicam QE; Roper Scientific) run by IPLab Spectrum software (Scanalytics). Images were processed for contrast and brightness using Photoshop CS2 (Adobe). Because of slight bleed-through from the fluorescein into the cyan channel in (panel 4), the fluorescein image was subtracted from the cyan image using ImageJ software (NIH). FISH was performed as previously described (), except that cells were grown on untreated coverslips and no dextran sulfate or vanadylribonucleoside complex was included with 40% formamide for hybridization with a 10-ng probe. Probes were synthesized and labeled with either Cy5 or Cy3 (GE Healthcare). In some cases, immunostaining was done after FISH, using the routine protocol, except that cells were fixed again but not permeabilized. Heterokayon assays were performed essentially as described previously (). Cell fusions were induced by the addition of 50% polyethylene glycol 8000 for 10 min, and 100 μg/ml of cycloheximide was added 1 h before and during fusion. The cells were fixed 15 min and 4 h after fusion and labeled with NAF1 antibodies followed by DNA stain with 5 μg/ml DAPI for 2 min. For the NAF1-LacI tethering assay, the 2-6-3 U2OS cell line () was transiently transfected with NAF1-LacI and analyzed by indirect double immunofluorescence with the indicated antibodies and by FISH. In vitro–transcription/translation and immunoprecipitation assays () and Western blots () were performed as described previously, except the blot in , which was probed with NAF1 rabbit serum (CRY6 at 1:100), NAP57 IgG (RU8 at 0.6 μg/ml), and tubulin monoclonal antibodies (DMA1 at 1.1 μg/ml; Sigma-Aldrich), developed with infrared conjugated goat anti–rabbit IgGs (Alexa fluor 680; Rockland Immunochemicals) and goat anti–mouse IgGs (IR800; Invitrogen), and scanned on an infrared imaging system (Odyssey; LI-COR Biosciences). RNase A/T1 protection analysis was performed as described previously (), with minor changes. Antisense probes were synthesized in the presence of α-[P]CTP using the Maxiscript kit (Ambion) and gel purified. 1–5 μg of total RNA (prepared with TRIzol [Invitrogen]) were mixed with probe (50,000 cpm) for hybridization and digestion with 40 μg/ml RNase A (Roche) and 2 μg/ml RNase T1 (Calbiochem). Typically, half of the samples were analyzed in 6% polyacrylamide–8M urea sequencing gels followed by autoradiography. ChIP was performed as described in detail by the Farnham Laboratory (). Cells used for ChIP were transfected with the indicated GFP-tagged constructs using FuGene 6 (Roche) and induced simultaneously with 6 μg/ml doxycycline for 24 h. The GFP fusion proteins were precipitated with anti-GFP agarose beads (MBL International Corporation).

The precellularization stages of embryogenesis entail 13 rapid nuclear divisions within a common cytoplasm. The first nine of these nuclear divisions take place deep within the interior of the embryo to produce roughly 300–400 nuclei by the end of the ninth division. Development continues with the migration of these nuclei toward the periphery of the embryo. Once at the periphery, nuclei undergo four additional rounds of division (nuclear cycles 10–13) in the stage known as the syncytial blastoderm. Except for the pole cells located at the posterior end of the embryo, all nuclei in the syncytial blastoderm embryo occupy a common cytoplasm. The sharing of a common cytoplasm ceases when each nucleus becomes individually encased in plasma membrane during interphase of nuclear cycle 14. This cellularization event produces >6,000 primary epithelial cells (; ). In most animal cell types, the organization of the ER and Golgi apparatus is intimately tied to the localization of the nucleus and astral microtubules derived from centrosomes. The ER extends off the nuclear envelope along microtubules as a tubule network (; ), whereas the Golgi apparatus localizes adjacently to centrosomes or to ER export sites (). Interestingly, this arrangement of organelles is absent in the preblastoderm embryo. There, maternally derived ER and Golgi membranes are localized distinctly at the embryo periphery (; ), whereas nuclei and centrosomes are found deep within the embryo interior (; ). A result of this arrangement is that when nuclei and centrosomes arrive at the embryo periphery, ER and Golgi membranes must somehow become segregated among the thousands of syncytial nuclei. In this study, we address how the secretory membrane system partitions among nuclei in preparation for cellularization in the syncytial blastoderm. Toward this end, we have used GFP-tagged ER, Golgi, and plasma membrane markers in a variety of biophysical-based experiments to examine membrane continuity and organellar dynamics in living embryos. We report that distinct nuclear-associated secretory units of ER and Golgi emerge across the embryo in the absence of plasma membrane boundaries during the syncytial blastoderm stage. The secretory units are shown to mediate localized protein delivery to the plasma membrane and to require centrosome-derived astral microtubules for their maintenance. We discuss how this organization helps to equivalently partition ER and Golgi into daughter cells at cellularization, and we propose potential roles for it in the establishment and maintenance of localized gene and protein expression patterns within the early embryo. To visualize ER in the early embryo, we used transgenic embryos expressing the ER marker Lys-GFP-KDEL (), which contains an NH-terminal signal sequence () and a COOH-terminal KDEL sequence (). Confocal imaging of live embryos expressing Lys-GFP-KDEL before nuclear arrival at the embryo periphery revealed that ER membranes were concentrated in a band at the cortex ( and Video 1 for 3D image; available at ). The membranes extended up to 30 μm beneath the plasma membrane, with no detectable pool in the embryo interior. Large, autofluorescent yolk granules (, red structures; z = 6 μm) filled the space between ER elements. As previously mentioned, the morphology of the ER in most animal cell types is dependent on microtubules (; ). In the preblastoderm embryo, however, centrosomes are bound to nuclei deep within the embryo. Nevertheless, there are nonpolymerized and short polymerized forms of tubulin localized immediately beneath the plasma membrane to a depth of 10 μm (). These short microtubules are organized around spherical particles, which are likely cortical granules and/or yolk platelets (). To investigate whether microtubules in the cortex of the preblastoderm embryo contributed to the observed organization of the ER, we examined the distribution of tubulin in GFP-tubulin–expressing embryos (see Materials and methods). Pronounced GFP-tubulin fluorescence was seen ∼10–20 μm beneath the plasma membrane () and included filamentous, tubulin-rich structures surrounding spherical particles that excluded fluorescence (, arrows), presumably representing yolk particles with associated microtubules as reported in other systems (; ; ). These observations led us to conclude that short microtubules were indeed present in the embryo cortex before nuclear migration. We then examined whether these short microtubules played a role in ER organization in the preblastoderm embryo. Nocodazole (a microtubule-disrupting agent) was microinjected into preblastoderm embryos expressing Lys-GFP-KDEL, and changes in the distribution of Lys-GFP-KDEL were examined. Notably, in the region of the embryo in which nocodazole was injected, both the spherical cluster and tubule network patterns of the ER observed before nocodazole injection (, left) disappeared and were replaced by long, parallel strands of ER (, right). Injection of DMSO alone did not result in any obvious morphological change of ER clusters (unpublished data). These results indicated that cortical microtubules are necessary for the ER to become organized into spherical clusters and tubule networks during the preblastoderm stage. We next asked whether ER membranes in the preblastoderm embryo existed as a single, interconnected system. To test this, we performed FRAP experiments in embryos expressing Lys-GFP-KDEL. A small region of ER at the embryo cortex was photobleached using a high intensity laser beam, abolishing GFP fluorescence in this area. Low power imaging was then used to assess fluorescence recovery into the photobleached region. As predicted for a continuous ER system, a rapid and complete recovery of Lys-GFP-KDEL fluorescence into the photobleached area was observed with no gross structural changes in ER membranes (). Further evidence suggesting that the ER existed as a single, continuous membrane system was obtained from repetitive photobleaching experiments (fluorescence loss in photobleaching [FLIP]; ). In these experiments, a region of the ER in preblastoderm embryos expressing Lys-GFP-KDEL was repeatedly photobleached, and fluorescence from distant areas was examined. If the ER exists as a continuous membrane system at this stage, fluorescence at distant areas should decrease during the FLIP protocol as a result of fluorescent molecules in these areas diffusing into the FLIP region of interest (ROI) and becoming photobleached. Consistent with this scenario, a 50% drop in Lys-GFP-KDEL fluorescence was observed from all ER areas in a 20-μm radius of distance from the FLIP ROI center within 5 min of photobleaching (, and see quantification in C). Hence, the ER behaves as a continuous system in preblastoderm embryos. We next examined whether the morphology and continuity of the ER membrane changes upon nuclear migration to the embryo cortex, which occurs during interphase of nuclear cycle 10 (). In time-lapse sequences, no change in the total level of Lys-GFP-KDEL fluorescence occurred before or immediately after nuclear arrival (Video 2, available at ), strongly suggesting that nuclei do not bring a significant amount of their own ER during this period. Nevertheless, the organization of already existing ER was significantly affected by the migrating nuclei ( and Video 2). In particular, the spiral clusters of the ER membrane appeared to unwind as nuclei arrived at the cortex. Interestingly, the membrane from these clusters, together with other tubule ER elements, then reassembled around individual nuclei. To test whether the observed changes in the ER during and after nuclear arrival were mediated by nuclear-associated microtubules, we imaged tubulin and ER simultaneously in Lys-GFP-KDEL–expressing embryos that were microinjected with rhodamine-labeled tubulin. Reorganization of the ER around an individual nuclei coincided with the appearance of nuclear-associated centrosomes and astral microtubules (seen as rhodamine-labeled microtubular arrays emanating from one side of the nucleus; ). To test whether this required microtubules, we microinjected nocodazole, which caused microtubules to disassemble at the site of injection (unpublished data). ER membranes at the site of injection were no longer tightly organized around individual nuclei and appeared as a loose network (). DMSO injection alone resulted in no change in the morphology of nuclear-associated ER membranes (unpublished data). Hence, microtubules are important for the reorganization and maintenance of the ER around nuclei in the syncytial blastoderm embryo. A model for how ER membranes reorganize as nuclei and their associated centrosomes migrate to the embryo cortex based on our results is depicted in . In this scheme, the existing tubulin at the cortex becomes incorporated into centrosome-derived astral microtubules as nuclei and their associated centrosomes migrate to the cortex. Growth of the astral microtubules recruits and retains ER membrane to areas specifically surrounding nuclei. Time-lapse images of a Lys-GFP-KDEL–expressing embryo during the syncytial blastoderm stage revealed the ER to be tightly organized around individual nuclei during interphase and mitosis ( and Video 3, available at ). This is in accordance with earlier observations of ER dynamics during the syncytial mitoses (). Upon photobleaching a small area of ER adjacent to one nuclei, rapid recovery into this area occurred as a result of fluorescent molecules outside the bleached box redistributing into the bleached area (). The fluorescent redistribution appeared to occur primarily from nonbleached ER surrounding the one nuclei and not from ER surrounding other nuclei. This raised the possibility that ER membranes that surround individual nuclei at this developmental stage exist as segregated units that do not exchange components. We investigated this possibility by performing FLIP in a small region of ER associated with one nucleus (). Notably, a 40% drop in fluorescence from the ER surrounding the nucleus adjacent to the bleached area occurred within 2–3 min of photobleaching (which is the duration of the examined interphase). Fluorescence in areas of ER that were associated with other nuclei was minimally affected during this period, with a 5% drop seen in areas extending 20 μm away from the FLIP ROI center (see for quantification). Given that the photobleaching experiment encompassed the whole interphase, these results suggested that the diffusion of Lys-GFP-KDEL molecules during this interval was restricted to areas of ER surrounding individual interphase nuclei. Hence, the ER around individual nuclei behaved as a compartmentalized system that did not exchange its components with ER surrounding other nuclei. To determine whether this property of ER depended on astral microtubules, we performed a parallel FLIP experiment in Lys-GFP-KDEL–expressing syncytial blastoderm embryos microinjected with nocodazole. Upon repetitive photobleaching of a small region of ER, fluorescence in areas extending 20 μm away from the FLIP ROI center now showed a significant drop in fluorescence (∼40%; ) compared with the slight drop observed in untreated cells (i.e., ∼5%; ). Hence, microtubules were necessary for the ER to behave as isolated units surrounding individual nuclei. It is known that the plasma membrane in the syncytial blastoderm partly invaginates around each nuclei (; ). To address whether this could explain the restricted exchange of ER proteins between ER associated with different nuclei, we examined the depth of such plasma membrane invaginations using the plasma membrane marker Spider-GFP, a casein kinase I encoded by the gene that associates with the plasma membrane and secretory vesicles destined for the plasma membrane (see Materials and methods; ). In syncytial blastoderm embryos expressing Spider-GFP, plasma membrane invaginations were seen extending down ∼5 μm from the embryo surface (, Spider-GFP). In contrast, ER membranes observed in Lys-GFP-KDEL–expressing embryos extended far deeper (∼15 μm; , Lys-GFP-KDEL). Therefore, if ER compartmentalization was dependent on plasma membrane invaginations, no compartmentalization should exist at depths >∼5 μm below the embryo surface. To test for this, we performed a series of FLIP experiments in Lys-GFP-KDEL–expressing embryos in which regions directly over nuclei (where plasma membrane invaginations are present) or well below nuclei (where plasma membrane invaginations are absent) were repeatedly photobleached. When a region over nuclei was repeatedly photobleached (), 40% of fluorescence was lost from the ER extending well below the same nuclei (up to 20 μm below the embryo surface) compared with only 10% from the ER surrounding adjacent nuclei. When a region below nuclei (10–20 μm below the surface) was repeatedly photobleached (), 70% of fluorescence was lost from the ER above the nuclei, but only 10% was lost from the ER adjacent to the photobleached region below nuclei. Because the restricted diffusion of ER proteins to ER surrounding an individual nucleus occurred at all depths of the ER, we concluded that plasma membrane invaginations could not explain the observed compartmentalized character of ER around an individual nucleus. We also performed FLIP experiments using Lys-GFP-KDEL in preblastoderm embryos whose ER had not yet become compartmentalized. The rate of fluorescence loss from regions lateral to the bleach area was significantly faster than that observed in the syncytial blastoderm embryo (), with a 60% drop in fluorescence (vs. 10% in the blastoderm) from a region of comparable distance from the bleached box within the same time period. Together, these data suggested that compartmentalization of ER in the embryo only occurs after nuclei have migrated to the cortex, does not rely on plasma membrane invaginations, and requires microtubules. Given our observation that the ER reorganizes into compartmentalized, nuclear-associated units in the syncytial blastoderm, we asked whether other organelles that exchange components with the ER also exhibit such compartmentalization. The Golgi apparatus receives all soluble and membrane-bound cargo exported out of the ER and sorts these components either back to the ER or to the plasma membrane or endosomes (). To gain insight into Golgi distribution and whether any organizational changes of the Golgi occur in the developing embryo, we imaged transgenic embryos expressing galactosyltransferase (GalT)-GFP (GalT-GFP), a resident Golgi enzyme. The same transgenic lines have been used to look at Golgi structure in ovaries (). Before and after nuclear migration, the Golgi appeared as several thousand punctate structures located at the periphery of the embryo (), as previously reported (; ). The dispersed Golgi puncta were localized to areas of the embryo enriched in ER and, in time-lapse videos, exhibited a jostling motion throughout successive nuclear cycles (Video 4, available at ). The appearance of these dispersed Golgi puncta resembled that of Golgi puncta found in plants (), sea urchin embryos (), and mammalian cells after treatment with nocodazole (; ). In the case of plant and nocodazole-treated mammalian cells, the Golgi puncta were shown to be localized next to ER exit sites (). An advantage of this type of Golgi distribution is that it allows the Golgi both to receive ER-derived secretory cargo and to recycle proteins back to the ER in the absence of long-range vesicular transport (). Therefore, a similar function might be served by the localization of Golgi puncta near ER in the embryo. To further understand the properties and behavior of the Golgi in the syncytial blastoderm, we asked whether the movement of Golgi puncta was restricted around individual nuclei. For this, we visualized puncta behavior in a syncytial blastoderm expressing GalT-GFP by confocal microscopy. Viewed from the embryo surface, Golgi structures were seen jostling around an individual nucleus but never moving between nuclei (Video 5, available at ). This could also be seen in images of embryos viewed from the side (Video 6). During their movement, Golgi structures often merged with other structures (, follow arrow in sequence), but merging only occurred between Golgi structures surrounding the same nuclei (Video 5). Finally, when a Golgi structure separated into two distinct structures, these two puncta always remained around the same nuclei (, follow arrow in sequence; and Video 5). These results indicated that the movements of Golgi puncta were restricted to areas surrounding a given nucleus. Previous studies have shown that Golgi enzymes undergo constitutive cycling through the ER (; ), a property that allows Golgi structures to continually modify their size and distribution within cells (). Given this characteristic, we investigated whether Golgi protein exchange between the ER and Golgi was restricted to ER and Golgi structures surrounding a particular nucleus in the syncytial blastoderm. We began by examining the rate of GalT-GFP exchange between Golgi puncta in embryos at the preblastoderm stage, in which the ER exists as a single continuous system and nuclei are deep within the embryo. Repetitive photobleaching of Golgi puncta in a preblastoderm embryo expressing GalT-GFP revealed a 40% drop in fluorescence from all areas in a 30-μm radius of distance from the FLIP ROI center within 3–5 min of photobleaching (; red box is the photobleached area, and blue box is the area being monitored; B shows quantification). As an individual Golgi puncta did not move laterally across the embryo for >30 μm (not depicted) and the extent of the lateral loss of Golgi fluorescence was similar to that observed in FLIP experiments with the ER marker (), the data were consistent with Golgi enzymes cycling through an interconnected ER membrane network. We next examined GalT-GFP exchange between Golgi puncta during the syncytial interphases of nuclear cycles 11–12 after nuclei have migrated to the embryo cortex (). A ROI containing several Golgi puncta was repeatedly photobleached (, red outline), and fluorescence was monitored in the area adjacent to the bleached region (, green outline) and the areas surrounding a nearby nuclei (, blue outline). Notably, fluorescence loss was observed only from puncta in the immediate ER-nuclear unit that contained the area being photobleached, showing a 35% drop in fluorescence (, green curve) with a minimal effect on puncta in the area of surrounding ER-nuclear units, which showed only a 10% drop in fluorescence (, blue curve). This pattern of fluorescence loss was similar to that observed with the ER marker at this stage of the embryo (). To address whether the compartmentalized exchange of GalT-GFP between Golgi puncta extended to deeper areas within the embryo, we repetitively photobleached an area located below several nuclei (, red box). Significant fluorescence loss was observed in Golgi puncta directly above the bleached region (, green box) but not from Golgi puncta lateral to the bleached region (, blue box). Quantification revealed a 60% loss of fluorescence occurring in the region of Golgi puncta above the bleached area compared with a 5–10% drop in the region lateral to the bleached area (). Thus, both the ER and Golgi behave as compartmentalized units surrounding individual nuclei in the syncytial blastoderm. Given the compartmentalized character of the ER and Golgi around individual syncytial nuclei, we wondered whether their secretory products were delivered to restricted areas of the plasma membrane near an individual nuclei. To test this possibility, we investigated whether blocking the secretory pathway in one area of the embryo resulted in the reduced delivery of secretory cargo to the plasma membrane only in that area or in other areas as well. To perturb the secretory pathway in these experiments, we used the drug brefeldin A (BFA), which blocks the transport of secretory cargo from the ER to the Golgi apparatus (; ). To monitor secretory transport, we observed the invagination of the plasma membrane during cellularization, which is dependent on newly synthesized membrane moving through the secretory pathway (; ). In embryos expressing Spider-GFP to label the plasma membrane, the injection of BFA resulted in a dramatic slowdown in membrane invagination during cellularization (). Notably, the slowing occurred only in the area of the embryo where BFA was injected, with sites far from the injection site invaginating their plasma membrane normally (i.e., up to 35 μm as found in wild-type embryos; unpublished data). Although the effect of BFA on plasma membrane invagination at cellularization has already been described (), the localized effect we observed is new. It suggested that delivery of material to the plasma membrane could not be compensated by surrounding secretory units and implied, therefore, that membrane insertion at the plasma membrane occurred in a localized manner in the embryo. To further test for localized secretory membrane insertion at the plasma membrane, we performed FRAP in embryos expressing Spider-GFP during early cellularization. A region encompassing only the plasma membrane pool of Spider-GFP (, blue box) and a region spanning both plasma membrane and intracellular pools of Spider-GFP (, red box) were photobleached simultaneously. If membrane insertion at the plasma membrane occurred in a localized manner from an adjacent intracellular pool, the plasma membrane fluorescence associated with the red box () should be unable to efficiently recover (as its adjacent intracellular pool was bleached), whereas the plasma membrane fluorescence associated with the blue box () should recover (as its adjacent intracellular pool was not photobleached). Consistent with this prediction, we found that as cellularization progressed, a more complete and uniform recovery occurred in the region of the blue box (). Because the fluorescence in the blue box () recovered simultaneously across all areas of the plasma membrane rather than initially from the edges of the bleach box, recovery was not caused by Spider-GFP diffusing laterally across the plasma membrane. Rather, the fluorescent pool responsible seemed to be derived from the area directly beneath the plasma membrane because when this area was bleached (as occurred in the red box; ), much less plasma membrane fluorescence recovered. The use of transgenic lines expressing GFP-tagged markers in the early embryo allowed us to follow and investigate the organizational and morphological changes in organelles of the secretory membrane system during the precellularization and cellularization stages of embryogenesis in real time. We found that before nuclear migration to the embryo cortex, the ER existed as a single, interconnected system through which proteins freely diffused. Upon nuclear arrival at the cortex, the ER and Golgi system became compartmentalized around individual nuclei. As a consequence, ER and Golgi resident proteins now moved only within the ER and Golgi associated with single nuclei. The carriers mediating the transport of secretory products to the plasma membrane also exhibited compartmentalized behavior, delivering their contents to the plasma membrane only in areas adjacent to the ER/Golgi system from which they were generated. The organization and continuity of ER membranes has been extensively studied during oocyte maturation and early embryogenesis of other organisms, including starfish (; ), sea urchin (), mouse (), and frog (). In these studies, ER membranes labeled by injecting either the ER lipophilic dye DiI or mRNA encoding for an ER lumenal protein marker (ssGFP-KDEL) were shown to exist as an interconnected network of membrane sheets and tubules in the cortex of the fertilized egg, as is the case in the fertilized preblastoderm embryo. In addition, prominent ER clusters directly beneath the plasma membrane were described in the cortex of the mouse and frog embryos (; ). These were postulated to have a role in the generation of transient calcium waves or in the propagation of calcium oscillations. The ER clusters that we observed in the preblastoderm embryo may have a similar role, which the recent study on the existence of distinct calcium microdomains during syncytial divisions () would support. Because the ER exists as a continuous system in the preblastoderm, the embryo must have a mechanism for partitioning this organelle among nuclei so that at cellularization, the newly formed cells have equivalent amounts of ER membrane. Our data suggest that this occurs by a microtubule-driven process that causes the ER to be divided up among the nuclei, resulting in each interphase nucleus becoming surrounded by a single ER membrane system that is separate from adjacent ones. This was demonstrated in FLIP experiments in which resident proteins of the ER and Golgi were seen to be rapidly circulating only within ER and Golgi membrane that associated with a particular nuclei during the four rounds of nuclear division at the periphery. Microtubule depolymerization by the microinjection of nocodazole resulted in the loss of ER and Golgi compartmentalization around a given nucleus, indicating that an intact microtubule network is essential to keep ER and Golgi structures close to individual nuclei in the syncytium. The significance of microtubules in the biogenesis and maintenance of the ER network has been unequivocally demonstrated (; ), so our data reinforce the role of intact microtubules in organizing ER membranes around individual nuclei. Based on these data, we propose a model for the organization and distribution of distinct and separate ER and Golgi membranes with individual nuclei, as illustrated in . The structural reorganization of the ER after nuclear migration reported in this study is likely to serve many important functions in the early embryo. One key function could be for cellularization. Membrane proteins are synthesized during the syncytial stages (; ; ) as well as during cellularization (; ). The partitioning of ER and Golgi upon nuclear migration ensures that equivalent amounts of ER and Golgi are packaged with nuclei in preparation for cellularization. During cellularization, such partitioning could additionally serve to facilitate plasma membrane synthesis by ensuring that it is homogenous across the embryo. Indeed, when we disrupted secretory trafficking locally by injecting BFA into a specific region of the embryo during cellularization, plasma membrane synthesis was inhibited only in this region, indicating that secretory vesicles do not randomly circulate in the embryo but must be locally produced near the plasma membrane. Another function for the compartmentalization of ER and Golgi in the embryo could be to help establish and maintain localized gene and protein expression patterns (; ). Although zygotic transcription and protein synthesis increases dramatically during cellularization, transcription and protein synthesis has been reported to start as early as nuclear cycles 8–10 (; ; ), resulting in highly localized expression patterns well before cellularization. One simple and logical way that compartmentalization of the ER may affect gene expression patterning could be by the segregation of maternal mRNAs and proteins. In ascidians, some maternally loaded mRNAs have been shown to associate with the rough cortical ER and relocalize with the ER as it moves (; ). Likewise, some maternal mRNAs in have been shown to be anchored on ER membranes (). Indeed, recent work has shown that mRNA anchors to specific transitional ER–Golgi units in the oocyte (), allowing specific sorting and secretion to take place. If maternally loaded mRNAs anchored on the ER become compartmentalized around individual nuclei upon nuclear migration, as our results would suggest, then these transcripts are likely to be locally expressed. Thus, an already existing polarity in the distribution of maternal material would not only be preserved but would also be further maintained during cortical divisions. To this end, ER and Golgi compartmentalization might provide a mechanism for spatially and temporally restricting maternally derived material during the early stages of embryogenesis. In summary, our data suggest that in the absence of plasma membrane boundaries surrounding nuclei but with the requirement of an intact microtubule network, the embryo is able to differentiate the secretory endomembrane system (i.e., ER and Golgi) into segregated nuclear-associated units. In a volume occupied by thousands of nuclei, this capacity for apportioning ER and Golgi among nuclei is likely to be vital for cellularization and for the establishment and maintenance of localized gene and protein expression patterns. The fact that many other organelles are organized by microtubules (i.e., endosomes, lysosomes, and mitochondria) further suggests that it is possible to have the functional equivalents of cells despite the complete absence of plasma membrane boundaries within a syncytium. pUASp:Lys-GFP-KDEL and pUASp:GalT-GFP transgenic lines have been previously described (). The nanos-Gal4:VP16 driver () was used to express UASp transgenes in the early embryo. Spider-GFP flies were a gift from A. Debec (Universite Pierre et Marie Curie, Observatoire Oceanologique, Villefranche-sur-mer, France). They were generated using a “protein trap” methodology () and can be obtained from the European Drosophila Stock Center in Szeged (). GFP-tubulin flies were a gift from A. Spradling (Carnegie Institution, Baltimore, MD). Embryos were collected on apple juice–agar plates, dechorionated in 2% bleach, and placed flat or upright on a Lab-Tech chambered coverglass (Nunc). Chambers were then filled with Ringer's solution (Tübingen and Düsseldorf). Confocal microscope images of live embryos were captured on an inverted microscope (510 Meta or ConfoCor-2; Carl Zeiss MicroImaging, Inc.) using the 488-nm line of an Ar laser with a 505–530 emission filter for GFP and a 543-nm HeNe laser line with a 560–615 emission filter for rhodamine. To image yolk autofluorescence, a two-photon laser (Chameleon; Coherent) at 820 nm was used with a 435–485 infrared emission filter. Images were captured with a C-Apochromat 1.2 NA 40× water immersion objective (Carl Zeiss MicroImaging, Inc.). Images were analyzed with Image software (W. Rasband, National Institutes of Health [NIH], Bethesda, MD) and Image Examiner software (Carl Zeiss MicroImaging, Inc.) and prepared by Adobe Photoshop 7.0. Embryos were collected for 30 min, aged on collection plates for 50 min, and dechorionated in 2% bleach. Dechorionated embryos were microinjected as previously described (). Rhodamine-tubulin was obtained from Cytoskeleton, Inc. BFA (Sigma-Aldrich) was microinjected at 5 mg/ml in DMSO. For nocodazole injections, embryos were dechorionated, microinjected with nocodazole (Sigma-Aldrich) at 10 mg/ml in DMSO, mounted in chambers, and placed on ice for 10 min to depolymerize microtubules. Embryos were then imaged at 25°C. FRAP and FLIP were performed by photobleaching a small ROI and monitoring fluorescence recovery or loss over time as described previously (; ). To create the fluorescence recovery or loss curves, the background-corrected fluorescence intensities were transformed into a 0–1 scale and were plotted using Microsoft Excel X. Video 1 is a 3D rendering of ER membranes at the surface of a Lys-GFP-KDEL–expressing embryo. Video 2 shows nuclear migration during nuclear cycle 10 interphase in a Lys-GFP-KDEL–expressing embryo as viewed from the embryo surface. Video 3 shows divisions of ER membranes during cycles 10–13 in a Lys-GFP-KDEL–expressing embryo as viewed from the embryo surface. Video 4 shows the dynamics of Golgi puncta during cycles 11–13 at a cross section of a GalT-GFP–expressing embryo. Video 5 shows the movement of Golgi structures around individual syncytial interphase nuclei in a GalT-GFP–expressing embryo as viewed from the embryo surface. Video 6 shows the same as Video 5 at a cross section of the embryo. Online supplemental material is available at .

Previous studies implicate necrotic cell death in devastating human pathologies such as stroke and neurodegenerative diseases (; ). In , specific mutations in several genes that encode ion channel subunits and regulators trigger the degeneration of specific sets of neurons (for review see ). Dying neurons exhibit macroscopic and ultrastructural characteristics that are reminiscent of the excitotoxic neuronal death that occurs during stroke in mammals (; ; ). Thus, vertebrates and share a death mechanism that involves the hyperactivation of ion channels. These observations are consistent with the hypothesis that a threshold of ion influx is needed to initiate the degenerative process. Perturbation of cellular ionic homeostasis contributes decisively to necrotic neuronal death (). In addition to ion homeostasis, intracellular pH has emerged as an important modulator of necrosis in . Cytoplasmic acidification develops during necrosis, whereas the vacuolar H-ATPase, which is a pump that acidifies lysosomes, is required downstream of cytoplasmic calcium overload to promote necrotic cell death (). Interestingly, similar acidosis accompanies necrotic cell death after stroke in mammals (; ). Moreover, the examination of postmortem human brains associates neuronal pH alterations with several pathological and neurodegenerative states (). Investigations in both nematodes and mammals converge to implicate specific calpain and aspartyl proteases (cathepsins) in the execution of necrotic cell death (; ). Calpain proteases are normally dependent on calcium for activation, whereas aspartyl proteases require a highly acidic environment for full activity and are primarily confined to lysosomes and other acidic endosomal compartments (; ). Studies in primates indicate that damage to the lysosomal membrane is inflicted enzymatically by activated calpains. Calpains localize to lysosomal membranes after the onset of ischemic episodes, with subsequent spillage of cathepsins to the cytoplasm (). This observation led to the formulation of the “calpain–cathepsin hypothesis,” whereby the calcium-mediated activation of calpains results in the rupture of lysosomes and leakage of killer cathepsins that eventually dismantle the cell (; , ). Although these observations collectively indicate that lysosomes participate actively in the process of cell death, their contribution is poorly understood. We examined the role of lysosomes in a well defined model of necrotic cell death in the nematode. We show that the alkalization of endosomal and lysosomal compartments protects against necrotic cell death that is induced by mutations in several ion channels, as well as by prolonged hypoxia. We investigated the effect of mutations that alter lysosome biogenesis in necrotic cell death and found that mutations resulting in the accumulation of large lysosomes exacerbate necrosis, whereas mutations that impair lysosome biogenesis are protective. Conditions that counterbalance intracellular acidification enhance suppression of neurodegeneration by aspartyl protease deficiency, indicating that aspartyl proteases are activated by low pH conditions, which develop during necrosis. By monitoring lysosomes during necrosis in vivo, we show that lysosomes coalesce around the nucleus and dramatically enlarge during the early and intermediate stages of necrosis, although, ultimately, lysosomal definition is lost. Together, these results point to a decisive role for lysosomes in the execution of necrotic cell death. Recent data suggests that vacuolar H-ATPase–mediated intracellular acidification is required downstream of cytoplasmic calcium overload to promote necrotic cell death, plausibly by enhancing the activity of the low pH–dependent proteases that dismantle the cell (). To emulate impaired lysosomal acidification in degenerating neurons, we treated animals expressing a neurotoxic () allele encoding a hyperactive ion channel subunit that is normally required for mechanosensation with NHCl and acridine orange. These lysotropic weak bases are known to accumulate in lysosomes and other acidic subcellular compartments, neutralizing their pH (). Treatment ameliorated degeneration of the six touch receptor neurons of mutant animals (). Similarly, cell death inflicted by the toxic allele, encoding a hyperactive acetylcholine receptor calcium ion channel, or by overexpression of the hyperactivated Gα(Q227L) variant (α) was also suppressed after treatment with NHCl and acridine orange (). We assessed cell survival by scoring for expression of a touch receptor–specific ∷GFP reporter fusion in adult animal neurons. Fluorescent neuron number increased after treatment with alkalizing agents in adult mutant animals (309 ± 19 NHCl and 192 ± 13 acridine orange vs. 176 ± 11 fluorescent neurons in untreated mutants; = 100; P < 0.001, unpaired test). Prolonged hypoxia, which is a condition of low oxygen availability that emerges during ischemia and stroke, induces necrotic cell death in the nematode (Scott et al., 2002). We examined the effect of alkalization in hypoxia-induced cell death. Nematodes were treated with sodium azide, which inhibits complex IV (cytochrome oxidase) of the respiratory chain and simulates hypoxia. Treatment with NHCl and acridine orange reduced the hypoxic death of wild-type animals (). We considered whether suppression of necrosis is an indirect effect of probable alterations in animal growth and development caused by alkalizing agents. We assayed developmental timing after egg hatching and past the L1 stage, which is where we assayed for cell death. We also assayed for animal locomotion, pharyngeal pumping, and defecation. Treatment with NHCl and acridine orange, at the concentrations and under the conditions used, does not result in any discernible defects in animal growth and development that could influence the course of necrotic cell death. We conclude that dependence on acidified intracellular compartments is a common denominator of necrotic cell death triggered by diverse stimuli. To further evaluate the lysosomal role in necrotic cell death, we examined mutants defective in lysosomal biogenesis. We examined necrosis in () mutant animals. Cells of mutants contain increased numbers of enlarged acidic lysosomes. encodes the mucolipin-1 homologue that is implicated in mucolipidosis type IV, which is a lysosomal storage disease that results in severe developmental neuropathology in humans (; ; ). Neurodegeneration inflicted by and is exacerbated in mutants (). We confirmed the reduced cell survival by scoring for expression of a touch receptor–specific ∷GFP reporter fusion in adult animal neurons. Fluorescent neuron number decreased in double mutants, compared with mutant animals (109 ± 16 versus 176 ± 11 fluorescent neurons in mutants; = 150; P < 0.001, unpaired test). In addition, mutants showed increased sensitivity to hypoxia compared with wild type (16.5 ± 5.3% wild-type survival versus 5.3 ± 4.1% animal survival. > 200; P < 0.005, unpaired test). Knockdown of , which encodes for a subunit of the vacuolar H-ATPase, by RNAi in double mutants abolished –mediated enhancement of cell death (). This suggests that the enhanced cell death observed in double mutants is caused by increased lysosome-mediated acidification. To confirm the effect of deficiency on cell death, we outcrossed mutant animals in an effort to minimize the possibility that the effects on neurodegeneration we observed were caused by unlinked spurious mutations in the genetic background of the allele. Furthermore, we used RNAi to specifically target . Both outcrossed derivatives and RNAi-mediated knockdown of resulted in enhanced –induced neurodegeneration (). In a reciprocal approach, we examined necrosis in mutants. The gene encodes a predicted Rab GTPase that is similar to proteins implicated in the biogenesis of specialized lysosome-related organelles (). mutant alleles were recovered in a screen aimed at identifying genes involved in the formation of birefringent gut granules, which are lysosome-related organelles (). mutants are defective in the biogenesis of lysosome-related gut granules, show little or no staining with lysosomal markers, and lack detectable expression of the vacuolar H-ATPase subunits VHA-17 and VHA-11 in intestinal precursor cells (). We found that all three alleles ameliorate necrotic cell death triggered by (). This suggests that the reduced number of lysosomes in touch receptor neurons of double mutants results in reduced intracellular acidification and, consequently, in reduced necrotic cell death. Specific calpain and aspartyl proteases are implicated in the execution of necrotic cell death in both nematodes and mammals (; ), and the importance of calpain and aspartyl protease activation in acute cell injury and necrotic cell death triggered by calcium influx has been previously established (for review see ). Calpain proteases become activated upon the abrupt increase of intracellular calcium that occurs in response to diverse necrosis-initiating stimuli, whereas aspartyl proteases function optimally under the highly acidic conditions present in the lumen of lysosomes and other acidic endosomal compartments (; ). We assessed the effect of lysosome-mediated intracellular acidification on the requirement for aspartyl proteases in necrosis. mutants maintain aspartyl protease activity that is 90% lower than in wild-type animals (). –induced neurodegeneration is attenuated in double-mutant strains (). RNAi-mediated knockdown of diminishes cell death inflicted by ( and ; ). Cell death was further reduced in mutant animals by RNAi-mediated knockdown of (). Two aspartyl proteases, ASP-3 and -4, contribute the bulk of protease activity required for neurodegeneration inflicted by diverse genetic insults in (). Similarly, knockdown of or by RNAi in double mutants augmented survival of the six receptor neurons, compared with single mutants (). In contrast, reduced V-ATPase activity did not further enhance suppression of necrosis by calpain protease deficiency (). We conclude that suppression of necrosis by aspartyl protease deficiency is enhanced by conditions that impede intracellular acidification. We considered the contribution of additional cellular pH homeostasis mechanisms in necrosis. Two other major mechanisms have been implicated in cytoplasmic and subcellular organelle pH regulation; first, the sodium–hydrogen exchanger (NHX), and second, the cation transporter P-type ATPase. These mechanisms operate both on the plasma membrane and at the membranes of subcellular organelles, such as mitochondria, to facilitate proton trafficking and pH homeostasis. Multiple genes are encoded in the genome (more than 9 isoforms; ; ). is the gene encoding the nematode P-type ATPase homologue. Knockdown of , -, and -, which are three isoforms expressed in neurons, did not alter the extent of neurodegeneration induced by (137 ± 19 , 134 ± 16 , and 134 ± 21 versus 139 ± 14 vacuolated neurons in mutants; = 100). To address the potentially redundant function of these genes in the nervous system, we assayed cell death in animals in which all three isoforms were knocked down by RNAi. We did not observe significant suppression of necrosis in these animals (131 ± 28 versus 139 ± 14 vacuolated neurons in mutants; = 100.). Similarly knockdown of did not affect necrotic cell death (: 141 ± 17, versus 139 ± 14 vacuolated neurons in mutants; = 100). Therefore, neurodegeneration is not affected by compromising these other, nonlysosomal pH homeostasis mechanisms. The importance of lysosomal membrane permeabilization in cell death has previously been established (). Approaches combining electron microscopy and immunodetection show that calpains concentrate on lysosomal membranes during ischemic stroke in primates (; ). However, information on lysosome fate and lysosomal system alterations during necrosis in vivo is lacking. We monitored the distribution and morphology of lysosomes in vivo during necrotic cell death in . To visualize lysosomes and late endosomes, we fused GFP at the COOH terminus of LMP-1, which is the only protein bearing a lysosomal targeting sequence (GYXXΦ; Φ, large hydrophobic amino acid residue) at its COOH terminus (). LMP-1 shows similarity to the vertebrate lysosome-associated membrane protein LAMP/CD68 (; ), and it is widely used as a lysosomal marker (; ; ). We examined lysosomal distribution and morphology in touch receptor neurons of wild-type and animals expressing an LMP-1∷GFP fusion. In the neurons of wild-type animals, lysosomes appear scattered throughout the cytoplasm (). In contrast, during the early stages of neurodegeneration, lysosomes enlarge and localize close to the nucleus (, i). As neurodegeneration progresses, lysosomes fuse to surround an internally vacuolated structure (, ii–viii). This encapsulated vacuole is likely the swollen nucleus of the dying neuron (). To confirm the nuclear origin of the internal vacuole, we performed DAPI staining in animals expressing the LMP-1∷GFP fusion protein. As shown in , LMP-1–labeled internal membranes are positive for DAPI staining. In agreement with our observation, elevation of cytosolic Ca concentration can induce lysosomal fusion (). Consistent with previous studies (; ; ), a portion of LMP-1 localizes to the plasma membrane (, ii–viii). Lysosomes remain confined around the distended nucleus during necrotic cell death (, ii–v). As neurodegeneration proceeds, the nucleus migrates to the periphery of the cell and condenses (, iv–vi). In the advanced stages of neurodegeneration, GFP intensity decreases and lysosomal definition is ultimately lost (, vii and viii). Interestingly, a decrease in LMP-1∷GFP immunoreactivity is associated with neuronal degeneration in mammals (). Diffusion of the highly localized GFP staining during the late stages of neurodegeneration indicates lysosome rupture. To further confirm the loss of lysosomal integrity, we stained animals with acridine orange, which is an acidophilic dye that distinctively stains lysosomes. Acridine orange accumulates diffusely in the cytoplasm of neurons during the late stages of necrosis, indicating extensive cytoplasmic acidification (). This observation, coupled with the loss of specific LMP-1∷GFP localization, suggests that lysosome rupture and the spillage of acidic lysosomal contents mediate cytoplasm acidification during necrosis. We investigated whether alkalizing treatments result in altered lysosomal fate. We examined the effects of acridine orange and NHCl treatment on LMP-1∷GFP distribution in mutants. Although the number of unvacuolated cells with a wild-type pattern of lysosomal distribution increases, vacuolated neurons show the same pattern of lysosomal distribution as animals (). Thus, reduced acidification does not affect lysosome distribution. Our observations are consistent with findings in animals deficient for the ATP-binding cassette transporter P-glycoprotein-2, which is also expressed in neurons. Acridine orange and Lysotracker red staining is reduced in animals lacking P-glycoprotein 2, indicating defective acidification. Nevertheless, LMP-1∷GFP distribution remains unchanged (). We further examined lysosomes in the different mutant genetic backgrounds that either enhance or suppress necrosis. We generated transgenic animals harboring LMP-1∷GFP in , , and the aspartyl protease–deficient mutant background. The number of vacuolated neurons is decreased in and mutants and increased in animals ( and ). We find that lysosomal morphology and distribution is similar in neurons that do vacuolate and die (). In this study, we have examined the involvement of lysosomes in necrotic cell death using a well characterized model of neurodegeneration in . We show that both genetic and pharmacological manipulations that affect lysosomal biogenesis and function modulate necrosis in the nematode. By following lysosome fate during neurodegeneration in vivo, we found that lysosomes fuse and localize exclusively around a swollen nucleus. In the advanced stages of cell death, GFP-labeled lysosomal membranes fade, indicating lysosomal rupture. What is the cause of lysosomal rupture during necrosis? Interestingly, calcium, which is one of the major upstream death-initiating signals, has been implicated in this process (). Activated calcium-dependent calpain proteases have been found to localize to disrupted lysosomal membranes in hippocampal neurons of primates after acute ischemia (, ), leading to the hypothesis that calpains compromise the integrity of lysosomal membranes and cause leakage of their acidic contents into the cytoplasm. Calpains become activated after the abrupt increase of intracellular calcium concentration that signals the initiation of necrosis (). Excessive calcium influx through several channels and transporter-mediated routes leads to intracellular calcium overload and concomitant cell death (; ). Sodium influx amplifies acute neuronal swelling and facilitates calcium entry through voltage-gated channels and the Na/Ca exchanger (). Cell injury and death can also be induced by disturbances of calcium homeostasis in the ER (; ). The ER is the major calcium storage compartment of the cell. Sequestration of calcium into the ER is mediated by the sarcoendoplasmic reticulum Ca-ATPase, and release back to the cytoplasm is controlled by ryanodine, and 1,4,5-inositol trisphosphate receptors (). Within the ER, calcium binds to calcium-binding molecular chaperones such as calreticulin and calnexin (; ). Under conditions of extreme cellular stress, ER calcium stores are rapidly mobilized, boosting the massive increase of intracellular calcium concentration, which signals cell demise (). Pharmacological treatments or genetic mutations that inhibit calcium release from the ER have a strong protective effect against necrotic cell death (; ). In contrast, treatment with chemicals such as thapsigargin, which promotes the discharge of calcium from intracellular stores by specifically inhibiting the sarcoendoplasmic reticulum Ca-ATPase calcium pump, induces necrotic cell death (; ). We hypothesize that generalized osmotic destabilization of the cell during necrosis may also contribute to the bursting of lysosomes. The exclusive confinement of lysosomes around the nucleus during neurodegeneration may reflect damaged microtubule or actin motors, which mediate the movement of lysosomal organelles (). It is known that calpain proteases contribute to cell death by cleaving essential cytoskeletal proteins of neuronal axons (for review see ). Therefore, calpains may act on microtubules at early stages of neurodegeneration. However, we cannot rule out the possibility that lysosomes are specifically targeted to the periphery of the nucleus. We observed that reduction of necrotic cell death by a drop in aspartyl protease activity is enhanced by conditions that counterbalance intracellular acidification. Such synergy between aspartyl proteases and pH indicates that aspartyl proteases become activated by low pH conditions, which develop during necrosis and facilitate cellular destruction. In addition to aspartyl proteases, other proteases that function optimally at low pH may become activated by acidification during necrosis. Such proteases have been implicated in both apoptotic and necrotic cell death (). We suggest that preventing acidification suppresses neurodegeneration, in part, by lowering the activity of these enzymes. Alternatively, aspartyl proteases and acidification may independently contribute to cell death. However, to discriminate between these alternatives requires complete elimination of aspartyl protease or vacuolar H-ATPase activity, which results in embryonic lethality (; ; unpublished data). The totality of our observations denotes an essential and general role for lysosomes in necrotic cell death induced by various insults. Our study is the first to monitor lysosomal alterations during necrosis in vivo, in any organism. Our findings uncovered novel aspects of the cellular changes that transpire during neurodegeneration in the nematode. Such information could be effectively used toward identifying candidate common intervention targets in an effort to battle numerous pathological conditions in humans. We envision that alterations of lysosomal biogenesis and function by genetic mutations or pharmacological treatments modify the susceptibility of neurons to necrosis. However, once a threshold is exceeded and cell death commences the sequence of events is essentially unaltered. We used standard procedures for strain maintenance, crosses, and other genetic manipulations (). Nematodes were grown at 20°C. N2 was used as the wild-type strain. The following mutant alleles were used: , which is referred to in the text as ; , which is referred to in the text as ; [pGα(Q227L)pGFP], which is referred to in the text as α ; [pssGFP], which is referred to in the text as , , , , and . The alleles were provided by G. Hermann (Lewis and Clark College, Portland, OR). The following double and triple mutants were used: , , [pssGFP], [pssGFP], , , and . To generate pLMP-1∷GFP, we fused GFP at the COOH terminus of the LMP-1 protein. The translational fusion includes the entire LMP-1 coding sequence lacking the stop codon, a Gly-Ser-Ser-Pro-Gly-Leu-Ala-Lys-Gly-Pro-Lys-Gly linker, and GFP. The resulting chimera was expressed in touch receptor neurons under the control of the promoter. The plasmid carrying the reporter fusion was constructed in two steps. First, the promoter was amplified from N2 genomic DNA with the primers 5′ CGGGATCCGAATCGTCTCACAACTGATCC 3′ and 5′ AACTGCAGGTGACTACTTGAGACCTG 3′. A 1,900-bp PstI–BamHI fragment was cloned into the promoterless vector pPD95.77 (). Second, the LMP-1 coding region was amplified from genomic DNA using the primers 5′ CGGGATCCGACGCTGGCATATCCTTGTCTC 3′ and 5′ CGGGATCCAATTGAACTATGTTGAAATCG 3′. A BamHI PCR fragment was cloned downstream of the promoter on the pPD95.77 plasmid vector. For RNAi experiments, we used HT115(DE3) bacteria, which were transformed with plasmids that direct the synthesis of double-stranded RNAs corresponding to the genes of interest; they were then fed to animals according to a previously described methodology (). For RNAi, we used a 1.5-kb PCR-generated fragment derived from the locus using the primers 5′ GGGGTACCCCATGATTTCAGATGTCTCGC 3′ and 5′ GGGGTACCCCGAATGCAAAGAATGAGAACG 3′. The primers used for RNAi constructs are as follows: 5′ GCTCTAGACTCTTCACTGGCCTGTG 3′ and 5′ CCGCTCGAGATCAGTATGACTGCG 3′ for , 5′ AACTGCAGTTATGGACGATATCAAC 3′ and 5′ CCGCTCGAGCCACAAACTTCAGCCAC 3′ for , and 5′ GCTCTAGATGGTGTCCTGACTCTTC 3′ and 5′ CCGCTCGAGCTTCCACTCCAGACATC 3′ for . For we used the following PCR primers: 5′ AACTGCAGATTGAAACACTGACATC 3′ and 5′ CCGCTCGAGTACCTGAAACATTCCG 3′. RNAi plasmids for , aspartyl proteases and , and calpain have been previously described (, ). We assayed the effectiveness of RNAi by monitoring the expression of full-length GFP reporter fusions. Plasmid vectors for were provided by A. Fire (Stanford University School of Medicine, Stanford, CA). Degeneration of specific neuron sets in animals bearing , , and α alleles was quantified as previously described (). For alkalization assays, we treated young adult animals with lysotropic alkalizing agents (5 mM NHCl and 40–150 μM acridine orange; Sigma-Aldrich) in liquid cultures supplemented with bacteria for 12 h at 20°C. Neurodegeneration was assayed in the progeny of treated animals at the L1 stage of development. To simulate death-inducing hypoxic conditions, we treated nematodes at the L4 stage of development with sodium azide (0.5 M for 30 min at 20°C; Sigma-Aldrich; adapted with modifications from Scott et al.). Statistical analysis of data was performed using Excel (Microsoft). L1 stage animals expressing LMP-1∷GFP were stained with 1 μg/ml DAPI for 15 min after methanol fixation. DAPI-stained animals were observed using a 40× objective (Plan-Neofluar; Carl Zeiss MicroImaging, Inc.), NA 0.75, and a 365 ± 12–nm band-pass excitation/397-nm long-pass emission filter set. A microscope was used, and pictures were taken using a camera (AxioPlan and AxioCam, respectively; both Carl Zeiss MicroImaging, Inc.). For LMP-1∷GFP imaging, animals were scanned with a 488-nm laser beam, under a confocal microscope (Radiance 2000; Bio-Rad Laboratories), using the LaserSharp 2000 software package (Bio-Rad Laboratories). Images of emission from individual PLM and ALM touch receptor neurons were acquired using a 515 ± 15–nm band-pass filter and a 40× Plan-Neofluar objective, NA 0.75. Acridine orange staining of necrotic cells was done by treating early L1 larvae with 1 μM acridine orange for 20 min. To visualize stained cells, animals were scanned with a 543-nm laser beam. Images of emission from individual touch receptor neurons were acquired using a 590 ± 35–nm band-pass filter. Animals were mounted in a 2% agarose pad in M9 buffer containing 10 mM sodium azide and observed at room temperature. Bright field and epifluorescence images were merged using Photoshop (version 7.0.1; Adobe).

In endocrine and neuroendocrine cells, bioactive molecules are packaged into nascent vesicles called immature secretory granules (ISGs), which bud from the TGN and are destined for regulated secretion (; ). ISGs undergo a series of maturation steps, including acidification of the granule lumen, prohormone processing (; ), AP-1–dependent removal of proteins (; ; ), and ISG–ISG homotypic fusion (), to become mature secretory granules (MSGs). MSGs, which are also called large dense core vesicles, accumulate in cells until they undergo fusion with the plasma membrane by regulated exocytosis. An in vitro fusion assay that reconstitutes ISG–ISG fusion has revealed that ISG homotypic fusion is dependent on NSF and α-SNAP () and on the SNARE protein syntaxin 6 (Stx6), but not on Stx1 or SNAP-25 (). SNAREs are essential components of membrane fusion, but whether they are sufficient for fusion and/or ensuring targeting specificity remains under debate. Additional proteins, including the Rabs () and the synaptotagmins (Syts; ), may coordinate and regulate vesicle trafficking and fusion. The Syts are a family of proteins characterized by a short lumenal NH terminus, one transmembrane region, and tandem C2A and C2B domains (; ). Currently, it is thought that Syts participate in the regulation of various steps during membrane fusion, primarily at the plasma membrane. Syt I, which was the first isoform identified (), is involved in calcium-dependent exocytosis () and functions as the calcium sensor that stabilizes the opening of the fusion pore at the final steps of fusion (), at the docking step (, ), and during vesicle recycling from the plasma membrane (). Syt I binds to the SNARE proteins Stx1 and SNAP-25 (; ), and this binding is thought to be important for the function of Syt I in membrane fusion. A genomic analysis has identified 16 Syt isoforms in mammals (), so that, like SNAREs or Rabs, Syts constitute a large family of proteins, suggesting that they regulate multiple membrane events. In support of this, although still controversial, a differential distribution of Syt I, III, IV, and VII has been reported in neuroendocrine cells (; ; ). Moreover, a recent study in showed that Syt isoforms localize to nonoverlapping subcellular compartments (). Syt IV () was characterized as an immediate early gene induced by depolarization in PC12 cells and rat brain (). Syt IV knockout mice exhibit abnormalities in motor performance, suggesting a role in synaptic plasticity (). The function of Syt IV in vesicular trafficking, however, remains unclear. Overexpressed Syt IV is sorted to MSGs upon NGF differentiation or forskolin treatment of PC12 cells, and is involved in the regulation of exocytosis (; ; ; ). Different studies have found contradictory localizations; Syt IV has been shown to colocalize with Syt I on synaptic vesicles and MSGs in PC12 cells (), whereas others demonstrated that Syt IV has a juxtanuclear distribution (), is localized on ISGs and not MSGs in PC12 and AtT20 cells, and does not colocalize with Syt I (; ; ). The function of endogenous Syt IV in relation to its localization in nondifferentiated neuroendocrine cells has never been studied. In PC12 cells, Syt IV is found on ISGs, but not on MSGs, and we have investigated whether Syt IV could be involved in ISG–ISG homotypic fusion before its removal from MSGs. We show that the cytoplasmic domain (CD) of Syt IV, but not of Syt I, VII, or IX, inhibits ISG–ISG homotypic fusion in an in vitro homotypic fusion assay, and that this domain is recruited specifically to ISG membranes. The role of Syt IV in ISG homotypic fusion was confirmed in vivo after siRNA-mediated depletion of Syt IV. We also find that Syt IV binds Stx6, and that Syt IV interacts with Stx6 via both the C2A and C2B domains. We show in the in vitro homotypic fusion assay that addition of the Syt IV CD, together with an anti-Stx6 antibody, leads to an additive inhibition of fusion, and we speculate that Syt IV and Stx6 are involved in regulating different stages of ISG homotypic fusion. Furthermore, using the dominant-negative Syt IV CD, and by the reduction of Syt IV levels using siRNA, we demonstrate in vivo a reduction of secretogranin II (SgII) processing by prohormone convertase 2 (PC2). Finally, we show that in cells transfected with the dominant-negative Syt IV CD, PC2 is mostly in the unprocessed/inactive proform. Collectively, our data provide the first direct evidence that Syt IV is involved in an intracellular membrane fusion event and in the regulation of secretory granule maturation in neuroendocrine cells. We first determined the distribution of Syt IV using subcellular fractionation techniques to isolate and separate ISGs and MSGs. PC12 cells were [S]sulfate-labeled either for short periods (5 min pulse and 15 min chase) to label ISGs or overnight (1 h pulse, overnight chase) to label MSGs (). The postnuclear supernatant (PNS) from labeled cells was subjected to continuous velocity sucrose gradient fractionation and analysis. Syt IV was found mainly in the light ISG-containing fractions, as well as in Golgi-containing fractions at the bottom of the gradient (, top), in contrast to Syt I, which localizes mainly to MSGs and synaptic-like microvesicles (Perin et al., 1991) and is found mainly in the slightly heavier fractions (, bottom). The ISG and MSG fractions were further separated on discontinuous equilibrium sucrose gradients (). Syt IV was in the ISG-containing fractions, together with [S]sulfate-labeled SgII and Stx6, which were previously shown to be on ISGs and not MSGs (; ). However, Syt IV was not found in the MSG fractions, which were identified by the presence of [S]sulfate-labeled SgII together with Syt I (, bottom). By indirect immunofluorescence in PC12 cells, Syt IV displayed a juxtanuclear distribution () and colocalized with Stx6 (Fig. S1 A, available at ). To confirm that Syt IV is not on MSGs, we used a PC12 cell line stably expressing PC2 (PC12/PC2 cells; ). In PC12/PC2 cells, PC2 cleaves SgII at several dibasic residues and produces a product of 18 kD (p18), which accumulates in MSGs and can be detected using an antibody that recognizes only p18, but not full-length or partially processed SgII. As shown in Fig. S1 B, a nonoverlapping distribution of Syt IV and p18 or Syt I was observed in PC12/PC2 cells. We conclude, in agreement with others (Eaton et al., 2000; ), that Syt IV is localized to ISGs, and not to MSGs, in PC12 cells. Because Syt IV is localized on ISGs, but not on MSGs, we hypothesized that it could play a role in granule maturation. In PC12 cells, ISGs undergo homotypic fusion during maturation, followed by excess membrane removal, most likely via AP-1–containing clathrin-coated vesicles. To test whether Syt IV is involved in the maturation of ISGs, we used an in vitro fusion assay that reconstitutes ISG homotypic fusion (). Addition of increasing amounts of the purified recombinant Syt IV CD into the complete fusion reaction resulted in a dose-dependent inhibition of ISG homotypic fusion by ∼40% (). Addition of the denatured CD had no effect on ISG–ISG fusion (unpublished data). In addition to Syt IV, Syt I, VII, and IX are the most abundant isoforms in PC12 cells, with Syt VII found at lower levels (; ). Furthermore, Syt I and IX are localized on MSGs and are involved in the regulation of calcium-dependent exocytosis in these cells (; ). Syt VII was localized either to the plasma membrane () or to large dense core vesicles (), and was also shown to be involved in the regulation of calcium-triggered fusion with the plasma membrane (; ; ). We asked if ISG homotypic fusion could be regulated by the other Syt isoforms present in PC12 cells. Purified Syt I, VII, and IX CDs, which were previously shown to inhibit exocytosis (; ), had no effect on ISG homotypic fusion (). We conclude that the inhibition of ISG–ISG fusion is specific for Syt IV and does not involve Syt I, VII, or IX. Many membrane fusion events, including the fusion of secretory granules with the plasma membrane (), require calcium. However, it has never been determined whether ISG homotypic fusion is calcium dependent. We found that the addition of increasing amounts of 1,2-bis (-aminophenoxy)ethane-,,′,′-tetraacetic acid (BAPTA) into the complete ISG–ISG fusion assay resulted in a dose-dependent inhibition of fusion of up to 95% (Fig. S2, available at ). The addition of equimolar BAPTA and calcium partially restored ISG–ISG fusion, and the addition of BAPTA alone after the fusion reaction had little effect. These results clearly demonstrate that ISG homotypic fusion is dependent on calcium. It was assumed that rat Syt IV is a calcium sensor because of the presence of the predicted calcium-binding residues in the C2B domain (). However, a recent crystal structure revealed that changes in the orientation of the calcium-binding residues render the rat Syt IV C2B domain unlikely to bind calcium (). Thus, it is likely that Syt IV is not the calcium sensor protein regulating ISG–ISG fusion and that another calcium-binding protein is regulating this process. Further work is required to clarify and confirm this speculation. Because the Syt IV CD was able to inhibit ISG homotypic fusion, we asked whether the site of inhibition resides on these membranes using an in vitro ISG-binding assay, which we have used to demonstrate ADP-ribosylation factor 1 and AP-1 binding to ISG membranes (; ). The addition of increasing amounts of Syt IV CD to ISGs showed a dose-dependent binding (), demonstrating that ISG membranes contain binding sites for Syt IV. No recombinant protein was detected in the pellet in the absence of ISGs, establishing that the binding is not a result of nonspecific aggregation of the protein. To test whether the binding is specific, 7 nM Syt IV CD was added to 50 μl ISGs or MSGs. The Syt IV CD bound to ISG, whereas no binding to MSG was observed (). In contrast, Syt I CD was able to bind both ISG and MSG membranes (). There was more binding of the Syt I CD to ISGs compared with MSGs, which might be attributed to the presence of synaptophysin-positive synaptic-like microvesicles in the ISG fraction (). These results show that Syt IV binds specifically to components that are present on ISGs and absent from MSGs, and further suggest that the site of inhibition of ISG–ISG fusion by the Syt IV CD is on the ISG membrane. In liposome-binding assays, some Syts such as Syt I and IX, but not Syt IV, are able to bind phospholipids via their C2 domains in a calcium–dependent manner (). However, others showed that Syt IV is able to bind liposomes containing only negatively charged phospholipids (). Although we do not know the lipid composition of ISG membranes, it is unlikely that they are composed solely of negatively charged phospholipids. To determine whether the recruitment of the Syt IV CD is mediated through protein or lipid components, we repeated the binding assays after pretreatment of ISG membranes with increasing amounts of trypsin, followed by the addition of trypsin inhibitor. We found that the recruitment of Syt IV CD to ISGs is trypsin sensitive (). When trypsin inhibitor was added before trypsin, the recruitment to ISG membranes of Syt IV was not reduced. The lumenal protein SgII was not degraded by the trypsin treatment (, bottom), demonstrating that ISGs remained intact during the assay. Although we cannot exclude the activation of phospholipases by trypsin, which would alter the composition of the membrane, our results, together with the recent confirmation of Syt IV's inability to bind phospholipids (), suggest that protein components on ISGs are necessary for the recruitment of the Syt IV CD. Our results suggest that the Syt IV CD is able to inhibit ISG–ISG homotypic fusion via recruitment to ISGs. Syts bind to SNARE proteins to regulate membrane fusion. Syt I, for example, is able to bind Stx1 and SNAP-25 (), and it is thought that this binding is an important facet of the ability of Syt I to regulate exocytosis (). We hypothesized that Syt IV may be part of a protein complex involved in an ISG homotypic fusion that contains Stx6, so we looked for an interaction between Stx6 and Syt IV. In GST pull-down assays from transfected human embryonic kidney 293 (HEK293) cell lysates, we found that myc-Stx6 binds to GST-Syt IV CD and not to GST alone (). Furthermore, we were able to coimmunoprecipitate Syt IV with Stx6 from PC12 cells, showing that endogenous Syt IV and Stx6 interact (). Next, to test whether Syt IV binds directly to Stx6, GST-Syt IV CD was incubated with increasing amounts of recombinant Stx6 CD. Syt IV CD was able to interact directly with Stx6 in a dose-dependent manner (). Together, these results suggest that Syt IV and Stx6 are part of the same protein complex. Previous data showed that Syt VIII, which is another isoform that was initially characterized as a calcium binding–deficient Syt, was able to interact with Stx2 in a calcium-dependent manner (). Therefore, we tested the possibility that Syt IV, which was also thought not to bind calcium (), binding to Stx6 might be promoted by calcium. We tested the binding of GST-Syt IV CD with recombinant Stx6 CD in the presence of either 2 mM EGTA or 1 mM Ca. Syt IV binding to Stx6 was similar in all conditions (), suggesting that the interaction is calcium independent. Furthermore, to investigate which Syt IV domain is required for binding to Stx6, we performed coimmunoprecipitation experiments with lysates prepared from HEK293 cells expressing myc-Stx6 and either HA-Syt IV-C2A, HA-Syt IV-C2B, or HA-Syt IV CD. HA-Syt IV CD could coimmunoprecipitate myc-Stx6 (), as could both HA-Syt IV- C2A and HA-Syt IV-C2B domains (). Similarly, myc-Stx6 could coimmunoprecipitate both HA-Syt IV-C2A and HA-Syt IV-C2B domains (). This result demonstrates that both domains are involved in the binding to Stx6. The binding efficiency of the individual Syt IV domains to myc-Stx6 was lower compared with the binding efficiency of the complete CD, suggesting that the C2A and C2B domains cooperate to bind Stx6, as is the case for Syt I binding to SNAREs (). After our findings that Syt IV binds Stx6 and that both are involved in ISG–ISG fusion, we investigated the effect of inhibitory reagents in the homotypic fusion assay. Our previous results (), along with the results in , show that the inhibition by each reagent individually was at most 50%. Combining reagents in the fusion assay increased inhibition of ISG homotypic fusion to 80% (). The addition of anti-Stx1 antibody, which was previously shown to have no effect on fusion, combined with the Syt IV CD, had no additional effect. Likewise, the addition of anti-Stx6 antibody with Syt I CD had no additional effect, demonstrating that the additive inhibitory effect is specific for Syt IV and Stx6 (). Preincubation of the fusion reaction with anti-Stx6 antibody for 30 min on ice before the addition of Syt IV CD (and vice versa) resulted in the same degree of inhibition of ISG homotypic fusion as when the two components were added simultaneously (unpublished data). To test directly the role of Syt IV in ISG maturation in vivo, and in homotypic fusion in particular, we depleted Syt IV from PC12/PC2 cells using siRNA and asked if ISG–ISG fusion was affected. Using a Syt IV–specific antibody, we observed a reduction of up to 95% in Syt IV levels in >80% of the siRNA-treated PC12/PC2 cells by both indirect immunofluorescence and immunoblotting, whereas Syt I distribution and levels remained unchanged in the absence of Syt IV (). ISG maturation in vivo is characterized by an increase in the size of ISGs on velocity-controlled sucrose gradients that occurs between 15 and 75 min after ISGs are formed from the TGN (). This increase in size reflects ISG–ISG fusion (). Therefore, we tested whether Syt IV was required for the fusion of ISGs in vivo by assaying for an increase in the size of ISGs (in particular the [S]sulfate-labeled granule protein SgII in the ISGs) during a chase of up to 60 min (). As shown in , depletion of Syt IV resulted in an inhibition of the time-dependent size increase of ISGs. During secretory granule maturation in endocrine or neuroendocrine cells, SgII, like many other secretory granule proteins and prohormones, is processed by PC2 in a pH-dependent reaction, which starts in the TGN and continues into the MSG (). We investigated whether PC2 processing of SgII was affected by the inhibition of fusion because of the absence of Syt IV by observing the processing of newly synthesized SgII in Syt IV siRNA-treated PC12/PC2 cells. In PC12/PC2 cells, after exit from the TGN and during the chase period, SgII is processed at several sites by PC2 to generate 38-, 28-, and 18-kD (p18) products (). This processing can be detected after a short pulse of [S]sulfate to label the TGN pool, followed by a chase with excess sulfate. In the absence of Syt IV, we observe reduced processing of the full-length 86-kD SgII (). The extent of this inhibition was determined from the ratio of p86 to p38 (). Before 30 min of chase, very little processing is detectable in this pulse-chase analysis, whereas maximum inhibition occurs between 45 and 60 min of chase, when in the mock-treated cells the rate of processing is highest. We conclude from these results that in PC12 cells Syt IV is required for fusion of ISGs and efficient processing of SgII. Next, we investigated the role of Syt IV in ISG maturation in vivo using the Syt IV CD, in particular, to determine if the reduction in SgII processing was a direct result of an inhibition of maturation or an alteration in PC2 sorting. As mentioned in the previous section, SgII in PC12/PC2 cells is processed to p18. p18 can be detected with an anti-p18 antibody recognizing only p18 and not larger SgII precursors. We used wild-type PC12 cells transfected with PC2 to eliminate the background signal from p18 stored in MSGs existing before addition of the dominant-negative HA-Syt IV CD, and asked if the appearance and accumulation of p18 in MSGs was affected by expression of the Syt IV CD. Transfection of the HA-Syt IV CD together with PC2 into PC12 cells resulted in an inhibition of p18 accumulation, whereas transfection of the Syt I CD together with PC2 or of PC2 alone, had no effect on p18 appearance (). Quantification of the intensity of the p18 signal in an equivalent number of cells cotransfected with HA-Syt IV CD and PC2 revealed a threefold decrease in p18 in these cells compared with cells transfected with PC2 alone or with Syt I CD and PC2 (). Because we observed a decrease in SgII processing using the Syt IV CD, we tested whether excess full-length Syt IV increased p18 levels. Using FACS analysis, there was no difference in p18 in untransfected PC12/PC2 cells or cells transfected with full-length Syt IV or I (unpublished data). We speculate that the lack of a stimulatory effect on maturation by excess Syt IV may be attributable to a limitation of Syt IV effectors that are necessary for granule maturation, and which may also need to be overexpressed. We have also tested whether the effect of Syt IV on p18 production is calcium-dependent by mutating the first and second aspartate residues within the C2B domain to asparagine (D318/324N); these mutations were previously shown to block glutamate release in glial cells (). Expression of this double mutant or full-length Syt IV together with PC2 did not affect p18 production in PC12 cells (unpublished data), suggesting that the dominant-negative effect on SgII processing observed with HA Syt IV CD is independent of calcium. PC2 is found as a 75-kD precursor in the TGN, which is then endoproteolytically cleaved to the 65-kD mature form in a pH-dependent manner (). PC2 activation depends on propeptide cleavage in the presence of bound 7B2, which is a member of the granin family shown to behave as a chaperone for pro-PC2 (). We asked if inhibition of SgII processing by the dominant-negative Syt IV CD might be a result of failure to activate PC2. Therefore, we looked at PC2 in PC12 cells cotransfected with the HA-Syt IV CD and PC2. To do this, the doubly transfected cells and PC12/PC2 cells were FACS-sorted using anti-HA and anti-PC2 antibodies. In the doubly transfected cells expressing HA-Syt IV, there was a higher ratio of pro-PC2 to PC2 than in PC12/PC2 cells (). The processing defect of pro-PC2 is not a result of its failure to enter granules because in the doubly transfected cells PC2 colocalized with the MSG marker chromogranin B (CgB; ). Furthermore, when PC12 cells expressing the HA-Syt IV CD were examined by EM after immunolabeling with an anti-HA antibody, compared with wild-type cells, the Golgi was not fragmented and secretory granule formation was not perturbed in these cells (unpublished data). In this study, we used a specific anti–Syt IV antibody () and subcellular fractionation approaches in undifferentiated PC12 cells to demonstrate that Syt IV is found on ISGs and is absent from MSGs, confirming and extending previous studies (Eaton et al., 2000; ). The CD of Syt IV inhibited in vitro ISG homotypic fusion, suggesting that Syt IV is involved in secretory granule maturation. siRNA depletion demonstrated that Syt IV is required for ISG–ISG fusion in vivo. A previous study has shown that microinjection of recombinant Syt IV in endocrine β cells did not effect calcium-evoked insulin secretion (). Similarly, catecholamine release from “cracked” PC12 cells was not changed by the addition of recombinant Syt IV (). Collectively, these data suggest that the ISG-localized Syt IV is not a regulator of calcium-triggered exocytosis in undifferentiated PC12 cells, but, rather, is involved in regulating ISG–ISG membrane fusion. Interestingly, we found that the MSG-localized isoforms Syt I, IX, and VII did not have any effect on the regulation of ISG–ISG fusion. Our data are in agreement with studies showing that Syt IV displays a different distribution to Syt I, VII, and IX in PC12 cells (; ). Moreover, Syt IV lacks the ability to heterooligomerize with Syt I (; ) and does not copurify with Syt IX (). Similarly, Syt VII was found to form complexes with Syt I and IX, but not with Syt IV (; ). Syt I, VII, and IX were shown to be involved in calcium-regulated exocytosis in PC12 cells, leading to the idea that these isoforms regulate different membrane fusion events from those involving Syt IV. The in vitro binding assay showed that the Syt IV CD is recruited to ISG membranes, which suggests that the site of inhibition is on these membranes. Syt IV CD bound specifically to ISGs and this binding is strictly via protein components, as the recruitment was completely inhibited after the pretreatment of ISG membranes with trypsin. Our experiments also confirm the data showing that Syt IV, in contrast to Syt I, is not a phospholipid-binding isoform (). The proposal that the Syt IV CD functions on the membrane concurs with data on the mode of inhibition of the Syt I CD in vesicle fusion with the plasma membrane (; ), where the association of the added Syt I CD with Stx1 and SNAP-25 inhibited the assembly of the endogenous Syt I–SNARE complexes. In the same way, the CD of Syt IV might bind to ISG SNAREs, thereby inhibiting the formation of an endogenous Syt IV–ISG SNARE complex, leading to inhibition of ISG–ISG fusion. The recruitment of the Syt IV CD to ISG membranes, as well as its inhibitory effect on fusion, led us to ask if Syt IV and Stx6 colocalize and interact. Immunostaining of PC12 cells with anti-Stx6 and anti–Syt IV antibodies showed colocalization mainly in the juxtanuclear area, which is unlikely to be endosomes because, unlike Stx6 (), Syt IV is not found on endosomes (). The immunoprecipitation of endogenous Syt IV with Stx6, and direct binding of the recombinant proteins to each other in vitro, suggested that these two proteins are in the same complex and could be regulating the same fusion event. We found that both Syt IV-C2A and C2B domains bind Stx6, and that they most likely cooperate to bind to Stx6. Syt I was also found to bind to SNAP-25 and Stx1 through its C2A and C2B domains (; ), and both domains are thought to cooperate to regulate fusion (). It would be interesting to investigate whether an individual domain or both domains are required for ISG homotypic fusion. Unfortunately, because of the insolubility of the GST-C2A and C2B domains, we could not investigate whether individual domains have an inhibitory effect. It is unlikely that the formation of a Syt IV–Stx1–SNAP-25 complex (; ) is involved in ISG fusion, as ISG fusion is not sensitive to the addition of botulinum neurotoxins (). To understand the mechanism by which Syt IV is regulating ISG homotypic fusion, and, more specifically, whether inhibition with the Syt IV CD is synergistic with the inhibition by Stx6 antibodies, we added both reagents to the fusion assay, which resulted in an additive inhibition of ISG–ISG fusion. One explanation for this result is that the fusion-competent ISGs are at different stages when isolated, and, therefore, some of the ISGs will be past the stage blocked by one or the other reagent. Our results support the notion that Syt IV and Stx6 function at different stages, rather than synergistically at the same stage of ISG–ISG fusion. We found that the binding of Syt IV CD to ISG membranes was not inhibited by the addition of anti-Stx6 antibody (and vice versa; unpublished data). This result suggests that the Syt IV–binding site in Stx6 is different from the epitope for the anti-Stx6 antibody. Our data cannot exclude the possibility that Syt IV and Stx6, although in the same complex, could function at different steps during tethering, docking, and fusion of ISGs. Most importantly, we were able to demonstrate a role for Syt IV in the maturation of ISGs in vivo. First, using siRNA depletion of Syt IV we found that Syt IV was required for ISG fusion and for efficient processing of newly synthesized SgII by PC2. Second, the dominant-negative Syt IV CD inhibited SgII processing by exogenous PC2 in PC12 cells. The greatest reduction in processing occurred when the processing rate was maximal in untreated controls at 45–60 min. of 45 min for fusion and maturation of ISGs in PC12 cells. Recently, showed that expression of a soluble form of Stx6 in INS-1 β cells slowed proinsulin processing and concluded that this reduction in processing was caused by an indirect effect on biosynthetic traffic. The inhibition of SgII processing that we observe is probably a direct result of an inability to activate PC2 in the maturing ISGs because PC2 was found mainly in the unprocessed proform in cells transfected with the dominant-negative Syt IV CD. We have been unable to test PC2 activation after siRNA treatment because of a lower than usual transfection efficiency after siRNA treatment, which precludes expressing PC2 in the Syt IV–deficient PC12 cells. As PC2 autocatalytic activation is dependent on acidification of the granules (), which occurs during the maturation from ISGs to MSGs, we speculate that the reduction in pro-PC2 processing could be caused by a failure to acidify the ISGs. Although the mechanism of secretory granule acidification is not known, it has been proposed to occur after an increase in the density of the vacuolar H-ATPase, as well as a decrease in the H permeability (). Perturbation of the vacuolar H-ATPase has also been shown to reduce processing in intermediate pituitary cells (). We propose that the Syt IV CD, through an inhibition of homotypic fusion, which may also limit membrane remodeling, may inhibit acidification. Further biochemical and morphological analyses will be used to study the regulation of the secretory granule pH in vitro, and should provide further information on how secretory granule biogenesis occurs. PC12 cells (clone 251) and PC12/PC2 cells stably expressing PC2 were described previously (). HEK293A cells were purchased from Invitrogen, BAPTA was obtained from Calbiochem, and [S]sulfate was purchased from GE Healthcare. All other reagents were obtained from Sigma-Aldrich. Polyclonal anti–Syt IV antibody was a gift from M. Fukuda (Institute of Physical and Chemical Research Institute, Saitama, Japan). Monoclonal anti-p18 and polyclonal anti-Stx6 antibody were previously described (; ). Anti-PC2 antibody was a gift from B. Eipper (University of Connecticut, Farmington, CT). Monoclonal anti-Stx6 antibodies were purchased from BD Biosciences and Abcam PVC (3D10), and monoclonal anti–Syt I antibodies were obtained from Synaptic Systems GmbH. Mouse anti-HA and anti-myc antibodies were developed in-house (Cancer Research UK, London, UK). Rat anti-HA was obtained from Roche, and rabbit anti-myc was obtained from Abcam PVC. GST-Syt I CD was obtained from G. Schiavo (Cancer Research UK). GST-Syt VII and GST-Syt IX CDs were obtained from M. Fukuda. Rat Syt IV CD (aa 37–425), Syt IV-C2A domain (aa 150–261), and Syt IV-C2B domain (aa 281–245) were amplified by RT-PCR from RNA extracted from PC12 cells. GST-Syt IV CD was made by inserting the DNA into the SmaI–NotI sites of pGEX-2T expression vector, and the expressed proteins were purified using glutathione beads (GE Healthcare). To eliminate negatively charged bacterial contaminants, the GST fusion proteins were further purified on a size exclusion column, and then on a cation exchange column (). HA-Syt IV CD and HA-C2A and HA-C2B domains were inserted into pcDNA 3.1 (Invitrogen). PNS was prepared from [S]sulfate-labeled PC12 cells and then loaded successively on a continuous velocity gradient, followed by a discontinuous equilibrium sucrose gradient to separate ISGs and MSGs, as previously described (; ). ISG–ISG homotypic fusion was performed as previously described (). In brief, complete fusion reactions are comprised of the following: 100 μl [S]sulfate-labeled PNS from PC12 cells, 10 μl ISGs purified from PC12/PC2 cells, an ATP-regenerating system and purified recombinant Syt IV, I, VII, or IX CDs, affinity-purified polyclonal anti-Stx6 antibody, or BAPTA as indicated in the figure legends. Fusion was carried out for 30 min, followed by the processing reaction for 90 min at 37°C. The product of PC2 cleavage of SgII, which is S-labeled p18, was immunoprecipitated and subjected to SDS-PAGE and autoradiography. The amount of p18 was quantified using ImageJ (National Institutes of Health) analysis software. Binding assays were performed as previously described (). The amount of bound Syt IV or I CDs was detected using anti–Syt IV and anti–Syt I antibodies, and was quantified using ImageJ analysis software. PC12 cells were seeded on poly--lysine–coated coverslips and transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were fixed 24 h after transfection with 3% paraformaldehyde, permeabilized with 0.2% Triton X-100, and stained with the appropriate primary and secondary antibodies diluted in PBS with 0.2% gelatin. Images were acquired by confocal microscopy using a confocal microscope and software (LSM510; Carl Zeiss MicroImaging, Inc.). For the quantification of p18 levels, confocal images were taken with identical acquisition parameters and the mean of intensity of p18 was measured in individual cells using Photoshop 7.0 software (Adobe). To obtain a homogenous population of cells expressing both HA-Syt IV and PC2, or PC2 alone, FACS-sorting was used. Transfected cells were fixed and stained the same as the PC12 cells, and sorted using the Moflo FACS sorter (DakoCytomation). 150,000 HA/PC2-positive cells (starting from 2 × 10 cells) and 300,000 PC2-positive cells were pelleted, lysed in sample buffer, and subjected to SDS-PAGE analysis. To monitor SgII processing, PC12/PC2 cells were used. To control for sorting and budding from the TGN, PC12 cells were used. Pulse-chase [S]sulfate-labeling of PC12 cells and analysis of SgII were performed as previously described (). Fig. S1 shows by indirect immunofluorescence that Syt IV colocalizes with Stx6, but not with p18, in PC12/PC2 cells. Fig. S2 shows that BAPTA inhibits ISG homotypic fusion. Fig. S3 shows that Golgi morphology and secretory granule biogenesis are normal in cells treated with Syt IV-siRNA. Online supplemental material is available at .

Terminal differentiation within the interfollicular epidermis ultimately leads to the formation of a functional skin barrier that protects organisms from water loss, infections, and various insults. This barrier is continuously renewed throughout the animal's postnatal life span as a result of the presence of stem cells that are capable of self-renewal and of producing transiently amplifying progenitor cells, which subsequently exit the cell cycle and embark on a terminal differentiation pathway as they migrate toward the skin surface (). This process is recapitulated during embryogenesis, when a single layer of multipotent surface ectodermal cells develops into a stratified epidermis. Keratins 5 (K5) and 14 (K14), which are expressed in the proliferating basal layer of the mature epidermis, are among the earliest epidermal markers that are activated in the single-layered ectoderm of the embryonic skin (; ). As development proceeds, the presumptive suprabasal layers arise and express differentiation-specific keratins K1 and K10 in the intermediate spinous layers as well as loricrin, a major component of the future cornified envelope (), in the upper granular layers. Associated with a spatiotemporally ordered change of gene expression are morphological transformations and biochemical events culminating in the formation of a barrier at the outermost layer of the epidermis that is impermeable after embryonic day (E) 17 (). Transcriptional regulation is key to a successful epidermal development/differentiation program (for review see ). Among transcription factors that control the balance between proliferation and differentiation of keratinocytes, the c-myc proto-oncoprotein and the inhibitor of differentiation (Id) family of proteins surfaced as positive regulators of a cycling and nondifferentiating progenitor state. However, little is known about how the expression of these factors is regulated in skin. Although existing studies have provided insights into how homeostasis is achieved in mature epidermis, few examine the genetic pathways and molecular mechanisms that govern the growth and differentiation of stem/progenitor cells of the developing epidermis (; ). is an evolutionally conserved family of genes encoding CH zinc finger transcription factors in animals. Functional studies in , , and mice suggest that this gene family plays important roles in the development of epithelial tissues and germ cells (; ; ; ; ). Genetic and biochemical studies suggest that at least two members of this gene family, and , act downstream of the Wnt– β-catenin–lymphoid enhancer factor/T cell factor signaling pathway (; ). Recently, was identified as a downstream target of the TGF-β/BMP7–Smad4 signaling pathway, a growth-inhibitory pathway in keratinocytes (). Therefore, the gene family members appear to be important integrators of upstream developmental signals and key regulators of epithelial development and differentiation. , the first mouse that was functionally characterized, is expressed in multiple somatic epithelial tissues, including skin (hair follicles and interfollicular epidermis) and kidney, as well as in the male germinal epithelium (). -deficient mice showed ruffled hairs, cystic kidneys, and defective spermatogenesis (). In this study, we describe a functional requirement for in epidermal development. Specifically, we show that is required to restrict the proliferation potential of embryonic epidermal progenitor cells in vivo and in vitro. We also present molecular evidence indicating that Ovol1 represses the expression of c-myc by direct binding to its promoter, providing a possible mechanism by which Ovol1 regulates the proliferation arrest of developing epidermal cells. The initial characterization of -deficient mice was performed in a 129Sv × C57BL/6 (B6) mixed (50:50) genetic background (). Upon close examination, we noticed that the epidermis of these mutant animals was often slightly thicker than that of the wild type (unpublished data). intercrosses. We note that previously described phenotypes, including ruffled hairs and cystic kidneys, persisted in this new background and that a subset of the -deficient pups died perinatally, with the surviving ones exhibiting flaky skin (unpublished data). During normal epidermal development, presumptive suprabasal cells appear at ∼E15.5 and are morphologically distinct from the underlying presumptive basal cells (). Different from those in mature skin, these developing suprabasal cells express differentiation marker K1 but retain their proliferative potential for another 2–3 d (see below; ; ) and are, therefore, embryonic epidermal progenitor cells. epidermis at E15.5, the morphological distinction between the presumptive suprabasal and basal layers was not apparent in many areas (), and more mitotic figures were seen than the wild type (, arrow; also see below). By E16.5, the epidermis, where a morphological stratification had now become obvious, was considerably thicker than the controls, resembling acanthosis described in human patients (). This defect was not caused by a transient delay in development, as it was also observed at later stages (). Furthermore, there was impaired enucleation, flattening, and compaction of the developing granular cells in mutant epidermis (), suggesting subtle, late differentiation defects. epidermis at all stages examined (unpublished data). mutant epidermis was caused by an expansion of the presumptive basal or suprabasal layers, we stained the developing epidermis for K14, K1, and loricrin. As development proceeded from E15.5 to the newborn stage, mutant epidermis acquired the expected spatial arrangement of K14, K1, and loricrin- positive layers (). However, subtle abnormalities were seen such that at E15.5, the loricrin-positive layer was closer in space to the K14-positive basal layer than that in the wild type (; compare with C and E). More importantly, beginning at E16.5, the mutant K1-positive layers were significantly expanded compared with their wild-type counterparts (). A slight expansion of the loricrin-positive layers was also observed (), but this was unlikely caused by a general expansion of the presumptive granular layers, as no expansion was observed for cells that were positive for transglutaminase 3 (TG3), another marker for granular cells (). No consistent expansion of the K14-positive layer was observed in the mutant. The mutant defects were not associated with any apparent change in the expression of K6 (unpublished data), which is generally regarded as a wound-healing keratin (). Collectively, these data demonstrate that the K1-positive embryonic progenitor cell population was expanded in the absence of , which underlies the thickening of the mutant interfollicular epidermis. epidermis led us to wonder whether the developmental transition between proliferation and differentiation was affected. As measured by BrdU incorporation, epidermal proliferation in control embryos slows down as development proceeds (, compare E15.5 with E16.5; ). In its extreme, very little proliferation was observed on the dorsal side of the E16.5 wild-type epidermis (), which is consistent with the fact that the barrier first forms in this region (). In the mutant, an increase in the number of BrdU-positive cells was already evident at E15.5 ( and not depicted). By E16.5, mutant epidermis still retained a considerable level of proliferation both dorsally and ventrally (), with the difference between wild type and mutant most prominent in the suprabasal compartment of the dorsal region (; compare with B). Using an antibody that detects phosphorylated histone H3 (H3-P), a marker for cells in mitosis (), we observed a higher average mitotic index in the mutant than the wild type, particularly at later developmental stages ( and not depicted). In keeping with these in vivo findings, the initial plating of comparable numbers of primary keratinocytes isolated from newborn pups consistently yielded more attached and growing keratinocytes for the mutant when compared with their wild-type control littermates (). A priori, multiple possibilities may account for the observation of an increased number of proliferating keratinocytes in mutant epidermis. Among these are the precocious activation of a population of presumably slow-cycling embryonic epidermal stem cells, enhanced rate of proliferation of transit-amplifying progenitor cells, or impaired growth arrest of progenitor cells. expression was observed in the presumptive suprabasal layers but was barely detectable in the presumptive basal layer () where embryonic epidermal stem cells presumably reside, making the first possibility highly unlikely. FACS analysis identified a similar percentage of cells that are positive for α6 integrin, a marker for basal keratinocytes (), in keratinocyte preparations from wild-type and -deficient newborn epidermis (unpublished data). Furthermore, a growth analysis of actively proliferating keratinocytes of secondary passages in the absence of any growth-inhibitory treatment revealed no difference between the wild type and mutant (). Together, these results argue against an involvement of in embryonic epidermal stem cell activation and in regulating the rate of keratinocyte proliferation itself. Several treatments have been shown to induce the exit of proliferation and possibly terminal differentiation of cultured keratinocytes. These include high concentrations of Ca (), LiCl, which is thought to mimic activated canonical Wnt signaling by inhibiting GSK3β (), and TGF-β (). Therefore, we examined the ability of -deficient keratinocytes to exit proliferation in response to these reagents. First, wild-type and mutant keratinocytes were cultured in the presence of low (∼0.09 mM) or high (1.2 mM) Ca, and their BrdU-labeling index was determined. Although a time-dependent decrease in the number of BrdU-labeled cells was observed for the wild type upon Ca treatment, the BrdU-labeling index remained high in -deficient keratinocytes even 24 h after Ca addition (). Second, LiCl, which induced efficient growth arrest of wild-type keratinocytes (as indicated by the <10% confluency of treated plates at the time when untreated replicate plates reached 100% confluency), failed to do so with the mutant cells (). Finally, a concentration of TGF-β that caused a significant reduction in the BrdU-labeling index in wild-type cells failed to cause -deficient keratinocytes to stop cycling (). Collectively, our results indicate that the loss of renders the proliferating keratinocytes in culture less sensitive to extrinsic growth-inhibitory signals. The inability of -deficient epidermal progenitor cells to exit proliferation implies that normally functions in the developing epidermis to ensure the growth arrest of these cells. As it has been shown that a down-regulation of c-myc expression in suprabasal cells is important to maintain a postmitotic status (; ), we hypothesized that Ovol1 may function by down-regulating c-myc expression. To test this hypothesis, we performed in situ hybridization experiments on developing epidermis using a c-myc cRNA probe. Although hybridization signals were observed in all layers of the interfollicular epidermis of both wild-type and -deficient E15.5 embryos, a slight reduction in signal intensity was often apparent in the newly formed presumptive suprabasal layers of the wild type but not mutant (). As development proceeded to E18.5, the intensity of c-myc hybridization signals became weaker overall, with only a few scattered basal cells showing detectable expression (). In the mutant, however, many suprabasal cells showed clearly detectable hybridization signals, including those that are close to the skin surface (, arrows). The number of c-myc–expressing basal cells was also significantly higher. Additionally, our analysis of c-myc protein expression in embryonic skin revealed differences between the wild-type and mutant epidermis that are similar to those observed at the RNA level (). Of particular note is that although no staining was observed in the wild-type suprabasal cells at E18.5, those in the mutant epidermis retained strong nucleolar signals that are characteristic of the c-myc protein (, arrowheads; ; ). To better quantify the difference in c-myc expression, we performed Northern blot analysis on RNA isolated from E16.5 embryonic skin and observed an ∼1.7-fold higher level of c-myc transcripts in the mutant than the wild type (). The expression of Id2, which was previously shown to be a target of c-myc transcriptional activation in skin and a target of Ovol1 transcriptional repression in the testis (; ; ), was also up-regulated by ∼1.5-fold in -deficient skin (). Although a failure to down-regulate c-myc expression in the mutant presumptive suprabasal cells could be a primary consequence of the loss of as it is expressed in these cells, the increase of c-myc expression in basal cells that normally do not express an appreciable amount of is curious and may be a secondary effect. To begin to distinguish between primary and secondary changes, we isolated primary keratinocytes from wild-type and mutant newborns and performed real-time PCR analysis to determine the level of endogenous c-myc transcripts in these isolated epidermal cells. No statistically significant difference was observed between wild-type and mutant keratinocytes that were basal-like when they were cultured under undifferentiating conditions (unpublished data). However, after these cells were treated with differentiation-inducing Ca, a significantly higher level of c-myc transcripts was detected in the mutant preparations (). To determine whether the increase in the c-myc transcript level was a result of increased transcription, we cloned a 2.3-kb human c-myc promoter fragment upstream of a luciferase reporter gene (see ) and transfected the reporter construct into wild-type and mutant keratinocytes. Again, Ca-treated mutant keratinocytes allowed higher promoter activity than the wild type, whereas no difference was observed in untreated cells ( and not depicted). Collectively, our results strongly suggest that is required to down-regulate c-myc transcription during epidermal development in the growth-restricted suprabasal cells. Does Ovol1 protein directly repress c-myc transcription, or are intermediate factors involved? To address this issue, we turned to study the nucleotide sequence determinants of Ovol1–DNA interaction. Previously, we showed that Ovol1 is able to bind to a Ovo consensus sequence (). Since then, we have identified mouse genomic sequences to which Ovol1 binds in vitro and arrived at a putative consensus motif, CCGTTA (unpublished data). Although single nucleotide mutation in this motif (CCGTTA: C→A, C→T, G→T, T→G, T→G, or A→C) resulted in diminished Ovol1 binding (), deletion or scramble of this hexamer motif totally abolished binding (). On the other hand, the insertion of a CCGTTA sequence into an oligonucleotide to which Ovol1 does not bind (; ) was able to confer binding (). Together, these results indicate that a CCGTTA sequence is necessary and sufficient to confer high-affinity Ovol1–DNA interaction in vitro. We next examined the sequences of gene-regulatory regions of both mouse and human c-myc genes and found that both contained a single CCGTTA motif, the position of which is also conserved (). In gel shift assays, recombinant Ovol1 bound to oligonucleotide , which contains human c-myc sequences including the CCGTTA motif (, lanes 2–6). The identity of the Ovol1–DNA complex was confirmed by the observation of a band supershift when anti-Ovol1 antibody was added to the binding reaction (, lane 7), and binding was abolished when the CCGTTA sequence was mutated (, lanes 9–13). Together, these results identify a bone fide Ovol1-binding site in the c-myc promoter. We next used chromatin immunoprecipitation (ChIP) assays to determine whether Ovol1 was engaged at the endogenous human c-myc promoter (see Materials and methods). The anti-Ovol1 antibody-precipitated DNA was subject to PCR amplification with five pairs of primers spanning five long sequence blocks of the c-myc promoter that are conserved between mice and humans (). Ovol1 occupancy was detected with primer pairs Cm-2, Cm-3, Cm-4, and Cm-5 but not with Cm-1 (). No signal was observed when normal rabbit IgG was used for precipitation, showing the specificity of this assay. Quantitative analysis revealed two peaks of Ovol1 binding: one encompassed the CCGTTA sequence (primer set Cm-2), and the other was located in the proximal promoter region, which overlaps the sixth conserved block (primer set Cm-5; ). These results indicate that Ovol1 physically associates with the endogenous c-myc promoter at the predicted distal site inside cells; however, they also reveal additional sites in the proximal promoter region that are occupied by Ovol1. We performed gel shift experiments on overlapping fragments encompassing this entire proximal region but observed no evidence of binding of recombinant Ovol1 to any fragment in this in vitro assay (unpublished data), raising the possibility that an alternative mechanism (e.g., assistance from auxiliary factors) exists in vivo to bring Ovol1 protein to this region (see Discussion). Having established that Ovol1 binds to the c-myc promoter in vitro and in cells, we next used reporter assays to investigate whether Ovol1 directly represses c-myc transcription. The 2.3-kb human c-myc promoter fragment was able to direct active transcription in both UG1 mouse keratinocytes () and 293T cells (). Cotransfection of an expression vector repressed reporter expression in both culture systems in a dosage-dependent manner (). To confirm that the observed repression is an Ovol1 protein–dependent event, we created the chimeric protein VP16-Ovol1 in which Ovol1 was fused to a strong, well-characterized transactivation domain from VP16 with the assumption that this activation domain might override the intrinsic transcriptional regulatory activity of Ovol1 and result in an activator. Indeed, VP16-Ovol1 activated the c-myc promoter in a dosage- dependent manner (). In contrast, the VP16 domain alone had no effect at all concentrations tested, indicating that the fusion protein was recruited to the c-myc promoter by its Ovol1 moiety. The specificity of the effect of Ovol1 was further demonstrated by the finding that a truncated Ovol1 protein (d15-Ovol1) lacking the first 15 amino acids at the NH terminus, which resembles the known repression domain SNAG (), failed to efficiently repress the c-myc promoter activity (). To determine whether repression depends on the DNA binding ability of Ovol1, we generated a construct expressing a mutant form of the Ovol1 protein in which the cysteine amino acids in the first three zinc fingers were replaced by alanine (ZnFC2A). This mutant protein, which is no longer able to bind DNA (not depicted), failed to repress the c-myc promoter (). To further explore the dependence of repression on DNA binding, we generated mutant promoters in which upstream sequences were deleted or mutated. Although the deletion of sequences from −2.3 to −1.6 kb had no effect (not depicted), a partial release of repression was observed when a 373-bp sequence containing the CCGTTA site was removed (). Moreover, replacing the CCGTTA motif with a non-Ovol1 binding sequence, ATGCGC, led to a similar reduction in repression by Ovol1 (promoter construct mut-1.6P in ), confirming that this site is indeed required for mediating Ovol1 repression. This said, considerable residual repression was still observed, implicating the contribution of other cis-elements. Analysis of additional deletion constructs mapped the minimum Ovol1 responsive region to within the smallest promoter fragment tested (the 0.1P construct; ; and not depicted). The position of this region coincides with that of the proximal Ovol1-binding site identified by the aforementioned ChIP assays, implying that Ovol1 repressed this minimum promoter by binding to it. Collectively, our data suggest that Ovol1 represses c-myc transcription by binding to its promoter. The apparent expansion of loricrin-positive layers in -deficient skin in the absence of any concomitant expansion of the TG3-positive layers raises the possibility that loricrin expression is aberrantly activated in intermediate suprabasal cells when is ablated. We next used Western blot analysis to determine loricrin protein levels in wild-type and mutant skin from different developmental stages. Although mutant skin started out expressing a slightly lower level of loricrin protein (75% of that in wild type at E15.5 after normalization against actin levels), possibly because of a transient delay in the proliferation–differentiation switch, it produced a much higher level of the protein at later stages (approximately fourfold higher than the wild type at E18.5; ). Using semiquantitative RT-PCR, we detected a twofold increase in loricrin RNA levels in the mutant skin taken from E16.5 embryos (), confirming that increased loricrin expression occurred at a transcriptional level. To better understand the effect of ablation on loricrin expression, we examined the effect of Ovol1 protein on loricrin promoter activity using reporter assays. We cloned a mouse loricrin promoter fragment () upstream of a luciferase reporter gene and found that this fragment directed active transcription in both UG1 and 293T cells (). This observation is consistent with previous findings that a core loricrin promoter fragment directs expression in not only granular cells but also basal and spinous cells of transgenic mice (). Cotransfection of an expression vector repressed loricrin promoter activity in both cell systems (). However, at high Ovol1 concentrations, we often observed a release of repression, which was likely a result of “squelching” caused by Ovol1 binding to specific trans-acting factors that might be limiting in these cells. To confirm that the observed regulation is an Ovol1 protein–dependent event, we tested VP16-Ovol1, assuming that this chimeric activator would likely bypass the requirement for those limiting factors needed for repression. Indeed, VP16-Ovol1 activated loricrin promoter–luciferase reporter expression in a clearly dosage-dependent manner, whereas the VP16 activation domain alone had no significant effect (). A CCGTTA sequence was found in the loricrin promoter; however, this site is very close to the transcription start so that its mutation abolished promoter activity, preventing us from examining the DNA site dependence of Ovol1 repression (unpublished data). Nonetheless, our data demonstrate that Ovol1 is capable of repressing loricrin expression in a cell-autonomous manner. Our studies identified a novel function of in epidermal development. The expansion of K1-positive layers and increase in the number of actively proliferating cells in the developing epidermis of -deficient mice are consistent with a hyperproliferative defect. A priori, hyperproliferation may occur as a primary cell-autonomous consequence of ablation in the proliferative compartment or as a secondary consequence of abnormal terminal differentiation and/or barrier defects, as frequently observed in the study of adult epidermis. Several lines of findings support the former possibility. First, increased proliferation was already evident at E15.5, which is before late-terminal differentiation events occur and a barrier forms. Although a transient delay in barrier acquisition was observed in -deficient embryos, all were fully impermeable to dye penetration at the age of E18.5–newborn (unpublished data), suggesting that barrier development is largely normal (). Late-differentiation defects and delayed barrier formation were also observed in loricrin knockout mice, yet loricrin-deficient epidermis is not acanthotic (), indicating that these defects are not sufficient to trigger a compensatory hyperproliferative response in the embryo. Second, the expression of K6, which is typically up-regulated in repair-associated hyperproliferation (), was not affected in the developing mutant epidermis. Finally, the most direct evidence came from our observation that keratinocytes isolated from mutant animals could not be efficiently induced to exit the cell cycle in response to extrinsic growth-inhibitory signals such as Ca, LiCl, and TGF-β. As these cells were cultured independently of feeders, this analysis allowed a direct assessment of the intrinsic proliferative capacity of the mutant epidermal cells. Based on these results and the onset of expression in early presumptive suprabasal cells (), we propose that is required for the growth arrest of embryonic progenitor cells during epidermal development (). It is worth mentioning that similar to the epidermis, -deficient male germ cells were sluggish in exiting mitosis (). Therefore, might play a general role in down-regulating the proliferation of developmental progenitor cells in tissues that require its function. While this study was in revision, published that epidermal stratification during mid–late gestation, where p63 is critically involved (; ), entails asymmetric divisions of the ectodermal cells, yielding a basal cell and a suprabasal cell with a short-lived proliferation potential. expression is activated immediately upon stratification in these transit-amplifying suprabasal layers (). Although stratification occurred in the absence of , the newly formed mutant suprabasal cells appeared different in size and shape from their wild-type counterparts despite their ability to express K1 ( and ). Is it possible that is an intrinsic molecular “clock” built into the suprabasal daughter cell that somehow makes it different from its long-lived parent and its postmitotic successors in terms of proliferation potential, so that in its absence, this daughter cell is improperly programmed upon stratification and, therefore, carries out just a few more rounds of proliferation than it should? What signals lie upstream of ? Existing evidence tentatively places downstream of Wnt as well as TGF-β1/BMP signaling. The suprabasal hyperproliferative phenotype of -deficient epidermis is similar to that observed in Ikkα-deficient mice (; ) and in mice with the repeated epilation () mutation, which is a mutation in 14-3-3σ (). However, loricrin expression is completely blocked in the absence of Ikkα or in mice but only slightly delayed in the absence of , suggesting that lies downstream of Ikkα and 14-3-3σ in the epidermal differentiation process. It is tempting to speculate that might be a key integrator of upstream developmental signals/molecular triggers like Wnt, TGF-β/BMP, and Ikkα in negative growth regulation during embryonic development. As little is known about the signaling and transcriptional network regulating the proliferation to differentiation transition during epidermal development, our elucidation of an in vivo role for , a target of well-known signaling pathways, offers interesting new angles to understand these cellular and developmental processes. How does Ovol1 down-regulate the proliferation of embryonic epidermal progenitor cells? We probed this important question by characterizing the DNA binding specificity of the protein and looking for possible downstream targets. These studies led us to the discovery of c-myc as a direct Ovol1 target. Repression of the c-myc promoter by Ovol1 in reporter assays as well as results of ChIP assays detecting a physical association of Ovol1 to the endogenous c-myc promoter in its chromatin context indicate that Ovol1 can directly repress c-myc expression. This provides at least one possible molecular mechanism by which Ovol1 down-regulates proliferation (). The up-regulation of c-myc expression in -deficient suprabasal cells as well as the phenotypic parallel, namely abnormal suprabasal proliferation, between -deficient mice and transgenic mice that overexpress c-myc under the suprabasal-expressing involucrin or loricrin promoter provide in vivo validation for this model (; ; ). The underlying mechanism by which Ovol1 is recruited to the c-myc promoter appears complex, as we found two regions in the promoter that mediate Ovol1 repression and are bound by Ovol1 inside cells, yet only one region contains a detectable in vitro binding site. Alternative mechanisms, such as the use of DNA-binding partners to enhance binding affinity/specificity or via protein–protein interactions (), likely exist as additional means to recruit Ovol1 to its target promoters in vivo. Future work is necessary to systematically explore these “hidden” cis-elements to fully understand the biochemical mechanism of Ovol1 repression. It is unlikely that c-myc serves as the only Ovol1 target to mediate its negative effect on proliferation. Our previous study on the role of in male germ cell differentiation identified Id2 as a direct target of the Ovol1 protein (). The observation of increased Id2 expression in -deficient skin suggests that Id2 is also repressed by Ovol1 during epidermal development, probably both directly by Ovol1 binding to its promoter and indirectly because of increased c-myc gene products, as it has been shown that c-myc induces Id2 expression in epidermis (). As both c-myc and Id2 have been implicated in tumorigenesis, these findings also raise the possibility that might play a negative role in malignant growth. Although the biological function of genes has been studied in various organisms, our study reports the first identification of candidate molecular targets of , namely c-myc and Id2, two key positive regulators of proliferation and negative regulators of differentiation, that bear relevance to the cellular process that they regulate. Is also required for terminal differentiation itself in the developing epidermis? The expression of differentiation markers such as K1, loricrin, and TG3 is detected in the mutant, indicating that the epidermis is able to execute a largely normal terminal differentiation process in the absence of a functional gene. This said, morphological defects in the granular layers as well as a premature activation of loricrin expression in the intermediate layers were observed. These abnormalities are subtle and are apparently not translated into severe functional impairment of the skin, as mutant embryos acquire a functional barrier with only a transient delay of ∼1 d or so (unpublished data). The premature activation of loricrin expression in the absence of together with the observation that Ovol1 represses the activity of the loricrin promoter in reporter assays led us to propose that Ovol1 might normally act to transiently repress loricrin expression in late presumptive spinous layers (). As the developing epidermis switches from a growing to a differentiating mode during late embryogenesis, a proposed involvement of in preventing premature terminal differentiation events at the critical cross-road might be important to ensure an orderly progression of the terminal differentiation-associated gene expression program. Alternatively, the up-regulated loricrin expression in -deficient epidermis from E16.5 onward might be a secondary consequence of the mutation. Clearly, future studies are necessary to distinguish between these possibilities. mice in a 129 × B6 mixed genetic background were backcrossed with B6 mice for 8–10 sequential generations, which, under no selection, is expected to generate a genetic background that is ∼99.9% B6. mice with an enriched B6 genetic background were then intercrossed to produce homozygous mutant progeny for study. Embryos or backskin samples were fixed in Bouin's fixative for 12–24 h at room temperature or in 4% PFA overnight at 4°C. 5 μm of paraffin or 5 μm of frozen sections were prepared and stained with hematoxylin and eosin or the appropriate polyclonal antibodies as described previously (): rabbit K1 (1:500; Covance), rabbit K14 (1:1,000; Covance), guinea pig K14 (1:50; a gift from D. Roop, Baylor College of Medicine, Houston, TX; ), rabbit loricrin (1:50; ), rabbit TG3 (1:100; a gift from L. Milstone, Yale University School of Medicine, New Haven, CT), rabbit K6 (1:200; a gift from P. Coulombe, Johns Hopkins University School of Medicine, Baltimore, MD; ), mouse c-myc (1:500; Abcam), and rabbit phosphorylated histone-H3 (1:1,000; Upstate Biotechnology). Images were acquired with a microscope (Eclipse E600; Nikon). Pregnant females were injected intraperitoneally with BrdU (Sigma-Aldrich) at a dosage of 50 μg/g of body weight, killed 2 h after injection, and embryos were dissected and fixed in 4% PFA. Embryos were then washed in PBS and frozen in optimal cutting temperature (Tissue-Tek). Frozen sections were treated with 50% formamide in 2× SSC at 65°C for 2 h followed by two brief 5-min rinses in 2× SSC and were incubated in 2N HCl at 37°C for 30 min. Samples were neutralized by incubation in 0.1M boric acid, pH 8.5, for 10 min, rinsed briefly in PBS, and endogenous peroxidase was quenched by incubating in freshly prepared 3% HO for 15 min. After three 5-min washes in PBS, samples were subjected to immunohistochemical analysis using a mouse monoclonal anti-BrdU antibody (Roche) according to the manufacturer's instructions. Keratinocytes were isolated from newborn backskin of mutant and wild-type littermates using an established protocol (). About 4–6 × 10 cells were recovered from each mouse and were plated at comparable cell densities (passage 0). The total number of attached cells obtained from each mouse was counted after 5 d of the initial plating and subsequently normalized against the total number of cells plated. For growth curve analysis, 3 × 10 cells of passage 1 were plated in replicate wells of six-well plates, and the total number of cells at 2, 3, and 4 d after plating were counted. For determination of the BrdU-labeling index in culture, wild-type and mutant keratinocytes were seeded in chamber slides precoated with collagen (36.9 μg/ml in PBS) and fibronectin (5 μg/ml in PBS) and were allowed to grow overnight. CaCl (final concentration of 1.2 mM), LiCl (final concentration of 20 mM), or TGF-β (final concentration of 1 ng/ml; Research Diagnostic) was added to the culture, and samples were fixed at various time points after the addition as indicated in the figures. BrdU was added 1 h before fixation at a final concentration of 10 μM, and fixation was in 100% methanol at −20°C for 10 min followed by three 5-min washes with 1× PBS. Keratinocytes were subsequently treated with 1N HCl for 30 min at 37°C followed by brief washes in PBS and were subject to immunohistochemical analysis as described in the previous section. Electrophoretic mobility shift assays (EMSAs) were performed using different amounts of partially purified recombinant His6-Ovol1 (final concentrations of Ovol1 were in the range of 47–600 nM) and ∼20 fmol (∼3 × 10 cpm) of gel-purified, 5′ P end-labeled double-stranded oligonucleotides. Typically, binding reactions were performed in a 20-μl volume containing 20 mM Hepes, pH 7.9, 75 mM KCl, 2.5 mM MgCl, 2 mM DTT, 1 mM EDTA, 12% glycerol, and 1 μg of poly(dI-dC) for 30 min at room temperature. In competition experiments, a 100-fold molar excess of unlabeled competitor was used. The protein–DNA complexes were resolved on 6% nondenaturing polyacrylamide gels and visualized by autoradiography. UG1 keratinocytes were cultured and transfected as previously described (). 293T cells were seeded in 24-well plates and transfected at 12–15% confluence with calcium phosphate as described previously (). A typical transfection mixture contained a total of 0.5 μg of plasmids, including 0.05 μg of a promoter construct (pGL3–c-myc, in which a 2.3-kb human c-myc promoter fragment [a gift from G. Radziwill, University of Zurich, Zurich, Switzerland] drives the luciferase reporter [], or pGL3-loricrin, in which a 1.3-kb mouse loricrin promoter fragment encompassing both the upstream and downstream transcription start sites drives the luciferase reporter []), with varying amounts of pCB6-Ovol1, an expression vector (), and 0.04 μg of a β-actin–β- galactosidase construct or a total of 0.5 μg of plasmids, including 10 ng of a promoter construct with varying amounts of the VP16-Ovol1–expressing vector and 0.04 μg of a β-actin promoter–β-galactosidase construct. pCB-6 (+) (empty vector containing the cytomegalovirus promoter) was used as stuffer DNA. Luciferase activity was measured in whole cell extracts using the Luciferase Assay System (Promega), and β-galactosidase activity was measured as previously described (). For transfecting primary keratinocytes, cells isolated from each newborn were plated in two 35-mm plates, one of which was treated with CaCl (final concentration of 1.2 mM) 24 h after plating. 3 h after calcium addition, each plate was transfected with 700 ng pGL3–c-myc and 350 ng β-actin–β-galactosidase using the helium-driven gene gun system (Biolistic PDS-1000; Bio-Rad Laboratories). Cells were collected 24 h later, luciferase activity was measured using the Luciferase Assay System (Promega), and β-galactosidase activity was measured using the Galacto-Light system (Tropix). 293T cells (a human kidney epithelial cell line) were seeded in 10-cm plates, and each plate was transfected with 6 μg pCB6-Ovol1. PCR amplification of the chromatin immunoprecipitates, prepared using the ChIP Assay Kit (Upstate Biotechnology) and anti-Ovol1 antibody () according to the manufacturer's instructions, was performed using the following primers containing sequences of the human c-myc promoter: 1F, 5′-AAGGAACCGCCTGTCCTTCC-3′; 1R, 5′-GCAACCAATCGCTATGCTGGA-3′; 2F, 5′-GGGAAAGAGGACCTGGAAAGG-3′; 2R, 5′-AGAGACAAATCCCCTTTGCGC-3′; 3F, 5′-ATCCAATCCAGATAGCTGTGC-3′; 3R, 5′-AAGAAGGGTATTAATGGGCGC-3′; 4F, 5′-ATCCTCTCTCGCTAATCTCCG-3′; 4R, 5′-TATTCGCTCCGGATCTCCCTT-3′; 5F, 5′-CCGCCTGCGATGATTTATACT-3′; and 5R, 5′-TTCTTTTCCCCCACGCCCT-3′. The following PCR program was used: 94°C for 1 min followed by 30 cycles of 94°C for 45 s, 60°C for 45 s, and 72°C for 1 min followed by a final extension at 72°C for 7 min. The percent input value was calculated for each specific primer set as follows: (PCR band intensity from anti-Ovol1 immunoprecipitate – PCR band intensity in IgG sample)/PCR band intensity from the input sample before immunoprecipitation. Total RNA was extracted from embryonic skin, and Northern analysis was performed as described previously () using the following cDNA probes: a 304-bp fragment containing sequences corresponding to 222–535 (5′-UTR) of c-myc mRNA (GenBank/EMBL/DDBJ accession no. ) and a 407-bp fragment containing sequences corresponding to 1,062–1,468 of Id2 mRNA (GenBank/EMBL/DDBJ accession no. ; a gift from S. Sinha, State University of New York at Buffalo, Buffalo, NY). In situ hybridizations were performed as described previously () using digoxigenin-labeled antisense and sense cRNA probes synthesized from a c-myc EST clone (Invitrogen). For RT-PCR, 5 μg RNA was reverse transcribed into cDNA using Superscript II RNase H reverse transcriptase (Invitrogen). PCR reactions were performed using the following primer set for loricrin: 5′-GTTCCTATGGAGGTGGTTCCAGCTG-3′ and 5′-TCCGTAGCTCTGGCACTGATACTGT-3′. For real-time PCR analysis, total RNA was extracted from primary keratinocytes cultured under low Ca conditions or treated for 24 h with CaCl (final concentration of 1.2 mM) and reversed transcribed into cDNA. PCR reactions were set up using the iQ SYBR Green Supermix (Bio-Rad Laboratories) and gene-specific primer pairs for c-myc (F: CTCGCTCTCCATCCTATG; R: CAAGTAACTCGGTCATCATC) and glyceraldehyde-3-phosphate dehydrogenase (F: CCTGCCAAGTATGATGAC; R: GGAGTTGCTGTTGAAGTC). Reactions were completed on a real-time PCR machine (iCycler; Bio-Rad Laboratories) according to the manufacturer's recommendations.

MAPK signal transduction pathways are among the most ubiquitous mechanisms of cellular regulation. The JNK family of MAPKs is classically activated by stresses leading to transcriptional regulation and apoptotic cell death (). JNKs are highly enriched in the nervous system, where they play additional roles as essential regulators of morphogenesis during early development (; ; ). Interestingly, the identity of JNK effectors regulating brain development is not clear. The best-characterized target, c-Jun, appears not to be responsible, as c-Jun–deficient mice show no overt brain abnormality (). In addition to the well-documented proapoptotic roles of JNKs (), JNK1-deficient mice display disrupted anterior commissure formation and loss of axonal microtubule integrity (). Microtubules are highly enriched in the brain, where they are major determinants of cell shape, intracellular transport, and cell polarity. Consistent with this, the list of proposed JNK functions that depend on microtubule structure is growing. We have previously shown that JNK activity regulates neuronal architecture (; ), and studies in suggest that the JNK pathway may regulate protein targeting (). Nonetheless, the molecular mechanisms of JNK action on the microtubule cytoskeleton remain elusive and the JNK substrates mediating such functions are not clear. Recently identified candidates include MAP2 and doublecortin (DCX). These proteins are phosphorylated by JNK; however, it has not been established whether they mediate JNK action on microtubules (; ; ). We therefore took a proteomics approach to identify effector molecules that may explain JNK action on neuronal shape. Affinity purification identified SCG10 as a major, brain-derived, JNK-interacting protein (JIP). SCG10 belongs to a family of proteins known as stathmins, which includes stathmin, SCG10, SCLIP, RB3, RB3′, and RB3″. Biochemical and structural studies show that stathmin proteins each bind two tubulin heterodimers and negatively regulate microtubule stability in vitro (; ; ). This function is inhibited upon phosphorylation of up to four residues or by mutations that simulate phosphorylation (). Although stathmin is ubiquitously expressed, SCG10, SCLIP, RB3, and RB3′ are neuron-specific proteins. Nonetheless, SCLIP, which displays 70% identity to SCG10, is anomalously up-regulated in a variety of human cancers (). Interestingly, stathmin- deficient mice display axonopathy in the nervous system (), as do JNK1 knockouts (). The absence of a more major phenotype in these mice likely reflects functional redundancy among other stathmin family proteins in the nervous system. , on the other hand, possesses only a single gene encoding phosphoproteins bearing homology to stathmins. Disruption of this gene using RNAi produces serious anomalies in nervous system development, with migration and commissure defects in particular (). We identify SCG10 as a neuron-specific in vivo substrate for JNK and demonstrate that S73 phosphorylation of SCG10 is reduced in brains from mice lacking JNK1. Phosphorylation of SCG10 by the cytoplasmic pool of JNK controls its ability to influence axodendritic architecture of cortical neurons. Moreover, by monitoring fluorescent tubulin recovery after photobleaching, we establish that inhibition of JNK significantly reduces microtubule polymerization in cultured neurons. Similarly, expression of the SCG10-AA mutant, which cannot be phosphorylated by JNK, disturbs microtubule dynamics with identical kinetics to JNK inhibition, suggesting that JNK1 activity on SCG10 regulates microtubule plasticity. Consistent with this, in embryonic brain, the spatial distribution of JNK site–phosphorylated SCG10 colocalizes with that of active JNK in regions undergoing intense neuronal differentiation. These results indicate that negative regulation of SCG10 microtubule depolymerizing activity by JNK1 contributes to microtubule homeostasis and axondendritic growth during brain development. To identify JNK binding partners from brain, we isolated GST-JNK1–interacting proteins from forebrain homogenate. Affinity-purified interacting proteins were separated by SDS-PAGE and visualized by silver staining. Bands that associated with GST-JNK1 and not with GST alone were analyzed by matrix-assisted laser-desorption/ionization time-of-flight mass spectrometry. The known JIPs GST-Pi, β-arrestin2, and ATF2 (; ; ) were identified as interacting with GST-JNK1 (). The 19-kD JNK1-associating protein was identified as SCG10. SCG10 belongs to a homologous family of proteins that share a COOH-terminal domain that is 70–80% conserved and an NH-terminal sequence that is 45–70% homologous. Given the overall sequence similarity, we anticipated that other stathmin family proteins would interact with JNK. To investigate this, we coexpressed GFP-tagged stathmin, SCG10, SCLIP, RB3, and RB3′ with GST-JNK in COS-7 cells. JIP1 (1–277; JNK binding domain [JBD]), a fragment of JIP1 that shows high-affinity interaction with JNK (), was used as a positive control. GFP-SCG10, -SCLIP, -RB3, and -RB3′ bound to GST-JNK1 after high salt washing (). The extent of interaction with GST-JNK1 was comparable to that of GFP-JBD. Parallel blots of the supernatants after pull down showed a corresponding depletion of JIPs, whereas GFP-stathmin levels remained elevated. Notably, GFP-stathmin, which lacks an NH-terminal extension, did not bind to GST-JNK1. Given the homology of the domain in this family, it is likely that the NH-terminal extension is required for interaction with JNK1. To determine whether SCG10 interacted stably with JNK under physiological conditions, we generated SCG10 antiserum for coimmunoprecipitation assays. Antibody specificity was tested using COS-7 cells expressing GFP-tagged stathmins (). Minor cross-reactivity with GFP-RB3 was detectable when used for immunoprecipitation but not for immunoblotting. The diffuse band visible on these immunoblots is antibody heavy chain leached from the Sepharose beads (, bottom). Using this serum, we demonstrated coimmunoprecipitation from brain extract of endogenous SCG10 and JNK1, the physiologically active form of JNK in brain (; ; ). Additional JNK isoforms also copurified with SCG10, which is indicated by an extra 54-kD JNK band (). It is notable that potential JNK phosphorylation motifs (SP) exist in stathmin, SCG10, and SCLIP, raising the possibility that these proteins are JNK substrates. It was previously shown that stathmin family proteins are phosphorylated in vitro by MAPKs (; ; ). However, which MAPK shows the highest activity toward these proteins and whether they phosphorylate SCG10 in vivo was not known. To more closely examine MAPK phosphorylation specificity, we analyzed the kinetics of stathmin family protein phosphorylation (). We first determined JNK isoform–dependent phosphorylation of stathmin family proteins, using the well-characterized JNK substrate c-Jun for comparison. Although GST-stathmin and -SCG10 were similarly phosphorylated by JNKs (1, 2, and 3) at high substrate concentrations (1.6 μM), GST-SCG10 was the preferred JNK1 substrate at lower concentrations (0.4 μM; ). As expected, GST-RB3 that lacks consensus JNK phosphorylation motifs (SP/TP) was not phosphorylated by JNK. A comparison of JNK phosphorylation kinetics toward GST-stathmin, -SCG10, and -SCLIP revealed that among the stathmin proteins, only GST-SCG10 is a good in vitro substrate for JNK (). Phosphorylation of SCG10 has not been studied as extensively as stathmin, though ERK, p38, and JNK were shown to phosphorylate bacterially expressed SCG10 in vitro (). To determine which MAPK showed preferential phosphorylation of SCG10, a kinetic analysis was undertaken (). JNK phosphorylated GST-SCG10 with fast kinetics, showing 40-fold higher V than did ERK or p38. Inefficient phosphorylation of GST-SCG10 by ERK and p38 was not due to lower activity of these kinases, as the relative activities of JNK, p38, and ERK toward GST–c-Jun, -ATF2, and -Elk were comparable (). We next compared the relative efficiency of GST-SCG10 phosphorylation to known JNK substrates. JNK1 showed higher specificity toward GST-SCG10 than it did toward the known physiological substrates GST–c-Jun and -Bim (), indicating that SCG10 is likely to be phosphorylated by JNK in vivo even when concentrations are limiting. To establish whether SCG10 could be a genuine JNK target in intact cells, COS-7 cells were transfected with GFP-SCG10 in the presence or absence of the JNK cascade activator GFP-MLK3. Activation of JNK resulted in increased phosphorylation of GFP-SCG10. This phosphorylation was prevented by incubation with 3 μM (an effective concentration in intact cells) of the JNK inhibitor SP600125 (; ). The traditional JNK activators UV, anisomycin, and MEKK1 also induced GFP-SCG10 phosphorylation; however, as these stimuli also activated ERK and p38 pathways (not depicted), we did not use them routinely. GFP-SCG10 phosphorylation was moderately induced by coexpression of GFP-MLK3. We postulated that the reason for such a moderate increase resulted from near saturation of GFP-SCG10 phosphorylation because of the presence of GFP-MLK3 before metabolic labeling. We therefore compared the relative phosphorylation of HA–c-Jun in vivo using the same approach (). Coexpression of MLK3 resulted in a similar induction of HA–c-Jun phosphorylation (). The extent of GFP-SCG10 phosphorylation by JNK in intact cells compared favorably to that of the well-characterized JNK target c-Jun. As kinases and phosphatases are not among the known JNK substrates, these data suggest that SCG10 is a direct target of JNK phosphorylation in vivo. We identified the JNK phosphorylation sites on SCG10 (S62 and S73) from in vitro–phosphorylated protein using mass spectrometry (unpublished data). To establish whether these sites were phosphorylated by JNK in vivo, GFP-SCG10 wild type (WT), or GFP-SCG10 with candidate phosphorylation sites, S62 and S73 mutated to alanine were expressed in COS-7 cells together with GFP-MLK3. After metabolic labeling, GFP-SCG10 was immunoprecipitated and digested with trypsin. Resulting peptides were separated in two dimensions by thin layer chromatography (). GFP-SCG10 phosphopeptides resolved as three spots. By individually mutating S62 and S73, we identified that S73 was more intensely phosphorylated by JNK (, D [middle] and E). The less intensely phosphorylated spots corresponded to a single peptide containing S62 (, right). Mutation of both sites resulted in complete loss of GFP-SCG10 phosphorylation, indicating that no additional sites on GFP-SCG10 are phosphorylated by JNK (). The JIP family of scaffold proteins can facilitate JNK target phosphorylation by recruiting select upstream regulators, including MLK3 (). To investigate whether SCG10 phosphorylation by JNK was enhanced by reconstitution of the JNK scaffolding machinery, we repeated the earlier experiments with the additional expression of GFP-JNK and -JIP1 (). GFP-SCG10 phosphorylation was enhanced threefold by addition of GFP-JIP1 (). Together, these results show that GFP-SCG10 phosphorylation by GFP-JNK is amplified but is not dependent on the expression of exogenous GFP-JIP1. If JNK plays a mandatory role in phosphorylating SCG10 and thereby controls its activity, JNK site phosphorylation of SCG10 should be reduced in JNK-deficient mice. To examine this, we generated antiserum against the major in vivo JNK phosphorylation site on SCG10, S73. After affinity purification, this antibody (PSCG10) recognized only the phosphorylated form of SCG10 ( [left] and Fig. S1, available at ) and showed acceptable specificity by immunoblotting (, right). To examine the importance of JNK for SCG10 phosphorylation in developing brain, cortex from JNK-deficient mice was immunoblotted with antibodies detecting SCG10 and PSCG10 (). JNK site phosphorylation of SCG10 was reduced by ∼50% in brains from JNK1−/− mice () and to a lesser extent in JNK2- and JNK3-deficient mice. A parallel analysis of JNK activity shows that JNK1 is the dominant active form in developing brain (). If SCG10 is a genuine JNK target in brain, its regional expression profile should overlap at least partially with that of JNK1, the constitutively active form of JNK in the brain. SCG10 expression is maximal during embryonic development (), although its precise location and function in developing brain was unknown. Examination of SCG10 and JNK1 immunoreactivity in embryonic brain showed a strikingly similar expression pattern that contrasted markedly to that of the glial cell intermediate filament protein (GFAP; ). JNK1 and SCG10 expression was mid to low throughout the midbrain, with more prominent staining in the telencephalon (developing cortex), the midbrain roof, the olfactory epithelium, the inferior colliculus, and the medulla oblongata (; see Fig. S2, available at , for magnified views). Close inspection of the telencephalon revealed concentrated JNK1 and SCG10 staining in the intermediate zone, which has the highest density of postmitotic neurons (). Similarly, phospho-JNK (PJNK) and JNK1 site–phosphorylated SCG10 localized in the same region as did class III β-tubulin, a marker for early differentiating neurons (). In contrast, the ventricular zone, the location of neuronal precursors, showed only an occasional ribbon-like staining pattern for PJNK, JNK1, and SCG10, possibly corresponding to radially migrating early differentiated neurons (). These observations were confirmed by microdissection followed by immunoblotting (). At the subcellular level, JNK1, PJNK, SCG10, and S73-phosphorylated SCG10 immunoreactivity were cytosolic (). This suggests that JNK signaling has functions in the cytoplasm during brain development in addition to the previously described regulation of nuclear events (). We previously reported that neuronal JNK1 activity is constitutively elevated and resides in the cytoplasm of neurons differentiating in culture (; ). We show that SCG10 and JNK1 associate tightly in vivo and in vitro ( and ). To examine whether S73-phosphorylated SCG10 localized with active JNK, cortical neurons in culture were stained with PJNK, SCG10, PSCG10, and tubulin antibodies (). SCG10 resided in the Golgi apparatus and growth cones but also displayed conspicuous punctate staining in the cytoplasm as previously shown (; ). PSCG10 and PJNK showed strikingly similar staining in the neurites and, unlike SCG10, PJNK and PSCG10 immunoreactivity were not elevated at the Golgi compartment. To examine whether S73-phosphorylated SCG10 and PJNK indeed colocalized in vesicles, membrane fractions were prepared from postnatal day 7 brain. Although total brain JNK was present in both soluble and membrane fractions, active JNK, SCG10, and S73-phosphorylated SCG10 concentrated in the P2 fraction, containing predominantly endosomal and Golgi vesicles (). To test the influence of cytosolic JNK on neuritic architecture without interference from nuclear JNK, we analyzed the efficacy of a cytosol-targeted JNK inhibitor. GFP-NES-JBD expression reduced the phosphorylation of a cytoplasmic JNK target (NES-Jun) in cortical neurons () and blocked JNK phosphorylation of recombinant GST–c-Jun (). Conversely, GFP-NES-JBD did not inhibit transcriptional activation of GAL4-Jun after withdrawal of trophic support (). Together, these data indicate that NES-JBD selectively blocks JNK from phosphorylating cytoplasmic targets. Tight regulation of microtubule dynamics is an essential determinant of axonal length. The known function of SCG10 as a microtubule disassembly factor made it a good candidate for mediating JNK regulation of neuronal shape. As active JNK concentrated in cortical regions undergoing differentiation, we examined the effect of JNK on neurite length in cortical neurons expressing CFP-tagged SCG10 variants (). The JNK site–phosphorylation mutants CFP-SCG10-AA (S62A/S73A), which cannot be phosphorylated by JNK, and CFP-SCG10-DD (S62D/S73D), which mimics JNK-phosphorylated SCG10, were used. The expression levels of SCG10 variants did not differ between samples ( and Fig. S3, available at ). Mutation of S62/S73 to alanine enhances tubulin depolymerizing activity, whereas mutation to phosphomimicking aspartates inhibits it (). Consistent with this, we observed a 25% reduction in axonal length in cortical neurons expressing SCG10-AA for 48 h (). Under the same conditions, total process length decreased by 30%, indicating that axonal and dendritic compartments were both affected. This is consistent with the proposed axodendritic expression of SCG10 (). Cells expressing CFP-SCG10-WT grew axons of normal length that did not differ significantly from cells expressing CFP alone. This was as expected because of the elevated JNK activity in cortical neurons (; ), which phosphorylates S62 and S73 (), leading to inhibition of its activity (). Similarly, cells expressing the functionally inactive, pseudophosphorylated mutant (CFP-SCG10-DD) extended processes of normal length. Notably, inhibition of cytoplasmic JNK using NES-JBD reduced neurite length, similar to CFP-SCG10-AA. These data suggest that endogenous SCG10 may mediate the determination of axodendritic length by JNK. To further test the implication that other JNK effectors were responsible, we expressed CFP-SCG10-DD together with NES-JBD. Inhibition of JNK with NES-JBD failed to influence neurite length in the presence of the pseudophosphorylated CFP-SCG10-DD mutant. This excludes the possibility that other JNK effectors were responsible. The ability of CFP-SCG10-DD to rescue the phenotype induced by NES-JBD suggests that JNK regulation of neurite length is mediated by SCG10. If JNK negatively regulates SCG10 function, then inhibition of cytosolic JNK or expression of an active SCG10 mutant (CFP-SCG10-AA) should have a similar effect on microtubule dynamics. To test this, we measured microtubule turnover in cortical neurons (). To monitor microtubule dynamics in living neurons, we expressed α-tubulin fused to Venus, a variant of YFP (). Neurites of cells expressing Venus–α-tubulin were photobleached to the extent that 80% of the initial fluorescence was destroyed. Recovery of Venus–α-tubulin into the bleached region was measured. Coexpression of the JNK inhibitor GFP-NES-JBD reduced Venus–α-tubulin recovery, as did expression of SCG10-AA. Kinetic modeling revealed a 50% decrease in the mobile fraction of Venus–α-tubulin upon JNK inhibition or SCG10-AA expression (). This is consistent with our proposal that SCG10 mediates the effect of JNK on microtubule dynamics. #text Rat stathmins and α-tubulin were isolated by PCR from rat cerebellar granule neuron cDNA and inserted into pEGFP-C1, pECFP-C1, pEYFP-C1 (CLONTECH Laboratories, Inc.), Venus-encoding vector (a gift from A. Miyawaki, Institute of Physical and Chemical Research, Saitama, Japan), pEBG (a gift from B. Mayer, Children's Hospital, Boston, MA), and pGEX-KG (a gift from J. Kyriakis, Massachusetts General Hospital, Boston, MA) vectors as described previously (). ECFP-tagged SCG10 phosphorylation site mutants S73A, S62A, S62A/S73A, S62D, S73D, and S62D/S73D were prepared by insertional overlapping PCR using mutagenic and flanking primers. The pEBG-JNK1, JNK2, JNK3, pEGFP-MEKK1Δ, pcDNA3-GAL4-Jun(5–105), and pGL3-G5E4▵38 constructs were described previously (, ; ). pEGFP-JIP-JBD, NES–c-Jun(1–146), and MLK3 were constructed by PCR-based methods from pcDNA3-mJIP1a, pMT108, and HeLa cDNA, respectively. c-Jun(5–89), JNK1, ATF2(1–109), and Elk1(205–428) in pGEX, ERK, and p38 in pEBG, pEF-EE-Bim-EL, and pCMV were gifts from J. Woodgett (Ontario Cancer Institute, Toronto, Canada), R. Davis (Howard Hughes Medical Institute, Worcester, MA), A. Sharrocks (University of Manchester, Manchester, UK), B. Mayer, D. Huang (The Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia), and S. van den Heuvel (Massachusetts General Hospital, Boston, MA), respectively. The coding sequence for Bim was cut from pEF-EE-Bim-EL and inserted into pGEX-KG. Recombinant c-Jun, ATF2, Elk, Bim, and stathmins were prepared from bacterial extracts as described previously (; ). Recombinant active JNK1, JNK2, JNK3, p38, and ERK were prepared as described previously (). For affinity purification of JIPs, postnatal day 7 rat cortex was homogenized in lysis buffer (20 mM Hepes, pH 7.4; 2 mM EGTA; 50 mM β-glycerophosphate; 1 mM DTT; 1 mM NaVO; 1% Triton X-100; 10% glycerol; 50 mM NaF; 1 mM benzamidine; 1 μg/ml aprotinin, leupeptin, and pepstatin; and 100 μg/ml PMSF) with an ultra-turrax (IKA) and a 27-gauge needle and centrifuged for 15 min at 10,000 at 4°C to remove insoluble material. 10 mg/ml of extract was incubated with 100 μg of GST coupled to glutathione–Sepharose for 1 h at 4°C. Precleared extract was rotated overnight at 4°C with 100 μg of GST-JNK1 conjugated to glutathione–Sepharose (Sigma-Aldrich). Immobilized proteins were washed in 2× lysis buffer, 3× LiCl buffer (100 mM Tris, pH 7.6, 0.5 M LiCl, 0.1% Triton X-100, and 1 mM DTT), and 2× lysis buffer. Equal proportions were separated by SDS-PAGE and visualized by silver staining. For identification of JNK1-interacting proteins, in-gel digestion was performed as described previously () and analyzed by matrix-assisted laser-desorption/ionization time-of-flight using a Voyager DE PRO. Peptide mass fingerprints were analyzed with Mascot (Matrix Science). Polyclonal antibodies were raised against bacterially expressed GST-SCG10 and against the phosphopeptide EAPRTLAS(PO)PKKKDLSLEE. Phosphospecific antibodies were affinity purified sequentially on phospho- and dephosphopeptide columns. Transfected COS-7 cells were serum starved overnight (0.1% FCS) and incubated in phosphate-free Eagle's minimum essential medium (Sigma-Aldrich) for 2 h. ATP pools were isotopically labeled by incubating cells with 0.5 mCi/ml [P]phosphate for 2 h at 37°C and 5% CO. Cells were washed with PBS, lysed in lysis buffer, and centrifuged at 13,000 rpm and 4°C for 10 min. Supernatants were immunoprecipitated using cross-linked SCG10 antiserum (5 μl) or with anti-HA (5 μl; Santa Cruz Biotechnology, Inc.). Phosphorylated proteins were visualized by autoradiography. Phospho- SCG10 was excised from the gel and trypsinized overnight at 37°C. Tryptic digests were separated by 2D-TLC with electrophoresis, pH 1.9, followed by ascending chromatography, pH 3.5. TLC-separated phosphopeptides were quantified by phosphorimaging. For in vitro analysis, active recombinant JNKs were used to phosphorylate bacterially purified GST fusion proteins using γ-[P]ATP as described previously (; ). Kinetic analyses were performed as described previously (). Cortical neuron cultures were prepared as described previously (). Cells for morphological analysis were plated at a density of 0.5 × 10. 24 h after plating, cells were transfected using Lipofectamine 2000 (Invitrogen) with 25% of total DNA as pEYFP-CAAX, 25% pECFP-C1, or ECFP-tagged SCG10 WT, S62A/S73A, or 62D/S73D and 50% pEGFP-NES-JBD as indicated. Cells were fixed 24 h later, and fluorescence was examined with a microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.) using a 20× objective. Digitized images (∼15 cells per coverslip) of EYFP fluorescence were acquired using a camera (ORCA II ERG; Hamamatsu) and Wasabi software (Hamamatsu). Overlapping images were taken where necessary to encompass the entire process length. Neurite length was measured from size-calibrated images using MetaMorph 6.1 imaging software (Universal Imaging Corp.). COS-7 cells were transfected using Lipofectamine. For analysis of protein–protein interactions, COS-7 cells were transfected as follows: 50% of total DNA as EGFP-tagged stathmin, SCG10, SCLIP, RB3, or RB3′ together with 50% of pEBG-JNK. Cells were lysed 48 h after transfection in lysis buffer. Precleared supernatants were incubated with -hexylglutathione–Sepharose at 4°C. Pull downs were washed (2× lysis buffer, 3 × LiCl wash, and 1× lysis buffer), suspended in laemmli buffer, and immunoblotted. Fig. S1 shows characterization of PSCG10 antibody specificity for immunostaining. Fig. S2 shows magnified views of JNK1, SCG10, and GFAP staining in E15 embryo. Fig. S3 shows composite images of YFP-CAAX/CFP-SCG10 mutant expression in cortical neurons. The supplementary text describes reporter gene expression analysis, immunohistochemistry, immunofluorescence, subcellular fractionation, and photobleaching procedures. Online supplemental material is available at .

Tuberous sclerosis complex (TSC), caused by loss of function of either the or - tumor suppressor genes, is an autosomal dominant disorder that leads to mental retardation, seizures, and the formation of tumors in various organs, including the brain, kidney, heart, and skin (; ; ). The gene encodes the 130-kD protein hamartin (), and the gene encodes the 198-kD protein tuberin (). Hamartin contains two coiled-coil domains, which have been shown to mediate binding to tuberin (), forming a stable, functional tumor suppressor heterodimer within cells (; ). Lesions that develop in TSC patients are histologically diverse; however, the tumors that arise as a result of loss of function of either TSC1 or -2 share common features, suggesting that hamartin and tuberin function within the same pathways to regulate cell cycle, cell growth, adhesion, and vesicular trafficking (; ). Recent studies have indicated that the hamartin–tuberin heterodimer regulates cell growth and proliferation as a downstream component of the phosphoinositide 3-kinase (PI3K)–protein kinase B (PKB/AKT) signaling pathway, which modulates signal transduction through target of rapamycin (TOR) in both and mammalian cells (; ). Several distinct yet complementary genetic and biochemical studies collectively show that tuberin is a GTPase-activating protein (GAP) for the small GTPase Ras homologue enriched in brain (Rheb), which activates TOR and its downstream targets, such as the ribosomal S6 kinase (RSK; ). Although loss of tuberin promotes cell growth and tumorigenesis, cells expressing tuberin must also be able to relieve tuberin repression of mammalian TOR (mTOR) signaling during conditions of mitogenic sufficiency. In this regard, tuberin contains multiple sites for AKT, MAPK, RSK, and extracellular signal–regulated kinase phosphorylation (; ; ; ; ; ; ). Although it is clear that activation of AKT blocks tuberin inhibition of TOR signaling (; ; ), the mechanism by which AKT inactivates this tumor suppressor is unknown (). In addition, there are conflicting data regarding the subcellular localization of tuberin. For instance, independent studies report that tuberin can localize to the cytosol (), the membrane/particulate (100,000 ) fraction (), and even the nucleus () of cells. In this study, we sought to determine the mechanisms by which tuberin is regulated during cell growth. We found that tuberin is localized in membrane and cytosol fractions but not in nuclear fractions, and the translocation of tuberin from the membrane to cytosol is regulated by AKT signaling in response to growth factors. Phosphorylation of tuberin by AKT causes tuberin to become sequestered by 14-3-3 proteins in the cytosol. Mutation of two specific phosphorylation sites (S939 and S981) prevents the cytosolic translocation of tuberin from cellular membranes and results in a constitutively active protein that inhibits mTOR signaling. Importantly, tuberin phosphorylation by AKT does not affect its GAP activity toward Rheb in vitro but promotes Rheb-induced S6 kinase (S6K) 1 activation through increased Rheb-GTP loading in vivo. Therefore, it is likely that AKT phosphorylation inhibits tuberin as a result of 14-3-3 binding and cytosolic translocation rather than by impairing its catalytic GAP activity. To investigate the subcellular localization of tuberin, mouse fibroblast (NIH3T3, Swiss3T3), rat kidney epithelial (TRKE2), human embryonic kidney (HEK293), and human breast cancer (MCF7) cell lines were fractionated, revealing that tuberin was detected within both the cytosolic and membrane fractions but not within the nuclear fraction (). Interestingly, tuberin in the membrane fraction migrated faster than tuberin within the cytosolic fraction (). To determine whether this mobility shift was due to changes in phosphorylation, we treated these subcellular fractions with either a Ser/Thr or Tyr phosphatase (calf intestinal alkaline phosphatase [CIAP] or YOP protein tyrosine phosphatase, respectively). After CIAP treatment, the mobility of tuberin in the cytosol increased, resolving as a faster migrating band similar to tuberin purified from membrane fractions (). However, YOP did not affect tuberin mobility, indicating that the mobility shift was primarily due to phosphorylation of Ser/Thr residues. Treatment with CIAP resulted in both membrane and cytosolic tuberin migrating faster, indicating that tuberin within the cytosolic fraction is hyperphosphorylated. In addition, the serum-induced decrease in mobility of tuberin in both membrane and cytosolic fractions is likely driven by multiple phosphorylation events, as CIAP treatment (removing all Ser/Thr phosphorylation) increased the mobility of tuberin in both fractions. To determine whether growth factor stimulation could also alter tuberin phosphorylation and subcellular localization, Swiss3T3 or MCF7 cells were serum starved and then stimulated with serum or insulin-like growth factor-1 (IGF-1) before fractionation (; and Fig. S1, available at ), as were NIH3T3 (Fig. S1 A) and TRKE2 cells (Fig. S1 B). Like serum, IGF-1 increased the amount of tuberin in the cytosolic fraction relative to starvation conditions (). In serum-stimulated cells, phosphorylation of membrane-localized tuberin was also increased after 1 h (comparable to cytosolic tuberin in starved cells but less than cytosolic tuberin from serum- stimulated cells; ), suggesting that phosphorylation at specific residues, rather than total levels of phosphorylation, was determining localization. By 6 h, tuberin became predominantly membrane localized, which correlated with a decrease in AKT activation (). Translocation of tuberin from the membrane to the cytosol was blocked by the PI3K inhibitors wortmannin and LY294002 ( and Fig. S1 B), implicating PI3K signaling in tuberin localization to the cytosol. To determine whether AKT directed tuberin's subcellular localization, tuberin from membrane and cytosolic compartments of NIH3T3 cells treated with IGF-1 or EGF was immunoprecipitated and detected with a (S/T) phosphosubstrate antibody that recognizes the consensus phosphorylation site for AKT and RSK containing phospho-Ser/Thr with Arg at position −5 and −3 (RXRXXpS/T) (; ; ; ). With equal tuberin loading (), the (S/T) phosphosubstrate antibody predominantly recognized tuberin within the cytosolic fraction, and the PI3K inhibitor wortmannin significantly reduced recognition of phosphorylated tuberin, suggesting that cytosolic tuberin was phosphorylated by AKT (). To confirm that activation of AKT directed tuberin to the cytosol, MCF7 cells stably transfected with a constitutively active AKT (myr-AKT) were examined and found to contain more cytosolic tuberin relative to wild-type MCF7 cells (). Recently, RSK was shown to phosphorylate tuberin at S1798 (). To determine whether RSK phosphorylation of tuberin could regulate its localization, we generated a COOH-terminal deletion mutant of TSC2 at residue 1734 (referred to as Δ73) that lacks S1798 (). Δ73 and wild-type tuberin had a similar distribution within membrane and cytosolic fractions under normal growth conditions (Fig. S1 C) and in response to serum (Fig. S2 A, available at ). In addition, overexpression of RSK1 did not affect localization of wild-type tuberin (Fig. S2 B). Collectively, these data indicate that tuberin residing in the cytosol is phosphorylated by AKT, suggesting that AKT (but not RSK) directly controls tuberin's localization and possibly its activity. Tuberin contains multiple S/T phosphorylation sites (). Among these, S1254 has been shown to be phosphorylated by MK2 (), whereas S939 (; ), S981 (), S1130/S1132 (), and T1462 (; ) have been reported to be AKT phosphorylation sites in vivo and/or in vitro, and several of these also have the potential for binding to 14-3-3 (; ). S981 is of particular interest, as it lies within the alternatively spliced exon 25 of tuberin. To examine these as candidate sites for the regulation of tuberin localization, we mutated these residues to alanine to create TSC2 constructs in frame with an NH-terminal Flag epitope (Flag-TSC2): S939A, S981A, and 2A (S939A + S981A); T1462A and SATA (S939A + T1462A; ); and S1254A and S1130A + S1132A double mutant. Wild-type and mutant TSC2 constructs were transfected into HEK293 cells and subjected to subcellular fractionation. As shown in , S939A, S981A, and 2A mutants predominantly localized to the membrane, as did the double alanine SATA mutant that lacks the S939 site. However, the T1462A single mutant partitioned in the cell similarly to wild-type tuberin (), indicating that T1462 phosphorylation was not directing translocation of tuberin from the membrane to the cytosol. These data were confirmed with a phosphospecific T1462 antibody, which recognized tuberin in both the membrane and cytosolic fractions equally (unpublished data). Furthermore, phosphorylation at both S939 and S981 contributes to cytosolic localization, as phosphorylation at S939 (determined with a phospho-S939 specific antibody) of the S981A mutant was not sufficient to partition tuberin to the cytosol (Fig. S2 C). Other tuberin mutants, S1254A, S1130A/S1132A (), Δ73, or S1338A (Fig. S1 C), also distributed equally between the membrane and cytosolic fractions. Thus, S939 and S981 phosphorylation are critical determinants of membrane versus cytosolic localization of tuberin. 14-3-3 has been previously reported to directly interact with phosphorylated tuberin (; ). S939 and S981 are predicted AKT phosphorylation and 14-3-3 interaction sites. When phosphorylated and nonphosphorylated S939 and S981 peptides were used in competition assays to block GST–14-3-3 interaction with tuberin, as shown in and Fig. S3 (available at ), phosphorylated but not nonphosphorylated S939 and S981 peptides clearly competed for the interaction of tuberin with several 14-3-3 isoforms. In addition, the amount of 2A mutant tuberin affinity purified by 14-3-3 was dramatically reduced relative to wild-type tuberin (). Importantly, inhibition of PI3K signaling by wortmannin ablated this 14-3-3 tuberin interaction (). These data indicate that S939 and S981 are critical sites of interaction between tuberin and 14-3-3 proteins and that this interaction is mediated by PI3K/AKT phosphorylation. To demonstrate that tuberin binding to 14-3-3 was directly responsible for its translocation to the cytosol, we transfected HEK293 cells with the EGFP-R18 construct that expresses a peptide that disrupts 14-3-3 binding (). As shown in , the R18 14-3-3 decoy clearly repressed cytosolic localization of tuberin, establishing a direct link between 14-3-3 and translocation of tuberin to the cytosol. Hamartin possesses a predicted transmembrane domain and two coiled-coil domains that mediate its association with tuberin (, ). We found that in both human and mouse cells, hamartin was only detected in the membrane fraction (; ; unpublished data). To confirm that tuberin mutants that did or did not constitutively localize to the membrane retained their ability to bind hamartin, immunoprecipitation experiments were performed. Immunoprecipitation showed that, similar to wild-type tuberin, tuberin S939A, S981A, S1338A, Δ73, and 2A mutants retained their ability to interact with hamartin (Fig. S4, A, B, and C, available at ). To determine whether hamartin played a role in the subcellular localization of tuberin, we transfected wild-type, S939A, and S981A Flag-TSC2 constructs into HEK293 cells with or without Myc- or Flag-TSC1. Although S939A and S981A mutants were primarily membrane localized, coexpression of hamartin increased membrane retention of both mutant and wild-type tuberin in a dose-dependent manner (). These data suggest that the interaction of hamartin with tuberin facilitates its localization to the membrane, implying that the tuberin–hamartin heterodimer functions in this subcellular compartment. Importantly, only tuberin in the membrane fraction remained associated with hamartin: no hamartin could be coimmunoprecipitated with tuberin in the cytosolic fraction, indicating that translocation to the cytosol dissociated tuberin from hamartin (). Similarly, when Flag-TSC2 and Myc-TSC1 were cotransfected into MCF7 cells, tuberin and hamartin were observed by confocal microscopy to colocalize in a discrete, punctate pattern (Fig. S4 D). Cells that expressed wild-type Flag-TSC2 in the absence of Myc-TSC1 exhibited a more diffuse staining pattern than cells that overexpressed both Myc-TSC1 and Flag-TSC2 (Fig. S4 D, comparing γ and γ′ with β and β′). However, the 2A mutant, which was constitutively membrane localized as shown by cell fractionation, retained this punctate localization pattern even in the absence of exogenously expressed hamartin (Fig. S4 E). Tuberin's GAP target Rheb is farnesylated and predicted to be membrane localized (), and recent data indicate that Rheb is localized in endomembranes (). To determine whether phosphorylation at S939/S981 affected tuberin's intrinsic GAP activity for Rheb, we first determined Rheb's subcellular localization. Rheb was detected only in the membrane fractions from HEK293 () and MCF7 (Fig. S5 A, available at ) as well as NIH3T3, Swiss3T3, and TRKE2 cells (not depicted). Treatment with EGF or wortmannin did not affect this localization over a period of 10 min to 3 h (Fig. S5 A and not depicted). Recognition of the 21-kD band in the membrane fraction using this anti-Rheb antibody was specific for Rheb, as antibody binding was ablated when Rheb RNAi was used to knock down Rheb (Fig. S5 B). When untagged /Flag- and Myc- were cotransfected into HeLa cells, tuberin and Rheb were observed with confocal microscopy to colocalize in a discrete, punctate pattern in the absence of serum (). In contrast, stimulation with IGF-1 resulted in partitioning of wild-type tuberin away from Rheb (), indicating that in response to growth factor stimulation, tuberin is no longer retained in physical proximity to its downstream target, Rheb. Collectively, these data indicate that the localization pattern of tuberin is modulated by both hamartin and growth factor signaling and that Rheb is in physical proximity to membrane-localized tuberin and hamartin within the cell. In addition, these data, along with confocal microscopy using 2A mutant tuberin and biochemical fractionation, suggest that S939 and S981 are crucial sites for determining the subcellular localization and, thus, function of tuberin in response to growth factor signaling. To elucidate the relationship between tuberin localization and function, we examined the effect of wild-type, S939A, S981A, and 2A tuberin mutants on mTOR-S6K activity. T389 is a known rapamycin-sensitive phosphorylation site of S6K that correlates with activation by mTOR (). In cells transfected with HA-tagged S6K, cotransfection of wild-type and expression constructs diminished phosphorylation of exogenously expressed S6K (). However, cotransfection of the 2A mutant with resulted in an even more dramatic reduction in phospho-S6K levels (). Similar experiments were performed to determine the effect of S939A and S981A single and double mutants on phosphorylation of endogenous S6K. In the presence of serum, wild-type Flag-tuberin was located in the membrane and cytosol, whereas S939A, S981A, and 2A Flag-tuberin mutants remained primarily at the membrane (, top). AKT was equally activated by serum in cells expressing wild-type or mutant tuberin constructs, and endogenous hamartin remained membrane localized (). In cells transfected with wild-type tuberin, phosphorylation of endogenous S6K at T389 increased in response to serum, coincident with translocation of tuberin to the cytosol. However, in cells transfected with S939A, S981A, or 2A tuberin mutants, phosphorylation of S6K was inhibited, with the 2A mutant most effectively blocking S6K activation (). Thus, tuberin mutants that were retained at the membrane were constitutively active and exhibited enhanced ability to repress S6K activation. These data suggest that AKT phosphorylation of tuberin regulates its activity by decreasing the amount of tuberin located at the membrane, thereby reducing its inhibitory effect on the Rheb–mTOR–S6K signaling pathway. Physical sequestration of tuberin from its target Rheb suggested a mechanism whereby AKT modulated tuberin by partitioning it away from the membrane rather than altering its intrinsic GAP activity for Rheb. We compared the relative RhebGAP activity of wild-type TSC2 and 2A mutant (S939A and S981A) in cells during activation of the PI3K–AKT pathway (). To do this, we quantified the ratio of GTP and GDP Myc-Rheb (percentage of GTP bound Myc-Rheb) when these TSC2 constructs were coexpressed during a time course of insulin stimulation. The 2A mutant enhanced the GTPase function of Rheb more effectively than wild-type tuberin, as indicated by impaired accumulation of the active GTP form of Rheb after 15 min of insulin stimulation (32% GTP bound) when compared with wild-type TSC2 (42% GTP bound; ). As an increase in GTP bound Rheb would be predicted to enhance mTOR-mediated cell signaling, we measured the activity of HA-S6K1 in these cells (). As expected, insulin-induced activation of S6K1 was markedly impaired in cells expressing the 2A mutant with the 2A mutant blocking insulin-induced activation of S6K1 by 50% (after 15 min) when compared with wild-type TSC2 (), reflecting the reduced levels of active GTP bound Rheb (). To determine whether phosphorylation at S939 and S981 had an impact on tuberin's intrinsic GAP activity, we analyzed the ability of wild-type or mutant tuberin proteins to activate Rheb GTPase in vitro. RhebGAP assays were conducted on immunoprecipitated tuberin–hamartin heterodimers containing either the wild-type or AKT-phosphorylation mutants of TSC2 (). Tuberin–hamartin complexes containing wild-type tuberin from both unstimulated and insulin-stimulated cells enhanced the intrinsic GTPase activity of Rheb at comparable rates. Induction of tuberin phosphorylation by insulin was confirmed in these lysates by detection of phospho-T1462 (, bottom). Furthermore, tuberin mutants lacking these sites (S939A, S981A, and 2A) also possessed similar RhebGAP activity to wild-type tuberin in vitro, even after insulin treatment (). These findings imply that AKT-mediated phosphorylation of tuberin does not directly alter the rates at which tuberin enhances the GTPase activity of Rheb, at least in vitro. However, these data support a mechanism whereby AKT phosphorylation suppresses tuberin function by translocating tuberin to the cytosol away from its membrane-associated binding partner, hamartin, and its downstream target, Rheb. Indeed, as the efficiency with which tuberin functions as a RhebGAP in vitro is significantly reduced in the absence of hamartin, physical separation from both hamartin and Rheb may be contributing to reduced tuberin activity (; ). Based on our data, we propose a model where in the absence of growth stimulatory signals, hamartin facilitates localization of hypophosphorylated tuberin to membranes in physical proximity to Rheb. The membrane-associated tuberin–hamartin complex binds Rheb and acts as a GAP to inactivate membrane Rheb signaling by stimulating GTP hydrolysis (). However, during mitogenic sufficiency (i.e., after growth factor stimulation), activation of PI3K signaling leads to activation of AKT, which then directly phosphorylates membrane-associated tuberin. In response to AKT phosphorylation, 14-3-3 proteins bind tuberin and sequester it in the cytosol. The deficiency of membrane-associated tuberin results in the accumulation of GTP bound Rheb, which leads to increased mTOR signaling, enhanced cell growth, and proliferation (). In this study, we provide evidence that AKT inhibits the tumor suppressor function of tuberin by altering its subcellular localization. Previously, AKT was shown to directly phosphorylate and inhibit tuberin function upon stimulation with growth factors (). However, it was not known how AKT phosphorylation of tuberin regulated its function or how hamartin contributed to tuberin regulation of Rheb and mTOR (; ; ). We have shown that although tuberin is found in both the membrane and cytosolic fractions of cells, cytosolic tuberin accumulates upon growth factor stimulation and is hyperphosphorylated. The translocation of tuberin from the membrane to the cytosol can be blocked by PI3K inhibitors, and AKT-phosphorylated tuberin is predominantly found in cytosolic fractions. The AKT phosphorylation sites S939 and S981 are crucial residues that affect tuberin localization, with localization of tuberin to the membrane accounting for the ability of tuberin to inhibit mTOR signaling via its GAP activity for Rheb. We show that two tuberin phosphorylation sites, S939 and S981, are responsible for interaction with 14-3-3 in a phosphorylation-dependent manner. 14-3-3 proteins are highly acidic dimeric intracellular proteins that chiefly bind to phosphoserine motifs (). They play a key regulatory role in many cellular processes, including signal transduction, apoptosis, and cell cycle checkpoint control (; ; ; ). In many instances, 14-3-3 binding leads to altered subcellular localization of target proteins, which modulates their function. As this appears to be the case for tuberin as well, we propose a mechanism whereby growth factor stimulation activates AKT, leading to tuberin phosphorylation and mislocalization within the cell. Although some studies have suggested that the tuberin–hamartin heterodimer is destabilized upon AKT-mediated phosphorylation of tuberin (; ), our data and those of others indicate that AKT-mediated phosphorylation of tuberin does not change its affinity for hamartin (; ; ). An alternative hypothesis is that in response to phosphorylation of S939 and S981 by AKT, 14-3-3 binds to these phosphorylated motifs to localize tuberin in the cytosol, physically sequestering tuberin away from hamartin. The presence of two adjacent 14-3-3 binding sites at S939 and S981 of tuberin may be important, as they would be predicted to stabilize binding of tuberin in the central channel of the 14-3-3 dimer more effectively than would a single binding site. In this regard, it is interesting that loss of either the S939 or S981 14-3-3 binding sites reduces localization of tuberin to the cytosol. As one of these binding sites is in the alternatively spliced exon 25 of tuberin, it suggests that alternative splicing may allow for wider or more varied regulation of this tumor suppressor. In splice isoforms lacking exon 25, the absence of the S981 binding site could result in retention at the membrane, potentially enhancing tuberin activity. This could provide a mechanism via alternative exon splicing for cells and tissues to regulate tuberin localization and function. It might be noted that in previous studies that failed to identify the S939 site within tuberin as a 14-3-3 binding site, the TSC2 construct used lacked the alternatively spliced exon 25 and the S981 14-3-3 binding site (). Hamartin has been shown to significantly enhance tuberin's GAP activity toward Rheb GTPase (; ), and our observations also emphasize the importance of the interaction between tuberin and hamartin as a functional unit. We found that hamartin directly interacts with tuberin and promotes tuberin colocalization at the membrane in proximity to its GAP target Rheb and enhances tuberin's ability to repress mTOR signaling. Both endogenously and exogenously expressed hamartin were detected in the membrane fraction of cells, even when cells were treated with serum or growth factors, data that are consistent with the previous reports that hamartin is a membrane protein (; ). The colocalization of tuberin in proximity to its activation partner hamartin and its GAP target Rheb at the membrane could explain why increased retention of tuberin at the membrane enhances its ability to inhibit phosphorylation of S6K. Given that the S939 and S981 residues lie distal from the COOH-terminal GAP region (residues 1517–1674) of tuberin and mutation of these sites does not affect the Rheb GAP activity of tuberin, we conclude that phosphorylation of S939 and S981 does not directly regulate the GAP function of TSC2. However, mutation of these sites significantly inhibits Rheb-induced S6K1 activation through decreased Rheb-GTP loading, suggesting that AKT regulates tuberin function by causing tuberin translocation to the cytosol, rather than by directly inhibiting its intrinsic GAP activity toward Rheb. The following antibodies were used: tuberin, hamartin, lamin A/C, HA, and Myc (Santa Cruz Biotechnology, Inc.); Rheb, phosphotuberin (T1462), S6K, phospho-S6K (T389), AKT, phospho-AKT (S473), phospho-AKT (T308), and phospho-(S/T) AKT substrate (Cell Signaling Technology); Flag M2 and Flag M2 immobilized agarose beads (Sigma-Aldrich); LDH (Chemicon International); EGFP (Abcam); and β1-integrin (CLONTECH Laboratories, Inc.). The following reagents were used: EGF, insulin, and wortmannin (Sigma-Aldrich) and IGF-1 (R & D Systems). Tuberin peptides were synthesized by W.M. Keck Biotechnology Resource Center. Full-length human TSC1 and -2 cDNAs (supplied by J. DeClue, National Cancer Institute, Bethesda, MD) were subcloned into pcDNA3.1 (Invitrogen) and pCMV-Tag2 (Stratagene) expression vectors with NH-terminal Myc or Flag epitopes, respectively. COOH-terminal truncation mutant of TSC2 (Δ73) (residues 1–1734) was created by EcoR V digestion followed by religation. TSC2 mutations were generated by site-directed mutagenesis (Stratagene). Other constructs used in this study were generously provided as follows: Flag-TSC1 and HA-S6K from J. Blenis (Harvard Medical School, Boston, MA), HA-RSK1 from J. Avruch (Massachusetts General Hospital, Boston, MA), GFP-R18 from T. Pawson (Mount Sinai Hospital, Toronto, Canada), and the Flag-TSC2 SATA and T1462A from B. Manning (Harvard School of Public Health, Boston, MA). GST–14-3-3 constructs were described previously (). Cell lines were grown as follows: MCF7 cells were grown in improved minimum essential medium (Biosource International). TRKE2 cells were grown in DF8 complete medium (). HeLa cells were grown in MEM, and HEK293, Swiss3T3, and NIH3T3 cells were grown in DME (Life Technologies, Inc.). All media contained 10% FBS (Hyclone) unless otherwise noted. Myr-Akt MCF7 cells were cultured as previously described (). Transfections were performed using the Lipofectamine 2000 reagent (Invitrogen) according to manufacturer's instructions. Cells were lysed using PBS containing 0.1% SDS, 1% NP-40, 0.5% deoxycholic acid, 1 μM PMSF, 20 μg/ml aprotinin, 10 μM leupeptin, and 1 μM NaVO. For immunoprecipitation, cell lysates were immunoprecipitated with the indicated antibodies and protein A– or protein G–Sepharose beads (GE Healthcare) and washed with buffer (10 mM Tris-HCl, pH 7.5, 1% NP-40, 1% Triton X-100, 100 mM NaCl, 50 mM NaF, 2 mM EDTA, 1 mM PMSF, and Complete protease inhibitor cocktail [Roche]). Immunocomplexes were subjected to SDS-PAGE and Western blotting. Cells (70–80% confluent in 15-cm plates) were washed and collected by scraping into ice-cold PBS, pelleted by centrifugation at 4°C, resuspended in hypotonic buffer (10 mM Hepes, pH 7.2, 10 mM KCl, 1.5 mM MgCl, 0.1 mM EGTA, 20 mM NaF, and 100 μM NaVO), and disrupted using a Dounce homogenizer. Crude nuclei and unbroken cells were then pelleted by centrifugation at 3,000 rpm at 4°C for 5 min. The postnuclear supernatant was separated by ultracentrifugation at 100,000 for 1 h at 4°C. The supernatant, or cytosolic fraction, was removed and the pellet was lysed in 1× lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM NAP0, 1 mM β-glycerophosphate, 1 mM NaVO, and 1 μg/ml leupetin). The insoluble fractions were removed by centrifugation at 14,000 rpm for 10 min, and the supernatant was collected as the membrane fraction. Crude nuclei were resuspended with hypotonic buffer and homogenized using a Dounce homogenizer. After centrifugation at 4°C for 5 min, the pellet was further washed with wash buffer (10 mM Tris-HCl, pH 7.4, 0.1% NP-40, 0.05% sodium deoxycholate, 10 mM NaCl, and 3 mM MgCl) and lysed in high-salt lysis buffer (20 mM Hepes, pH 7.4, 0.5 M NaCl, 0.5% NP-40, and 1.5 mM MgCl). All lysis and wash buffers contained 1× Complete protease inhibitor cocktail. The nuclear, membrane, and cytosolic lysates were normalized using the BCA Protein Assay kit (Pierce Chemical Co.) and subjected to SDS-PAGE and immunoblot analysis. For phosphatase treatment, equal amounts of membrane and cytosolic lysates were treated with CIAP serine/threonine phosphatase or YOP tyrosine phosphatase at 30°C for 1 h. GST pull-down assays were performed as previously described (). For peptide competition assays, incubations were performed in the presence or absence of 100 μM phospho-Ser or Ser tuberin peptide (SGSGFRARSTS LNERPK) or phospho-Ser or Ser tuberin peptide (SGSGFRCRSIS VSEHVV). Anti-Flag M2 antibody was used to immunoprecipitate Flag-tuberin from HEK293E cells. Immunocomplexes of TSC1/TSC2 were used for in vitro RhebGAP assays as previously described (). α-[P]GTP and α-[P]GDP were eluted from Rheb and resolved by thin layer chromatography on PEI cellulose (Sigma-Aldrich) with KHPO. The relative levels of radiolabeled GTP and GDP were quantified with a phosphorimager. S6K1 activity assays on immunoprecipitated HA-S6K1 were performed as previously described () using recombinant GST-S6 (32 COOH-terminal amino acids of ribosomal protein S6) as substrate. Analysis of Rheb guanine nucleotide binding in cells was determined as previously described (). Intracellular localization of wild-type tuberin and 2A mutant and colocalization studies was determined by immunofluorescence analysis of HeLa cells. Transfected cells were plated onto glass chamber slides in the absence of serum. After 12 h of serum starvation, transfected cells were treated with either 20% FBS or 30 ng/ml IGF-1 for 1 h. Cells were then fixed in 50% ethanol plus 10% acetic acid for 1 h at 4°C, and nonspecific antigens were blocked for 1 h in PBS containing 7.5% BSA at 37°C. Rabbit anti-Flag (Sigma-Aldrich) and/or mouse anti-Myc (Santa Cruz Biotechnology, Inc.) primary antibodies were incubated in PBS/7.5% BSA at 37°C for 2 h. Primary antibodies were detected with FITC-conjugated goat anti–rabbit (Abcam) and Cy3-conjugated donkey anti–mouse (Jackson ImmunoResearch Laboratories) antibodies, respectively. Fluorescence images were analyzed either on a confocal microscope (Fluoview Scanning Laser Biological Microscope IX 70 system; Olympus) equipped with two lasers (Ar 488 and Kr-Ar 488–564) using UplanFl 100× oil-immersion (NA 1.30) objective or a conventional microscope with fluorescence attachment (BX40 attached with BX-FLA; Olympus) using UplanFl 40× (NA 0.75) objective. Fluoview version 1.26 software (Olympus) and MagnaFire version 2.1C (Olympus) were used for image acquisition from confocal microscopy and conventional microscopy, respectively. Photoshop 8.0 software (Adobe) was used for minor adjustments and processing of images. Fig. S1 shows that AKT-mediated phosphorylation of tuberin leads to subcellular translocation. Fig. S2 shows that S939 and S981 are required for altered subcellular localization of tuberin. Fig. S3 demonstrates that 14-3-3 proteins bind phosphorylated S939 and S981 residues. Fig. S4 demonstrates that tuberin mutants retain their ability to interact with hamartin. Fig. S5 shows that Rheb is retained in the membrane fraction in the presence or absence of growth factor treatment. Online supplemental material is available at .

The and loci encode a variety of receptor isoforms, in addition to the canonical full-length tyrosine kinase receptors (; ). Although several kinase-deficient Trk receptor isoforms have been identified over the years for both the and the genes, only TrkBT1 and the truncated TrkC isoform, which we call TrkCT1 in this study (also known as TrkCTK [; ; ], TrkCNC2 [], and TrkCic158 []), are believed to play important roles in vivo. The cytoplasmic tails of these truncated receptors are encoded by separate exons, which are evolutionarily conserved. Their protein products are present in both the embryo and in the adult animal, and their expression is dynamically regulated during development (; ). To date, the main function attributed to the kinase-deficient truncated Trk isoforms is inhibition of the kinase-active receptor isoforms, which is achieved by acting as a dominant-negative inhibitor of the full-length receptor or by a ligand-sequestering mechanism, which limits the neurotrophic factor available to bind the kinase-active receptor (; ). However, the high degree of sequence conservation of the intracellular domains of truncated receptors among species suggests the potential for other functions, such as interaction with cytoplasmic adaptor proteins and activation of signaling pathways (; ). Indeed, it has recently been reported that brain-derived neurotrophic factor induces the production of calcium waves in astroglia through the truncated TrkBT1 receptor, and that TrkBT1 can alter astrocytic morphology via the regulation of Rho GTPase activity (; ). To date, no molecules have linked truncated TrkCT1 receptors to intracellular signaling pathways. Moreover, there are no data on direct biological functions, per se, although it has been reported that TrkCT1 with p75 can induce neural crest cell differentiation, and in animal models of glaucoma truncated TrkCT1 is overexpressed concomitantly with retinal ganglion cell death (; ). We present the identification of a new signaling pathway activated by the kinase-deficient TrkCT1 receptor that employs the scaffold protein tamalin (; ), the cytohesin-2–Arf nucleotide-binding site opener (ARNO), the ADP-ribosylation factor 6 Arf6), and the Rac1 GTPase. We show that neurotrophin-3 (NT3) activation of this signaling cascade by TrkCT1 causes Arf6 translocation to the membrane, followed by actin reorganization and membrane ruffling. Thus, we have identified a new pathway that provides a mechanism by which NT3 can control cell morphology, shedding light on the elusive role of abundantly expressed truncated Trk receptors in development. Moreover, it provides the only completely defined growth factor–activated pathway leading to Arf activation. The unique COOH terminus of TrkCT1 is encoded by two exons (13b and 14b in human; ; ). Exon 14b is the most conserved among species such as mouse, human, rat, and chicken. Therefore, we used a yeast two-hybrid system to screen an adult mouse brain cDNA library with the 13-aa-long exon 14b (38 aa) as bait for interacting proteins (see Materials and methods). This approach yielded several candidate genes, including four independent clones for GRASP/tamalin (; ). These clones initiated at proline 19, alanine 22, arginine 68, and arginine 80. Full-length cDNA for tamalin was not isolated. Next, we analyzed the specificity of interaction between tamalin and TrkCT1 by assessing yeast two-hybrid β-galactosidase activity in liquid assays with a series of COOH- and NH-terminal deletions of tamalin and exon 13b and/or 14b of TrkCT1. All tamalin plasmids isolated from the brain cDNA library contained an intact Psd-95/Dlg/ZO1 (PDZ) domain, suggesting that this domain may be involved in the interaction with TrkCT1. Indeed, significant β-galactosidase activity can only be detected when the tamalin PDZ domain is intact (). Furthermore, the phylogenetically conserved exon 14b of TrkCT1 is both necessary and sufficient to promote association with the tamalin protein. Interestingly, full-length tamalin generates the lowest β-galactosidase activity when compared with the other constructs that have an intact PDZ domain. Most likely, the full-length fusion protein does not localize to the nucleus to activate reporter expression. This could also explain why no full-length tamalin cDNA clones were isolated in the original screen. To further confirm the interaction between tamalin and TrkCT1 observed in yeast, we performed in vitro GST pull-down assays. We tested whether a full-length, COOH-terminal, HA-tagged tamalin could bind different GST–TrkCT1 fusion proteins (). GST alone or GST–TrkCT1 fusion fragments containing the juxtamembrane domain or exon 13b failed to pull down tamalin. However, GST fusion proteins containing exon 14b were able to interact with tamalin. The deletion of even a single aa from the exon 14b COOH terminus was sufficient to abolish binding to tamalin, suggesting that the COOH-terminal tail is responsible for binding to the tamalin PDZ-binding domain (unpublished data). These data confirmed that the interaction between tamalin and TrkC requires an intact exon 14b, which is unique to the TrkCT1 isoform. Furthermore, the results suggest that the juxtamembrane region, which is common to both the TrkCT1 and the TrkC kinase–active (TrkC-kin) isoforms, is not involved in the interaction. Additional GST pull-down experiments performed with GST fusion proteins, including the kinase domain, also failed to pull down tamalin (unpublished data), confirming that the interaction between tamalin and TrkC receptors is specific to the truncated TrkCT1 isoform. To verify that TrkCT1–tamalin interaction is physiologically possible in the animal, we analyzed their pattern of expression in the adult mouse brain. In situ hybridization experiments, using specific tamalin and TrkCT1 antisense riboprobes, showed that this scaffold protein gene and the truncated TrkC receptor have an overlapping pattern of expression in several areas of the mouse brain. For example, both genes are highly expressed in the hippocampus, cortex, striatum, and olfactory bulb (). Furthermore, immunofluorescence staining of human embryonic kidney 293 (HEK293) cells transfected with TrkCT1 and tamalin-HA shows that they colocalize at the plasma membrane, indicating that the subcellular localization of these proteins allows their association (). Next, we used coimmunoprecipitation (coIP) experiments to investigate if tamalin interacts with TrkCT1 in mammalian cells. Lysates obtained from HEK293 cells expressing tamalin-HA and the truncated TrkCT1 isoform were immunoprecipitated with an anti-TrkC serum and immunoblotted with an anti-HA antibody. As shown in , the anti-TrkC–specific antibody can coIP tamalin-HA only when both tamalin and TrkCT1 proteins are present. Therefore, TrkCT1 does interact with tamalin in mammalian cells. To gain insight into the physiological relevance of TrkCT1–tamalin binding we next investigated whether the TrkC ligand NT3 would affect this interaction. TrkCT1 and tamalin were coexpressed in HEK293 cells. Before lysis, cells were serum starved for 4 h, followed by treatment with NT3 for 5 min (). Interestingly, the amount of tamalin-HA associated with TrkCT1 was significantly increased by NT3 (). Quantitation from multiple coIP experiments showed that the addition of NT3 led to a threefold increase in the amount of tamalin bound to TrkCT1 (). These data show that the ligand NT3 promotes interaction of TrkCT1 with tamalin and support the physiological relevance of the TrkCT1–tamalin association. The finding that NT3 promotes the interaction of TrkCT1 with tamalin in vitro prompted us to expand this observation in a physiological context. Mouse hippocampal neurons, which in the adult express the highest levels of both tamalin and TrkCT1 (), were isolated from embryos, differentiated for 10–14 d in culture to allow neurons to mature and endogenously express truncated TrkCT1 and tamalin, and treated with NT3. Confocal microscopy analysis of control neurons stained with a specific antibody recognizing the tamalin COOH terminus showed that this protein has a diffuse, mainly cytoplasmatic pattern of expression (). Conversely, TrkC receptors visualized with a mouse monoclonal antibody recognizing the extracellular region of TrkC appear to be present uniformly along the plasma membrane, as previously described (), with little or no colocalization with tamalin (). Interestingly, after NT3 treatment, tamalin distribution becomes more punctuate, especially along the neuronal axons, and appears to colocalize with the TrkC receptors. Because this experiment shows that tamalin redistributes in response to NT3 and colocalizes with TrkC receptors, we repeated this analysis in neurons from a TrkC mutant mouse that specifically lacks the TrkC kinase isoforms but retains the truncated ones (; data not depicted). Immunostaining of neurons lacking the kinase TrkC isoform reveals a specific punctuated TrkC pattern of expression that is consistent with the previously reported pattern of expression of TrkCT1 (). Remarkably, neuronal treatment with NT3 shows dotted neuronal redistribution of tamalin that matches the TrkCT1 cellular pattern of expression (, and insets therein). Collectively, these data demonstrate that in neurons endogenous tamalin redistributes in response to NT3 and colocalize with TrkCT1. Tamalin forms a protein complex with cytohesin-2–ARNO (; ), which is a member of the guanine nucleotide exchange factors for the Arf1 and Arf6 GTP-binding proteins. ARNOs have been proposed to be part of signaling cascades based on the finding that growth factor or insulin treatment of cells results in the translocation of ARNOs to the cell edge (; ). These results motivated us to examine the effects of NT3 binding to the TrkCT1–tamalin complex on Arf subcellular localization and cell morphology. Because HEK293 cells express tamalin (although at a low level; unpublished data), cytohesin family exchange factors, and Arfs endogenously (; ), expression of TrkCT1 should reconstitute the NT3-activated pathway. After the generation of a stable line expressing TrkCT1, we assessed the translocation of epitope-tagged Arf6 to actin-rich ruffles and protrusions as a measure of Arf6 activation in response to NT3 (, ; ; ; ; ; ). In both the parental HEK293 and TrkCT1 cells, Arf6 associated to a small extent with the edge of the cell, as well as with a tubular endosomal compartment (). Treatment of parental HEK293 cells with NT3 had no significant effect on the distribution of Arf6 (); in contrast, treatment of the HEK293-TrkCT1 cells resulted in the redistribution of Arf6 from the tubular endosomal compartment to the cell edge in actin-rich ruffles ( [arrows] and Y), whereas Arf1 had a perinuclear distribution that was not affected by NT3 (not depicted). This translocation from the tubular endosomal compartment to the cell edge is a well defined consequence of Arf6 activation (; ). Moreover, tamalin was necessary for Arf6 activation through TrkCT1 because transfection of two independent dominant-negative forms containing the PDZ domain of tamalin blocked these effects (). In contrast, the response of TrkCT1 cells overexpressing tamalin to NT3 was more robust than that of the nontransfected cells, with Arf6 moving from a tubular endosomal compartment () to the cell edge in ruffles ( [arrows] and Y) and large protrusions (, arrowheads) colocalizing with tamalin. Because we could not exclude that dominant-negative tamalin might block NT3 signaling by masking the TrkCT1 docking site encoded by exon 14b, we next investigated whether a catalytically inactive form of cythohesin-2–ARNO would affect NT3-mediated ruffling (). E156K-ARNO has a mutation in the Sec7 domain that blocks the ability of ARNO to promote GTP exchange at Arf6 but does not disrupt its ability to bind to other proteins (). Transfection of E156K-ARNO into HEK293-TrkCT1 cells blocked actin polymerization (), whereas wild-type ARNO enhanced NT3-mediated actin polymerization at the cell edge (). Thus, these data strongly suggest that NT3 mediates ruffling through the TrkCT1–tamalin–cythohesin-2–ARNO pathway. Arfs function as GDP/GTP-regulated switches in the pathways that stimulate actin reorganization and membrane ruffling (). Part of this effect is mediated by Rac1, which is a well characterized Rho GTPase located downstream of Arf6 (; ). Therefore, we investigated whether NT3 signaling trough TrkCT1–tamalin leads to Rac1 activation (; ). We found that NT3 treatment of HEK293 cells transfected with TrkCT1 and tamalin caused a 4.1 ± 0.47–fold increase in Rac1 activity (, compare lanes 5 and 6). Moreover, only expression of a dominant-negative form of Arf6 could block Rac1 activation, whereas expression of inactive Arf1 had no effect on Rac1 activation (, lanes 7 and 8). These data demonstrate that NT3 signaling through TrkCT1–tamalin can modulate Rac1 through Arf6 activation. In this study, we describe the identification of a new signaling pathway activated by an evolutionarily conserved truncated isoform of the TrkC receptor. This pathway links NT3 to downstream molecules affecting the regulation of actin cytoskeleton and membrane trafficking through TrkCT1 and the scaffold protein tamalin (). The interaction of TrkCT1 with tamalin suggests a novel mechanism by which neurotrophins affect the development, maintenance, and function of the mammalian nervous system. Tamalin forms a protein complex with multiple postsynaptic and protein-trafficking scaffold proteins, suggesting that NT3 could be an extracellular trigger of these cellular processes by causing the recruitment of tamalin to TrkCT1 (). Among the several postsynaptic proteins with which tamalin interacts are the group 1 metabotropic glutamate receptors (mGluR1 and mGluR5) that affect ion channels and intracellular second messengers through G proteins (; ). Glutamate receptors play a key role in neural development, especially in neuronal plasticity (; ). It has been suggested that these mGluR1 functions may be mediated by tamalin, which provides a direct link to cytohesin-2–Arf GTP-binding proteins, which are involved in membrane trafficking, cytoskeletal reorganization, and phospholipase D activation (; ). Thus, it is possible that the interaction of tamalin with TrkCT1 may provide a molecular pathway similar to the one activated by mGluR1 and by which neurotrophins can affect neuronal plasticity in mammals, a function that is still not well understood. Several lines of evidence suggest that characteristic axonal and dendritic morphologies throughout the nervous system are determined, in part, by local patterns of expression of neurotrophins and neurotrophin receptors (; ; , ; ). For example, transfection of explant cultures of dorsal root ganglia neurons with the TrkC kinase or the TrkCT1 receptor revealed that specific ratios of receptor isoforms are associated with different axonal morphologies. Specifically, neurons overexpressing tyrosine kinase containing TrkC showed enhanced elaboration of major axonal processes, whereas the truncated isoform reduced elaboration of major processes and increased branching (). Our data suggest that the effect on truncated TrkCT1 by NT3 may affect neuronal morphology, not by affecting the TrkC full-length receptor by a dominant-negative mechanism, but, rather, by activating Arfs and Rac1 GTPases that are known modulators of cell morphology. In summary, this newly discovered neurotrophin-activated pathway may provide a means by which NT3 can exert some of its critical functions in development and might explain why mice lacking all TrkC receptor isoforms, including the TrkCT1, display a more severe phenotype than mice lacking the full-length TrkC tyrosine kinase receptors exclusively (). Importantly, these results identify a novel signaling pathway upstream of Arf6 in which an activated receptor binds directly to an ARNO-binding protein to stimulate ARNO and consequent nucleotide exchange on Arf6. The PCR-amplified COOH-terminal region of the TrkCT1 isoform (corresponding to residues 575–612 of the rat TrkCT1 [] or exon 14b of the human TrkC-truncated receptor []) was subcloned in frame to the GAL4-binding domain of the bait plasmid pGBD.C2. This fusion bait plasmid and an adult mouse brain cDNA library fused in frame to the GAL4 activation domain (Mouse Brain MATCHMAKER cDNA Library; CLONTECH Laboratories, Inc.) were sequentially transformed into the yeast strain PJ69-4A and subjected to colony selection and liquid β-galactosidase assays. Different tamalin deletion prey and TrkCT1 bait plasmids were generated by cloning PCR-generated cDNA fragments into pACT2 (CLONTECH Laboratories, Inc.) and pGBD.C2, respectively, and were used for subsequent two-hybrid analysis. HEK293 cells were grown in DME (Invitrogen) supplemented with 10% FBS, 100 U/ml of penicillin/streptomycin at 37°C, and 5% CO. Cells were transfected using Fugene 6 (Roche) or Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Mouse embryonic hippocampus neurons were dissected from 18-d-old mouse embryos by the standard technique and seeded on coverslips pretreated with poly--lysine, collagen, and laminin. Cells were grown for 2 wk on DME supplemented with 2% B27, 5% fetal calf serum, 100 U/ml of penicillin/streptomycin, and 4 μM AraC; the media was changed every 3–4 d. GST–TrkC fusion proteins were generated by cloning the appropriate PCR products containing different rat TrkCT1 or TrkC full-length regions () in pGEX-4T-1 (GE Healthcare). The expression of the GST fusion proteins in strain BL21 was induced with IPTG, and soluble cell extracts were generated by sonication in lysis buffer containing PBS, pH 7.4, 1 mM EDTA, 1% Triton X-100 and complete mini EDTA-free protease inhibitor (Roche). The soluble proteins were immobilized on glutathione–Sepharose CL-4B beads (GE Healthcare) for 1 h at RT, washed five times with lysis buffer (PBS, pH 7.4, 1 mM EDTA, 1% NP-40, and complete mini EDTA-free protease inhibitor), and incubated with HEK293 cell extracts expressing the full-length COOH-terminal HA-tagged tamalin (tamalin-HA) protein in 500 μl of lysis buffer for 10–12 h at 4°C. The beads were washed five times with lysis buffer on a MicroSpin column (GE Healthcare). The bound proteins were eluted by centrifugation after heating at 95°C for 3 min in 2× SDS-PAGE loading buffer, separated by SDS-PAGE on 4–12% acrylamide gels (Invitrogen), and immunoblotted for detection with an anti-HA mouse monoclonal peroxidase-conjugated antibody (High Affinity 3F10; Roche). A HEK293 line expressing TrkCT1 (HEK293-TrkCT1) was generated by transfection with a pcDNA3.1 puromycin-resistant plasmid containing the TrkCT1 cDNA. For protein expression experiments, a tamalin-HA plasmid was transiently transfected into naive HEK293 cells or the HEK293-TrkCT1 line, and lysates were prepared for protein analysis. A goat anti-TrkC (Upstate Biotechnology) antibody was used for coIP experiments. Immunoprecipitates were resolved by SDS-PAGE (4–12% gel) and immunodetected with the anti-HA antibody and the rabbit anti–TrkC 656 antiserum (raised against aa 484–501 of the juxtamembrane region; a gift from P. Tsoulfas, University of Miami, Miami, FL). Rac1 activation was measured using a nonradioactive Rac activation assay kit (CHEMICON International, Inc.) following the manufacturer's recommendations. In situ hybridization protocols using the antisense full-length tamalin () or the TrkCT1-specific sequence (corresponding to exon 13b and 14b and the subsequent 286 bases of the 3′ un-translated region) riboprobes were performed as previously described (). For immunocytochemical experiments, cells were transfected with the pcDNA3.1 (+) expression vector, containing the open reading frame for FLAG-tagged wild-type human Arf6 and ARNO (wild-type ARNO), Myc-tagged wild-type tamalin and mutant E156K-ARNO, and HA-tagged mutant (aa 1–189) tamalin. 10 μg of wild-type human Arf6 or 7 μg of wild-type ARNO and E156K-ARNO were used in single transfection; 5 μg of wild-type human Arf6 and 8 μg of wild-type or mutant tamalin were used in cotransfection experiments. Transfected cells were plated on 100-mm culture dishes precoated with poly--lysine (20 μg/ml for 10 min at 25°C). 24 h later, they were reseeded on fibronectin-coated coverslips in serum-free medium (OPTI-MEM; Invitrogen) and incubated for 12 h at 37°C. To determine their responses to NT3, cells or primary neurons were starved for 4 h and incubated with 100 ng/ml NT3 (Upstate Biotechnology) for 5 min at RT before fixation. HEK293 cells were fixed in 2% paraformaldehyde in PBS for 10 min at RT, washed three times with PBS and once in PBS containing 10% FBS and 0.04% sodium azide, and permeabilized with staining buffer (PBS containing 10% FBS, 0.2% saponin, and 0.04% sodium azide) for 10 min at RT. Mouse hippocampal neurons were fixed for 30 min in 4% paraformaldehyde, washed three times in PBS, and blocked overnight with mouse IgG blocking reagent (Vector Laboratories) in PBS. After two washes in PBS, they were incubated for 30 min in 0.1% Triton X-100, 5% mouse protein concentrate (Vector Laboratories), and 5% normal goat serum (S1000; Sigma-Aldrich) in PBS. After incubation with the primary antibodies in staining buffer for 1–3 h at RT, cells were washed three times for 10 min with staining buffer and incubated with FITC- or rhodamine-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) for 1 h at RT. After three 10-min washes, glass coverslips were mounted on slides with fluorescent mounting medium (Dakocytomation). Cells were visualized at RT by confocal microscopy with a confocal system (LSM510 NLO) and with an inverted microscope (Axiovert 200M; both Carl Zeiss MicroImaging, Inc.) and operating with a 25-mW argon laser tuned to 488 nm and a 1-mW HeNe laser tuned to 543 nm. Cells were imaged with a 63×, 1.4 NA, oil immersion objective for HEK293 cells and Plan-Neofluar 40×, 1.3 NA, oil differential interference contrast objective for neuron analysis. Using AIM software (Carl Zeiss MicroImaging, Inc.), images were collected using a multitrack configuration where the FITC and rhodamine signals were sequentially collected with a bp 500–550-nm filter and a bp 565–615-nm filter after sequential excitation with the 488- and 543-nm laser lines, respectively. Antibodies were used at the following concentration in PBS: polyclonal anti-TrkC 656 (1:100), polyclonal anti-HA (1:5,000; Roche), monoclonal M5 anti-FLAG (1:1,000; Sigma-Aldrich), polyclonal anti-Myc (1:100; Santa Cruz Biotechnology, Inc.), polyclonal rabbit anti-tamalin (1:1,500, against the TEAREQALCGAGLRKTKYRSFR epitope of the mouse tamalin COOH terminus), and mouse monoclonal TrkC 2B7 (1:1,500, raised against epitope ESTDNFILFDEVSPTPPI of the human TrkC extracellular domain).

Distinct signaling pathways have been demonstrated to mediate estrogen (E2) action and to directly affect its function. Examples include the regulation of normal mammary development and breast cancer growth. E2 is known to be coupled with growth factor–signaling networks to promote enhanced cell growth in human breast cancer. Several growth factors and their receptors are known to participate in E2 signaling, amongst which the EGF receptor (EGFR) family of receptor tyrosine kinases are of particular interest because of their critical involvement in human cancer (; ). Indeed, aberrant expression and activation of EGFR is frequently observed in various tumors, especially of the breast and ovary, where it correlates with a poorer patient prognosis (; ). In addition, up-regulation of EGFR signaling is thought to be an important mechanism that confers antiestrogen resistance of breast cancer, resulting in a failure of endocrine therapy (; ). Multiple lines of evidence have suggested that the interaction of EGFR with E2 signaling can occur at various levels. E2 primarily acts on nuclear estrogen receptors (ERs), leading to regulation of gene expression, which was traditionally deemed the genotropic action of E2. Many E2-responsive genes are indeed key signaling molecules that participate in EGFR signaling (for review see ). Alternatively, a cell membrane–associated form of ER (mER) has been reported to couple with and activate various G proteins, and thereby mediate the EGFR transactivation, serving as a nongenotropic effect of the ER (; ). More recently, an orphan G protein–coupled receptor (GPCR), GPR30, has been suggested to be an intracellular receptor of E2 that specifically binds E2 with a high affinity and promotes various rapid E2 signaling events, such as Ca mobilization and activation of Akt cascades (; ). In addition, reported that E2-induced EGFR transactivation was mediated via GPR30, suggesting a model of EGFR transactivation by E2 similar to that induced by other well documented GPCR ligands (). However, as GPR30 was found to be uniquely localized to the endoplasmic reticulum (), whether this intracellular receptor coupled with G proteins can directly transactivate EGFR and the physiological function of GPR30 remains to be investigated. Since first described by , the transactivation of EGFR by GPCR ligands has been considered an important model of cellular signal transduction. Several GPCR ligands, including lysophosphatidic acid, thrombin, angiotensin II, and endothelin-1, have been documented to transactivate EGFR, leading to activation of survival or mitogenic pathways (). Sphingosine 1-phosphate (S1P), which is a recently identified GPCR ligand (), has also been shown to induce EGFR transactivation through S1P receptors (; ). We recently demonstrated that E2 serves as a potent activator of sphingosine kinase-1 (SphK1), which is a key enzyme that catalyzes the formation of S1P (). We also demonstrated that the activation of SphK1–S1P signaling participates in the nongenomic action of E2, including intracellular Ca mobilization and ERK1/2 activation (). Moreover, SphK1 activity has been shown to regulate neoplastic cell growth of breast cancer in response to E2 stimulation at both an in vitro and an in vivo level (; ), suggesting an important role of SphK1 in the transmission of E2 signaling in breast cancer cells. In this study, we provide evidence that not only demonstrates the capacity of S1P to stimulate EGFR transactivation in its own right through the S1P receptors in breast cancer cells but also reveals a critical role for SphK1 in mediating E2-induced EGFR transactivation in an S1P receptor–dependent manner. Furthermore, these findings illustrate a novel signaling mechanism, called criss-cross transactivation, which is triggered by SphK1 activation that signals between three individual ligand-receptor systems (i.e., E2, S1P, and EGF). Treatment of MCF-7 cells with S1P resulted in significant increases in tyrosine phosphorylation of EGFR in a concentration-dependent manner (). A significant response to S1P was commenced at 1 nmol/liter and peaked at ∼100 nmol/liter, which fits well within the range of reported binding affinities to S1P receptors (). In parallel, S1P treatment caused a significant increase in ERK1/2 phosphorylation, which is a key downstream signaling event of EGFR activation, in a similar concentration-dependent pattern to the S1P-induced EGFR phosphorylation (). Time course studies showed that S1P induced both EGFR and ERK1/2 phosphorylation that peaked at 10–15 min and decreased thereafter, but was still evident at 240 min after stimulation (). Collectively, these results demonstrate an ability of S1P to induce EGFR activation in MCF-7 cells, which was consistent with the observations previously reported in vascular smooth muscle cells () and fibroblasts (). We have recently reported that E2 was capable of inducing SphK1 activation and S1P formation that participated in E2 nongenomic signaling (). The ability of E2 to induce EGFR transactivation has been previously demonstrated (; ). We sought to determine whether S1P could mimic E2 to stimulate EGFR transactivation. Treatment of MCF-7 cells with E2 resulted in a rapid tyrosine phosphorylation of EGFR and ERK1/2 phosphorylation similar to that observed in the S1P-treated cells (). Both E2- and S1P-induced activation of EGFR and ERK1/2 were blocked by pertussis toxin (PTX), which is a Gi-specific inhibitor (). In contrast, PTX had no effect on EGF-stimulated phosphorylation of EGFR and ERK1/2 (). As transactivation of EGFR relies on its internal tyrosine kinase activity (), we examined whether the tyrosine kinase activity is required for E2- or S1P-induced EGFR transactivation. In the presence of AG1478, which is a specific EGFR tyrosine kinase inhibitor, both E2- and S1P-induced activation of EGFR and ERK1/2 were abolished completely (). Serving as a control, EGF-stimulated autophosphorylation of EGFR was completely inhibited by AG1478, supporting its specific effect on EGFR activity in MCF-7 cells. The Src family of kinases has been reported to play a signaling role in GPCR-mediated transactivation of EGFR (). Src was also suggested to be required for E2-induced EGFR transactivation (; ). Consistent with these previous studies, both E2- and S1P-stimulated activation of EGFR and ERK1/2 were significantly inhibited by the Src-specific inhibitor PP2 (), supporting a role for Src in mediating either E2- or S1P-induced transactivation of EGFR. The shedding of heparin-binding EGF (HB-EGF) upon matrix metalloprotease (MMP) activation has also been recognized as an important mechanism in mediating EGFR transactivation by GPCR ligands () or by E2 (; ). Therefore, we examined whether HB-EGF shedding was involved in E2- or S1P-induced EGFR transactivation. We subjected MCF-7 cells to an acid-wash step, to reduce background autocrine stimulation, and pretreated the cells with the MMP inhibitors phenanthroline or GM6001. Both E2- and S1P-induced EGFR and ERK1/2 activation were blocked by these two MMP inhibitors (). In contrast, the MMP inhibitors had no effect on EGF-induced autophosphorylation of EGFR. Furthermore, depletion of HB-EGF production from the culture media by EGF-neutralizing antibodies resulted in a significant inhibition of both E2- and S1P-induced EGFR activation (). Efficiency and specificity of the neutralizing antibodies were demonstrated by the inhibition of EGF-stimulated EGFR tyrosine phosphorylation. These data suggest that HB-EGF shedding and release are required for both E2- and S1P-induced EGFR transactivation. As S1P was able to mimic the effect of E2-stimulated EGFR transactivation, and E2 was capable of stimulating S1P production upon SphK1 activation, we hypothesized that the E2-induced EGFR transactivation could be mediated by SphK1 activation. To test this hypothesis, we used stably transfected MCF-7 cell lines overexpressing wild-type SphK1 (SphK1), dominant-negative SphK1 (SphK1), or empty vector alone. Previously, we demonstrated that the baseline SphK activity in SphK1-transfected cells was ∼10-fold higher than in control cells (). E2 stimulation resulted in a rapid increase in SphK activity of approximately twofold more than the basal level in both SphK1-transfected and control MCF-7 cells (). In contrast, the SphK1-transfected cells had a similar basal SphK activity to the control cells, whereas E2-stimulated SphK activity was completely abolished (). Interestingly, although E2-stimulated tyrosine phosphorylation of EGFR and ERK1/2 phosphorylation were enhanced in SphK1-transfected cells, the stimulatory effect of E2 was abrogated in the SphK1 transfectants (). There were no significant differences in total EGFR and their cell-surface expression levels between these transfected cell lines (). In contrast, neither EGF nor S1P-stimulated EGFR phosphorylation was significantly influenced by SphK1. Thus, these results suggest a specific role for SphK activity in the E2-induced EGFR transactivation. Two human SphK isoforms, SphK1 and SphK2, have been identified, and both isoforms account for total cellular SphK activity (; ). To define which isoform (if not both) is responsible for the transactivation of EGFR, as well as the role of endogenous SphK, we used an siRNA strategy to down-regulate each isoform's expression levels in MCF-7 cells. Endogenous SphK1 and SphK2 levels were reduced by 86 and 67%, respectively, after treatment with SphK1- or SphK2-specific siRNA, compared with cells treated with a scrambled siRNA (). The specificity of these siRNAs was demonstrated by their inability to inhibit the alternative isoform of SphK and the control gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Both SphK1- and SphK2-siRNA decreased the baseline levels of SphK activity by ∼50%. Whereas SphK1-siRNA significantly attenuated SphK activity in response to E2 stimulation, the extent of the E2-induced increases in SphK activity was not changed by SphK2-siRNA (), suggesting that SphK1 is the key isoform responsible for the E2-induced SphK activity. Furthermore, by down-regulating SphK1, SphK1-siRNA significantly attenuated the E2-induced EGFR and ERK1/2 phosphorylation to an extent that was similar to SphK1-transfected cells (). In contrast, SphK2-siRNA had no effect on the E2-induced phosphorylation of EGFR and ERK1/2. Again, serving as controls, neither EGF- nor S1P-induced EGFR and ERK1/2 phosphorylation were inhibited by SphK1- or SphK2-siRNA (). Collectively, these data suggest a critical role for endogenous SphK1, but not for SphK2, in mediating E2-stimulated transactivation of EGFR. Our previous work suggested that the E2-induced SphK1 activation was likely to be mediated by mER in a G protein–dependent manner (). As GPR30, which is an orphan GPCR, has been more recently identified as an E2-specific GPCR (; ), we sought to define the role of GPR30 in E2-induced SphK1 activation. To this end, we used GPR30 antisense oligonucleotides (AOs) that specifically down-regulated GPR30 expression in MCF-7 cells (). The cells treated with AO-GPR30 resulted in a significant reduction of the E2-induced increases in SphK activity (), suggesting a critical involvement of GPR30 in the E2-stimulated SphK1 activation. Serving as a control, AO-GPR30 had no effect on EGF-induced SphK activity (). Consistent with our previous study (), treatment of cells for 18 h with ICI 182780, which down-regulated ERα expression (), resulted in a reduction of the E2-induced SphK activity similar to that observed in the AO-GPR30–treated cells (not depicted). Consequently, the E2-induced EGFR and ERK1/2 phosphorylation were significantly inhibited by either AO-GPR30 or ICI 182780 (). In contrast, neither AO-GPR30 nor ICI 182780 had effects on the S1P- or EGF-induced EGFR and ERK1/2 phosphorylation. Collectively, these data suggest that both GPR30 and ERα are capable of mediating SphK1 activation, and the resultant EGFR transactivation in response to E2 stimulation. The biological function of SphK1 relies on its product, S1P, which functions chiefly as a ligand for the Edg family of GPCR receptors (; ). Therefore, we sought to determine the role for S1P and its receptors in the SphK1-dependent EGFR transactivation induced by E2. We first examined whether S1P is released upon SphK1 activation in cells responding to E2 stimulation. As shown in , in parallel with the elevated intracellular content of S1P, S1P levels were increased by 86% (P < 0.01) in conditioned media (CM) collected from the E2-stimulated MCF-7 cells in comparison to that from unstimulated cells. No increase in S1P levels were detected after E2 stimulation in CM from the SphK1-transfected cells (), indicating that SphK1 activation is responsible for the elevated S1P production and release from the E2-stimulated cells. Correspondingly, CM derived from the E2-treated cells exhibited a substantially greater capacity to stimulate EGFR tyrosine phosphorylation compared with the CM from untreated cells (). Furthermore, treatment with CM derived from the E2-treated SphK1-transfected cells that contained high levels of S1P () resulted in a further increase in EGFR phosphorylation, whereas CM derived from the SphK1 transfectants had no detectable effect on the EGFR phosphorylation (). These results suggest that the ability of the CM to stimulate EGFR activation was dependent on its cellular SphK1 activity and the amount of S1P release. To explore this notion further, we used two strategies: (a) we lipid stripped CM to remove S1P, and (b) before CM stimulation, we treated cells with PTX that has been reported to block the majority of S1P receptors (). Either lipid-stripped CM or CM pretreated with PTX completely abolished the CM-induced EGFR activation (), supporting a critical involvement of S1P and its receptors in the E2-stimulated transactivation of EGFR. According to , Edg-3, which is a PTX-sensitive GPCR, is the predominantly expressed S1P receptor in MCF-7 cells. To evaluate the potential role of Edg-3 in E2-induced EGFR transactivation, we used the antisense strategy to knockdown endogenous Edg-3 expression. Cells transfected with AO-Edg3 resulted in a significant down-regulation of Edg-3 expression levels (∼80%; ). Correspondingly, S1P-induced EGFR tyrosine phosphorylation was also blocked by AO-Edg3 (). Moreover, treatment of MCF-7 cells with AO-Edg3 caused a significant reduction in E2-induced EGFR tyrosine phosphorylation, whereas EGF-induced EGFR autophosphorylation was retained (). Furthermore, as a functional consequence, E2-induced cell growth was significantly inhibited by AO-Edg3 to a similar extent as that observed in cells treated with the EGFR inhibitor AG1478 (). Collectively, these findings suggest that the S1P receptor Edg-3 is required for the E2-induced EGFR transactivation and cell growth in MCF-7 cells. The current understanding of cell signaling has grown broadly, from individual ligand-receptor systems, such as those controlled by GPCR or receptor tyrosine kinases, to an interdependent network that is capable of communicating across individual signaling systems. One particular example is that of EGFR, which can be transactivated by several GPCR ligands (). Although the mechanism that controls this transactivation remains largely unknown, EGFR transactivation has been recognized as an important pathway in the regulation of complex biological processes, such as cancer development. In this study, we demonstrate that E2, acting on its own receptors (GPR30 and/or mER), results in the activation of the S1P-specific receptor Edg-3 via SphK1 activation, leading to EGFR transactivation (summarized in ). To the best of our knowledge, this is the first work to describe such a signaling phenomenon, i.e., a given GPCR ligand–mediated (S1P) EGFR transactivation is driven by another independent ligand (E2), which suggests a new model of criss-cross transactivation between three individual ligand-receptor systems. As described in this study, SphK1, which is the enzyme that catalyzes S1P formation, plays an essential role in this criss-cross transactivation phenomenon. We have previously shown that E2 stimulates SphK1, resulting in both a rapid, transient response and a delayed but prolonged activation (). Although the latter response relies on ER transcriptional activity, the E2-induced rapid activation of SphK1 appears to be necessary for E2 cytoplasmic signaling, such as intracellular Ca mobilization and ERK1/2 activation (). In addition, cellular SphK activity has been functionally linked to the E2-dependent mitogenic and carcinogenic action in human breast cancer (; ), suggesting an important signaling role of SphK1 in the biological function of E2. Indeed, SphK1, serving as an agonist-activated signaling enzyme, has been implicated in a wide spectrum of agonist-driven cellular responses, including cell survival, motility, proliferation, and differentiation (). This pleiotropic action of SphK1 is attributed to its product, S1P, which functions as both an intracellular second messenger and a ligand for cell-surface receptors (; ). S1P receptors belong to the Edg family of GPCR, which consists of Edg-1 (also called S1P), -3 (S1P), -5 (S1P), -6 (S1P), and -8 (S1P; ). The identification of S1P as a ligand for GPCR has provoked exploration of a potential role for S1P in the transactivation of receptor tyrosine kinases. recently reported that S1P was capable of inducing transactivation of EGFR and the platelet-derived growth factor β receptor in vascular smooth muscle cells. Consistent with this finding, we are able to confirm that S1P transactivates the EGFR in MCF-7 breast cancer cells, in both a time- and dose-dependent manner (). S1P, like many other GPCR ligands, induces EGFR transactivation via Gi activation of the GPCR and the intrinsic kinase activity of EGFR (), as demonstrated by the inhibitory effects of PTX and AG1478 on S1P-induced transactivation of EGFR (). The ability of GPCR ligands to activate MMP, resulting in HB-EGF shedding and release, has been demonstrated as a necessary event in GPCR-mediated transactivation of EGFR (). Although we have not directly determined the MMP activities and HB-EGF production in this study, MMP activation is likely involved in S1P-induced EGFR transactivation, as the transactivation of EGFR was blocked by either the MMP inhibitors (phenanthroline or GM6001) or EGF-specific neutralizing antibodies (). Additional investigations are needed to determine the effect of S1P on MMP activation. E2, which is a steroid hormone that functions primarily through its nuclear receptors (ERα and ERβ), has recently been shown to elicit a variety of rapid nongenotropic effects, including intracellular Ca mobilization; activation of adenylyl cyclase, Raf-1, c-Src, and ERK1/2; and EGFR transactivation (; ). These rapid actions of E2 are believed to be mediated by its membrane-associated receptors. Studies examining the identity of these receptors are ongoing, with evidence to suggest that they may be related to nuclear ER () or the orphan GPCR GPR30 (), or form part of a GPCR–ER complex (). More recently, GPR30 has been suggested as an E2-specific intracellular receptor with a high affinity and a specific binding site for E2 (; ). In keeping with these findings, we found that down-regulation of GPR30 expression in MCF-7 cells by AO-GPR30 attenuated the E2-induced SphK activity (). On the other hand, we have previously reported that down-regulation of ERα in MCF-7 cells by long-term treatment with ICI 182780 resulted in a similar inhibition of E2-induced SphK activity (), suggesting the involvement of ERα in this signaling event. Consequently, E2-induced EGFR transactivation was significantly inhibited by either AO-GPR30 or ICI 182780 in MCF-7 cells (). These results suggest the capacity of both GPR30 and mER, and perhaps of their cooperative actions, to mediate E2-induced SphK1 activation and the resultant EGFR transactivation. However, it remains to be defined if and how these receptors function cooperatively in transmitting E2 signaling in breast cancer cells. The role of SphK1 activation in the coupling of E2-induced EGFR transactivation was further demonstrated by the following series of observations: (a) CM obtained from E2-stimulated cells that contained higher levels of S1P were capable of inducing EGFR activation; (b) removal of S1P from CM by either lipid stripping or the pretreatment of cells with PTX before CM stimulation completely abolished the ability of CM to stimulate EGFR phosphorylation; (c) abrogated SphK1 activation by the expression of SphK1 resulted in an attenuation of the E2-stimulated EGFR transactivation, whereas S1P-induced EGFR transactivation was preserved; (d) down-regulation of endogenous SphK1, but not of SphK2, by their specific siRNA caused a significant inhibition of both SphK1 activation and EGFR transactivation in response to E2 stimulation; and furthermore, (e) by down-regulating endogenous Edg-3, which is a specific receptor for S1P, AO-Edg3 profoundly inhibited the E2-induced EGFR transactivation. Thus, we have provided compelling evidence to suggest that an autocrine or paracrine S1P signaling loop, triggered by SphK1 activation, plays a critical role in transactivating EGFR through the S1P receptor Edg-3 in response to E2 stimulation. Despite Edg-3 being previously reported as (), and shown to be, the predominant receptor that accounts for the receptor-dependent action of S1P in MCF-7 cells, we are unable to rule out the possibility that other members of the S1P receptor family expressed in these cells may also be subsidiarily involved in the EGFR transactivation. This requires further investigation. It is noteworthy that by blocking E2-induced SphK1 activation without alterations in the baseline SphK activity, SphK1 attenuated the E2-stimulated EGFR transactivation. Moreover, although both SphK1- and SphK2-siRNA caused a decrease in the basal SphK activity, only SphK1-siRNA that inhibited E2-induced SphK1 activity was able to block the EGFR transactivation. In contrast, SphK2-siRNA had no effect on E2-induced SphK1 activity or EGFR transactivation. These results not only suggest a specific role for the SphK1 isoenzyme but also strongly indicate that the activation of SphK1, rather than its baseline activity, is critical for the E2-induced EGFR transactivation. In fact, the enzymatic function of SphK1 has been suggested to act at two levels: (a) the constitutive basal activity is involved in the catabolism of cellular sphingolipids and, therefore, may play a housekeeping role (); and (b) the agonist-induced elevated activity is fundamental for its signaling role in the regulation of many biological functions, including cell survival, proliferation, differentiation, and oncogenesis (, ; ). Although the detailed mechanism by which E2 induces SphK1 activation is currently unknown, E2 was able to stimulate SphK1 phosphorylation in an ERK1/2-dependent manner (Fig. S1, available at ). This is consistent with the findings of and suggests that ERK1/2-promoted phosphorylation is required for E2-induced SphK1 activation. Indeed, by inhibiting ERK1/2 activity, the ERK kinase-specific inhibitor U0126 not only blocked the E2-induced SphK1 phosphorylation but also significantly attenuated the SphK1-mediated EGFR transactivation in response to E2 stimulation (Fig. S2). Interestingly, in addition to the role of ERK1/2 in initiating SphK1 activation and the resultant EGFR transactivation, inhibition of SphK1 activity by either SphK1 or SphK1-siRNA resulted in a significant attenuation of ERK1/2 activation by E2. The E2-induced ERK1/2 activation was also inhibited by a blockade of EGFR transactivation in MCF-7 cells (), which is in agreement with previous studies (; ). Collectively, these observations suggest that ERK1/2 could be placed upstream or downstream of the SphK1 signaling and has a dual role in the initiation and amplification of a positive-feedback signaling loop across E2, SphK1, and EGFR in breast cancer cells. However, one question that has been raised by these observations is, how does E2 induce an “initial” activation of ERK1/2? Recent studies have demonstrated that membrane ERα was able to assemble a signalsome complex with various signal molecules, such as c-Src (), the p85 subunit of phosphoinositide 3 kinase (), or caveolin-1 (). Whether such complexes directly initiate ERK1/2 activation and the methodology to detect the initial signal require further investigation. Nevertheless, as ERK1/2, SphK1, and EGFR all possess potent mitogenic signals, this positive-feedback loop could contribute to the aberrant signaling associated with neoplastic cell growth. Indeed, inhibition of the SphK1–S1P pathway by expression of SphK1 () or AO-Edg3 () resulted in a significant inhibition of breast cancer cell growth in response to E2 stimulation, similar to that previously observed in experiments with the ERK1/2- or EGFR-specific inhibitors (; ). In summary, we have demonstrated for the first time that SphK1 plays a prominent role in mediating E2's nongenomic signaling across three membrane-spanning events including GPR30/mER, Edg3, and EGFR. Pathways triggered by receptor tyrosine kinases have been strongly implicated in the pathogenesis and progression of a variety of cancers, such as breast cancer (; ). Indeed, the retention, up-regulation, and transactivation of EGFR in endocrine-resistant or ER-negative tumors have been demonstrated to be associated with a more aggressive phenotype, high disease recurrence rates, and decreased patient survival (; ). Our previous studies have shown an oncogenic potential for SphK1 that is not only able to transform rodent fibroblasts and form tumors in nude mice (; ) but also able to potentiate the carcinogenic effects of an oncogene, (), in addition to that of E2 (). Thus, the findings reported here may represent a specific example of a general system in which SphK1 plays a coordinating role between multiple oncogenic signaling systems. This not only elucidates the molecular mechanism responsible for the carcinogenic potential of SphK1, but may also provide a potential target to create new therapeutic strategies for cancer treatment by blocking the SphK1 signaling pathway. Human MCF-7 (ERα/β) breast cancer cells were obtained from the American Type Culture Collection and cultured in phenol red–free DME (CSL Biosciences) containing 10% FBS. Constructs of SphK and SphK, and stably transfected MCF-7 cell lines overexpressing SphK, SphK, or empty vector alone, were previously described (; ). Chemically synthesized siRNA duplexes with 3′-fluorescein modification were purchased from QIAGEN. The siRNA targeted sequences were as follows: AAGAGCTGCAAGGCCTTGCCC (SphK1), AACCTCATCCAGACAGAACGA (SphK2), and AATTCTCCGAACGTGTCACGT (scrambled control siRNA). The following 18-mer phosphothioate oligonucleotides were synthesized by Geneworks: Edg-3 antisense 5′-CGGGAGGGCAGTTGCCAT-3′ and sense 5′-ATGGCAACTGCCCTCCCG-3′; and GPR30 antisense 5′-TTGGGAAGTCACATCCAT-3′ and sense 5′-GATCTCAGCACGGCAAAT-3′. For transfection, Lipofectamine 2000 reagent (Invitrogen) was used, and cells were seeded into 6-well plates at a density of 50,000 cells per well 1 d before the experiment. After a 36–48-h transfection, the targeted gene expression levels were detected by RT-PCR and/or Western blot. Total RNA was extracted from cell cultures using TRIzol (Invitrogen). First-strand cDNA was synthesized from 1 μg total RNA using Omniscript reverse transcriptase (QIAGEN) and oligo-dT primer (Geneworks), in a 20 μl total volume. SphK1, SphK2, and EdgG-3 were amplified on a PTC-100 programmable thermal controller (Bio-Rad Laboratories) with an internal GAPDH control. The primers used to amplify were as follows: SphK1 sense 5′-TTGAACCATTATGCTGGCTATGA and antisense 5′-GCAGGTGTCTTGGAACCC; SphK2 sense 5′-GCTCAACTGCTCACTGTTGC and antisense 5′-GCAGGTCAGACACAGAACGA; Edg-3 sense 5′-GCCCTCTCGTGGATTTTGG and antisense 5′-CGCATGGAGACGATCAGTTG; and GPR30 sense 5′-CTGGGGAGTTTCCTGTCTGA-3′ and antisense 5′-GCTTGGGAAGTCACATCCAT-3′. The amplified products were visualized by electrophoresis on 1.5% agarose stained with ethidium bromide. Images were captured on a gel documentation system (UVitec). Cells were harvested and lysed by sonication in lysis buffer containing 50 mM Tris/HCl, pH 7.4, 10% glycerol, 0.05% Triton X-100, 150 mM NaCl, 1 mM dithiothreitol, 2 mM NaVO, 10 mM NaF, 1 mM EDTA, and protease inhibitors (Roche). Aliquots of cell lysates were resolved by 8–12% SDS-PAGE and transferred to Hybond-P membranes (GE Healthcare). The membranes were then probed with the appropriate antibodies, according to manufacturer's standard method. The immunocomplexes were detected with an enhanced chemiluminescence PLUS kit (GE Healthcare). Densitometry was performed on a mode imager (Typhoon 9410; GE Healthcare) using the ImageQuant program (Molecular Dynamics). SphK activity was routinely determined by incubating the cytosolic fraction with 5 μM --sphingosine dissolved in 0.1% Triton X-100 and γ-[P]ATP for 30 min at 37°C, as described previously (). The enzyme activity was defined as the amount of S1P formation (picomoles per minute per milligram of protein). For measurement of S1P, an enzymatic method was used as previously reported (). After E2 stimulation for 30 min, cells and conditional media were collected separately and lipids were extracted by alkaline mixture of chloroform and methanol. The basic aqueous fractions containing S1P were incubated with alkaline phosphatase for 30 min at 37°C and lipids were extracted twice with 1 ml of acidic chloroform. Pooled organic fractions containing newly generated sphingosine were dried by vacuum-spin, resuspended in 100 μl of reaction buffer, and incubated with recombinant SphK1 and γ-[P]ATP for 30 min at 37°C. The generated [P]S1P was then resolved by TLC and quantified as described previously (). MCF-7 cells were transfected with Edg-3 sense or AOs and incubated with or without 10 nM E2 in phenol red-free media containing 1% charcoal-treated FBS for 5 d. Cell number was then quantified by the MTS assay as described previously (). The absorbance intensity of the MTS product is directly proportional to the number of viable cells in culture when cell number is between 2,000 and 200,000, otherwise the exponential dependence was determined. Total cell numbers were calculated based on calibration curves. Unpaired tests were used for comparison between two groups. For multiple comparisons, results were analyzed by analysis of variance, followed by the Dunnett's test. A value of P < 0.05 was considered statistically significant. Fig. S1 shows that E2 stimulated SphK1 phosphorylation in an ERK-dependent manner. Fig. S2 shows that inhibition of ERK activity by the ERK kinase-specific inhibitor U0126 resulted in a significant attenuation of EGFR phosphorylation in response to E2 stimulation. Online supplemental material is available at .

Epstein-Barr virus (EBV) is the causative agent of infectious mononucleosis and is associated with several human malignancies (). The most abundant of the few viral genes (4–11) expressed during EBV latency are the noncoding RNAs, EBV-encoded RNA 1 (EBER1) and EBER2, which are expressed at ∼5 × 10 per cell (). The EBERs, which are ∼170 nts in length, are transcribed by RNA polymerase III and assembled into nuclear ribonucleoprotein particles containing the La protein (). EBER1 also binds the ribosomal protein L22 and relocalizes a large fraction of the free cellular L22 to the nucleoplasm in EBV-positive cell lines (). The physiological function of EBERs has remained elusive. Although not necessary for EBV-mediated immortalization of B cells in vitro, EBERs promote cellular transformation in various systems (; ) and inhibit apoptosis that is induced by α interferon (; ). These activities have been attributed to the binding and inhibition of the double-stranded RNA–dependent protein kinase R (PKR; ; ; ), despite multiple studies that have found that EBERs are nucleoplasmic (; ), whereas PKR and its well documented effect on translation initiation are cytoplasmic (). Recent results (; ) indicate that EBERs do not inhibit PKR activity in vivo when cells are challenged with various PKR stimuli. The La protein is an abundant nuclear phosphoprotein that facilitates the correct folding and maturation of RNA polymerase III transcripts through its specific association with the short polyU sequence at their 3′ ends (). The human La protein has also been reported to play a role in the translational regulation of some messages (), including those that harbor unique terminal oligopyrimidine–rich motifs at their 5′ ends. Indeed, an unphosphorylated form of La has been detected that is specifically bound to terminal oligopyrimidine–containing mRNAs (). Previously, the idea that La actively shuttles between the nucleus and cytoplasm was supported only by observations of its localization in drug-treated cells (). We used heterokaryon and other assays to define the cellular trafficking of the EBERs and the La protein. We find that the EBERs are confined to the cell nucleus, whereas the endogenous La protein undergoes nucleocytoplasmic shuttling. As a control for the shuttling of small RNAs, we report that spliceosomal U1 small nuclear RNA (snRNA) does traffic to the other nucleus in human/mouse heterokaryons that are negative for EBER shuttling. We initially undertook heterokaryon-shuttling experiments () with the well characterized EBV-transformed suspension cell line, BJAB-B1. Because these cells did not adhere well to glass slides, we switched to the human HKB5cl8 cell line, which is a hybrid between human embryonic kidney 293S (HEK293S) cells and 2B8 cells, which are an EBV-positive Burkitt's lymphoma B-cell line (; ). HKB5cl8 cells not only attach to the glass slides but are morphologically superior in that the nucleus and cytoplasm can be readily distinguished. By RT-PCR analyses (unpublished data), HKB5cl8 cells establish type I latency () that is characteristic of Burkitt's lymphoma cells. We also performed Northern blot analyses and found that EBER1 and EBER2 are expressed in HKB5cl8 (, lane 1) at levels only two- to threefold lower than in BJAB-B1 cells (, lane 3). To test whether the endogenously expressed EBERs shuttle in and out of the nucleus, heterokaryons were formed by fusing human HKB5cl8 cells with mouse NIH3T3 cells (). The human cells had previously been transfected with plasmids expressing the shuttling heterogeneous nuclear ribonucleoprotein (hnRNP) A1-GFP protein (); heterokaryons were identified by the appearance of hnRNP A1-GFP in both the human and the mouse nuclei (, and ). Mouse nuclei were readily distinguished by punctate DAPI staining, which replicates the species-specific nuclear staining difference previously reported for Hoechst dye (). EBER1 and EBER2 were detected by in situ hybridization using DIG-labeled antisense DNA oligonucleotides. These probes were complementary to the 3′ half of the EBERs, but not to regions including conserved polymerase III promoter elements A and B (which may explain the unique report of cytoplasmic localization of EBERs []). As shown in (3–6), EBERs remained in the human nuclei and did not shuttle into the mouse nuclei during the 6-h incubation. HEK293 cells transiently expressing EBERs also did not exhibit shuttling (unpublished data); titration of the EBER-expressing plasmids showed that in situ hybridization signals would have been detected even with RNA levels <10% (as observed by Northern blotting; unpublished data). To ensure that the nucleocytoplasmic shuttling of RNA, as well as of protein molecules, could be observed in our assays, we examined U1 snRNA. We used a modified human U1 RNA, α2 U1 RNA, in which the first 20 nts are significantly different from either the human or mouse U1 snRNA (). This U1 RNA is functional in vivo () and, therefore, is expected to follow the wild-type maturation pathway, which involves export to the cytoplasm before assembly with Sm proteins and reimport into the nucleus (; ). For heterokaryon assays, we transfected an α2 U1 RNA–expressing plasmid into HKB5cl8 cells and visualized the RNA with probes that hybridize specifically to the modified region. We observed α2 U1 RNA in both the human and the mouse nuclei (, ), indicating that α2 U1 moves out of and back into the nuclei of somatic human cells. Importantly, in the same heterokaryons where U1 shuttling was observed, endogenous EBER1 was confined to the human nuclei (, ); the same result was obtained with a longer 12-h incubation (not depicted), as opposed to a 6-h incubation. In the RNA-shuttling assays, cycloheximide was omitted, ruling out the possibility that the lack of EBER1 shuttling is protein synthesis-dependent. EBER2 was also tested, but we were unable to find a hybridization temperature that would allow simultaneous detection of EBER2 and α2 U1 RNAs (unpublished data). The absence of EBER signals from mouse nuclei in heterokaryons could be attributable to the rapid cytoplasmic degradation of RNA once it is exported from the human nucleus. Therefore, we compared the turnover rates of EBER1 and other small RNAs; 7SL and Y1 RNAs are both cytoplasmic and transcribed (like EBERs) by RNA polymerase III, whereas U1 RNA is a nuclear RNA polymerase II product. After the addition of actinomycin D to HKB5cl8 or BJAB-B1 cells, EBER1 exhibited an apparent half-life of 25–30 h (), which is significantly greater than Y1 (apparent half-life of 7 h; ) and only slightly less than 7SL and U1 (). Because shuttling was observed for U1, but not for EBER1 (), and they are both extremely stable RNAs, rapid cytoplasmic degradation cannot explain the lack of EBER1 shuttling. To confirm nuclear retention in another system, we performed oocyte microinjection assays using in vitro–transcribed EBER1, U6, and tRNA. 2.5 h after injection, almost all of the positive nuclear export control, tRNA, was detected in the cytoplasmic fraction (, lanes 1, 4, and 5). In contrast, EBER1 remained in the nucleus, as did the negative export control, U6 RNA (, lanes 1, 4, and 5). To address whether La is responsible for the nuclear retention of EBER1, we repeated the microinjection assays using an EBER1 mutant lacking its 3′ polyU tail (required for stable La binding); the terminal nts were changed from UGUUUU to GAACAC. As expected, this EBER1 mutant exhibits eightfold reduced binding to La, based on immunoprecipitation using BJAB cell extracts (unpublished data). 2.5 h after injection, mutant EBER1 remained in the oocyte nucleus, whereas most tRNA was in the cytoplasm (, lanes 6, 9, and 10). Therefore, it is unlikely that La is responsible for the nuclear retention of EBER1. Finally, to probe why EBERs are not exported, we performed in vitro exportin 5 (Exp5)–binding assays. Exp5 mediates nuclear export of premicroRNAs and adenovirus noncoding RNA VAI by binding to a terminal stem (, ; ; ; ), which is also proposed to exist in EBER1 (). Using an electrophoretic mobility shift assay, we performed competition experiments to ask if EBER1 can displace the VARdm RNA () from recombinant Exp5. Although unlabeled VARdm efficiently competed with the Exp5-bound substrate (, lanes 2–4), neither EBER1 (, lanes 5–7) nor the negative control U6 RNA (, lanes 8–10) significantly displaced the probe, even at 200-fold excess. The same EBER1 preparation was active in binding its protein ligand L22 (). Thus, lack of binding to an export receptor may explain why EBER1 is not exported from the nucleus. Moreover, it is unlikely that EBERs function by interfering with host cell microRNA biogenesis, which is consistent with observations (unpublished data) that the level of let-7 microRNA is not altered by the presence of EBERs. Our strategy in investigating the cellular trafficking of EBERs included testing if its obligatory protein partner La undergoes nucleocytoplasmic shuttling. A typical EBV-infected cell harbors ∼5 × 10 copies of each EBER (), whereas most human cells express ∼2 × 10 molecules of La protein (). Thus, even though EBERs do not shuttle, the La protein could. To examine La protein shuttling, HKB5cl8 cells were transfected with plasmids expressing either the shuttling hnRNP A1-GFP or the nonshuttling hnRNP C1-GFP as controls. After fusion with mouse NIH3T3 cells for 4 h, endogenous human La protein was detected using a monoclonal anti-La antibody that does not cross react with mouse La protein (unpublished data; Wolin, S., personal communication), demonstrated by the lack of nuclear staining of unfused mouse cells, labeled m in (panels 7, 8, 10, and 11) and (panels 1, 2, 4, and 5). clearly shows that La shuttled from the human nucleus into the mouse nucleus (, panel 2), mimicking the shuttling of hnRNP A1-GFP in the same heterokaryon (, panel 3). Inclusion of cycloheximide during the fusion period ruled out the possibility that newly synthesized human La protein was imported into mouse nuclei. Although the nonshuttling hnRNP C1-GFP remained in the human nucleus (, panel 6), the human La protein moved into the mouse nucleus (, panel 5). We then confirmed that La nucleocytoplasmic shuttling is not cell-type specific by repeating the experiments with nonvirally infected human cells, HeLa or HEK293. Again, the nonshuttling hnRNP C1-GFP remained in the human nuclei and the human La protein shuttled into the mouse nucleus in both kinds of heterokaryons (, panels 8 and 9 and 11 and 12, respectively). We conclude that La, which is predominantly nuclear in multiple types of mammalian cells (), has the capacity to exit and return to the nucleus. Next, we asked whether La protein is exported via the Crm1 nuclear export receptor because a human La protein lacking its putative nuclear retention element had been reported to accumulate in the cytoplasm, but to be retained in the nucleus in the presence of the Crm1 inhibitor leptomycin B (LMB; ). To ensure that LMB inhibits Crm1 in heterokaryons of HEK293 cells and NIH3T3 cells, we included as a control PP32, which is a known shuttling protein whose nuclear export is Crm1-dependent (). We transfected HEK293 cells with a plasmid-expressing Flag-PP32 and, as expected, observed that both La and Flag-PP32 shuttled from the human to the mouse nucleus (, panels 2 and 3, respectively). In the presence of 30 ng/ml LMB, Flag-PP32, but not La, movement was inhibited (, panels 5, 6, 9, and 10). In this experiment, hnRNP A1-GFP, which does not require Crm1 for nuclear export (), was coexpressed to identify the hybrid cells (, panels 7 and 11). Because inhibition of Crm1 blocked the shuttling of Flag-PP32, but not of intact La protein, we conclude that the nuclear export of full-length La is either Crm1 independent or that La is exported by more than one pathway. Further studies are needed to resolve the pathways and whether the phosphorylation state of La regulates its shuttling activity (). Because EBERs do not exit the nucleus of either human cells () or oocytes (; even in the absence of a La binding site), it is not the La protein, but rather some other feature of their RNA structure, that retains the EBERs in the nucleus of EBV-infected cells. We tested the prediction, based on the presence of a terminal stem, that EBERs might bind and interfere with the activity of Exp5 (), which is limiting in the case of premicroRNA export (). Our findings suggest that EBERs do not function in this way, but instead participate in some other exclusively nuclear process that enhances the expression of several growth factors, including insulin-like growth factor I, interleukin-9, and interleukin-10 (; ; ) in EBV-transformed cells. Whether these consequences represent an active function of the EBER particles or arise through partial sequestration of La, ribosomal protein L22, or some other protein partner remains to be determined. 10 HKB5cl8 cells were transfected with 2 μg hnRNP A1-GFP, hnRNP C1-GFP (both gifts from G. Dreyfuss, University of Pennsylvania School of Medicine, Philadelphia, PA), or pα2U1 () plasmid using 6 μl TransIT reagent (Mirus) for ∼40 h on coverslips. 10 HeLa cells were transfected with 2 μg hnRNP C1-GFP plasmid using 6 μl Lipofectamine reagent (Invitrogen) for ∼40 h on coverslips. 10 HEK293 cells were transfected with 2 μg hnRNP C1-GFP plasmid, 2 μg Flag-PP32 plasmid, or 1 μg each of hnRNP A1-GFP and Flag-PP32 plasmids using 6 μl TransIT reagent (Invitrogen) for ∼40 h on coverslips. For heterokaryon assays, 10 mouse NIH3T3 cells were added to the transfected human cells described in the previous paragraph and allowed to seed on coverslips for 3 h. 100 μg/ml cycloheximide and 30 ng/ml LMB, as indicated in the figure legends, were added to the medium to block protein synthesis for 30 min, and the cells were fused using 50% PEG 3350/PBS for 2 min at RT. Cells were then washed in PBS three times and incubated in medium containing 100 μg/ml cycloheximide and 30 ng/ml LMB, as indicated in the figure legends, for 4–7 h to allow shuttling. The lack of signals in mouse nuclei for hnRNP C1-GFP and for Flag-PP32 when LMB was added indicates that cycloheximide effectively shut down translation. Cells were fixed in 4% formaldehyde/PBS and were processed for either in situ hybridization or immunofluorescence, as described in the following sections. Light microscopy and the appearance of shuttling proteins in the mouse nuclei were used to identify heterokaryons. Fluorescence images were photographed using a digital charge-coupled device camera (model C4742-95-12; Hamamatsu) through a microscope (Axioplan II; Carl Zeiss MicroImaging, Inc.) with a 40×, 1.3 NA, oil objective (Plan-Neofluar; Carl Zeiss MicroImaging, Inc.). Images were captured using Openlab imaging software (Improvision) and incorporated into figures using Photoshop CS and Illustrator CS software (both Adobe). Fixed cells were washed with PBS twice for 5 min, permeabilized with 0.5% Triton X-100/PBS on ice for 10 min, and washed with PBS once and 2× SSC twice at RT. Cells were prehybridized with Phil's hybridization solution at 37°C for 1 h and hybridized with 2 ng/ml EBER1R152 or EBER2R134 probe (complementary to EBER1 nts 130–152 or EBER2 nts 106–134; see Northern blot analysis section for sequences) in Phil's hybridization solution () overnight at 37°C. These probes were conjugated with DIG label using the 3′-DIG labeling kit (Roche) and were detected by incubation with a 1:200 dilution of rhodamine-conjugated anti-DIG antibody (Invitrogen) in PBS at RT for 1 h. Cells were washed three times with PBS at RT for 10 min each and once with 0.2 μg/ml DAPI/PBS solution at RT for 10 min, and then mounted for fluorescence microscopy. Alternatively, when EBER1 and α2 U1 were simultaneously probed at RT, the following oligonucleotides replaced the DIG-labeled probes and anti-DIG antibody: for EBER1, NEB1R148, 5′-XCTGGTACTTGACCGAAGACGGCAGAAA-3′; for α2 U1, NHA2U1A, 5′-XCTGCTTGTGTTAGATTATGTGGAT-3′; and for α2 U1, NHA2U1B, 5′-XCCCCTGCTTGTGTTAGATTATGTGGAT-3′. X denotes the 5′-amino group attached to a six-carbon linker. The 5′-amino group allowed conjugation of the Alexa Fluor 488 dye onto NEB1R148 and of the Alexa Fluor 594 dye onto NHA2U1A and NHA2U1B, using Alexa Fluor Oligonucleotide Amine labeling kits (Invitrogen). Fixed cells on coverslips were washed with PBS twice for 5 min, permeabilized with 0.4% Triton X-100/1% normal goat serum (Invitrogen) in PBS on ice for 10 min, and washed with 1% normal goat serum/PBS three times at RT for 10 min each. The cells were then incubated with primary antibodies in 1% normal goat serum/PBS at RT for 1 h. A mouse monoclonal anti–human La (a gift from M. Bachmann, Technical University Dresden, Dresden, Germany) and rabbit polyclonal anti-Flag (Sigma-Aldrich) antibodies were used at 1:100 dilutions. The coverslips were washed three times with 1% normal goat serum/PBS at RT for 10 min each, incubated with Alexa Fluor 594–conjugated (red) goat anti–mouse (for La) or Alexa Fluor 488– (green) or 680–conjugated (infrared) goat anti–rabbit (for Flag-PP32) antibodies for 1 h, washed three times with PBS at RT for 10 min each and once with 0.2 μg/ml DAPI/PBS solution at RT for 10 min, and mounted for fluorescence microscopy. Actively growing HKB5cl8 and BJAB-B1 cells at 4 × 10 cells/ml were treated with 10 μg/ml actinomycin D. At indicated time points, 2 × 10 cells were removed and pelleted by centrifugation. Total RNAs were analyzed by Northern blotting. Total cellular RNA was purified using Trizol reagent (Life Technologies) and 5 μg of RNA (), subjected to 7 M urea gel electrophoresis, transferred to Zeta-blot (Bio-Rad Laboratories), and cross-linked to the membrane by UV irradiation. The immobilized RNA was hybridized with the indicated probe, and the signal detected and quantified with a PhosphorImager (Molecular Dynamics). In , EBER levels were normalized to the signal obtained for cellular U6 snRNA; the probe was produced from plasmid pT7U6 () that was linearized with EcoRI and transcribed in the presence of α-[P]UTP. Other RNA sequences were detected by Northern blotting using the following γ-[P]–labeled DNA oligonucleotide probes: EBER1R152, 5′-CCAGCTGGTACTTGACCGAAGAC-3′; EBER2R134, 5′-ATTAGAGAATCCTGACTTGCAAATGCTCT-3′; U1R96, 5′-AATCGCAGGGGTCAGCACATCCGGAG-3′; HY1R60, 5′-GTTCGATCTGTAACTGACTGTGA-3′; and 7SLR99, 5′-GCATAGCGCACTACAGCCCAGAA-3′. The wild-type EBER1 coding sequence was cloned into the pUC19 vector (). Using this plasmid as template, EBER1 3′ polyU mutant was generated by PCR amplification with the primers ECORIT7, 5′-CGCGAATTCTAATACGACTCACTATAG-3′ and EBER1PML, 5′-GCCGGATCCCACGTGTTCTGCGGACCACCAGCTGGTACTTGA-3′. To generate tRNA plasmid, oligonucleotides PHE5P, 5′-CGCGAATTCTAATACGACTCACTCTAGGCGAAATAGCTCAGTTGGGAGAGCGTTAGACTGAAGATCTAAAGG-3′, which contains a T7 promoter, and PHE3P, 5′-GCCGGATCCCAGCTGGTGCCGAAACCCGGGATGGAACCAGGGAC- CTTTAGATCTTCAGTCTAACGCTCTCCC-3′ were hybridized and filled in with Klenow polymerase. The construct for producing the VARdm substrate () was generated by PCR using pAdEasy (American Type Culture Collection) as a template with primers TopVA, 5′-GACCGAATTCTCGGGACGCTCTGGCCGGTCAGG-3′, which contains a T7 promoter, and VA D4M, 5′-CGCGGATCCAGTACTAGGAGCACTCCCCCGTTGTCTGACGTCGCACACCTGGGTTATCACGGCGGACGGCCGGATACGG-3′. All DNA fragments were inserted into the pUC19 vector using the EcoRI and BamHI restriction sites. The EBER1 plasmid and pT7U6 were linearized with DraI, the mutant EBER1 plasmid with PmlI, and the tRNA plasmid with PvuII. RNAs were in vitro transcribed in the presence of α-[P]UTP (GE Healthcare), gel purified, and injected into the germinal vesicles of whole oocytes. 0.5–1 fmol each of U6 RNA, tRNA, and EBER1 were injected in a volume of 9.2 nl with 20 mg/ml blue dextran as a marker. Oocytes were incubated at RT in OR2 buffer (5 mM Hepes, pH 7.8 with KOH, 82.5 mM NaCl, 2.5 mM KCl, 1 mM NaHPO, 1 mM MgCl, and 1 mM CaCl) for 0.5 or 2.5 h and manually dissected in cold isolation buffer (80 mM KCl, 17 mM NaCl, 6 mM NaHPO, 3.5 mM KHPO, and 10 mM MgCl). Five to six oocytes were fractionated into nucleus and cytoplasm () and pooled for each time point. Proteinase K digestion and phenol/choloform extraction were performed, and 0.5 oocyte equivalents were run on an 8% urea polyacrylamide gel. The La-specific antiserum used was ON, which was provided by J. Hardin (Yale University, New Haven, CT). Whole-cell sonicates (16 μl) from 2 × 10 BJAB cells () were incubated with 40 fmol (∼2 × 10 cpm) in vitro α-[P]UTP–labeled wild-type or mutant EBER1 and 2 μg tRNA as a nonspecific competitor in 20 μl for 30 min at RT. The reactions were immunoprecipitated with either anti-La or anti-L22 () attached to protein A–Sepharose beads (GE Healthcare), or beads alone at 4°C for 1–2 h. The beads were then washed five times with NET-2 (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.05% NP-40) at 4°C. RNA was extracted with phenol/chloroform/isoamyl alcohol (25:25:1), ethanol-precipitated, and electrophoresed in an 8% urea polyacrylamide gel. Equal amounts of wild-type and mutant EBER1 were immunoprecipitated in control reactions with anti-L22, suggesting proper folding of the EBER1 mutant. Expression and purification of Exp5 was performed as previously described (), except that the purified protein was dialyzed against buffer A (20 mM TrisHCl, pH 7.5, 100 mM KCl, 2 mM MgCl, 2 mM DTT, and 10% glycerol). RanQ69LGTP was prepared as previously described (plasmid provided by I. Macara, University of Virginia, Charlottesville, VA; ; ). The VARdm substrate () was generated from the VARdm plasmid linearized with ScaI by in vitro transcription in the presence of α-[P]UTP. Binding reactions (10 μl) containing 4.5 fmol VARdm RNA, 0.1 μM Exp5, 0.5 μM RanQ69LGTP, and the indicated amounts of competitor RNAs were incubated for 40 min at 30°C in RNA-binding buffer containing 20 mM TrisHCl, pH 7.5, 100 mM KCl, 2 mM MgCl, 2 mM DTT, 10% glycerol, and 2 pmol of the T7 terminator DNA oligonucleotide 5′-GCTAGTTATTGCTCAGCGG-3′ to reduce nonspecific binding. Before loading, 1 μl of a 0.6 mg/ml heparin and 0.2 mg/ml Bromophenol blue mixture was added to each sample. The samples were loaded on a preelectrophoresed (30 min) 6% native gel in 0.5× TBE buffer (45 mM Tris borate and 1 mM EDTA). Electrophoresis was performed at 12 V cm for 1 h at RT.

Many cells adapt to changes in extracellular nutrient levels by altering the expression of genes encoding transporters for these molecules. This is accomplished through signal transduction pathways that respond to the binding of extracellular nutrients to plasma membrane–sensor proteins. In some cases (for review see ), including plants and animals (for reviews see ; ), the sensors share significant amino acid sequence similarity with transporters of the nutrient in question. The plasma membrane of contains transporter-like sensors for amino acids (Ssy1p) and glucose (Snf3p and Rgt2p), but these proteins apparently do not transport their respective nutrients (; ; ; ). Sensing of extracellular amino acids through Ssy1p leads to transcriptional induction of genes encoding amino acid transporters (, ; ; ; ) and some other genes through activation of the transcription factors Stp1p and Stp2p (for review see ). How nutrient sensors convert changes in extracellular nutrient concentrations into appropriate signals is key to the understanding of nutrient homeostasis. Kinetic and biochemical studies (for reviews see ; ) and recent structural data (; ) indicate that carrier-type transporters have a single substrate binding site (for symporters, a single site for each ligand molecule transported in a cycle) that is exposed to either side of the membrane, depending on the conformational state of the transporter. Observations that mutations in can increase the apparent affinity for sensing of amino acids and concomitantly confer an increased basal level of signaling (; ) suggests that these mutations alter an equilibrium between a signaling and a nonsignaling conformation of Ssy1p in the absence of ligand. In fact, ligand-independent occurrence of a signaling conformation has previously been found among the very different 7TM-type receptors (). We propose, then, that extracellular amino acids are sensed because of their ability to bind to and stabilize a signaling conformation. Given the structural similarity of Ssy1p to transporters, existence of states such as I (inward facing), O (outward facing), and O·L (outward facing, ligand bound; ), interconverting through reactions 1 and 2, would provide a formal model for the initial step of sensing and would, at the same time, interpret hyperresponsive and constitutive mutants in a simple way, namely, as being affected in the equilibrium constant for reaction 1. The additional idea that a cytoplasmic ligand might bind to state I (i.e., existence of reaction 3 and state I·L [inward facing, ligand bound]) represents an extended model for transporter-like sensors, which has the salient feature that ligand binding inhibits the conformational shift (i.e., reaction 4 is not efficient). Reaction barriers for conformational changes are actually common in transporters. For example, in the case of the anion carrier in erythrocytes, reaction 1 is at least 10,000-fold less efficient than reaction 4 (; ), a fact that explains the strong antiport function of the carrier for chloride and bicarbonate. We decided to test our model by investigating whether the signaling potency of extracellular ligand is influenced by the cytoplasmic ligand concentration. To study the effect of intracellular amino acids on signaling by Ssy1p, we chose to manipulate the cytoplasmic concentration of leucine, as extracellular leucine is the most potent known elicitor of signaling through Ssy1p (; ). To increase the cytoplasmic level of leucine, cells were grown in minimal medium with increasing levels of extracellular leucine. This resulted in an increase in the cytoplasmic concentration of leucine from 0.5 μmol/g dry weight (DW) without addition of leucine to 26 μmol/g DW for cells grown in medium with 1 mM leucine (, experiments 1 and 2), allowing us to measure the effect of intracellular leucine levels on Ssy1p signaling. After removal of leucine from the growth medium by washing the cells, signaling was induced by addition of extracellular amino acid at various concentrations and measured as cleavage of the transcription factor Stp1p. The dose–response relationship () allowed determination of the median effective concentration ( , apparent dissociation constant). that was four times higher than that of cells grown without leucine ( and , experiments 1 and 2). The value of 13 μM for cells grown without leucine agreed with data previously determined (,). Thus, increasing the cytoplasmic leucine concentration led to an increase in the apparent dissociation constant in the sensing of extracellular leucine. for the sensing of phenylalanine (, experiments 10 and 11). We tested in several ways the possibility that changes in the apparent affinity of extracellular ligand might be a consequence of the history of signaling, rather than a direct consequence of the cytoplasmic leucine concentration. In one experiment, cells were grown in media containing 50 or 125 μM leucine, amounts that are sufficient to almost fully induce signaling but which did not strongly increase the cytoplasmic leucine concentration. close to that obtained with cells grown without leucine (, compare experiments 3 and 4 with 1), suggesting that the signaling history was irrelevant. Next, we tested a leucine uptake-defective strain (M5568), which lacked the broad-spectrum amino acid transporter Agp1p () and the leucine transporters Bap2p, Tat1p, and Bap3p (). In the control experiment, i.e., after growth in medium without leucine, this strain exhibited a normal for leucine (, compare experiment 5 with 1 and 7). of leucine only increased to 27 μM (, experiment 6), as compared with 53 μM for the uptake-proficient strain (, experiment 8). follows intracellular leucine rather than history of exposure, and it can occur independently of whether the cells were treated with leucine for many hours or 2 h. In a third experiment, the cytoplasmic leucine concentration was perturbed in a manner that did not involve feeding of leucine from the outside. Organisms that synthesize leucine do so by a series of four reactions, using 2-oxoisovalerate as a precursor. In , encodes the major isoform of the enzyme catalyzing the first step (), which is subject to feedback inhibition by leucine (). The mutation () confers reduced or eliminated sensitivity to leucine, leading to an increased intracellular concentration of leucine. Introduction of the mutation caused an increase of the cytoplasmic leucine pool from 0.5 to 12 μmol/g DW (, compare experiment 9 with 1). toward extracellular leucine increased from 13 to 30 μM. Thus, this independent method of increasing cytoplasmic leucine produced the same effect on signaling as growth in medium with a high leucine concentration. by dissolving into the lipid bilayer of the plasma membrane and changing its characteristics. However, there was no significant effect of adding isoamyl alcohol at relevant concentrations (unpublished data). We have considered equilibrium equations for reactions 3 (K = [I][L]/[I·L]), 1 (K = [O]/[I]), and 2 {K = [O·L]/([L][O])} in , where [O], [O·L], [I·L], and [I] are the concentrations (amounts) of the four depicted forms of the sensor. The cytoplasm may contain several compounds that can appreciably bind to I but, for a start, we consider a single compound, L, which may or may not be identical to the offered extracellular ligand, L. The concentrations of the ligand-free forms of the sensor, [I] and [O], can be eliminated from the three equations to yield a single equation:where [I·L] and [O·L] are the concentrations of the ligand-containing forms of the sensor and K = K·K·K (). If ligand bound forms of both conformations are strongly predominant, [I·L] will be the concentration of nonsignaling Ssy1p and [O·L] will be that of signaling Ssy1p. Then, the ratio of signaling to nonsignaling Ssy1p is proportional to the concentration ratio across the plasma membrane; i.e., the sensor output is determined by the ligand concentration ratio. , of an extracellular ligand to Ssy1p in a dose–response analysis. will be approximately proportional to [L]. becomes approximately proportional to Σ [L]/K, a weighted sum of cytoplasmic ligand concentrations, where each ligand is referred to by an index integer, , and K is the equilibrium constant of reaction 3 () characteristic for the cytoplasmic ligand in question. of an extracellular ligand will vary linearly with [Leu]. Indeed, our data were consistent with linearity (). Best fit (R = 0.80) to = [Leu] + yields = 1.57 μM/(μmol/g DW). An analogous relation can be made in which the abscissa is the cytoplasmic leucine concentration, rather than amount per DW. With an approximate value of 2 ml/g DW for the specific cytoplasmic volume (), this relation has a slope of 0.003 (analogous to , but dimensionless). At sufficient concentrations of cytoplasmic and extracellular leucine, this relationship means that the leucine concentration ratio across the plasma membrane is sensed, rather than the absolute extracellular leucine concentration. This reflects a sensing principle that would not be straightforward to obtain with nontransporter-like sensors. We interpret the intercept with the ordinate in to reflect binding to the inward-facing form of Ssy1p of cytoplasmic ligands other than leucine, presumably other -α-amino acids. of the highest point in ), considering the relatively high amounts of some cytoplasmic amino acids measured; e.g., we see rather constant levels of ∼120 μmol/g DW (i.e., ∼60 mM) of glutamate, which is fourfold higher than the leucine concentration at the highest point in . However, this is consistent with the possibility that the relative affinities of the various amino acids for binding to outward-facing Ssy1p () are fully or partially conserved when inward-facing Ssy1p is considered, i.e., that leucine is the strongest binding amino acid, also from the inside. We present a model in which binding of amino acids to transporter-like sensors from outside or inside stabilize signaling and nonsignaling conformations, respectively. We tested the model by looking for an influence of the concentration of a cytoplasmic amino acid on Ssy1p-mediated sensing. , of signaling to correlate linearly with the concentration of cytoplasmic leucine. It will be of interest to determine whether our model can account for sensing by other transporter-like sensors, including the glucose sensors Snf3p and Rgt2p (). The model makes obvious sense in terms of intracellular nutrient homeostasis. It can also account for the function of sensors that can transport their respective solutes, such as the general amino acid permease (Gap1p; ) if the rate constants for reaction 4 are different from those for reaction 1. Yeast strains were derived from strain M4054, originating () from S288C via X2180-1A by a spontaneous, low-reverting mutation and a deletion of most of . Genetic fusion of the ZZ tag to Stp1p was as described previously (), and deletions were introduced () by standard techniques. Strain M5446 was made by integrating a PCR fragment with from strain XK14-15D, provided by G.B. Kohlhaw (Purdue University, West Lafayette, IN), into strain M4054, followed by introduction of the ZZ construct. Yeast cells were grown aerobically batch-wise in shake flasks overnight to a turbidity (OD) of 0.25–0.6, corresponding to 3–6 × 10 cells/ml, in glucose- and ammonium-based minimal medium supplemented with uracil () and buffered with 85 mM succinic acid + 150 mM NaOH (SD medium). As indicated, some cultures were inoculated in medium with additional leucine, and others contained additional leucine during the last 2 h of cultivation. To determine the cytosolic pool of amino acids, 200 ml of culture with a known DW of cells were mixed with crushed ice (100 g), harvested by centrifugation, and washed twice with ice-cold water. For specific release of cytosolic amino acids, the plasma membrane was selectively permeabilized with the cupric ion method as described previously (), using the following specific protocol: cells were resuspended in 1.5 ml of permeabilization buffer (5 mM MES, pH 6.0, and 0.4 mM CuCl), incubated at 30°C for 10 min, and centrifuged. The supernatant was pooled with the supernatant obtained after a step of washing (0.75 ml 5 mM MES buffer, pH 6.0) and dried in a SpeedVac. For some experiments, the aforementioned volumes were scaled a few times up or down. The sample was dissolved in 100 μl of water, acidified with 10 μl of 3% sulfosalicylic acid, and centrifuged at 20,000 for 20 min at 4°C. The supernatant was neutralized with 10 μl of 1 M NaOH. After evaporation, the volume was adjusted, and part of the sample was applied to the amino acid analyzer (Biochrom 20; GE Healthcare), using ninhydrin for detection. The quantification of signaling by Ssy1p is based on the findings that the transcription factor Stp1p is proteolytically activated by removal of a 10-kD NH-terminal inhibitory part and that it has such a short half-life that monitoring of signaling is possible irrespective of the physiological history of the cells (). Stp1p was expressed as a fusion protein (Stp1-ZZ) containing a bacterial IgG binding domain, allowing monitoring of proteolytic processing by Western blotting as described previously (,). Leucine or phenylalanine at appropriate concentrations was added to aliquots of the culture to induce signaling. After 10 min, proteins were extracted and separated by electrophoresis, and Stp1-ZZ in processed and unprocessed form was quantified. The median effective concentration ( , apparent dissociation constant) was determined by measurement of Stp1p processing ( = /[ + ], where is the amount of processed form and is the amount of unprocessed form) at 0.001–1,000 μM of leucine or 0.01–10,000 μM of phenylalanine. cannot be negative and + cannot be >1.

Melanocytes (MCs) are pigment cells responsible for the pigmentation of animals. MCs synthesize pigment granules within a special organelle termed a melanosome, where several enzymes that are involved in melanin biosynthesis, such as tyrosinase (Tyr) and dopachrome tautomerase (Dct), are assembled and thereby transfer pigment granules into keratinocytes to form pigmented skin or hairs. MCs provide an attractive system to study the molecular basis of cell regulation, as genetic alterations involved in MC regulations are easily identifiable as coat-color mutants. In fact, >120 different loci have been identified as coat color mutants (). Of these loci, mutations in genes that are particularly critical for MC development, such as , , and , are characterized by severe hair-pigmentation defects. In the hairy region of the skin, MCs are exclusively localized in hair follicles (HFs), where proliferation and maturation of MCs are strictly regulated by the hair cycle. MCs at the hair matrix (HM) appear and proliferate during the growth phase of the HF and thereafter differentiate into mature MCs to generate pigmented hairs. Subsequently, all differentiated MCs are eliminated by apoptosis during the regression phase of the HF. It has been demonstrated that this cyclic appearance of MCs in the HM is maintained by a small pool of undifferentiated MC stem cells (MSCs) localized at the lower permanent portion (LPP) of the HF (). During embryogenesis, MC precursors, melanoblasts (Mbs), arise from the neural crest and migrate through the epidermis toward newly developing HFs. Once in the follicles, they are segregated into two populations: MCs, which localize at the HM and contribute to the initial wave of melanogenesis during HF morphogenesis, and MSCs, which colonize at the LPP and constitute the MC system in subsequent hair cycles (). In nonhairy regions, Mbs stay immature and remain on the basement membrane of the epidermis, where they undergo differentiation into mature MCs upon stimulation from keratinocytes. It has been proposed that the homeostatic regulation of Mbs is maintained by the keratinocytes through cell–cell interactions (), although the precise molecular interactions are largely unknown. Notch comprises a family of highly conserved receptors, whose activation is induced by their specific ligands, Delta and Jagged, through cell–cell interactions (). Once activated, the Notch intracellular domain (NICD) is cleaved by γ-secretase, which leads to translocation of the NICD into the nucleus. Subsequently, NICD is associated with the transcription factor RBP-J to generate the transactivation complex, which initiates transcription of target genes such as the () transcriptional repressors (). Notch signaling is involved in various aspects of cellular regulation, including stem cell maintenance (; ); however, the exact molecular mechanisms remain unclear. In this study, we demonstrate the role of Notch signaling in the maintenance of Mbs and MSCs. In addition, we demonstrate the critical role of Hes1 in ensuring the survival of Mbs by preventing apoptosis. Thus, our data provide new insights to the molecular mechanisms underlying the homeostatic regulation of Mbs. During gene expression profiling of embryonic Mbs, we noted that several Notch-related genes were expressed in the Mb population (unpublished data). To examine the implication of Notch signaling in Mbs, we performed double immunostaining of embryonic skin using specific antibodies against activated Notch1 (NICD1) and Dct, which can mark all the subsets of MC lineage. Consistent with previous observations (; ), nuclear localization of NICD1 was observed in the suprabasal and basal layers of the epidermis that include Dct Mbs (). In parallel with Notch1 activation, the Notch ligand is expressed in the basal layer of the epidermis (). To identify Notch target genes in Mbs, we performed transcript analysis using highly enriched Mbs (Fig. S1, available at ). The initial RT-PCR screening detected , , and transcripts in Mbs (). Further quantitative analysis shows that, of these genes, expression is the most prominent, whereas expression of and is much less intensive (). These data suggest that, among family genes, represents a predominant target of Notch signaling in Mbs. To further confirm transcription in Mbs, we used () mice, in which GFP expression mimics endogenous promoter activation (). Parallel to Notch1 activation, the promoter is also activated in Dct Mbs (). Hence, these data indicate that both Notch1 and Hes1 are activated in Mbs, implying that Notch signaling plays a role in the regulation of Mbs. To assess the role of Notch signaling in Mbs, we conditionally ablated the gene in an MC lineage. For this purpose, we crossed mice carrying floxed alleles (; ) with transgenic (Tg) mice expressing under the control of an MC-specific promoter (; ), which is stringently activated in Mbs from embryonic day (E) 11.5 onward. By crossing mice with the Cre-reporter strain (), we also confirmed specific Cre-mediated recombination in Mbs in a postnatal day (P) 1 mouse (Fig. S2, available at ). mice showed severe coat-color dilution in the initial hairs, followed by loss of hair pigmentation after the first hair molting (). Detailed analysis of the hairs of mice revealed that the first hairs developed from the mice were a mixture of pigmented and unpigmented, whereas the hairs grown in the subsequent hair cycle were almost all unpigmented (). In addition, a histological analysis of skin showed a dramatic reduction of MCs in the HFs at P4 () and the virtually complete absence of MCs in the HFs at P32 (see ). These data indicate that melanogenesis is severely impaired in mice, demonstrating the indispensable role of Notch signaling in the MC system. Given that the initial wave of melanogenesis during HF morphogenesis is directly derived from embryonic Mbs (), we reasoned that the severe reduction of MCs in mice is due to a defect in the maintenance of embryonic Mbs in which Notch signaling is shown to be activated. To examine the effect of disruption in Mbs, we performed whole-mount immunostaining of embryonic skin using an antibody against c-Kit (ACK4) that allows identification of embryonic Mbs (). As compared with the control, numbers of Mbs were apparently decreased at E16.5 and P0 in mice (). Hence, these data indicate that Notch signaling is critical for the maintenance of Mbs. As an alternative avenue for analyzing the function of Notch signaling in Mbs, we adopted an embryonic skin organ culture system by blocking Notch activation with a pharmacological inhibitor for γ-secretase (). Skin fragments obtained from E13.5 embryos were cultured for 4 d in the presence of the γ-secretase inhibitor DAPT ([3,5-difluorophenylacetyl]--alanyl--2-phenylglycine t-butyl ester) and then grafted onto immunodeficient (nude) mice to examine the effect on melanogenesis. In contrast to the control skin, which emerged with normal black hair, the hair grown from the DAPT-treated skin was unpigmented from the initial hairs and subsequently through several hair cycles (). In fact, histological analysis of DAPT-treated skin demonstrated the virtually complete loss of Mbs in the epidermis (), indicating that all the subsets of Mbs that contribute to the postnatal melanogenesis were ablated by DAPT treatment. Shortly after the DAPT treatment (24 h), a proportion of Mbs were positive for TUNEL staining and activated caspase 3 (), demonstrating the induction of apoptosis in these Mbs. Thus, it is evident that Mbs are eliminated from DAPT-treated skin by apoptosis. By combining these results with the phenotype of mice, our observations suggest a direct role for Notch signaling in the promotion of Mb survival by inhibiting the initiation of apoptosis; however, we cannot rule out the possibility that DAPT treatment indirectly affects the survival of Mbs. Immunostaining of DAPT-treated skin with NICD1 indicated a down-regulation of Notch signaling (unpublished data). In addition, quantitative PCR (Q-PCR) demonstrated a dramatic reduction of transcript in DAPT-treated Mbs (). These data indicate that DAPT treatment abrogates Notch signaling through in Mbs. To test whether the specific expression of in Mbs would be sufficient to rescue the loss of Mbs after DAPT treatment, we generated Tg mice in which the gene was expressed under the control of an MC-specific promoter (; ; ). In the Tg mice, the majority of Mbs remained in the epidermis even after DAPT treatment (). In addition, hairs grown from the DAPT-treated skin showed hair pigmentation (), demonstrating that Mbs, which contribute to postnatal melanogenesis, are restored in skin. These observations provide strong evidence for the direct involvement of in the promotion of Mb survival at the downstream of Notch signaling. Because functions as a transcriptional repressor (), the data suggest that represses the initiation of apoptosis by preventing the gene expression required for apoptosis in Mbs. Immunohistochemical staining showed that Mbs at the LPP were positive for NICD1 (). Further, analysis of follicles also indicated that transcription was activated in the Mb population at the LPP, whereas it was inactive in the differentiated MCs at the HM (). These observations indicate that Notch signaling through is also activated in MSCs, supporting the notion that Notch signaling may function in MSCs. To demonstrate this idea, we examined the distribution of immature Mbs in HFs during postnatal hair cycles in mice. Besides an apparent reduction of MCs at the HM (), some Dct-positive Mbs are identifiable at the LPP of HFs in mice at P4 (). However, at P12, loss of these Mbs at the LPP was evident (), whereas Mbs were retained at the LPP in control mice (). Looking at the next hair cycle at P32, HFs lack both immature Mbs at the LPP and differentiated MCs at the HM (), indicating the elimination of the MSCs that reconstitute the MC system in subsequent hair cycles after the initial HF morphogenesis. These data, along with the extensive graying phenotype, demonstrate a defect in the maintenance of MSCs in mice during the early stage of HF morphogenesis. Thus, it is shown that Notch signaling is also critical for the maintenance of MSCs. We demonstrate that Notch signaling, acting through a Hes1 transcription factor, plays a predominant role in the maintenance of Mbs, including MSCs. Given expression in the basal layer of the embryonic epidermis and the outer root sheath of the developing HFs (), Notch signaling in Mbs is likely to be activated through interaction with the surrounding keratinocytes. Thus, our data suggest that Notch signaling represents a fundamental component of the homeostatic regulation of Mbs that may be mediated by Mb–keratinocyte interactions. Taking account of the intensive migration capacity of Mbs, one possible explanation for the physiological role of Notch signaling in the regulation of Mbs is that, in collaboration with c-Kit signaling (), it may play a role in confining Mbs to the epidermis by allowing their survival and proliferation only under the control of epidermal keratinocytes. Notch signaling is implicated in the maintenance of the stem/progenitor cells of a variety of stem cell systems (). In addition, recent reports also have evidenced that Notch signaling regulates stem cell fate in collaboration with other signaling pathways, such as Wnt and BMP (; ). In such a situation, it is important to define the exact molecular linkages at the downstream of Notch signaling under physiological conditions. Taking advantage of the MC system, where genetic alterations affecting stem/progenitor cell maintenance can be easily identified by coat-color defects, MCs would provide an avenue for elucidating the exact molecular networks involved in stem cell regulation by Notch signaling, as we have shown here that Hes1 acts at the downstream of Notch in Mbs. C57BL/6 mice and nude mice (BALB/C Slc-) were purchased from Japan SLC, Inc. , mice were bred in our animal facility. All animal experiments were performed in accordance with the guidelines of the RIKEN Center for Developmental Biology for animal and recombinant DNA experiments. Tg mice were generated by injecting the Tg constructs that allow to be expressed under the control of the promoter (provided by I. Jackson, Western General Hospital, Edinburgh, UK; ) into fertilized eggs. Our organ culture system was adapted from . Skin specimens were prepared from the dorsal coat of E13.5 embryos derived from wild-type or Tg mice and cultured in the presence or absence of 1 μM of the γ-secretase inhibitor DAPT (Peptide Institute, Inc.). After 4 d of skin organ culture, skin fragments were transplanted onto nude mice to assess the effect of DAPT treatment on melanogenesis. Total RNA was harvested from the FACS-isolated Mbs () by using an RNeasy mini kit (QIAGEN). Reverse transcription was performed with oligo dT primer using SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's protocol. To examine expression of family genes in the isolated Mbs, RT-PCR analysis was performed with 40 cycles of amplification. To confirm specific transgene expression in Tg mice, a specific primer pair was designed at the 3′ region of the promoter and the 5′ region of cDNA, and PCR was performed with 35 cycles of amplification. Q-PCR was performed using a QuantiTect SYBR Green PCR kit (QIAGEN) according to the manufacturer's protocol. To present relative expression values to that of , the absolute expression value of each gene was calculated from each external calibration curve generated by serially diluted control vectors containing each gene. Then, each absolute expression value was divided with that of . PCR primers used in this study are shown in Table S1 (available at ). Immunohistochemistry using cryosections was performed as described previously (). The following antibodies were used as primary antibodies: cleaved Notch1 (Cell Signaling), GFP (Invitrogen), cleaved caspase 3 (Cell Signaling), Pax3 (provided by G. Grosveld, St. Jude Children's Research Hospital, Memphis, TN; ), Keratin 5 (Covance), and Dct (Santa Cruz Biotechnology, Inc.). Staining was performed using specific secondary antibodies conjugated to Alexa 488 or 546. TO-PRO3 iodide (Invitrogen) was used for nuclear staining. For whole-mount immunostaining, epidermis sheets were peeled off the dorsal skin of E16.5 and P0 mice. Immunostaining using ACK4 was then performed as described previously (). In situ hybridization was performed as described previously (). TUNEL staining was performed using an ApoAlert DNA fragmentation assay kit according to the manufacturer's protocol (BD Biosciences). Gross images were acquired using a digital camera (model C-5050; Olympus). Light microscopy images were taken using a microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) with a 20×/0.6 plan-APOCHROMAT objective by a cooled charge-coupled device camera (AxioCam HRc; Carl Zeiss MicroImaging, Inc.) controlled by AxioVision 3.0 software (Carl Zeiss MicroImaging, Inc.). Confocal imaging was performed using a microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.) with 40×/1.3 plan-NEOFLUAR and 63×/1.4 plan-APOCHROMAT objectives equipped with a confocal microscope system (Radiance 2100; Bio-Rad Laboratories) controlled by LaserSharp 2000 software (Bio-Rad Laboratories). Images were processed using Photoshop 7.0 (Adobe). Fig. S1 shows flow cytometric isolation of Mbs from the E16.5 embryo epidermis. Fig. S2 shows the MC lineage–specific Cre-mediated recombination in mice. Table S1 shows PCR primers used in this study. Online supplemental material is available at

Keratins are polypeptide components of the epithelial intermediate filament (IF) cytoskeleton that contribute to tissue homeostasis as important protective factors to combat various types of stress (; ). The intrinsically rigid IF system must adapt to changing structural requirements in cells undergoing reshaping during development, cell division, migration, or metastasis. In contrast to microtubules and actin filaments, however, comparatively little is known about the molecular mechanisms governing the dynamic properties of IFs. Their assembly from nonpolar tetrameric building blocks and their propensity to spontaneously assemble without additional cofactors in vitro () suggest unique organizational principles. Molecular details of the in vivo assembly and turnover are understood even less, although it is generally agreed that phosphorylation is important (). Imaging of living cells producing fluorescent keratins has provided evidence for the importance of phosphorylation () and has led to novel concepts about the spatial and temporal specifications of keratin filament (KF) dynamics. Thus, it was shown that KF network formation and turnover are dictated by the cell periphery with nascent KF precursors (KFPs) appearing in close proximity to the actin-rich cell cortex, followed by KFP elongation and their transport toward the cell interior before network assembly (; ). This process is independent of cellular differentiation and may also be relevant to other IFs (; ). The dynamic properties of KFs and other IF systems are further determined by their association with the microtubule and actin filament systems (for reviews see ; ). This linkage becomes particularly apparent in situations when either of the other systems is destroyed, thereby uncovering two types of KFP motility: a comparatively slow, continuous, and inward-directed actin-dependent movement and a fast, intermittent, and bidirectional motility that is dependent on an intact microtubule system (). Considering the predominant formation of KFPs in proximity to the actin-rich cortex and the transport of KFPs in an actin-dependent fashion, we decided to examine the possibility that KFP formation is linked to the peripheral actin system and its associated adhesion structures. To examine the interrelationship between the actin system and the formation and transport of KFPs, we focused on lamellipodia, because they are rich in cortical actin, actin stress fibers, and actin-anchoring focal adhesions (FAs). Immortalized mammary epithelium–derived EpH4 cells were used, given their tendency to form large lamellipodia (). Cells were doubly transfected with cDNAs coding for human keratin 18 (HK18)–YFP and actin–red fluorescent protein (RFP) to simultaneously monitor the keratin and actin system. Fluorescence micrographs were recorded together with phase-contrast pictures of spontaneously forming lamellipodia ( and Video 1, available at ). Typically, emerging lamellipodia contained abundant actin, but, at least initially, no particulate keratin fluorescence. The first KFPs appeared only a few minutes after expansion (, 7.5 min, arrows). Subsequently, more KFPs formed, which were transported inward before fusion with each other and establishment of a new network extension in the protrusion (, 35 min). We conclude that lamellipodia are regions that induce the formation of novel KF networks and, hence, contain the regulatory mechanisms and signals favoring this process. To characterize the nature of the continuous inward transport of KFPs, we examined their mobility in relation to actin stress fibers, which are often very prominent in lamellipodia. Double-fluorescence recording of HK18-YFP and actin-RFP revealed a remarkable codistribution (). Most KFPs were seen to migrate along actin fibers (, arrows) in a continuous, inward-directed, and comparatively slow (∼500 nm/min) fashion (Video 2, available at ), which is characteristic for actin-dependent transport (; ). Overlays of all frames of the keratin and actin fluorescence recordings () resulted in both parallel and partially overlapping tracks, as would be expected for linked transport phenomena. In rare instances, however, individual KFPs that were away from stress fibers exhibited a discontinuous, bidirectional, and rapid motility (, arrow), which are the signets of microtubule-dependent transport (). Cells were treated with the actin polymerization inhibitor latrunculin B to further assess the importance of actin filaments for KFP dynamics. KFP transport ceased immediately, whereas KFP formation, elongation, and fusion continued ( and Video 3). We conclude that the great majority of lamellipodial KFPs are transported along actin fibers. Comparable transport modes are probably also relevant to other IFs because cortical actin–dependent transport of newly synthesized neurofilament subunits occurs in axonal shafts (; ) and growth cones (). Recruitment of newly formed KFPs usually occurred at the most peripheral tips of actin stress fibers ( and Video 2), which is where they are anchored to the plasma membrane via FAs that also attach cells to the extracellular matrix (; ; ). In addition, FAs act as important signaling platforms that also affect the microtubule system (). To explore whether FAs are also determinants of KFP formation, EpH4 cells were doubly labeled with HK18-YFP and FA components such as RFP-zyxin and paxillin-DsRed2. Overall, abundant KFP formation was noted in lamellipodia containing abundant FAs, but was low in the areas between lamellipodia (, and Video 4, available at ). KFPs appeared remarkably close to FAs. Thus, 87% ( = 31) of the KFPs that formed during the 50-min recording of Video 4 were first detected within six pixels (corresponding to 0.66 μm) of a zyxin-labeled FA ( and Fig. S1). The high magnification images in provide an example for the tight spatial configuration of emerging KFPs and FAs. Similar images could be recorded using the early FA marker paxillin ( and Video 5). In this instance, 35 emerging KFPs were identified, 75% of which were first seen less than seven pixels away from a FA (Fig. S1). It should be kept in mind that we may not be able to detect the initially forming KFPs, but we may be able to see enlarging and, hence, increasingly fluorescent KFPs only after release from their presumptive nucleation sites. Also, cytoskeletal linker molecules are capable of bridging considerable distances of several hundred nanometers and could therefore be responsible for the detection of associated molecules outside of their anchoring platforms. To elucidate the temporal relationship between FA formation and KFP assembly, the appearance of paxillin- and keratin-positive structures was examined in emanating lamellipodia. Dual-color time-lapse images revealed that paxillin-containing FAs were established in expanding lamellipodia before KFPs appeared in close proximity ( and Video 5). To find out whether FA-associated KFP formation is determined by cell-specific factors, analyses were extended to nonepithelial SW13 cells lacking cytoplasmic IFs. These cells form extended cytoplasmic KF networks when transfected with fluorescent protein–tagged HK8 and HK18 in stable cell clone SK8/18-2 (; ). Because the entire keratin system is solely composed of fluorescent polypeptides, SK8/18-2 cells are particularly suited for detection of very small keratin particles. In perfect agreement with the results obtained in epithelial cells, multicolor time-lapse imaging of HK18-YFP and RFP-zyxin revealed a tight correlation between KFP formation and FAs ( and Video 6, available at ). Visual assignment of all KFP initiation events ( = 83) onto the zyxin images further demonstrated that 77% were positioned within six pixels of each other (, and Fig. S1). The stability of FAs in these cells allowed us to prepare multidimensional representations in which the surface of the labeled FAs drawn in time space were correlated with the two-dimensional fluorescence patterns at different time points (). Animation of these time series (Video 7) further highlights the emergence of forming KFPs from FAs and their subsequent integration into the peripheral KF network. This process repeats itself multiple times, with single FA sites serving as platforms for the formation of several KFPs ( and Video 8). It was recently shown that KF network formation is inhibited in cells producing mutant keratins that are known to cause the skin disorder epidermolysis bullosa simplex, although KFPs are still formed in the cell periphery (). These KFPs, however, fail to elongate, and, instead, enlarge into short-lived spheroid granules that move continuously in an actin-dependent process toward the cell center and disintegrate into soluble material at a distinct transition zone. When we transfected paxillin-DsRed2 cDNAs into epithelial cell lines, producing the fluorescent HK14 mutants enhanced YFP(EYFP)-K14, we observed that paxillin-labeled FAs were prominent initiation sites from which KFPs emerged ( and Video 9, available at ). Overlay of visually assigned sites of KFP formation onto paxillin-DsRed2 images further supported their tight spatial relationship () with 93% of initiation events ( = 264) detected within 6 pixels of FAs (Fig. S1). Repeated granule formation from the same FAs was noted every 1.5–4 min (; Video 9). Cells were treated with shRNAs directed against FA components to directly evaluate the contribution of FAs to KFP formation. Talin-specific shRNAs induced retraction of the KF network around the nucleus and depletion of KFPs in peripheral regions, whereas cells transfected with control shRNAs did not show these alterations (). Similarly, cells synthesizing mutant keratins retained only the residual perinuclear filament system and lacked the characteristic peripheral keratin granules (). The specificity and efficiency of talin down-regulation was confirmed by talin immunofluorescence in each instance (Fig. S2, available at ). We have recently suggested that certain submembraneous sites are important for KF formation and turnover, thereby directing the assembly machinery into cell regions requiring major IF restructuring (; ). The current study identifies FAs as perfectly suited candidates to perform such tasks. FAs are abundant and prominent entities at the extracellular matrix–cell interface, providing complex platforms that mediate structural and signaling functions. The evidence presented in this study for a tight linkage of these sites to KF formation, in combination with well established knowledge about the crucial functions of FAs for the actin system (; ; ) and recent studies linking FAs to the microtubule network in an intricate reciprocal cross talk (; ), assign even more important functions to these sites than hitherto anticipated and place them as central regulators ensuring coordination of the entire cytoskeleton in situations of polarized cell shape changes. A finely adjustable program appears to start in lamellipodia once FAs have formed, beginning with stress fiber recruitment, followed by several poorly characterized transition states, each of which could be subject to feedback control with the intra- and extracellular environment, eventually ending up in the formation of a stabilizing IF cytoskeleton (). In this way, polarized reorganization of the cytoskeleton is accomplished, resulting in directed movement and relocation of the trailing part of the cell, which would have to undergo similarly coordinated disassembly processes. Such a hierarchical organization of the cytoskeletal systems is also supported by observations during epithelial sheet formation, when zippering of actin-anchoring punctate adhesions in fingerlike protrusions of epidermal keratinocytes precede keratin/desmosome-mediated clamping adhesion (). The uncovered tight relationship between KFPs and FAs reflects basic cellular properties, as it is detectable in cells of different origin and is also maintained in cells producing mutant keratins. The same phenomena were also noted using other keratins and cell lines (Video 10, available at ). Furthermore, we would like to suggest that the observed mechanisms are not restricted to keratin IFs, but also apply to other IF types. Evidence for this notion was provided in endothelial cells producing fluorescent vimentin and β3 integrin–labeled FAs (). The observed dynamic colocalization supports the possibility that FAs are sites of vimentin filament nucleation and/or assembly. In summary, FAs may be the long sought after IF-organizing centers. Cloning of a cDNA coding for fusion protein HK18-YFP was previously described (). To prepare hybrid cDNAs coding for HK14 fused to EYFP, K14 was amplified from a K14 cDNA (provided by V. Nimrich, German Cancer Research Center, Heidelberg, Germany) using the amplimers 00–52 5′-AAA AAG CTT ATG ACT ACC TGC AGC CGC CAG-3′ and 00–53 5′-AAA GGA TCC GGG TTC TTG GTG CGA AGG ACC TG-3′; the resulting fragment was cloned into the HindIII–BamHI sites of pEYFP-N1 (CLONTECH Laboratories, Inc.), thereby generating plasmid HK14-YFP. Plasmid HK14-YFP was created by substituting the HindIII–Asp718 fragment with the corresponding fragment of EYFP-K14 (). A RFP-zyxin–encoding construct was provided by A. Huttenlocher (University of Wisconsin, Madison, WI; ). For preparation of actin-RFP, the cDNA for monomeric RFP () was PCR amplified using primers 03–88 5′-AGA TCC GCT AGC CGA TAA GGA TCC GAT GGC C-3′ and 03–89 5′-AGC TCG AGA TCT GGC GCC GGT GGA GTG GC-3′. The NheI–BglII–cleaved PCR product was used to substitute the EGFP-encoding fragment in plasmid pEGFP-actin (CLONTECH Laboratories, Inc.). Construct paxillin-pDsRed2-N1 coding for paxillin-DsRed2 was provided by A. Horwitz (University of Virginia School of Medicine, Charlottesville, VA; ). For the shRNA cloning plasmid, pTER () was first modified by introducing a 1,379-bp SpeI–XbaI–limited, CMV promotor–driven, EGFP-encoding fragment that was PCR amplified from pEGFP-C3 (CLONTECH Laboratories, Inc.; prepared and provided by L. Griffin and E. Bockamp, Institute of Toxicology, Johannes Gutenberg University, Mainz, Germany). The EGFP-encoding fragment of the resulting plasmid, pTER-EGFP, was then exchanged for a 765-bp NheI–XbaI–cleaved mRFP fragment that was amplified from the actin-RFP–encoding plasmid (see previous section) with primers 03–88 and 04–72 5′-AAA GCG GCC GCT TAG GCG CCG GTG GAG TGG C-3′, thereby generating the plasmid pTER-mRFP. Synthetic oligonucleotides were subsequently inserted downstream of the H1 promotor into the BglII–HindIII sites. For talin-specific constructs, the oligonucleotides talin_1_sense 5′-GAT CCC GGC ACT CAC TGG AAC CAT TTT CAA GAG AAA TGG TTC CAG TGA GTG CCT TTT TGG AAA-3′ and control_1_antisense 5′-AGC TTT TCC AAA AAG GCA CTC ACT GGA ACC ATT TCT CTT GAA AAT GGT TCC AGT GAG TGC CGG-3′, and the control oligonucleotides, control_1_sense 5′-GAT CCC GGC ACT CAC TGG AAC CAT TTT CAA GAG AAA TGT TCC AGT GAG TGC CGG GAT CTA-3′ and control_1_antisense 5′-AGC TTA GAT CCC GGC ACT CAC TGG AAC ATT TCT CTT GAA AAT GGT TCC AGT GAG TGC CGG-3′, which differed only slightly from the talin oligonucleotides, were used. Complementary oligonucleotides (100 μM of each) were annealed by incubation for 5 min at 95°C, followed by incubation at 75°C for 15 min in annealing buffer (100 mM potassium acetate, 30 mM Hepes-KOH, pH 7.4, and 2 mM magnesium acetate). In some instances, the EGFP-encoding fragment of pTER-EGFP was removed with AgeI–PmeI and substituted with the corresponding ECFP-encoding fragment of plasmid pECFP-N1 (CLONTECH Laboratories, Inc.), resulting in plasmid pTER-ECFP. Subsequently, cloned shRNA-encoding oligonucleotides were excised with HindIII–EcoRI from the pTER-mRFP–derived plasmids and inserted into pTER-ECFP. The efficiency of shRNA-mediated inhibition of talin production was evaluated with specific talin antibodies (clone 8D4; Sigma-Aldrich) 48 h after plasmid transfection in formaldehyde-fixed cells. SK8/18-2 () and MT5K14-26 (provided by N. Werner and T. Magin, Universitätsklinikum, Bonn, Germany; ) were cultured as previously described. The spontaneously immortalized mouse mammary epithelial cell line EpH4 was provided by L. Huber (Universität Innsbruck, Innsbruck, Austria) and H. Beug (Institute of Molecular Pathology, Wein, Austria; ). It was maintained in DME with high glucose and 10% fetal bovine serum in a humidified incubator at 37°C with 5% CO. Cells were transfected with the help of Lipofectamine 2000 reagent (Invitrogen). For time-lapse live-cell fluorescence recording, cells were grown in glass-bottom Petri dishes (MatTek Corp.), and phenol-free Hank's medium (Invitrogen) was used during imaging. Pictures were recorded by epifluorescence microscopy using an inverse microscope (model IX 70; Olympus) and an attached slow scan camera (model IMAGO; TILL Photonics) as previously described (). The microscope was kept in a closed chamber at 37°C. A 60×, 1.4 NA, oil immersion objective was used, and fluorescence excitation with a monochromatic light of 500 or 570 nm was accomplished with a monochromator (TILL Photonics). At each time point of the time-lapse sequences YFP, RFP, and phase-contrast images were acquired successively within 1.5 s (TILLvisION software; TILL-Photonics). The resulting image sequences were edited with ImagePro Plus (Media Cybernetics). Raw data were cropped and lookup table–adjusted, and the fluorescence channels were color-coded and combined for optimal presentation. To remove residual bleed-through of the RFP signal into the YFP channel, unmixing was performed in some instances ( and Video 1) by subtracting the corrected (F = 0.12) red channel (R) from the green channel (G) according to UMIX = G – (R × F). Amira software (Mercury Computer Systems) was used to generate the 3D visualization of fluorescence in . The cropped and lookup table–adjusted fluorescence image stacks were edited using Amira software. The x, y, and t positions of the first appearance of KFPs were located by visual inspection at high magnification and manually tagged into the frames. A positive score was only assigned when fluorescence was significantly above the diffuse background (probably corresponding to the soluble keratin pool) in at least four adjacent pixels (one pixel covers ∼120 × 120 nm), and when particle growth could be unambiguously detected in subsequent frames. The manually tagged positions of KFP appearance were transferred into the RFP images depicting the location of labeled FA sites. The distances between the yellow fluorescent KFPs to the next red fluorescent FAs were measured and analyzed in a spreadsheet. The tags were projected into one frame and overlaid onto the projected RFP recordings. Most videos are represented, at least in part, in the figures. Video 1 () depicts the establishment of a novel KF network in a newly formed lamellipodium. Video 2 () demonstrates the transport of KFPs along actin stress fibers in a lamellipodium. Video 3 () records the peripheral keratin fluorescence in a cell treated with latrunculin B. Videos 4 and 5 () show the appearance of KFPs in close neighborhood to FAs that are either labeled with RFP-zyxin (Video 4) or paxillin-DsRed2 (Video 5). Videos 6–8 () document the emergence of KFPs from RFP-zyxin–labeled FAs in nonepithelial SK8/18-2 cells at low magnification as a tableau (Video 6), in an animated time-space reconstruction (Video 7), and at high magnification (Video 8). Video 9 () presents an image series of MCF7 cells producing fluorescent keratin 14 mutants together with paxillin-DsRed2. Video 10 is a composite of two recordings (no corresponding figures) of either HK14-YFP and paxillin-DsRed2 (top) or HK14-YFP and RFP-zyxin (bottom) in EpH4 cells. In addition, Fig. S1 provides quantitative data on the spatial relationship between KFP formation and FAs ( and Videos 4–6 and 9). Fig. S2 presents control data to the experiments depicted in . Online supplemental materials are available at .

Ribosome biogenesis is a very conserved process in the eukaryotic kingdom. In , the pathway begins with transcription of the 35S and 5S ribosomal RNA (rRNA) precursors by RNA polymerases I and III, respectively. The association of ribosomal proteins and pre-ribosomal factors with nascent pre-rRNAs gives birth to a 90S pre-ribosomal complex that undergoes various steps of maturation. The 90S complex separates into a pre-60S complex, which will generate the large ribosomal subunit containing mature 25S, 5.8S, and 5S rRNAs, and a pre-40S complex, which will generate the small ribosomal subunit containing 18S rRNA. The maturation of both particles follows two distinct pathways, first in the nucleolus and then in the nucleoplasm, and finally in the cytoplasm after Crm1-dependent export through the nuclear pores (; ; for review see ). Several factors are necessary for correct modification, cleavage, and processing of pre-rRNAs, positioning of ribosomal proteins, and export of the pre-60S and pre-40S particles (for review see ; ; ). For the large ribosomal subunit, apart from the 46 ribosomal proteins that are found in the mature particle, ∼100 pre-ribosomal factors have been identified to date. Except for a few factors with known enzymatic activity, the function of most of the pre-ribosomal factors biochemically associated with pre-60S complexes remains to be defined. Throughout the pathway of biogenesis, the composition of the pre-60S particles changes. At the beginning of the pathway, they contain many pre-60S factors; these proteins gradually dissociate from the complexes in the nucleolus, the nucleoplasm, or the cytoplasm. Even if other pre-60S factors load later on the complexes, the average number of pre-60S factors is significantly lower at the end of biogenesis than at the beginning (). The whole process is highly dynamic; as they dissociate from the precursors of the large ribosomal subunits, pre-60S factors can be recycled and participate in new rounds of biogenesis. In the late cytoplasmic events of 60S biogenesis, only a few pre-ribosomal factors are present on the pre-60S particles (for review see ; ). Some of these factors are shuttling proteins such as Tif6 (), Rlp24 (), Nmd3 (; ), and Arx1 (), which are loaded on the pre-60S complexes in the nucleus and undergo export toward the cytoplasm together with the particle. Other factors seem exclusively cytoplasmic, such as the GTPases Lsg1 () and Efl1 (). Existing data reveal that recycling shuttling factors and triggering the end of biogenesis are tightly intertwined events. Indeed, since the beginning of the 1980s, the pre-60S factor Tif6 is thought to act as an antiassociation factor between the small and large subunits (); the GTPase Efl1 triggers the recycling of Tif6 (). Recycling of the Nmd3 shuttling protein also requires the GTPase Lsg1 and the late-associating ribosomal protein Rpl10 (; ). The shuttling of pre-60S factors between the cytoplasm and the nucleus involves the general import machinery. Recycling of Nmd3 is mediated by the β-karyopherin Kap123 (), which also accounts for the import of various ribosomal proteins to the nucleus through the nuclear pore complexes (). At least three β-karyopherins participate in the import of ribosomal proteins and pre-ribosomal factors— Kap123, Kap121/Pse1, and Kap104 ()—with a significant overlap as far as their specificity for cargoes is concerned. In this study, we investigate late events in the maturation of the large ribosomal subunit in which a newly described, exclusively cytoplasmic pre-60S factor, Rei1/Ybr267w, plays a central role. Our results indicate that coordinated dissociation and return to the nucleus of shuttling pre-60S factors are essential for correct maturation of the large ribosomal subunits and participate in their activation for subsequent entry into translation. Most ribosomal proteins of the large subunit are loaded on 60S pre-ribosomal particles in the early nuclear steps of 60S biogenesis. However, a few proteins, such as Rpl10 and Rpl24, appear to associate later to the pre-60S particles (). Rpl24 is a late-associating ribosomal protein that shares significant homology with the essential pre-ribosomal factor Rlp24 (). Rlp24 is present on pre-60S particles from the nucleolus to the cytoplasm, where it is believed to dissociate when Rpl24 associates with the particle. To identify potential partners required for the loading of Rpl24 to the particles, a two-hybrid screen using Rpl24b as bait was performed. The most frequent prey selected in the screen was , which is found as seven distinct inserts. All of the inserts containing the ORF shared a minimal interacting domain from amino acids 270 to 342 (). Rei1 is a 393–amino acid–long protein that contains three conserved CH Zn finger motifs extending from amino acids 7 to 31, 162 to 187, and 215 to 239. The domain extending from amino acids 57 to 202 displays similarity with the PFAM E-MAP-115 domain characteristic of ensconsin, a microtubule-associated protein expressed in higher eukaryotes (). The minimal two-hybrid interaction domain between Rei1 and Rpl24b is located in the COOH-terminal part of the protein apart from all of these motifs. This region is quite conserved in Rei1 homologues among eukaryotes, mainly in the portion extending from amino acids 300 to 330 (unpublished data). The two-hybrid link between Rei1 and Rpl24 suggested that Rei1 might participate in molecular events involving the large ribosomal subunit together with the late-associating ribosomal protein Rpl24. This hypothesis was supported by previous tandem affinity purifications (TAPs), which had revealed the presence of Rei1 in several pre-60S complexes (; ). The clustering of these TAP data (for review see ) had already shed light on Rei1 as a putative late pre-60S factor. Therefore, we further investigated the possible role of Rei1 in the cytoplasmic events of 60S biogenesis. In addition to its cytoplasmic localization (), the interaction of Rei1 with Rpl24 prompted us to determine whether this factor was associated with ribosomal or pre-ribosomal complexes in the cytoplasm. We performed TAP experiments on strains producing a Rei1 protein fused to a 13Myc epitope and either Rlp24 or Rpl24b fused to the TAP tag. Rei1-13Myc was enriched in the Rpl24b-TAP–associated complexes compared with the Rlp24-TAP–associated complexes or to the strictly nuclear Ssf1-TAP–associated complexes used as negative controls (). In contrast, the presence of Nog1 was observed, as expected, in both Ssf1-TAP– and Rlp24-TAP–associated pre-ribosomal complexes but not in the Rpl24b-TAP–associated complex. These data suggest that Rei1 is specifically associated with Rpl24-containing cytoplasmic complexes. To analyze a potential association of Rei1 with mature translating ribosomes, total cellular extracts from cells producing Rpl24b-TAP were fractionated on a sucrose gradient. As expected, Rpl24b-TAP was found in fractions corresponding to the 60S, 80S, and polysomes. We observed that Rei1 was absent from the 80S and polysomal fractions but present in the 60S fraction (), indicating that in contrast to Rpl24, Rei1 is not associated with mature ribosomes during translation. Exclusive association with 60S particles supported the hypothesis that Rei1 is a late pre-60S factor that transiently binds to pre-ribosomal complexes in the cytoplasm after the release of Rlp24 from the particles. To determine whether Rei1 could be involved in ribosome biogenesis or in translation, the strain and the corresponding wild-type strain were used for sucrose gradient analysis. The strain displayed a slow growth phenotype at 30°C and was cold sensitive as previously observed in another genetic background (). In the strain, the polysome profile was clearly affected. At 30°C, the amount of 60S subunit decreased significantly (unpublished data). At 23°C, not only a decrease of the 60S subunit amount was observed, but also a drop in the average level of polysomes and the appearance of half-mers, which can be explained by abortive 48S preinitiation complexes being formed but remaining complexes being blocked on mRNA because of a lack of mature large ribosomal subunits (). The clearly affected polysome profiles in the strain correlated with a defect in rRNA maturation (). Ethidium bromide staining of agarose gels, Northern blots, and primer extensions were performed to determine the relative levels of various rRNA intermediates and mature species. The rRNA defect was already significant at 30°C (unpublished data) but increased at 23°C. At this temperature, we observed a drop in the levels of total 27S species, 27SA and 7S nuclear pre-rRNAs, as well as mature 5.8S rRNA normalized relatively to U2 small nuclear RNA levels. The 27SB/27SA pre-rRNA ratio was increased by 3.1 ± 0.4, and the 27SB/7S ratio was increased by 4.7 ± 0.5. Altogether, these data are typical of a defect in the ITS2 processing step (). They are in correlation with the drastic decrease in the 60S peak on sucrose gradients, which results in a stoichiometric imbalance between the small/large ribosomal subunit (). Curiously, although Rei1 appears to be a cytoplasmic protein, its absence displayed rRNA maturation defects at the nuclear level. To investigate whether Rei1 might be a shuttling factor, we monitored Rei1-GFP localization in a strain defective for the export of pre-60S particles (). We used a strain in which a version of the dominant-negative mutant was overexpressed compared with the same strain overexpressing a wild-type version (). As a control, we monitored the localization of the ribosomal protein Rpl25-GFP. As expected, Rpl25-GFP, which is normally mainly cytoplasmic, was accumulated in the nucleus when was overexpressed. In contrast, no accumulation was seen for Rei1-GFP, which appears to be an exclusively cytoplasmic pre-60S factor. We additionally tested the shuttling ability of pre-ribosomal factor Rlp24-TAP and of ribosomal protein Rpl24b-TAP (). In the dominant-negative mutant, Rlp24-TAP accumulated in the yeast nuclei compared with wild-type conditions. Meanwhile, Rpl24b-TAP remained exclusively cytoplasmic. This confirmed previous data () in favor of Rlp24 being a shuttling nucleocytoplasmic pre-60S factor, whereas Rpl24 appears to be a ribosomal protein that associates with pre-60S particles in the very last cytoplasmic steps of ribosome biogenesis. As a late pre-60S factor, Rei1 was a candidate for participating in the export of pre-60S particles from the nucleus to the cytoplasm. Therefore, we monitored the localization of an Rpl25-GFP reporter construct in a strain. Because no export defect was observed (unpublished data), we assume that Rei1 is not required for the export of precursors of the large ribosomal subunit. Considering the previous hypothesis, we investigated the possible role for Rei1 in the recycling of shuttling pre-60S factors such as Tif6 (), Rlp24 (), and Arx1 (). We looked at the localization at 23°C of the Arx1-GFP, Tif6-GFP, or Rlp24-TAP fusion proteins in a strain compared with a wild-type strain (). In the wild-type strains, Arx1-GFP and Tif6-GFP were mainly observed in the nucleus, and a small fraction of the proteins was found in the cytoplasm. In the absence of Rei1, Arx1-GFP or Tif6-GFP were mainly cytoplasmic. This confirmed our hypothesis that nuclear import of some shuttling factors could be impaired in the absence of Rei1. In contrast, no recycling defect was observed for Rlp24-TAP, which even appeared somewhat concentrated in the nucleolus of the strain compared with the wild-type strain, where it appeared more equally distributed between the nucleolus, nucleoplasm, and cytoplasm. These data offer a simple explanation for the nuclear rRNA maturation defect observed in the strain. Indeed, the absence of Rei1 affects the processing of ITS2 in a manner similar to the repression of Tif6 (). To better illustrate this, a strain in which was placed under the control of a promoter was shifted to glucose from 0 to 6 h, and pre-rRNA species were analyzed and quantified along these kinetics (). The rRNA intermediates ratio of the strain to the wild type were comparable with the effects of a repression for 4–6 h. We also tested the phenotype of an deletion, but no obvious rRNA maturation impairment was detected (unpublished data). Therefore, the 60S maturation defects observed in the strain are likely caused by a decrease of nuclear Tif6. We tested whether the overexpression of could compensate this phenotype, but overexpression appeared to be toxic in wild-type as well as in strains (Fig. S1, available at ). As Arx1-GFP and Tif6-GFP accumulated in the cytoplasm in the absence of Rei1, we wondered whether they were still associated with pre-60S complexes. To address this question, we separated extracts from a or wild-type strain expressing chromosomal or on sucrose gradients (). In the wild-type strain, we could detect Arx1-GFP in the 60S fractions of the gradient. In the absence of Rei1, Arx1 appeared to sediment closer to the top of the gradient, which suggested that in such conditions, Arx1-GFP was present in the cytoplasm as a small complex or as an oligomeric protein. In contrast, in the absence of Rei1, the Tif6-GFP fusion protein was still associated with particles of the size of 60S. Although the absence of Rei1 leads to a cytoplasmic retention of both shuttling pre-60S factors Tif6 and Arx1, Tif6 is still stably associated with 60S particles, whereas Arx1 accumulates as a small complex in the cytoplasm. The presence of Arx1 in a small cytoplasmic complex in the absence of Rei1 suggested the possibility that Arx1 could associate and act together with other proteins. We performed a two-hybrid screen using Arx1 as bait to identify the physical partners of Arx1. The most frequent prey we selected was , which is hereafter referred to as (A little brother 1). This prey was found as seven distinct inserts; all of the inserts shared a minimal interacting domain from amino acids 148 to 175 (). The deletion of had little effect on growth and displayed no obvious rRNA maturation impairment. In the , double mutant, the growth phenotype as well as the polysome profile resembled that of the single mutant (Fig. S3, available at ). To confirm the physical interaction between Arx1 and Alb1, we purified the complexes associated with Alb1 (). Arx1- and Alb1-TAP–associated complexes look very similar on the Coomassie staining compared with Rei1-TAP. We could detect Arx1 and Rei1 in the three complexes. These results corroborate previous affinity precipitations (; ) where Alb1 had been identified in both Rei1- and Arx1-associated complexes. They not only confirm that Arx1 and Alb1 are physically associated under physiological conditions but also suggest that Alb1 could be a pre-60S–associated factor. Note that Rlp24 was present in Arx1- and Alb1-TAP–associated complexes but was absent from Rei1-TAP complexes, which confirms that Rei1 loads onto the particles after the release of Rlp24. In addition, we observed that Alb1-TAP sedimented around the 60S peak but was absent from the polysomal fractions on a sucrose gradient (); therefore, it is likely to associate with precursors of the large ribosomal subunit. To determine whether any direct interaction could be detected between these factors, we performed in vitro GST pull-down experiments with GST-tagged baits (GST-Alb1, GST-Rei1, or GST) and the His6-Arx1 fusion protein as prey (). His6-Arx1 copurified with GST-Alb1 but not with GST or GST-Rei1 used as negative controls. Thus, we conclude that Arx1 and Alb1 are able to interact directly even in the absence of other yeast components. First, we tested whether Alb1 was an additional shuttling factor by monitoring the localization of an Alb1-GFP reporter construct in a wild-type or strain at 23°C (). In the wild-type strain, Alb1-GFP localization was similar to that of Arx1-GFP, with a strong nuclear–nucleolar signal and a weak cytoplasmic signal. In the mutant, Alb1 was relocalized to the cytoplasm, as was the case for Arx1. Second, in a sucrose gradient analysis (), Alb1-GFP could be detected in the 60S fractions in the wild-type strain together with Arx1, whereas in the strain, it was found mainly in the same fractions as Arx1 at the top of the gradient. Thus, the association of Alb1 to nuclear pre-60S particles was impaired, as it was for Arx1, in a strain. We tried to further characterize the small complex observed in the absence of Rei1 by purifying the Arx1- and Alb1-TAP–associated complexes in these conditions (). The Coomassie staining showed a weak decrease of the levels of most proteins in the strain compared with the wild-type profiles, except for Arx1. In the absence of Rei1, immunoblots performed on the purified fractions revealed a loss of the pre-ribosomal factors Nog1 and Tif6. These results suggest that Arx1 and Alb1 can form cytoplasmic subcomplexes in the absence of Rei1 independently from pre-ribosomal particles. To confirm these results, purified complexes associated with Alb1-TAP were separated on a sucrose gradient in the absence of Rei1 (). These complexes sedimented in the first five fractions of the gradient and contained Arx1 and Kap121, as discussed in the following paragraphs. In contrast to the other Ponceau-stained bands, these two proteins could be depleted by preincubating the TEV eluate with calmodulin affinity resin, which retains Alb1-CBP, indicating that they are strongly associated with this bait. To gain more insights into the molecular events involved in the observed phenotypes in the absence of Rei1, we performed a high copy number suppressor screen with a strain deleted for grown at 20°C. In addition to itself, we selected the homologous gene . The overexpression of this gene was able to partially complement the cold-sensitive phenotype of () as previously shown (), yet we could not detect any defect in strains on polysome profiles nor on rRNA maturation (not depicted). Thus, we assume that although it shares partial functional overlap with Rei1, Reh1 is not limiting for the biogenesis of the large ribosomal subunit under physiological conditions. The third gene identified in our screen was (). It was selected as seven distinct DNA inserts containing the whole ORF. The smallest insert constituted a DNA fragment extending from 589 bp upstream of the initiation codon to 762 bp downstream of the stop codon. Kap121, also known as Pse1, is a β-karyopherin, which was shown to be involved in the import of various ribosomal proteins () as well as pre-ribosomal factors such as Nop1 and Sof1 from the cytoplasm to the nucleus (). The mislocalization to the cytoplasm of Arx1-GFP, Alb1-GFP, or Tif6-GFP in a strain could be compensated by overexpression of either or (). Both constructs were able to relocalize the three fusion proteins to the nuclear compartment compared with a strain transformed with an empty vector. Because overexpression of the β-karyopherin Kap121 was sufficient to overcome the import defect of all three shuttling factors in the strain, one might wonder whether Kap121 was the physiological importin for these proteins. Such a hypothesis was supported by the presence of Kap121 in Arx1- and Alb1-TAP–associated complexes ( and ; ; ) and by two-hybrid data (Fig. S4, available at ). In contrast, Kap121 was not detected in Tif6-TAP–associated complexes; therefore, we did not further assess the possibility that Kap121 is the specific importin for this factor. As a control, Rlp24 was as efficiently copurified with Tif6-TAP as with Alb1-TAP or Arx1-TAP compared with Rei1-TAP, which was associated with neither Kap121 nor Rlp24 (). Further evidence for Arx1 and Alb1 being Kap121 cargoes was provided by localization of the Arx1-GFP and Alb1-GFP fusion proteins in strains in which the expression of was repressed (). Both fusions were redistributed to the cytoplasm compared with wild-type strains, where we observed bright nuclear–nucleolar signals. This result suggested that the karyopherin Kap121 might be the physiological importin for Arx1 and Alb1. Altogether, these results show that the nuclear import of Arx1 and Alb1 is Kap121 dependent. In the absence of Rei1, the return of Arx1–Alb1 to the nucleus as well as that of Tif6 can be restored by overexpressing this karyopherin. The deletion of either or in the strain restored growth at 23°C almost up to the levels of or single mutant strains, respectively (). Therefore, the cytoplasmic accumulation of Arx1–Alb1 accounts for the observed cold-sensitive phenotype of the strain. According to our hypothesis, Arx1–Alb1 might also explain the cytoplasmic retention of Tif6. or double mutant strains (). In these mutants, Tif6 was mainly nuclear, as in the wild-type strain, compared with the mostly cytoplasmic signal in the strain. Thus, we conclude that in the absence of Rei1, Arx1 and Alb1 participate in a cytoplasmic subcomplex that prevents the adequate release of Tif6 from pre-60S particles and its correct return to the nucleus for a new round of biogenesis. Previous studies have shown that some of the latest pre-60S factors can be involved in recycling shuttling factors. Indeed, the GTPase Lsg1 was shown to participate in the release and recycling of Nmd3 (). An additional GTPase, Efl1, is required for release from the particles and recycling of the pre-60S factor Tif6 (). In this study, we characterized Rei1 as a new pre-60S–associated cytoplasmic factor at the center of maturation events taking place in the nucleus. We provide evidence for the role of this factor in the recycling of Arx1, Tif6, and a novel pre-60S shuttling factor, Alb1. Arx1 was first characterized as a component of pre-60S complexes by TAP experiments (; ). This nonessential protein, which is present both in the nucleus and cytoplasm of yeast cells, contains a domain that is characteristic of methionine amino peptidases. This study sheds light on an additional shuttling pre-60S factor, Alb1. Our data suggest that Arx1 and Alb1 can form a dimer because they interact in a two hybrid (), Arx1 is overrepresented in Alb1-TAP–associated complexes (), they interact directly in vitro (), and they can be copurified as a nonribosomal subcomplex (). In physiological conditions, both Arx1 and Alb1 participate in pre-60S complexes from the nucleus to the cytoplasm. Nevertheless, neither of them seems to be essential for the maturation of the large ribosomal subunit. Indeed, in the absence of either factor, we failed to detect any strong rRNA maturation, export, or polysome profile defect (Fig. S3 and not depicted). In the absence of Rei1, both factors are retained in the cytoplasm ( and ) in the form of a small complex ( and ); they are not imported into the nucleus and, consequently, are not incorporated into new pre-60S complexes. Ribosome biogenesis needs the import into the nucleus of many proteins through the nuclear pore complexes. In addition to ribosomal proteins, all of the factors that participate in the transcription of rRNA or in ribosome biogenesis must be imported into the nucleolus. Two main β-karyopherins have been shown to be involved in this directional transport: Kap123 and Kap121/Pse1 (). Each of these importins is responsible for the nuclear import of specific cargoes: Kap123 imports most of the ribosomal proteins, and Kap121 also carries nucleolar proteins such as Nop1 and Sof1 (). Although we did not detect any consensus Kap121-specific NLS in the sequence of Arx1 or Alb1, three independent pieces of evidence indicate that both factors are imported into the nucleus by a mechanism involving Kap121: the cytoplasmic accumulation of Arx1 and Alb1, released from pre-60S complexes in the absence of Rei1, can be overcome by the overexpression of Kap121 (); Arx1 and Alb1 remain in the cytoplasm in the absence of Kap121 (); and Kap121 can be found in Arx1- and Alb1-associated complexes (; ) and was found as a two-hybrid partner of Alb1 in a matrix assay (Fig. S4). Thus, one might assume that Arx1 and Alb1 are additional Kap121 cargoes. Nevertheless, we cannot exclude the possibility that the observed effects could be indirect. So far, the nature of the nuclear import signal for Arx1–Alb1, which is affected when they dissociate from pre-60S particles in the absence of Rei1, is still unclear. However, the binding with Kap121 is not affected because Kap121 is still copurified with Arx1 and Alb1 in such conditions (). recapitulates the conclusions of this study concerning the cytoplasmic events that take place at the end of the large subunit biogenesis in wild-type conditions or in the absence of Rei1. In the model that we propose, we only took into account the factors that had shown physical or functional links in our study. At the exit of the nucleus, pre-60S particles still contain shuttling pre-ribosomal factors such as Rlp24, Tif6, Arx1, and Alb1. At this stage in the pathway, the cytoplasmic pre-60S factor Rei1 and the late-associating ribosomal protein Rpl24 associate on the particle. So far, we could not determine which of the factors loads first; however, our data suggest that they associate with the particle after the dissociation of Rlp24 (). Arx1, Alb1, and Tif6 then leave the particle and return to the nucleus, where they can enter a new round of biogenesis. In the absence of Rei1, we showed that the Kap121-dependent recycling of Arx1–Alb1 was impaired and that the cytoplasmic accumulation of this duplex inhibited the release of Tif6 from pre-60S particles and subsequent return of this factor to the nucleus. Because Tif6 and its homologues were described as putative antiassociation factors in several eukaryotic species (; ), we believe that stalling of Tif6 on the particles at the end of the 60S biogenesis could result in the observed phenotypes. Because Efl1 was previously characterized as the GTPase that triggers the dissociation of Tif6 from the particle (), it was a good candidate for explaining the recycling defect observed for Tif6. However, Efl1 was always found as a free protein in sucrose gradients, whatever the presence of Rei1 (Fig. S2, available at ). Additionally, it was never found to be associated with Arx1 or with Alb1 in TAP experiments (unpublished data). Lastly, the overexpression of wild-type or mutant forms of Efl1 did not complement a strain (Fig. S2). Consequently, the retention of Tif6 on pre-60S particles in the absence of Rei1 is not likely caused by a lack of Efl1. Possibly, the Arx1–Alb1 cytoplasmic complex might sequester another, as yet unknown, factor required for the release of Tif6. Rei1 (required for isotropic bud growth) was previously described as a member of the mitotic signaling network (). This factor was proposed to be involved in the regulatory network around the Swe1 kinase by way of Nis1. The Swe1 kinase is the converging platform for several events that result in the negative regulation of cyclin-dependent kinases, which, in turn, control the switch between polar growth and isotropic growth at the bud neck. These data are not in contradiction with the present results and even offer the possibility that Rei1 might be involved in coupling ribosome biogenesis with cell cycle control. Indeed, other factors appear to be implicated in both pathways. For instance, Nap1, which belongs to this mitotic signaling network, also displays interactions with proteins involved in ribosome biogenesis and, in particular, with Rei1 (). According to these data, Rei1 could also act at the intersection between these two major cellular pathways. One way to address the question would be to confirm the existence of an E-MAP-115 domain in Rei1. This domain was characterized in ensconsin, a microtubule-associated protein of 115 kD. In higher eukaryotes, this protein was shown to bind microtubules in cell lines of epithelial origin (). To further speculate, we could imagine that Rei1, if it also interacted with microtubules, could orient newly synthesized large ribosomal subunits to the bud neck, where a high rate of protein synthesis is required for bud growth. An appealing hypothesis would be that Rei1 triggers the dissociation of Tif6 only once the 60S particles have reached active translation sites. The yeast strains used in this study are listed in Table S1 (available at ). They were generated by homologous recombination using PCR products to transform either MGD353-13D or BY4741/4742 (S288C background) strains (). The pBS1479 vector was obtained from B. Seraphin (Centre National de la Recherche Scientifique [CNRS], Gif-sur-Yvette, France). A pFA6a-TAP plasmid was obtained from F. Stutz (University of Geneva, Geneva, Switzerland). The SC0708 and SC0406 strains were obtained from Euroscarf and have been constructed by CellZome AG. Disrupted strains in the BY4741/4742 background with the KanMX4 marker came from the Euroscarf collection of deletion strains (). Plasmids for two-hybrid screens were obtained by Gateway cloning in pAS2 and pACTIIst destination vectors (gift from E. Bertrand, CNRS). The sequence of oligonucleotides used for Northern hybridization and primer extension analysis were previously described (). The yeast two-hybrid screens were performed using a cell-to-cell mating strategy as previously described (). The strain CG1945, transformed with either pAS2-RPL24 or pAS2-ARX1 baits, was mated with the strain Y187 transformed with an DNA library cloned in pACTIIst. Diploids were spread on minimal medium without leucine, tryptophan, and histidine. 55 and 36 positive clones were recovered in the Rpl24b and Arx1 screens, respectively. The BY4741 strain was transformed with a yeast genomic high copy number vector library constructed in pFL44L (provided by F. Lacroute, CNRS). The transformants were grown on solid synthetic minimal medium lacking uracil at 20°C. Colonies that had lost their cold-sensitive phenotype when compared with the same strain transformed with an empty vector were selected. Plasmidic DNA was recovered, and DNA inserts were sequenced. The suppressor plasmids were checked by retransformation of the strain; their growth phenotypes were compared by spotting transformants in 10 dilution steps on minimal medium without uracil at 23 and 30°C. Cells were broken with glass beads, and total RNAs were subjected to phenol-chloroform extraction. RNAs were resolved on 6% polyacrylamide-urea gels or on 1% agarose gels, transferred to Hybond-N+ membranes, and probed with various P-labeled oligonucleotides. Primer extensions were performed with P-labeled oligonucleotides; the products were then resolved on 5% polyacrylamide-urea gels. Quantifications were performed with ImageQuant software (GE Healthcare). Total protein extracts were prepared from exponentially growing yeast cells as previously described () and were separated on 10–50% sucrose gradients by centrifugation for 3 h at 190,000 (SW41 rotor; Beckman Coulter). In each fraction of the gradient, the proteins were precipitated with 10% TCA, separated on 10% polyacrylamide-SDS gels, and transferred to nitrocellulose membranes. TAP-tagged proteins were detected with a 1:10,000 dilution of the peroxidase–antiperoxidase (PAP) soluble complex (Sigma-Aldrich). Native, 13Myc-tagged, and GFP-tagged proteins were detected by indirect immunoblotting using specific polyclonal rabbit antibodies as primary antibodies at a 1:5,000 dilution, anti-cMyc mouse monoclonal IgG1 (Santa Cruz Biotechnology, Inc.) at a 1:2,000 dilution, or anti-GFP rabbit polyclonal IgG (Santa Cruz Biotechnology, Inc.) at a 1:400 dilution. TAP-tagged proteins were detected with PAP (Sigma-Aldrich) at a 1:10,000 dilution. Anti-Nog1 antibodies were described previously (). Antibodies against Kap121 and Tif6/Efl1 were obtained from J.D. Aitchison (Institute for Systems Biology, Seattle, WA) and F. Fasiolo (CNRS), respectively. Anti-Arx1 and anti-Rei1 antibodies were produced by the immunization of rabbits with specific immunogenic peptides and were affinity purified (Covalab). Secondary antibodies (goat anti–rabbit or goat anti–mouse HRP conjugate; Bio-Rad Laboratories) were used at a 1:10,000 dilutions. Visualization of the peroxidase activity was performed with the ECL+ chemiluminescence kit (GE Healthcare). Complexes were purified according to the standard TAP protocol (), starting from 4 liters of yeast culture. The TEV eluate was precipitated with 10% TCA and analyzed by Western blotting as described above. Interactions between the GST-tagged baits and the His6-Arx1 fusion prey were assessed as previously described (). Cells transformed with a centromeric plasmid expressing (gift from E. Hurt, University of Heidelberg, Heidelberg, Germany) or cells expressing chromosomal reporter constructs were cultured as indicated and mounted in liquid minimal culture medium on glass slides. Immunofluorescence detection of the TAP-tagged fusion proteins was performed as previously described (). The fluorescence of GFP or Cy3 was detected with an epifluorescence microscope (model DMRB; Leica) at room temperature with 100× NA 1.30–0.60 oil immersion objective lenses (Leica). Images were acquired with a dual mode cooled CCD camera (C4880; Hamamatsu) and the HiPic32 acquisition software (Hamamatsu). The pAJ368/544 vectors were provided by A.W. Johnson (University of Texas, Austin, TX). Table S1 recapitulates all yeast strains used in this study. Fig. S1 shows the results of TIF6 overexpression. Fig. S2 presents the data showing that a lack of Efl1 is likely not the cause of the recycling defect observed for Tif6 in the absence of Rei1. Fig. S3 displays the phenotypes on growth and polysome profiles of or single or double mutants. Fig. S4 shows the two-hybrid interactions between Arx1 and Alb1 and between Alb1 and Kap121. Online supplemental material is available at .

All transport events between the cytoplasm and the nucleus occur through a large channel in the nuclear envelope called the nuclear pore complex (NPC). Very little is known about the mechanism of NPC biosynthesis, particularly how the proteins composing this complex are assembled, inserted, or anchored into the nuclear envelope. In metazoans, there are at least two NPC assembly pathways. The first assembly pathway occurs upon the completion of mitosis, when membrane vesicles and nucleoporins (Nups) are recruited to chromatin during nuclear envelope reformation (). The second pathway occurs in interphase, during which time NPCs are continually synthesized and inserted into intact nuclear envelopes (; ). Although the targeting and assembly of membranes and Nups to postmitotic nuclei is somewhat understood, little is known about NPC biogenesis in interphase cells. Many unicellular eukaryotes, such as the yeast , rely exclusively on NPC insertion into intact double membranes because they undergo a closed mitosis with no nuclear envelope breakdown. Of the 33 Nups in the yeast three are pore membrane proteins (POMs), and two integral membrane Nups have been described in metazoans. Because of their membrane association, the POMs are attractive candidates for performing functions such as targeting soluble Nup subcomplexes to the nuclear envelope or inserting the NPC into the membrane. Several studies have been performed on the metazoan transmembrane Nups GP210 and POM121, although results pertaining to their function in NPC assembly are somewhat conflicting. For example, one study in extracts found that the inhibition of GP210 by antibody addition or the overexpression of its cytoplasmic tail domain resulted in distorted nuclear envelopes lacking NPCs (), and another study in the nematode found a similar nuclear envelope defect when GP210 was depleted by RNAi (). More recently (), both GP210 and POM121 were depleted from HeLa cells and extracts by siRNA and immunodepletion, respectively. It was found that when POM121 was depleted postmitotic nuclear envelope reformation was inhibited, and the localization of other Nups in the nuclear envelope during interphase was reduced (). However, no defects were seen after the depletion of GP210 from cells (; ). Of the three POMs in budding yeast, and are not essential for cell viability, whereas is an essential gene. Interestingly, Ndc1p localizes to both the NPC and the spindle pole body (SPB), which is comprised of two large protein complexes that span both bilayers of the nuclear envelope (). Most of the work on Ndc1p has focused on its role at the SPB, where it has been shown to be required for duplication of the inner plaque of that structure (). Recently, a conditional allele of was shown to prevent the incorporation of newly synthesized Nup49p into the NPC (). However, this mutant did not affect the steady-state localization of Nup49p or nuclear transport. To examine the role of transmembrane pore proteins in interphase NPC assembly and function in yeast, we depleted Ndc1p, Pom34p, and Pom152p, either alone or in combination. We found that the depletion of Ndc1p led to a partial mislocalization of NPC components and that this phenotype was exacerbated when Pom152p was also absent. However, lack of Ndc1p alone was sufficient to perturb NPC function and to cause mislocalization of nuclear import reporters. This suggests that Ndc1p is required both for efficient nuclear transport and for NPC assembly, although Pom152p can partially compensate for the loss of Ndc1p in NPC assembly. Interestingly, transmission electron microscopy (TEM) images demonstrate that the nuclear envelopes of cells lacking both Ndc1p and Pom152p contain pores with larger diameters than wild-type NPCs and without apparent protein material. In accordance with this, these POM-deficient cells also allowed passive diffusion of proteins that are normally above the exclusion limit of the NPC. These results suggest that the POMs Ndc1p and Pom152p are important for targeting soluble Nups to the nuclear envelope. To examine the functions of the three yeast POMs, Ndc1p, Pom34p, and Pom152p, we created strains in which one, two, or all three of the POMs were depleted. Cells with , , or both genes deleted at their genomic loci grew indistinguishably from wild type on either dextrose or galactose plates (, top four rows). Because is essential for cell viability (), it was depleted from cells using a transcriptional shutoff method. The gene was placed under the control of the promoter at its endogenous genomic locus (-). Thus, expression can be induced in the presence of galactose and repressed in the presence of dextrose. To monitor the expression and localization of Ndc1p, - was also tagged with at the COOH terminus. were inviable when grown on the repressive dextrose plates, but grew like wild type on galactose plates (, bottom four rows). - strain consistently displayed slightly increased growth on dextrose plates compared with the other - strains (, pom34Δ GAL1-NDC1). To determine the extent and time course of Ndc1p depletion, we monitored Ndc1p expression in - cells upon the shift to dextrose-containing media. During growth in galactose-containing media, all - strains expressed high levels of Ndc1p-GFP, but >95% of the Ndc1p-GFP had been depleted after 24 h of growth in liquid dextrose media, as determined by Western blotting (). We also examined cells from each of the - strains by fluorescence microscopy and observed that most of the cells lacked a detectable Ndc1p-GFP signal after 24 h of growth in dextrose (). Depletion of Ndc1p coincided with an increase in cell size (), and the - strains arrested as large-budded cells with only one intact SPB, which is illustrated by one bright spot labeling the SPB component Spc24p-RFP (). In contrast, wild-type cells with large buds had two spots of Spc24p-RFP signal (), confirming that the function of Ndc1p in SPB duplication had been perturbed (). Because is essential, transcriptional shutoff of the gene should inhibit cell division () and, ultimately, cause cell death. To determine whether the cells in our experiments were still viable after being depleted of Ndc1p for 24 h in dextrose, we stained each strain with two different vital dyes. Both Fun1 and MitoTracker dyes have been shown to stain only cells that are metabolically active (; ). Fun1 is actively taken up into live-cell vacuoles and stains intravacuole tubules, whereas MitoTracker is incorporated into actively respiring mitochondria. In each of our strains, both dyes were incorporated into cells, indicating that they remain metabolically active (). As a control, neither dye was incorporated into cells treated with sodium azide and 2-deoxyglucose, which stop cellular respiration (, left). Based on these experiments, we conclude that Ndc1p can be effectively depleted from yeast cells either alone or in combination with complete deletions of and . Furthermore, 24 h in dextrose is sufficient to effectively deplete Ndc1p and induce a cell cycle arrest, although cells remain metabolically active for at least an additional 24 h in the absence of Ndc1p (unpublished data). To examine NPC distribution in the absence of transmembrane Nups, each strain was transformed with plasmids expressing Nup159p-CFP, Nup59p-CFP, or Nup60p-CFP under their endogenous promoters. These three Nups were chosen because they each localize to a distinct region of the pore. Nup159p is found on the cytoplasmic filaments, Nup59p is found in the central core, and Nup60p is found on the nuclear side of the NPC (). All strains were grown in dextrose-containing media for 24 h, and live cells were analyzed by fluorescence microscopy (). Only cells lacking the Ndc1p-GFP signal were included in the analysis, and cells were scored as having normal NPCs if there was a visible accumulation of Nup-CFP signals around the nucleus. Deletion of , , or a combination of both had no significant effect on the localization of Nup159p-CFP (). deletion strains (). In contrast, the strain depleted of Ndc1p displayed a partial reduction in the percentage of cells with properly localized NPCs, with only 41 ± 6% of the cells showing a correct NPC distribution (). Similar results were obtained in the strain lacking both Ndc1p and Pom34p, in which 32 ± 5% of cells had a normal Nup159p-CFP localization pattern (). - cells, the loss of Nup159p NPC staining appeared to coincide with an increased pool of cytoplasmic Nup159p-CFP (). Strikingly, cells that lacked both Ndc1p and Pom152p displayed the most severe mislocalization defects (). - strain had no cells with normal Nup159p-CFP, and the triple-mutant strain ( - had only 2 ± 2% of cells with proper Nup159p-CFP localization (). Similar results were obtained when the central Nup59p-CFP and the nuclear Nup60p-CFP were visualized and scored as described in the first paragraph of this section (). In each experiment, Ndc1p depletion led to a partial NPC localization defect, and this effect was exacerbated by the loss of Pom152p. However, the loss of Nup59p from the NPC was not as dramatic as the loss of the peripherally associated Nup159p or Nup60p, suggesting that Nup59p is more stably associated with the core of the NPC. To rule out the possibility that the observed Nup mislocalizations were caused by a decrease in Nup-CFP protein levels, we performed Western blots on several of these strains after Ndc1p depletion. Although we consistently observed minor variations in Nup-CFP protein expression levels between strains and between experiments, most likely attributable to their plasmid-driven expression, all tested Nups were expressed at comparable levels in dextrose growth conditions (Fig. S1, available at , and not depicted). Importantly, there was no correlation between the observed NPC defects and the variations in protein levels. To determine whether the lack of perinuclear NPC signal in some POM mutant strains was indirectly caused by nuclear envelope defects, we used a reporter protein with a tetrapeptide ER-retention signal (HDEL) fused to dsRED (dsRED-HDEL). In wild-type cells, this reporter localizes to the ER lumen, which in yeast includes the space between the inner and outer nuclear envelopes, as well as the peripheral ER (). Although the Ndc1p-depleted cells occasionally had misshapen nuclear envelopes, the dsRED-HDEL reporter was able to properly localize to the lumenal space between the inner and outer nuclear envelopes. This result suggests that the NPC mislocalization is not caused by disruption of the nuclear envelope. The Nup mislocalization suggested that NPCs were not properly formed in POM-deficient strains. To determine whether the NPCs in these strains were functional for nuclear transport, we tested two different import pathways. First, we monitored the steady-state localization of a classical nuclear import reporter, consisting of the SV40 NLS fused to monomeric red fluorescent protein (mRFP). In wild-type cells, this protein accumulates in the nucleus in 99% of cells ( ). No change in localization of the reporter was seen when either or was deleted (). However, in cells depleted of Ndc1p there was an ∼50% reduction in the number of cells that displayed a concentration of the NLS-mRFP signal in the nucleus (). In contrast to the Nup localization experiment, no significant additional defect was seen when Ndc1p depletion was combined with . We next examined the localization of the yeast-shutting protein Npl3p, as well as a mutant version, Npl3. Although Npl3p-GFP is exclusively nuclear at steady state, with no detectable cytoplasmic signal in wild-type cells (), the Npl3 variant partially mislocalizes to the cytoplasm because of the absence of Ser411 phosphorylation (; ). We monitored the localization of these proteins in each strain, and scored individual cells as being exclusively nuclear (100% N), predominantly nuclear with some cytoplasmic signal (N > C), or as having no nuclear accumulation (N = C; ). Although the absence of , or both had no significant effect on Npl3p localization, each of the four - strains displayed a marked increase in cells with cytoplasmic Npl3p-GFP (both N > C and N = C; ). Similarly, all four - strains displayed a significant increase in the percentage of cells with cytoplasmic (N = C) Npl3p (). In addition to protein import, we also monitored mRNA export by performing in situ hybridizations against the poly(A) tails using an oligo dT probe to monitor nuclear accumulation of poly(A) RNA ( ). As a positive control, we shifted - cells to the nonpermissive temperature for 30 min, which is known to cause a severe mRNA export defect. Although the probe was strongly concentrated in the nucleus in - cells, we did not see nuclear accumulation in any of our POM-deficient strains (). - cells. As we observed by fluorescence microscopy, both the wild-type and mutant cells had intact nuclear envelopes. NPCs in wild-type nuclei were identified by a characteristic gap in the nuclear envelope containing an electron-dense plaque structure (, arrow). - mutant nuclei also contained detectable pores in the nuclear envelope, although there appeared to be fewer pores in each thin section than in the wild-type cells. - mutant strain. Only ∼10% of the pores that we observed were comparable to wild type. The majority of the pores fell into the following two phenotypic classes: (a) pores that were roughly the same diameter as wild-type NPCs, but lacked any detectable electron-dense material (), and (b) pores that were approximately the same diameter as a wild-type NPC, but contained electron-dense material with an abnormal, less compact morphology (, arrow). Another 10% of pores had the most dramatic morphological defect, having a considerably larger diameter than a wild-type NPC and lacking any electron-dense material (). We measured the diameters of at least 55 NPCs from each strain, and found that wild-type NPC diameters fell into a narrow range, with a mean diameter of 59 nm and a maximum of 75 nm. However, mutant pore diameters displayed a broader distribution, with a mean diameter of 82 nm and a maximum diameter of 240 nm (). Interestingly, in the double-mutant cells we occasionally observed ribosomes in the nuclear compartment (), whereas ribosomes were strictly cytoplasmic in wild-type nuclei (). This suggests that the channel size in these cells has increased enough to allow diffusion of intact ribosomes through the pore. - cells lack many normal nuclear pore components and may no longer provide a fully functional diffusion barrier. To test whether molecules above the size threshold for diffusion through a wild-type NPC are excluded from the nucleus in the mutant cells, we used the following two reporter proteins: (a) the LexA DNA-binding domain fused to a nuclear export signal (NES) and GFP (LexA-NES-GFP), with a molecular mass of 51.5 kD, and (b) a version of Pho4p lacking its NLS (Pho4) fused to GFP (Pho4-GFP), with a molecular mass of 61.8 kD. Previous studies have shown that similar NES-containing constructs with molecular masses as low as 36 kD are unable to passively diffuse through the NPC in wild-type cells (). Both of the reporters used were excluded from the nucleus in 80–90% of wild-type cells (, A and B, top). - cells (, A and B, bottom), indicating that the pores in many POM-depleted cells have an increased limit of passive diffusion. - cells have larger pores that are devoid of any detectable proteinaceous material, suggesting that these pores lack a functional central channel. - cells, suggesting that the diffusion limit in POM-deficient cells is variable and is significantly higher than that in wild-type cells. The assembly of nuclear pore proteins into a functional complex that spans both bilayers of the nuclear envelope is a poorly understood and complicated process. We have characterized the role of the three transmembrane nuclear pore components, Ndc1p, Pom152p, and Pom34p, in interphase NPC assembly and function in yeast. The results presented in this study clearly demonstrate a critical role for Ndc1p in these processes, but also imply a partially redundant function of Pom152p in pore formation, as only the loss of both of these proteins caused complete Nup mislocalization in our assays. In contrast, we were unable to detect any additive effect caused by the loss of Pom34p, suggesting that Pom34p is dispensable for NPC assembly and function. Although the depletion of Ndc1p and Pom152p had an additive effect with regard to Nup mislocalization, the nuclear import defect that could be observed after Ndc1p depletion was not significantly exacerbated by the absence of . This implies that, although Ndc1p-depleted strains show only a partial Nup localization defect by fluorescence microscopy, the NPCs in these strains are already functionally compromised and, thus, display a phenotype in sensitive transport assays. For example, it is possible that although Ndc1p depletion causes only the partial loss of the three Nups that we monitored, Nup159p, Nup59p, and Nup60p, this could be sufficient to affect the diffusion limit of the central NPC channel. - cells displayed such dramatic Nup and import reporter localization defects, it was originally surprising that we did not observe any mRNA export defects in this strain. However, TEM analysis and nuclear exclusion experiments suggested that POM depletion results in a widening of the central channel and a breakdown of the normal diffusion barrier. Because normal protein import and defective mRNA export both result in an accumulation of substrate inside nuclei, we postulate that NPCs lacking Ndc1p and Pom152p allow the diffusion of molecules that normally rely on facilitated nuclear transport mechanisms. Thus, these molecules are unable to move against their concentration gradients, which is consistent with the observation that neither nuclear import reporters nor mRNA accumulate in the nucleus. In addition, new transcription of mRNA appears to be significantly reduced upon depletion of Ndc1p and Pom152p (unpublished data), further preventing a buildup of nuclear poly(A) RNA. A critical role for transmembrane Nups has now been demonstrated in both yeast NPC assembly (; this study) and postmitotic assembly in metazoans (; ; ). However, there have been somewhat conflicting results regarding the contributions of the integral membrane proteins GP210 and POM121 in NPC assembly. Interestingly, homologues of yeast Ndc1p have recently been described in humans and other metazoans () and may be critical for postmitotic NPC assembly in those organisms as well. Based on our data in yeast, which reveal partial redundancy between Ndc1p and Pom152p in NPC assembly, it is conceivable that the conflicting results pertaining to the metazoan POMs may be explained by a similar type of overlapping function between POM121, GP210, or metazoan Ndc1. Previous studies in extracts have also demonstrated important roles for the small GTPase Ran and the Nup107/160 complex in metazoan postmitotic assembly (; ; ; ,; ,). However, it has been unclear whether these same proteins are important for NPC assembly into intact nuclear envelopes, which occurs in nonmitotic metazoan cells and throughout the cell cycle in . Studies in budding yeast have shown that mutations in the genes encoding Ran and the Nup107/160 orthologue, which is the Nup84/85 complex, disrupt the NPC and, therefore, may be required for yeast NPC biogenesis (; ). Although NPC assembly has not been tested directly in these mutants, the intriguing similarities between metazoan postmitotic assembly and yeast NPC assembly may indicate that the processes occur by a conserved mechanism. Our results in yeast imply that Ndc1p plays an important structural role at the NPC, although Pom152p may partially compensate for the loss of this transmembrane protein. Although the detailed mechanistic role of transmembrane Nups remains unclear, multiple possible mechanisms by which transmembrane Nups could function in NPC assembly can be envisioned. For example, transmembrane Nups could remodel the two nuclear envelope membranes to induce membrane curvature, thus forming a prepore-type structure that is needed for NPC insertion. Furthermore, they could be required to anchor or target certain Nup subcomplexes to the nuclear envelope. The dramatic mislocalization of three spatially distinct Nups (Nup60p, Nup59p, and Nup159p) is consistent with a model in which Ndc1p and Pom152p act at an early stage of NPC assembly, possibly in the recruitment of soluble Nup subcomplexes to the nuclear envelope. - cells have pores that appear to be lacking the bulk of their protein content. - cells is increased, further strengthening the conclusion that POM depletion results in pores that lack a functional central NPC channel. Based on the data presented in this study, we suggest that Ndc1p, together with Pom152p, acts as an essential tether between the membrane and soluble components of the NPC. Yeast media and strain constructions were performed according to established protocols. All plasmids used in this study are listed in , and all yeast strains are listed in . To delete at its genomic locus in our strain background, PCR was performed using oligos (forward, GATGAACAAAAAGATTTATA; reverse, GCTAATCATATGTAAAATAT) and genomic DNA from the Yeast Deletion Consortium strain () as templates. The PCR product was transformed into KWY165 (W303 ) and integrated by homologous recombination. To delete at its genomic locus, PCR was performed using oligos (forward, GATAAACGGATTATAGATTTATCATACCAGATACGTTTATCAGGGCAGATCCGCTAGGGATAA; reverse, TATATTATACATTACAATTGTACAAAAATATTGCGGGAAAGAATTCGAGCTCGTTTAAAC) and plasmid pAF6-- () as templates. The PCR product was transformed into KWY390 (W303 ) and integrated by homologous recombination. -- strains were created by first fusing to the COOH terminus of using oligos (forward, GTTTCTAGAAGTGTACGCCTCAGGCAACCCTAATGCTACGCGGATCCCCGGGTTAATTAA; reverse, CCGGAAACGATAAAGGTAGCTTTTTGCTCTTTTTGCTCATGAATTCGAGCTCGTTTAAA) and the plasmid pFA6a--hisMX6 (Longtine et al., 1998) as templates. The PCR was transformed into KWY165 (W303 ) and integrated by homologous recombination. The promoter was integrated 5′ of the -coding region by linearizing the integration plasmid pKW1368 (-) with NheI. HDEL-dsRED strains were created by linearizing pKW1803 with EcoRV and integrating it at the TRP1 locus. To generate plasmid pKW1368 (pRS306--), the promoter was cloned into pRS306 as a NotI–EcoRI fragment, and the NH terminus of was PCR amplified from genomic DNA using oligos (forward, AAAGAATTCATGATACAGACGCCA; reverse, AAACTCGAGCGAAAGTGATGAAGA) and inserted as an EcoRI–XhoI fragment. To generate the plasmid pKW1552 (pRS315--), was PCR amplified as an EcoRI–NotI fragment to replace the triple from pKW1220 (pRS313--) to get pRS313--. This was subsequently subcloned into pRS315 () to change the marker. To generate plasmid pKW1706 (pRS425--), , along with its promoter, was PCR amplified from yeast genomic DNA and subcloned into pRS425 using XhoI and NaeI sites. was PCR amplified as a NaeI–DraIII fragment and subcloned at the 3′ end of 9. pKW1752 (pRS425--) was generated by PCR amplifying and its promoter from yeast genomic DNA and subcloning it into pKW1706 using NaeI–DraIII, thus, replacing with . The plasmid pKW1219 (pRS425--) was created by replacing the from pKW674 (pRS426----) with mRFP. This construct was then subcloned into pRS425 using XhoI–SacI. To generate pKW1783 (pRS425--), , along with its promoter, was amplified from genomic DNA as a SacI–BamHI fragment and cloned into pKW1219, replacing the promoter and NLS. pKW1803 (YIPlac---) was generated from YIPlac-- (a gift from C. Reinke and B. Glick, University of Chicago, Chicago, IL) by cloning in the NatMX cassette as a HindII–SapI fragment. To generate pKW409 (pRS314---), a SacI–KpnI PCR product was amplified from pEG202 () and cloned into pRS314. The NES from PKI was inserted an EcoRI–BamHI fragment by annealing the oligos (forward, AATTCAATGAATTAGCCTTGAAATTAGCAGGTCTTGATATCAACAAGACAGG; reverse, GATCCCTGTCTTGTTGATATCAAGACCTGCTAATTTCAAGGCTAATTCATTG). Finally, GFP(S65T) was inserted as a BamHI–XhoI fragment. Plasmid pKW1898 (pRS315- -) was constructed from EBO383 (a gift from E. O'Shea, Harvard University, Cambridge, MA; ) to change the marker from URA3 to LEU2. A SalI–BamHI fragment was cut out of EBO383 and ligated into pRS315. Plasmids pKW551 and pKW776 were a gift from C. Guthrie, University of California, San Francisco, San Francisco, CA. Cells were grown at 30°C in galactose media overnight, washed, and diluted to OD = 0.1 in dextrose media. Cells were then grown at 30°C in dextrose for 24 h to deplete - cells of Ndc1p. The efficiency of Ndc1p depletion was verified by the visualization of Ndc1p-GFP by fluorescence microscopy. Cells were grown under the conditions indicated in the text and figure legends, and embedded in 0.5% agarose for visualization. Images were acquired with a digital camera (model CA742-98; Hamamatsu) controlled by the Metamorph software program (Universal Imaging Corp.). Images were processed using Photoshop CS (Adobe), and figures were assembled using Photoshop CS and Illustrator CS (Adobe). DNA was stained in live cells by adding either 2.5 μg/ml of DAPI (Sigma-Aldrich) or 5 μg/ml Hoechst dye (Sigma-Aldrich) directly to cell culture and growing for 30 min before visualization. The vital dyes Fun1 (Invitrogen) or MitoTracker red (Invitrogen) were added directly to cells in growth media at final concentrations of 4 μM and 100 ng/ml, respectively. Cells were grown for 30 min and then visualized by fluorescence microscopy. As a negative control, wild-type cells were treated with 10 mM sodium azide and 10 mM 2-deoxyglucose for 30 min to stop respiration, and were then dyed and visualized the same as the cells in growth media. Cells were grown for 24 h in dextrose before being fixed in paraformaldehyde and permeabilized with zymolyase T100 (MP Biomedicals). poly(A) RNA was detected by hybridization of a digoxigenin-labeled oligo (dT) probe and a rhodamine-labeled anti–mouse antibody, as described previously (). DNA was stained with DAPI (Sigma-Aldrich). The known mRNA export mutant - was used as a positive control, and was shifted to 37°C for 30 min before fixation. Yeast cultures were normalized by OD, and protein samples were prepared by spinning down the cells and resuspending them in 0.1 M NaOH for 5 min before spinning down and resuspending them in 1× SDS loading buffer. Protein samples were loaded onto a 10% polyacrylamide gel and separated by SDS-PAGE. Proteins were transferred to nitrocellulose membranes, which were cut in half, and the top half was blotted with a mouse monoclonal anti-GFP antibody (Roche) to visualize Ndc1p-GFP and either Nup159p-CFP, Nup59p-CFP, or Nup60p-CFP. The bottom half was blotted with a rabbit polyclonal anti-Dhh1 antibody () as a loading control. For Fig. S1, protein levels were quantified using ImageJ software (National Institutes of Health), and a ratio of GFP/Dhh1 signals was calculated. This ratio was normalized to the ratio in the wild-type strain. High pressure freezing of cells followed a protocol similar to that explained by . In brief, cells were harvested from liquid cultures by vacuum filtration onto filters (Millipore), transferred to sample chambers, and frozen with a high pressure freezer (model EM PACT2; Leica). Frozen samples were stored in liquid N2 and transferred to a freeze substitution apparatus (model AFS2; Leica) and fixed with 0.2% OsO and 0.05% uranyl acetate. Cells were then infiltrated and embedded with Epon resin. Serial sections were cut at 60-nm thick using an Ultracut E (Reichert). Sections were then picked up on formvar-coated grids and stained with 2% uranyl acetate in 70% methanol and Reynolds lead citrate. Cells were then observed and imaged on an electron microscope (model 1200 CX; JEOL) operating at 80 kV. Fig. S1 shows that Nup159p and Nup60p proteins are expressed in all POM-depleted strains, as determined by Western blotting. Online supplemental material is available at jcb.200506199/DC1.

Neuronal components are primarily synthesized in the cell body (; ) and move to supply the growing axon by the classic behaviors of fast and slow axonal transport (). In most neurons, it is thought that cytoskeletal proteins and organelles are transported by kinesins and dynein along a stationary cytoskeleton (; ) and that elongation occurs by tip growth (; for review see ). Although there have been numerous studies, both axonal transport and axonal growth () have been controversial for decades, in part because of significant differences in what seem to be direct observations. Slow axonal transport was classically envisioned as the movement of a coherent column from the cell body to the growth cone in a process described as axoplasmic flow (). Today, it is generally accepted that the framework of the cytoskeleton is stationary in cultured chicken dorsal root ganglion (DRG) neurons, as well as in vivo for developing zebrafish and grasshopper neurons (; ). Although an early study suggested that the coherent movement of microtubules may occur in the neurites of PC12 cells (), later work indicated that the framework of the cytoskeleton was also stationary (). In all but one case, isolated polymers or soluble subunits and organelles are now thought to move along a stationary framework to their site of addition at the growth cone (; ; ). In neurons, it is agreed that cytoskeletal framework markers move toward the growth cone, but why these neurons seem to have a unique mechanism for axonal transport has been a vexing question (; ). Although the reported differences between the outgrowth of neurons and all other neurons may be real, they may also reflect limitations of methodology. Photobleaching and photoactivation studies are limited by marked regions that dissipate because cytoskeletal polymers are dynamic (). Thus, over prolonged periods of time, uncertainty arises in regard to the position of the mark. Although it has been convincingly demonstrated that marked regions in non– neurons do not translocate at the rate of growth cone advance (), it is possible that en bloc anterograde movements in all other axons have been overlooked. A more tractable approach to understanding axonal transport is to focus on a component that can be tracked completely. Thus, we have concentrated on mitochondria because they are easily labeled and monitored and move by fast axonal transport, but also form stable associations with the axonal framework through docking (; ; ). Mitochondria are critical for neuronal function, and their possible involvement in Alzheimer's disease (), Huntington's disease (), and Parkinson's disease () makes their study clinically relevant. Along the axon, mitochondria alternate between being rapidly transported by motors with instantaneous velocities between ∼0.1 and 2 μm/s and docked states where the mitochondria bind actin filaments, neurofilaments, and microtubules (; ; ). Our previous study showed that fast-transported mitochondria stop (dock) along the axon (). This accounts for the decrease in fast mitochondrial transport along the axon (), but begs the question of how mitochondria are added to the newly elaborated axonal tip. Although there has been much work on axonal growth, no study has directly observed total transport along the full length of the axon during axonal elongation. To test the current models, we undertook a global study of mitochondrial transport in cultured chicken DRG neurons. Surprisingly, we found that the distal axon was continually being pulled forward, even when the growth cone was stationary, indicating that the axon was under tension that was independent of growth cone movement. These findings indicate that axonal growth in DRG neurons does not occur by tip growth, but rather by axonal stretching caused by a force-generating mechanism in the growth cone, coupled with intercalated addition through fast transport. We investigated mitochondrial transport in chicken DRG neurons grown on polyornithine/laminin-coated coverslips. Our procedure was designed to routinely follow transport and outgrowth for up to 1 h, with an image acquisition rate of 0.5 Hz, while minimizing photodamage (). To facilitate the long term visualization of mitochondrial transport, we (a) reduced the intensity of light used for illumination by using neutral density filters (25 or 50%) and switching the laser to standby mode during observations, (b) used a minimal incubation time and concentration of the MitoTracker CMX-Ros dye (1 min and 100 nM), (c) allowed an interval of several hours (2–4 h) between the time of dye incubation and observation, and (d) observed the neurons in a closed flow chamber to eliminate the accumulation of free radicals (). Images obtained in this way were assembled into kymographs to reveal en bloc movements that might have been overlooked or discarded by conventional particle-tracking routines or by the photobleaching and photoactivation techniques (; and Video 1, available at ). Although the image-capture protocol lowered the optical resolution in regard to submicron mitochondrion morphology, it permits the acquisition of image sequences with excellent temporal resolution and duration. For our global survey, we analyzed docked mitochondria for movement in the proximal (the first 150 μm continuous with the cell body), middle (the region where neither cell body nor growth cone were in view), and distal axon (the last 150 μm that includes the growth cone). Data were acquired from 2–3-d-old neurons in which the total axonal length exceeded 500 μm. In , a total of 19 separate kymographs, each from the axons of different neurons, are assembled to give an overview of the movements of docked mitochondria along the axon. We separated our analysis of observations of the growth cone into periods where the growth cone was either moving forward in the distal region () or paused () to determine if there was a difference in the pattern of mitochondrial transport. We found that docked mitochondria move anterogradely, with a low velocity that varies along the axon (). Taking the mean velocity of transport for the three regions of the axon revealed a zero velocity in the proximal axon (−2 ± 14 μm/h; mean ± SD; = 113 mitochondria analyzed in eight neurons), an intermediate velocity in the middle of the axon (12 ± 19 μm/h; mean ± SD; = 123 mitochondria analyzed in nine neurons), and the highest velocity in the distal axon (43 ± 26 μm/h; mean ± SD; = 80 mitochondria analyzed in seven neurons; ). Comparing the mean velocity of transport in the proximal and middle regions of the axon (−2 ± 14 vs. 12 ± 19 μm/h; mean ± SD; = 113 and 123 observations of mitochondrion movement; P < 0.0001, two-tailed test) and the middle and the distal axon during elongation (12 ± 19 vs. 43 ± 26 μm/h; mean ± SD; = 123 and 80 observations of mitochondrion movement; P < 0.0001, two-tailed test) revealed that the low velocity movements of the mitochondria significantly increased in rate along the length of the axon. If low velocity movement of docked mitochondria occurred during growth cone pauses, it would indicate that it occurred as the result of a transport process, instead of axonal stretching through growth cone advance. We found that, in the distal axon, low velocity transport (LVT) continued unabated during pauses (, F and H; 42 μm/h ± 23; mean ± SD; = 65 mitochondria from nine different neurons for the movement of docked mitochondria during growth cone pauses vs. 43 ± 26 μm/h; mean ± SD; = 80 mitochondria analyzed in seven neurons for the movement of docked mitochondria during growth cone advance; P = 0.66 from a two-tailed test comparing the mean velocity of transport over the last 120 μm of the distal axon for advancing and paused growth cones, but excluding the last 10 μm of the axon for paused growth cones). Thus, the low velocity movements of the mitochondria occur through a transport mechanism that is independent of growth cone advance. Although LVT was independent of growth cone advance ), it raised the question of whether the movements of the docked mitochondria reflected the coherent movement of the axonal framework or were moving via an asynchronous transport mechanism powered by motors moving individual cargoes in a stop-and-go manner. For example, the apparent low velocity movements could result from the coherent movement of the underlying axonal framework (; ). Alternatively, they could result from asynchronous mitochondrion transport by an unidentified slow kinesin or myosin, or from interaction with cytoskeletal filaments (; ; ) undergoing brief, short-distance stop-and-go transport (; ). To differentiate between these two modes, we analyzed the correlation between the changes in mitochondrial position over time (i.e., the first derivative) along the axon (). A high correlation would suggest movement of the axonal framework, but the lack of a correlation would suggest asynchronous transport. This test was more stringent than a simple correlation because it tested whether small back-and-forth movements were linked (), rather than occurring asynchronously, yet in the same overall direction. Mitochondrial positions from thresholded montages were determined in ImageJ using the Analyze Particles function. The coordinates of the positions were then smoothed using a running 30-s average (). The change in position over 1-min intervals was calculated () to examine the correlation between the movements of the mitochondria with each other (). Movements of mitochondria along the axon shaft were highly correlated (r = 0.51; P < 0.0001; = 62; five separate pairs of mitochondria were analyzed; for a representative example). Surprisingly, there was a much smaller correlation between the movement of mitochondrion in the growth cone and the axonal mitochondrion (r = 0.08; P = 0.03; = 62; ). This low correlation appeared to occur because LVT of the mitochondria along the axon persisted during growth cone pauses, yet phases of growth cone advance occurred more rapidly than LVT. The last 10 min of the kymograph in and Video 2 (available at ) demonstrate transport during a growth cone pause, and the time period between 35 and 50 min shows a rapid advance of the growth cone that is faster than LVT. These results demonstrate that LVT of docked mitochondria occurs because they interact with a coherent framework that is in motion, but that this motion is not tightly linked to growth cone motility. To test if stretching similar to that reported in neurons occurred, we examined how the distance between pairs of mitochondria changed over time. We analyzed the distances between 166 consecutive pairs of docked mitochondria from nine different axons over 10–15 min time periods in the distal axons where the growth cone was elongating. Using a two-tailed test between the observed mean and a hypothesized population with a mean of zero (), we found that the spacing between the mitochondria over the last 150 μm of the axon increased significantly, at a rate of 0.8 ± 0.7% per minute (mean ± SD; P < 0.005, against the null hypothesis that the distance between the mitochondria is constant). The SD was large because there was a variation in displacement along the axon (see example in and Video 2). In the proximal axon, we found that the rate of spacing change was 0.004 ± 0.025% per minute ( = 138 consecutive pairs from 11 different neurons). This was not significantly different from a 0% rate of change. These results suggested that low velocity mitochondrial transport occurred by a process similar to axonal stretching reported in neurons, but differed in that it only occurred distally (; ). Although we found clear evidence for LVT of docked mitochondria during growth cone pauses over 10–15 min time intervals, this movement could have been caused by a passive, time-delayed elastic property of the axon that abated over time (). Although in most cases pauses were relatively brief, in three examples we observed growth cones that remained paused for >30 min (see for individual frames from a time-lapse movie that lasted 50 min, with the accompanying kymograph in , and Video 3, available at ). In this example, LVT continued after the growth cone paused, and even increased in velocity (a similar example of this type of increase was found in the last 10 min of the kymograph in ). This provided strong and direct evidence that low velocity mitochondrial transport in the absence of growth cone advance was not a passive, time-delayed viscoelastic response caused by the advance of the growth cone, but a transport process. Fast transport has been considered the sole mechanism for moving mitochondria along the axon. Our findings suggest that it works in concert with LVT. To determine the pattern and flux of fast mitochondrial transport, we measured the flux by counting the mitochondria per hour that passed a single point in the center of kymographs from the proximal ( = 132 mitochondria from 10 neurons), middle ( = 102 mitochondria from nine neurons), and distal ( = 91 mitochondria from 13 neurons; pooled data; see next paragraph for advancing vs. paused growth cone breakdown) regions, at a rate of >0.2 μm/s (). Significant decreases in flux occurred between the proximal and distal regions for anterograde, retrograde, and net transport (). Globally, the number of fast-transported mitochondria was highest in the proximal axon and decreased significantly in the distal axon. When we examined fast transport during growth cone pauses, we found no significant differences in the flux in the distal axon between periods of axonal elongation and growth cone pauses for either anterograde (22 ± 16 vs. 18 ± 15 mitochondria/h; mean ± SD; = 7 mitochondria from six neurons; P = 0.39) or retrograde fast transport (7 ± 5 vs. 9 ± 5 mitochondria/h; mean ± SD; = 7 mitochondria from six neurons; P = 0.43). Although changes in fast mitochondrial transport have been linked to changes in axonal outgrowth (), our findings suggest fast mitochondrial transport is handled differentially during brief pauses seen during normal outgrowth. These results indicated that fast mitochondrial flux decreased along the axon, but raised the question of the relative contributions of fast flux and LVT to total mitochondrial flux. To determine the relative contributions of fast flux and LVT, we converted velocity to flux (mitochondria/h) by multiplying the rate of LVT (μm/h) by mitochondrial density (mitochondria/μm). Mitochondrial density was determined by counting the total mitochondrial number along measured lengths in the proximal, middle, and distal axon (). In each region, 10 axons were analyzed, and 849 mitochondria were counted in total. The densities of the mitochondria in the proximal, middle, and distal axon were 0.34 ± 0.11 (mean ± SD), 0.22 ± 0.07, and 0.27 ± 0.10 mitochondria/μm, respectively. Along the axon, our analysis demonstrated that flux from fast transport decreased, flux from LVT increased, and net flux was relatively constant (). Although this analysis did not allow an absolute measure of the contributions of the different modes of transport at each point along the axon, it provided a comparison that strongly suggested a model where fast transport and LVT worked together to transport mitochondria at a constant flux during axonal elongation (). These observations show that docked mitochondria in the distal axon slowly move anterogradely by a LVT process, and that the velocity increases as the mitochondria approach the growth cone (). This transport occurs both during periods of axonal elaboration and during growth cone pauses ( and ). Correlation analysis indicates that LVT occurs because docked mitochondria interact with a coherent axonal framework that is slowly transported toward the growth cone (). These movements are linked to increases in the intermitochondrial distance in the distal axon, which suggests that low velocity mitochondrial transport occurs through a conserved mechanism similar to that seen in neurons (). Analysis demonstrates that LVT, in combination with fast axonal transport, generates a constant flux along the axon during elongation (, , and ). We suggest that the axonal framework is transported at a low velocity in the distal axons of cultured DRG neurons by a force-generating mechanism that is independent of growth cone advance. In neurons, marks on microtubules and granules attached to the surface of the axon translocate coherently with a velocity profile that increases along the axon (; ). Because docked mitochondria are coherently transported along the axon with an increasing velocity profile that is distally associated with increases in the distances between pairs of docked mitochondria (– ), we suggest that frog and chick neurons have similar mechanisms for axonal outgrowth. These results are consistent with stretching of the axonal shaft, but because they occur independently of growth cone advance they reflect transport ( and ). Although this is en bloc transport of the cytoskeletal framework () and may contribute to slow axonal transport because it does not occur in the proximal axon (; ), it cannot entirely account for slow axonal transport (). To convey the notion that the movement we observe reflects transport, but is not slow axonal transport, we suggested calling the phenomena LVT. Although our conclusions differ from those of previous work, those studies focused on the middle or proximal axon, and did not rule out distal LVT. For example, demonstrated conclusively that photobleached marks in the proximal axon of cultured rat DRG neurons were stationary. demonstrated that axonal towing did not pull mitochondria from the cell bodies of neurons. We previously reported () that labeled microtubules in the cell body were not transported into axons. demonstrated that photobleached and activated marks along rat DRG axons did not advance at a speed equal to the rate of growth cone advance. Thus, this study and previous works differ in their conclusions, but their results are in agreement. Our results are also consistent with earlier findings that tension leads to en bloc viscoelastic displacement of the internal cytoskeletal framework (). When ligand-coated beads are held on the surface of stationary growth cones, internally generated tension results in viscoelastic displacement of organelles and microtubules (). When this tension is experimentally relaxed, the initial surge of microtubules and organelles recover their original position (). Our results extend these observations in two ways. First, they demonstrate that a force-generating mechanism leads to movements not only in the growth cone but also along the axonal framework (– ; ). Second, they show that force generation in the growth cone occurs during both growth cone advances and pauses ( and ; ). Force generation in the growth cone has been widely attributed to a myosin/actin-dependent process (; ). Although a myosin/actin-linked clutch may be engaged during axonal elaboration and disengaged during growth cone pauses (; ; ; ), we suggest the clutch for LVT is always activated during growth cone–mediated axonal elongation. This is consistent with the observation that myosin/actin-dependent activities are important for force generation, but are not absolutely required for axonal elaboration (; ; ; ). Although we have focused on mitochondria, based on the correlation analysis () it seems likely that the phenomena of LVT includes the movement of a coherent cytoskeletal framework. The observation of LVT in the distal, but not the proximal, axon suggests that forces generated in the growth cone are dissipated along the axon through transmembrane contacts with the substrate (). Furthermore, LVT velocity would be expected to increase distally simply because there is more space/opportunity for material to have been added behind increasingly distal markers (). Although this intuitively suggests a pushing mechanism, previous evidence indicates that application of tension at the growth cone leads to axonal lengthening and microtubule assembly (; ; ). If microtubule assembly along the axon drove LVT by pushing forces, it would be expected that the induction of microtubule assembly would lead to a lengthening of the axon. Instead, it is found that microtubule assembly by treatment with taxol slows elongation and increases axonal caliber (; ). Thus, we suggest that viscoelastic stretching, instead of pushing by motors or assembly, drives axonal elongation (). If stretching of the axon occurred without mass addition, the axon would thin and eventually break. Although we did not measure axonal diameter during elongation, two recent studies suggest that substantial lengthening without thinning occurs in response to tension-induced elongation. In one experiment, the internodal spacing between the nodes of Ranvier in rat sciatic nerve was measured after orthopedic leg-lengthening procedures. They found that doubling the length of the nerve doubled the internodal distance without a significant decrease in axonal diameter (). In the second experiment, tension-induced lengthening of the axon at rates of up to 8 mm/day for 20 d occurred with a concomitant 35% increase in axonal diameter (). Although older work suggested that axonal stretching in neurons is not real growth because the axonal diameter thins (; ), these more recent studies (; ) suggest that tension-induced lengthening of the axon is compensated over longer times by mass addition along the axon. Based on observations of mitochondrial docking along the axon (; ) and the proximal to distal decline in fast mitochondria transport (; ), we suggest that new mitochondria are added by docking () to compensate for gaps created by axonal stretching. Thus, we suggest tension generated at the growth cone leads to axonal elongation and then intercalated mass addition (). Our calculations indicate that there is a constant total flux of mitochondria along the axon, which may be important for maintaining a constant mitochondrial distribution over space and time during elongation. When the rate of flux varies, we suggest that addition occurs where flux slows () and depletion occurs where flux increases. The simultaneous large increase in LVT associated with stretching and the drop in fast mitochondria flux in the distal axon () is consistent with this region as the primary site of viscoelastic axonal stretching/creep and intercalated addition. As the pool of fast-transported mitochondria moves along the axon, flux would be expected to decrease because mitochondria dock in the gaps created during LVT and to replace damaged mitochondria (; ; ). This model of axonal growth () simultaneously explains the increasing flux profile of LVT ( and ) and the decrease in fast anterograde mitochondrial transport ( and ). As we and others have observed, axonal elaboration occurs by a process where components that have accumulated in the growth cone are consolidated in the axonal shaft ( and Video 2; ). Based on our observations that mitochondria undergoing LVT slow down as they enter the growth cone during growth cone pauses ( and ), we speculate that the motor complex generates pulling forces on the axonal shaft and pushing or contractile forces in the growth cone. Over time, the process of elaboration appears to involve periods of protrusion and engorgement, as is generally accepted (); however, our work indicates that both fast axonal transport and LVT move cytosolic components into the growth cone. As a speculative hypothesis for axonal elongation in terms of LVT, we propose that tension generation at the growth cone, with or without advance, leads to low velocity mitochondrial transport through viscoelastic stretching/creep of the axonal framework. The axonal thinning caused by viscoelastic stretching is then compensated for by intercalated addition of mitochondria and other cytosolic components (). In summary, in most growing neurons it is thought that the axonal framework is stationary. Our data show that in growing chick DRG neurons the axonal framework undergoes a LVT. This transport increases in velocity along the axon, is associated with viscoelastic stretching, and is independent of growth cone movement. Furthermore, low velocity stretching and fast axonal transport are reciprocally related such that there is a constant flux of mitochondria along a growing axon. We suggest that axonal growth occurs by stretching of the axon through a LVT that is independent of growth cone advance, followed by the intercalated addition of material delivered by fast axonal transport. The methods and equipment used were previously described (). Images were acquired on a confocal system (FV300 FluoView; Olympus) with built-in photomultiplier tubes on a microscope body (model IX70; Olympus) using the FluoView acquisition software, and a 60×, 1.4 NA, PlanApo objective (Olympus). Samples were illuminated with an argon/krypton Omnichrome 488/568/647 line laser set to standby mode, with the laser intensity set at 6%, the confocal aperture set at 5, and pixel dwell times of either 100 or 200 μs; the light was further attenuated with neutral density filters (either 25 or 50%). Cells were observed in an enclosed flow chamber in DME with high glucose, -glutamine, and pyridoxine HCl, and without sodium pyruvate (Invitrogen) that was then supplemented with 10% fetal calf serum, 10 ng/ml nerve growth factor (Sigma-Aldrich), and 1 mM sodium pyruvate. Temperature was maintained at 37°C using a forced air heater. Embryonic day 14 chicken DRG neurons grown on poly--ornithine/laminin coated glass coverslips were cultured and used either 2 or 3 d after plating. Mitochondria were labeled with 0.1 μM MitoTracker red CMX-Ros (Invitrogen) in DME for 1 min, and then allowed to recover for at least 2 h in fresh media. For the creation of kymographs, FluoView files were opened in ImageJ software (National Institutes of Health [NIH]) as stacks, rotated so that the axons in the images were horizontal, cropped, and resliced to switch the time axis and the y axis. These stacks were then z-projected using the Standard Deviation option, which resulted in 32-bit images, with time located on the y axis and distance on the x axis. The kymographs were then converted to 16 bits and saved. For analysis of low velocity movements (), raw kymographs were opened in Photoshop (Adobe), levels were adjusted to maximize bit depth, and images were converted to 8 bits, inverted in color, compressed fivefold on the time axis, thresholded using the Filter/Sketch/Photocopy function, and thresholded using the standard function (Image:Adjustments:Threshold). Compression of the image on the time axis, followed by thresholding, erased most of the fast-transport mitochondria. Any traces of fast-transported mitochondria and noise were erased by hand so that each docked mitochondria was an individual trace (, inset). These images were overlaid with the original kymographs to verify that they accurately reflected the low velocity movements of the mitochondria. The images were saved as uncompressed 8-bit tiff files. For the analysis of velocity, images were opened in ImageJ or NIH image. To determine the angle (velocity) and position of each mitochondrial trace, we used the analyze particle function (Analyze:Analyze Particle). The results were exported to Excel (Microsoft) and converted to velocities. For the analysis of fast mitochondrial transport, a single position in the center of the axon was picked and a line was drawn along the time axis of unfiltered unthresholded kymographs (). For a movement to be counted, the mitochondria were required to fully pass the line. The size of the mitochondria was not considered in the flux calculation. Each kymograph was from a separate neuron and was limited to the first 15 min of observation. Mitochondrial density was calculated from individual frames of time-lapse movies and included both fast-transported and docked mitochondria. Video 1 is a DIC and fluorescent observation of mitochondrial transport during axonal elongation of cultured DRG neurons. Video 2 shows mitochondrial transport during axonal elongation. Video 3 shows mitochondrial transport during a growth cone pause. Online supplemental material is available at .

Cell locomotion and adhesion play key roles during embryonic development, tissue regeneration, immune responses, and wound healing in multicellular organisms. Cell migration, changes in cells' shape, and adhesive properties are regulated by continuous remodeling of the actin cytoskeleton. Although multicellular organisms contain a wide array of actin filament assemblies, the actin structures that play fundamental roles in cell migration can be roughly divided into three categories: lamellipodial actin network at the leading edge of the cell, unipolar filopodial bundles beneath the plasma membrane, and contractile actin stress fibers in the cytoplasm (for review see ). The lamellipodium contains a network of short, branched actin filaments that produce the physical force for protrusion of the leading edge. The formation of new actin filaments at the leading edge is promoted by the Arp2/3 complex, which nucleates new filaments from the sides of preexisting filaments and thus induces the formation of a branched filament network (; ). The elongation of newly nucleated filaments is subsequently inhibited by capping proteins to maintain short, stiff filaments as well as to concentrate polymerization to the protruding region close to the plasma membrane (for reviews see ; ). Filopodia are thin cellular processes containing long parallel actin filaments arranged into tight bundles. Recent studies have demonstrated that filopodia are initiated from the dendritic lamellipodial actin network by uncapping and subsequent elongation of subsets of privileged barbed ends (). Ena/VASP family proteins and formins appear to play a central role in uncapping and elongation of filopodial actin bundles (; ). In contrast to relatively well characterized lamellipodia and filopodia, the assembly mechanisms of actin stress fibers are still poorly understood. Stress fibers are contractile actomyosin bundles, which are essential for cell adhesion to the substratum and for changes in cell morphology, specifically the retraction of the trailing edge (“tail”) during migration. Stress fibers are composed of relatively short actin filaments with alternating polarity (). These filaments are cross-linked by α-actinin and possibly also by other actin-bundling proteins. α-Actinin and myosin display periodic distribution along stress fibers typical also for other types of contractile structures, such as myofibrils of muscle cells. Animal cells contain at least three different categories of stress fibers: ventral stress fibers, transverse arcs, and dorsal stress fibers. Ventral stress fibers are contractile actin filament bundles that are typically associated at both their ends to focal adhesions. These structures are located at the ventral surface of the cell and play an important role in cell adhesion and contraction. Transverse arcs are curved actomyosin bundles that are not directly associated to focal adhesions at their ends. In motile cells, transverse arcs show typical flow from the leading edge toward the cell center. Dorsal stress fibers are actin bundles that insert into focal adhesions at the ventral cell surface, rise toward the dorsal section of the cell, and often terminate to a transverse arc at their proximal ends (; ; ). Stress fiber assembly is regulated by a signaling cascade involving the RhoA small GTPase (). The GTP bound form of RhoA activates Rho-associated kinase, which in turn promotes stress fiber formation by inhibiting actin filament depolymerization (through inactivation of actin depolymerizing factor/cofilins via LIM kinase) and by inducing contractility (through phosphorylation of myosin light chains [MLCs]). In addition, RhoA directly activates formins, which have been proposed to induce actin assembly during stress fiber formation (for review see ). However, the actin filament assembly pathways promoting the formation of three types of stress fibers and the mechanism of myosin incorporation into these structures has not been determined. We applied multicolor live cell microscopy and FRAP methods to explore how actin stress fibers are assembled in cells and show that actin filaments for stress fibers are derived from two different sources. For multicolor live cell microscopy of stress fiber assembly, we used previously characterized GFP fusion proteins of central stress fiber components: actin (), α-actinin (), and myosin II regulatory light chain () as well as a focal adhesion marker, zyxin (; see Materials and methods). These four proteins were also fused with the GFP spectral variants YFP and CFP to simultaneously examine the distribution and dynamics of two different fusion proteins during stress fiber assembly. To avoid possible overexpression artifacts, only cells displaying an intact actin cytoskeleton and expressing the lowest detectable amounts of fusion proteins were chosen for further analysis. It is also important to note that localizations of GFP fusions of actin, α-actinin, zyxin, and MLC were identical as compared with the subcellular localizations of endogenous proteins (unpublished data). The analysis of stress fiber assembly has been hampered by the fact that most nonmotile cell types contain thick, nondynamic stress fibers, whereas most motile cell types contain very few and thin stress fibers and are thus not suitable for live cell microscopy analysis. Therefore, we first screened several cell lines by phalloidin staining to visualize the actin cytoskeletons in fixed cells as well as by expressing GFP-actin to examine the dynamics of these actin structures. U2OS human osteosarcoma cell line was chosen for further analysis because these cells displayed thick stress fibers that were also relatively dynamic based on GFP-actin distribution in live cells ( and Video 1, available at ). We examined the assembly mechanisms of the three types of stress fibers in cultured U2OS cells by live cell microscopy methods and determined the dynamics of the most central stress fiber components by FRAP. Our analyses revealed the following: Dorsal stress fibers are generated through formin (mDia1/DRF1)–driven actin polymerization at focal adhesions. Myosin bundles can be subsequently incorporated into the α-actinin cross-linked actin bundles of dorsal stress fibers, although this incorporation typically occurs only after dorsal stress fibers have associated with transverse arcs at their proximal ends. In contrast to dorsal stress fibers, transverse arcs form by the end-to-end annealing of cortical Arp2/3-nucleated actin bundles and myosin bundles. Experiments using blebbistatin demonstrated a critical role of myosin II activity in the transverse arc assembly and maintenance. Ventral stress fibers, which are associated with focal adhesions at both ends, can be generated from the preexisting transverse arc/dorsal stress fiber network in U2OS cells. The latter also provides an explanation for how contractile actin bundles that are attached to substratum at their two ends can be generated in cells. Models for dorsal stress fiber, transverse arc, and ventral stress fiber assembly based on the data obtained here are presented in . However, it is important to note that in addition to the model presented in , ventral stress fibers may also be generated by other mechanisms, such as the annealing of short focal adhesion–attached actin bundles (). This is also supported by our observations showing that in Arp2/3 knockdown cells, two dorsal stress fibers growing from opposite sides of the lamella could fuse with each other to form a ventral stress fiber. Together, our data provide a plausible explanation for the previous inconsistencies in the mechanism of stress fiber assembly. Earlier live cell microscopy analyses provided evidence that the initial site of stress fiber assembly involves discrete spots near the cell edge, followed by a unidirectional growth/stretching of actin bundles from this site (). In contrast, other studies suggested that stress fibers are generated from preexisting short actin bundles, which assemble together to generate longer, contractile structures (; ). Our model proposing two distinct assembly mechanisms of stress fibers, formin-driven assembly at focal adhesions during dorsal stress fiber generation and endwise annealing of short myosin bundles and Arp2/3-nucleated actin bundles during transverse arc formation is in good agreement with these earlier, seemingly contradictory observations. Furthermore, our FRAP analysis provided direct evidence that dorsal stress fibers elongate through actin polymerization at focal adhesions and not through insertion of short actin filaments at the proximal ends of these structures as previously suggested (; ). Analysis of stress fiber component dynamics by FRAP demonstrated that actin filaments are relatively stable in all three types of actin bundles. However, our previous analyses demonstrated that stress fibers became more prominent and their turnover rates decreased when central regulators of actin filament depolymerization, actin depolymerizing factor/cofilin or cyclase-associated protein, were depleted from cells (; ). This observation suggested that, despite slow turnover rates, the association and dissociation of actin monomers to and from stress fibers is promoted by actin binding proteins. Interestingly, the association of α-actinin with stress fibers is highly dynamic. We propose that a dynamic cross-linking of actin filaments during the formation of stress fibers and in mature stress fibers may be essential for the following reasons: First, the dynamic association of α-actinin may be necessary for the incorporation of myosin into dorsal stress fibers. Rapid exchange of this actin cross-linker would allow myosin motor domains to interact with actin filaments because myosin and α-actinin binding to actin filaments are mutually exclusive. Second, dynamic cross-linking could allow rotation of actin filaments to convert a unipolar noncontractile structure to a bipolar actin structure, which, together with a myosin bundle, is capable of contracting. Consistent with this hypothesis, it was recently shown that expression of a mutant α-actinin with defects in PIP binding and slower association/dissociation rate to actin filaments in vivo results in the formation of abnormal stress fibers (). Finally, it is proposed that during stress fiber contraction, α-actinin must be progressively displaced to allow the movement of myosin toward the barbed ends on α-actinin cross-linked filaments (). Therefore, we suggest that dynamic cross-linking of actin filaments is also important for the contractility of stress fibers. In addition to stress fibers, multicellular organisms display other types of contractile actin filament bundles, such as myofibrils in muscle cells. The assembly of myofibrils is generated through intermediates, namely, premyofibrils and nascent myofibrils. Premyofibrils consist of actin, α-actinin, and nonmuscle myosin II and, interestingly, the assembly of premyofibrils visualized with α-actinin–GFP closely resembled the assembly of transverse arcs described in this study. Similar to transverse arcs, formation of premyofibrils begins at the spreading edges of the myocytes (). We speculate that a molecular assembly mechanism similar to the transverse arc assembly pathway revealed here may also contribute to the formation of the periodic contractile actin structures of muscle cells. Together, these studies form a framework explaining how contractile actin stress fibers are assembled in cultured U2OS cells. In the future, it will be important to examine whether stress fibers in other cell types are generated through similar or different mechanisms. Recent studies have also shown that mechanical stretching induces changes in the morphology and composition of stress fibers (; ). Therefore, how actin polymerization in focal adhesions and myosin incorporation into actin filament bundles are regulated during maturation and stretching of a ventral stress fiber provides important challenges for future research. Human osteosarcoma (U2OS) cells (a gift from T. Mäkelä and T. Vallenius, University of Helsinki, Helsinki, Finland) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Hyclone), 2 mM -glutamine, penicillin, and streptomycin (Sigma-Aldrich). For immunofluorescence, the U2OS cells were plated 3 h before fixation on coverslips precoated with 10 μg/ml fibronectin. Immunofluorescence was performed as described previously (). As an exception, for the mDia1/DRF1 stainings, cells were permeabilized with 0.05% Saponin. Vinculin was visualized with a monoclonal anti-vinculin antibody (dilution 1:140; Sigma-Aldrich); mDia1/DRF1 with monoclonal anti-mDia1 antibody (dilution 1:100; BD Biosciences); p34 with rabbit polyclonal anti-p34 antibody (dilution 1:50; Upstate Biotechnology); α-actinin with monoclonal α-actinin antibody (dilution 1:50; Sigma-Aldrich); myosin II with rabbit anti–nonmuscle myosin antibodies (dilution 1:60; Biomedical Technologies); and secondary antibodies conjugated to FITC, rhodamine, or Cy5 (Invitrogen). F-actin was visualized with Alexa 488 or 568 phalloidin (dilution 1:100; Invitrogen). Images were acquired through a charge-coupled device camera (DP70; Olympus) on a microscope (AX70 Provis; Olympus). For the image acquirement, the AnalySIS software (Olympus) and PlanApo 60×/1.40 (oil) or UPlanApo 100×/1.35 (oil) objectives (Olympus) were used. (−)-Blebbistatin was obtained from Sigma-Aldrich. Human GFP–β-actin plasmid () was a gift from M. Bähler (Westfalian Wilhelms-University, Münster, Germany). YFP–β-actin was purchased from CLONTECH Laboratories, Inc. Human nonmuscle α-actinin-1 cDNA cloned to the 5′-end of the humanized S65T version of GFP () was a gift from C. Otey (University of North Carolina School of Medicine, Chapel Hill, NC). The cDNA encoding α-actinin was amplified by PCR (primers 5′-GCCGCTCGAGATGGACCATTATGATTCTCAGCA-3′ and 5′-GCCGGAATTCCGAGGTCACTCTCGCCGTACAG-3′), and the PCR fragment was subcloned into the XhoI–EcoRI sites of the pECFP-N1 and pEYFP-N1 vectors (CLONTECH Laboratories, Inc.) to generate plasmids expressing α-actinin–CFP and –YFP fusion proteins. PCR fragment was verified by sequencing. GFP-MLC () was a gift from L. Peterson (University of North Carolina School of Medicine, Chapel Hill, NC). To create YFP-MLC, MLC cDNA was excised with XhoI and BamHI from GFP-MLC plasmid and ligated into the corresponding sites of the pEYFP-C1 vector (CLONTECH Laboratories, Inc.). Zyxin-GFP plasmid () was a gift from K. Rottner (German Research Center for Biotechnology, Braunschweig, Germany). The zyxin cDNA was transferred from zyxin-GFP into pECFP-N1 vector (CLONTECH Laboratories, Inc.) using the EcoRI and BamHI restriction sites to generate zyxin-CFP plasmid. To generate the GFP-mDia1 Ile810Ala mutation construct, the GFP-mDia1 construct (a gift from N. Watanabe, Kyoto University, Kyoto, Japan; ) was mutated by inverse PCR by using primers 5′-CCTTTTTGGGTTCATTCCGCATGCCCTA-3′ and 5′-CTGAGAGATTCTGCGCTGTCTTTGAATC-3′. This generated two nucleotide changes to the mDia1 target sequence (TCATTT sequence was mutated to TCATTT), which altered isoleucine 810 residue to alanine. mDia1 Ile810 is comparable to Bni1p Ile1341 (). Myc-tagged Scar1-W and -WA fragments were gifts from L. Machesky (University of Birmingham, Birmingham, UK; ). Transient transfection of U2OS cells was performed with FuGENE6 (Roche) using 25% of the recommended DNA and FuGENE amounts according to the manufacturer's instructions. For the siRNA experiments, 1,500 ng of preannealed 3′ Alexa Fluor 488–labeled mDia1/DRF1 siRNA (target sequence AAAGGCAGAGCCACACTTCCT []; QIAGEN) or p34 siRNA (target sequence AAGGAACTTCAGGCACATGGT) oligonucleotide duplexes were transfected into cells on 6-well plates by using GeneSilencer's siRNA-transfection reagent (Gene Therapy Systems) according to the manufacturer's recommendations. After 44 or 68 h, the cells were detached with trypsin-EDTA, diluted, and plated on precoated (10 μg/ml fibronectin) glass-bottomed dishes (MatTek) for the live cell imaging. Analysis of the cells was performed 46–52 h after transfection. The cells for Western blotting were washed three times with cold PBS, scraped, and lysed with PBS containing 1% Triton X-100 and 0.3 mM PMSF. Total protein concentrations were measured using Bradford reagent (Sigma-Aldrich). Anti-mDia1 antibody (BD Biosciences) was used at a dilution of 1:500, anti-p34 (Upstate Biotechnology) at a dilution of 1:500, and anti-actin AC-15 antibody (Sigma-Aldrich) at a dilution of 1:10,000. Transfected cells were replated on 10 μg/ml fibronectin–coated glass-bottomed dishes. Normal growth medium was used as imaging medium. The time-lapse images were acquired with an inverted microscope (IX-71; Olympus) equipped with a Polychrome IV monochromator (TILL Photonics) with the appropriate filters, heated sample environment (+37°C), and CO control. UApo 40×/1.35 (oil) objective (Olympus) with 1.6× magnification or PLAPON 60×O TIRFM 60×/1.45 (oil) objective (Olympus) were used. The software used for the image acquirement was TILL Vision 4 (TILL Photonics). Total internal reflectance fluorescence analysis was performed with the same microscope setup using a 488-nm laser line. The cameras used for the study were Imago QE (TILL Photonics) and Andor iXon (Andor). Time-lapse videos were deconvoluted by AutoQuant AutoDeblur 2D Blind Deconvolution (AutoQuant Imaging, Inc.). To analyze the association/dissociation rates of stress fiber components, U2OS cells expressing various GFP constructs were grown for 24 h on glass-bottomed dishes or 3–5 h on glass-bottomed dishes precoated with 10 μg/ml fibronectin. Normal growth medium was used as imaging medium. Confocal imaging was performed on a confocal microscope (TCS SP2 AOBS; Leica) equipped with Leica Confocal Software (Lite 2.61.1537), heating (+37°C), and CO control. For GFP imaging, a 488-nm line and a HCX PL APO 63×/1.4–0.6 (oil) objective were used. After three prebleach scans of an entire image, five bleaching scans (3.9 s each) with 100% intensity of 476 nm (15 mW), 488 nm (70 mW), and 496 nm (15 mW) laser lines over the region of interest (2.5 × 20 μm) were performed. After bleaching, the fluorescence recovery was monitored 10 times every 3.9 s and 15 times every 20 s. The recovery of the GFP intensity was measured by Leica Confocal Software. The intensity of the bleached area was normalized to a neighboring nonbleached area to diminish the error caused by normal photobleaching during the monitoring period. Bleached and control areas used for measurements were also outlined to contain only one stress fiber to diminish fast intensity recovery caused by the diffusion of soluble proteins. The values of intensity versus time (min) were charted in a scatter plot, the recovery half-time (t) was measured from the plots, and the k values (first-order rate constant) were calculated by using the equation k = 0.693/t. Stress fiber elongation and arc flow rates were measured from GFP-actin videos acquired using TILL Photonics imaging system. The length of dorsal stress fibers (μm) was measured from every 10th time-lapse frame (5 min) by Bitplane Imaris suite software (Bitplane AG). The length versus time was plotted in a chart, and k (length change in μm/min) was calculated from the linear trend line (Excel; Microsoft). Polymerization rates from focal adhesions were measured by bleaching the region of interest as described in the previous section, and the elongation of bright stress fiber region from the focal adhesion was measured similarly to the elongation of an entire stress fiber. Fig. S1 shows that the overexpression of a loss-of-function mDia1/DRF1 mutant induces abnormal dorsal stress fiber morphology and accumulation of α-actinin in these structures. Fig. S2 shows that expression of Scar-WA in U2OS cells leads to a loss of lamellipodia and transverse arcs. Video 1 displays actin dynamics in U2OS cells visualized by GFP-actin. Video 2 shows actin dynamics in U2OS cells visualized with GFP-actin. Video 3 shows ventral stress fiber assembly visualized with YFP-actin and zyxin-CFP. Video 4 shows dorsal stress fiber elongation visualized by zyxin-CFP and α-actinin–YFP in U2OS cells. Video 5 displays dorsal stress fibers and arcs visualized by α-actinin–YFP in U2OS cells. Video 6 shows dorsal stress fibers and arcs visualized by zyxin-CFP and α-actinin–YFP in mDia1/DRF1 siRNA–transfected U2OS cells. Video 7 shows dorsal stress fibers and arcs visualized by zyxin-CFP and α-actinin–YFP in mDia1/DRF1 siRNA–transfected U2OS cells. Video 8 demonstrates arc assembly from short α-actinin cross-linked actin filament bundles generated at the leading lamellipodium. Video 9 shows dorsal stress fibers visualized by GFP-actin in a p34 siRNA–transfected U2OS cells. Video 10 shows the effect of blebbistatin treatment in U2OS cells expressing GFP-actin. Online supplemental material is available at .

The ability of cancer cells to disseminate from primary tumors (and metastases) gives rise to a growing tumor burden that is distributed in multiple sites in the body, resulting in death for many cancer patients. Understanding the steps at the cellular level that are used by cancer cells during invasion and metastasis can form the basis for new diagnostic, prognostic, and therapeutic approaches that allow control of cancer malignancy. Microarray-based expression analysis of whole tumors has been used to identify genes involved in metastasis and patterns of gene expression that would give an indication of the likelihood of metastasis of breast tumors (; ). These studies have successfully identified patterns of gene expression that correlate with metastasis; however, these patterns have not been informative with regard to the mechanisms contributing to metastasis because the expression patterns of all cells of the tumor are averaged (). We have developed an in vivo invasion assay that provides an opportunity to collect primary tumor cells that are actively in the process of invasion. The in vivo invasion assay has been combined with array-based gene expression analyses to investigate the gene expression patterns of carcinoma cells in primary mammary tumors during invasion (). The expression of genes involved in cell division, survival, and motility were most dramatically changed in invasive cells, indicating a population that is neither dividing nor apoptotic but intensely motile (; ). In particular, the genes coding for the three end stage effectors (Arp2/3 complex, capping protein, and cofilin) of the minimum motility machine that regulates actin polymerization at the leading edge and, therefore, the motility and chemotaxis of carcinoma cells were up-regulated. Interestingly, in the cofilin pathway, LIM kinase 1 (LIMK1) and cofilin were coordinately up-regulated in invasive cells (). Chemotaxis to EGF is an important determinant in the haematogenous metastasis of mammary tumors (), and cofilin is required for chemotaxis to EGF (). Thus, it is essential to understand how the LIMK–cofilin pathway contributes to metastasis. LIMK1 is a member of a novel class of serine–threonine protein kinases that contain two tandem LIM domains at the amino terminus, a PDZ domain in the central region, and a protein kinase domain at the carboxy terminus (). Cofilin is the only known physiological substrate of LIMK1. LIMK1 phosphorylates cofilin, which inactivates it. This inhibits cofilin's actin-severing and depolymerization activities (; ; ). Cofilin activity is required for tumor cell motility and invasion. Local activation of cofilin by uncaging induces lamellipod formation and sets the direction of cell motility (). Inhibition of cofilin activity in carcinoma cells with either siRNA () or the expression of constitutively active LIMK domain () inhibits cell motility. The suppression of cofilin expression with siRNA reduces the invasion of carcinoma cells by reducing the assembly and stability of invadopodia (). The overexpression of cofilin increases the velocity of cell migration in () and in human glioblastoma cells (). The spontaneous overexpression of cofilin has been detected in the invasive subpopulation of tumor cells in mammary tumors (). Cofilin is overexpressed in the highly invasive C6 rat glioblastoma cell line (), and the amount of phosphorylated, inactive cofilin is decreased in cell lines derived from T lymphoma (Jurkat) and carcinomas from the cervix (HeLa), colon (KM12), liver (HepG2), and kidney (COS1; ). These results suggest that cofilin activity enhances, but LIMK activity inhibits, cell motility and metastasis. In previous studies, LIMK1 has been shown to affect cell motility and invasion, but there is confusion about whether LIMK increases or decreases invasion. The expression of a constitutively active LIMK1 that increases the amount of phospho-cofilin in vivo inhibits actin polymerization and motility in response to EGF in mammary carcinoma cells (). In addition, the overexpression of LIMK suppresses motility in neuroblastoma, and dominant-negative LIMK increases it (). However, the overexpression of LIMK1 in prostate epithelial cells increased their invasiveness in vitro () and the formation of osteolytic lesions in nude mice (). Furthermore, a reduction in LIMK1 expression in metastatic prostate cell lines decreased their ability to invade matrigel in vitro (). Clearly, these studies do not agree as to the consequences of LIMK1 expression on invasion. Therefore, it is likely that LIMK1 expression alone is not involved directly in determining the motility and invasion status of carcinoma cells. Cofilin, the substrate of LIMK1, is directly responsible for severing actin filaments and regulating actin polymerization and depolymerization during cell migration (). It is required for cell motility because reduced cofilin expression levels correlate with the inhibition of locomotion (), whereas increased expression levels correlate with enhanced motility (). Hence, cofilin is arguably the key effector in the cofilin pathway that determines cell migration and invasion, whereas LIMK1 may only serve to modify the level of cofilin activity. We propose that the output of the cofilin pathway—the generation of barbed ends during the first transient after EGF stimulation by cofilin, which is required for chemotaxis to EGF ()—is the key determinant of whether a carcinoma cell will be more or less motile and invasive. This hypothesis is based on results demonstrating that the activation of cofilin is sufficient to stimulate cell protrusion and motility and determine the direction of locomotion (), that cofilin activity is required for the chemotaxis of carcinoma cells to EGF (; ), and that invasive cells isolated from mammary tumors overexpress cofilin and its upstream regulatory proteins (). Therefore, if the activity status of the cofilin pathway can be altered by varying the expression of LIMK1, perhaps the metastatic potential of tumor cells might be related to LIMK1 expression. In this study, we test this hypothesis by varying the expression of LIMK1 and its dominant-negative kinase domain in the rat MTLn3 carcinoma cell line to examine the effects on the activity of the cofilin pathway and invasion and metastasis of primary mammary tumors grown from these cells. The full-length LIMK1 and dominant-negative kinase domain were introduced into MTLn3 cells, and stable cell lines were obtained. The output of the cofilin pathway (barbed end production in response to EGF) was then measured and correlated with invasion and metastasis. Although other research groups have suggested that LIMK1 may play important roles in regulating breast tumor growth independently of cofilin (; ) and that overexpression of LIMK1 causes increased invasion directly (), our results indicate that the effects of LIMK1 on invasion, intravasation, and metastasis of cancer cells can be explained most simply by LIMK1's regulation of cofilin activity and cofilin-dependent cell motility. We have used a model of breast cancer that allows the direct examination of the chemotaxis and invasion of individual carcinoma cells in live metastatic primary tumors (; ). The metastatic tumor cell line used to create these model tumors is the MTLn3 cell line. The invasive cells isolated from rat mammary tumors overexpress cofilin and its upstream regulatory proteins in a pattern suggesting increased activation of the cofilin pathway (). Therefore, we propose that the cofilin pathway is one of the key determinants of whether a carcinoma cell will be more or less invasive and metastatic. To test this hypothesis, we assessed the EGF-induced early barbed end transient in invasive cells, which are cells that were collected by means of the in vivo invasion assay using microneedles from the primary tumor (). The level of barbed ends in tumor cells after 1 min of EGF stimulation has been shown previously to depend on the output of the cofilin pathway (). Therefore, the number of barbed ends in invasive tumor cells was compared with that in FACS-sorted tumor cells from the primary tumor, a population composed primarily of noninvasive cells (). The results show that the EGF-induced cofilin-dependent early transient of barbed ends is increased in the invasive tumor cell population, indicating that the cofilin pathway regulating the early transient has increased activity (). To further test the hypothesis that the output of the cofilin pathway (the generation of barbed ends during the first transient of actin polymerization after EGF stimulation by cofilin) is the key determinant of whether a carcinoma cell will be more or less invasive and metastatic, four LIMK1 constructs were transfected separately into the parental MTLn3 cells. Stable transfectants were obtained for the full-length GFP-LIMK1 and dominant-negative GFP-KS (), which grew normally compared with control cells (MTLn3 cells transfected with pGreenLantern-1 vector). However, no stable cell lines were obtained from GFP-FS (inactive GFP-tagged LIMK1 Short), which is consistent with an involvement of LIMK1 in regulating cell cycle dynamics independently of cofilin phosphorylation (). In addition, no stable cell lines were obtained with GFP-K (constitutively active GFP-tagged kinase domain of LIMK1), which is consistent with the ability of the constitutively active LIMK domain to completely inhibit cofilin activity measured as cofilin-mediated barbed end formation and protrusion () and the requirement of cofilin activity for cell viability (). Western blot analysis () confirmed the expression of GFP-LIMK1 and GFP-KS at approximately two to four times the levels of the endogenous LIMK1. These cell lines will hereafter be called GFP (GFP-expressing control), F (full-length LIMK-expressing cells), and KS (LIMK dead truncated). Although the total cofilin levels remained unchanged in all of the transfectants (), the phosphorylation of cofilin was greatly increased in F cells, decreased in KS cells, and unchanged in GFP cells (). As discussed above, previous studies have come to contradictory conclusions as to the role of LIMK1 in tumor cell motility, invasion, and metastasis, claiming both increases and decreases in cell motility and invasion as a result of LIMK1 overexpression. Therefore, it is unlikely that LIMK1 expression alone directly determines the motility, invasion, and metastasis status of carcinoma cells. LIMK1 is part of the cofilin pathway that regulates barbed end formation during tumor cell chemotaxis in response to EGF (; ) and is one of the genes up-regulated, along with cofilin, Rho-associated coil-containing protein kinase, and PKCζ, in invasive tumor cells during invasion and metastasis (; ). These results suggest that it is the overall activity of the cofilin pathway and not that of LIMK1 alone that determines the invasive and metastatic phenotype of tumor cells. In this study, we tested this hypothesis by varying the expression of LIMK1 and its dominant-negative kinase domain in the rat MTLn3 carcinoma cell line to examine the effects of LIMK on the activity of the cofilin pathway and, consequently, on the invasion and metastasis of primary mammary tumors grown from these cells. Our results demonstrate that the effect of LIMK1 expression on invasion, intravasation, and metastasis of cancer cells can be most simply explained by the regulation and activity status of the cofilin pathway. When LIMK1 overexpression (F cells) decreases the activity of the cofilin pathway (the EGF-induced early barbed end transient), the invasion and metastasis of tumor cells is proportionately decreased, whereas if the activity of the cofilin pathway is unchanged or increased, as in the GFP and KS cells, the chemotaxis, invasion, and metastasis of the tumor cells is either unaffected or increased. Furthermore, our study indicates that the increased expression of LIMK1 in carcinoma cells significantly reduces their cell motility as the phosphorylation of cofilin by LIMK1 abolishes EGF-stimulated actin nucleation, protrusion, and chemotaxis in vitro and invasion, intravasation, and metastasis in vivo. The in vitro results in particular are consistent with previous studies linking the activity of the cofilin pathway to EGF-stimulated directional sensing, protrusion, and chemotaxis (; ; ). The results reported in this study are consistent with a previous study demonstrating that the expression of the kinase domain of LIMK1, which results in the near total phosphorylation of cofilin, inhibits EGF-stimulated actin nucleation, protrusion, and motility in tumor cells (). This phenotype is caused by cofilin inhibition because these effects could be rescued by the expression of S3A cofilin, which cannot be phosphorylated by LIMK1 (), whereas the similar results reported in this study for F cells could be rescued by the overexpression of wild-type cofilin. The results reported in this study are also consistent with the inhibition of cell motility by the overexpression of LIMK1 in Ras-transformed Swiss 3T3 fibroblasts (). It is interesting to note that in cells expressing KS, although barbed ends are slightly elevated, protrusion and invasion in vitro and in vivo are not significantly affected compared with GFP control cells. However, intravasation, metastasis, and morbidity are significantly increased in tumors derived from KS cells. This observation is consistent with a recent study () showing that cofilin affects the assembly and stability of invadopods that are required for intravasation. Furthermore, it is known that tumor cells exhibit two different types of cell motility in the primary tumor during invasion and intravasation: high velocity amoeboid movement guided by the linear geometry of the extra cellular matrix; and slow, directed invadopod-mediated extracellular matrix degradation at blood vessels where intravasation occurs (; ). Therefore, it is expected that metastasis is the sum of two different types of cell motility during invasion and intravasation. Cofilin can affect both types but may do so unequally. In particular, our results indicate that the slightly higher activity of cofilin in KS cells does not significantly affect protrusion and invasion of amoeboid cells but affects invadopod-dependent intravasation (). These results suggest that the effect of KS on metastatic phenotype is mainly on intravasation, which is the rate-limiting step for metastasis (; ). A key feature of the cofilin pathway is the generation of actin filament barbed ends after stimulation with EGF. Cofilin-generated actin filament barbed ends are required for chemotaxis to EGF (), a property essential for invasion and metastasis (; ). The important role of barbed ends generated by the cofilin pathway in chemotaxis and invasion makes it essential to examine the output of this pathway in cells that have been manipulated for the expression of LIMK1. Other studies in which the overexpression of LIMK1 in tumor cell lines was reported to increase their motility and invasiveness in vitro (; ) might be resolved with our studies by measuring cofilin-dependent barbed ends during chemotaxis to determine whether the activity of the cofilin pathway is elevated in their models. Furthermore, the stability of LIMK expression in these experimental models should be checked to determine whether LIMK1 expression was suppressed during invasion and metastasis as observed in our study. For example, the inhibitory effect of the overexpression of LIMK1 on metastasis is emphasized by our observation that successful metastatic lung tumors derived from F tumors overexpressing LIMK1 have been selected by the tissue microenvironment for decreased LIMK1 expression independently of tumor growth. This observation may explain why our results appear inconsistent with previous studies in which the overexpression of LIMK1 in tumor cells has been reported to increase their ability to metastasize to bone in an experimental metastasis model (). To resolve this inconsistency, it would be necessary to determine the expression status of LIMK1 in the bone metastases. Other research groups have shown that LIMK1 may play a role in regulating cell division in a prostate cancer cell line PC3 and mouse embryo fibroblasts in vitro (; ). However, our results demonstrate that tumor growth was not significantly affected by altering the expression of LIMK1 in F- and KS-derived mammary tumors during growth in vivo, indicating that the changes in invasive and metastatic potential resulting from F and KS expression were not caused by changes in tumor growth and/or mass. The different results from our group and others may result from the different cell types used in these studies (for example, mammary vs. prostate cancer cells) and differences between growth in the local tumor microenvironment in vivo as used in our study and growth observed in vitro in the fibroblast and PC3 studies (; ). Microarray analyses of invasive tumor cells collected in the in vivo invasion assay indicate that LIMK1 is more highly expressed in the most invasive cells compared with the less invasive cells that are residents of the primary tumor (). This observed increase in LIMK1 expression could lead to the incorrect conclusion that the overexpression of LIMK1 is sufficient for increased invasion and metastasis. The most important implication of our study is that looking at the expression status of a single gene can be misleading because it is the collective activity of the pathway of which that gene product is a part that determines the metastatic phenotype of a tumor. In fact, in the microarray study, the elevated expression of LIMK1 was coupled with increased levels of the expression of cofilin, which occurred spontaneously (). In this study, the experimental overexpression of both cofilin and LIMK is required for the activation of the cofilin pathway resulting in EGF-stimulated protrusion, cell motility, invasion, and metastasis. Our results indicate that in invasive cells collected from primary tumors, the stimulatory and inhibitory pathways regulating cofilin must be in balance in order for cells to be invasive and metastatic. The application of this idea to the analysis of microarrays of other pathways may lead to more accurate predictions of metastatic outcome in breast tumors. Our results also suggest that the phosphorylation status of cofilin in tumor cells within the primary tumor in situ may have value in predicting the metastatic potential of a breast tumor. Rat MTLn3 cells were maintained and grown to log phase before being used as described previously (; ). GFP-F (active GFP-tagged full-length LIMK1), GFP-FS (inactive GFP-tagged LIMK1 Short), GFP-K (active GFP-tagged kinase domain of LIMK1), and GFP-KS (inactive GFP-tagged kinase domain of LIMK1 Short) were described previously (; ). Rat nonmuscle cofilin cDNA was subcloned into the EcoRI/BamHI sites of pBabe-Puro, a retroviral vector with puromycin selection marker (). To account for any effects that might arise from the introduction of EGFP into cells, MTLn3 cells transfected with pGreenLantern-1 vector (Life Technologies) were used as controls. The different constructs were introduced into cells by transfection using FuGene (Roche). Transfected MTLn3 cells were grown in the presence of 500 μg/ml geneticin or 1 μg/ml puromycin (Invitrogen) for 3–4 wk to select stable expressers. Cells expressing higher amounts of LIMK1 were further selected by increasing the concentration of geneticin to 4 mg/ml. Cells were cultured at low density (30%) and starved free of serum for at least 3 h before stimulation with EGF to assure that the cells at time = 0 were resting. Total cell lysates were resolved in 10% SDS-PAGE and subjected to immunoblotting using goat polyclonal antibody against the COOH terminus of LIMK1 (Santa Cruz Biotechnology, Inc.) so that both long and short forms of LIMK1 could be detected. Chicken polyclonal antibody against rat full-length cofilin was used for total cofilin. To measure the absolute value of phosphorylated cofilin, IEF gels (pH 3–10; Bio-Rad Laboratories) were used to separate phospho-cofilin from cofilin. Anti–β-actin antibody (Sigma-Aldrich) was simultaneously used as an internal control for loading. Three separated experiments were performed for all of the Western blotting, and statistical analysis was performed with a test. It should be noted that this absolute method is different from the relative method used by and gave different numerical values. All pictures were taken of live cultures at 37°C and of all cells using a 60× NA 1.4 infinity-corrected objective on a microscope (IX170; Olympus) supplemented with a computer-driven cooled CCD camera (SensiCam QE; Cooke) operated by IPLab Spectrum software (VayTek). Digital images were linearly converted in ImageJ software (National Institutes of Health [NIH]) and analyzed using macro analysis. The in vitro invasion assay measures the ability of macrophages to induce the invasion of a dense collagen gel by tumor cells, mimicking a paracrine-induced invasion as observed in primary mammary tumors (). This assay was performed as described previously (). In brief, MTLn3-GFP ( = 80,000) were plated on a 35-mm dish (∼80 cells/mm; MatTek) in the presence or absence of 200,000 BAC1.2F5 cells in 2 ml α-MEM with 10% FBS and 36 ng/ml CSF-1. After 16 h, cells were overlaid with a 750–1,000-μm layer of 5–6 mg/ml collagen I, which was allowed to gel for 90 min before the addition of 1 ml α-MEM with 10% chemically defined lipid mix (Invitrogen) and insulin-transferrin-selenium (Invitrogen). After 24 h, the assay was fixed with 4% formaldehyde and analyzed by confocal microscopy; optical z sections were taken every 5 μm starting at the base of the dish and extending at least 50 μm into the collagen gel. To quantify the invasion of MTLn3-GFP cells, GFP fluorescence in the z sections from 20 μm into the collagen and above was added and divided by the sum of GFP fluorescence in all of the z sections. The data shown represent the analysis of ≥200 cells with data collected from at least three independent experiments. We used MTLn3-derived mammary tumors in rats and the in vivo invasion assay described previously () to study cancer cell invasion. In brief, the in vivo invasion assay uses microneedles filled with matrigel and growth factors (EGF in this case) to collect the invasive tumor cells from primary tumors. Microneedles are held in a clamping devise and positioned in the primary tumor with a micromanipulator. Cell collection takes 2.5–4 h, and the number of tumor cells collected was counted as described previously (). For barbed end staining, the needle-collected invasive tumor cells and total resident tumor cells collected from the primary tumor by FACS sorting of GFP were plated on collagen-coated dishes (MatTek) for ∼12 h and were starved for 4 h before EGF stimulation. Control experiments include the exposure of FACS-sorted tumor cells to EGF to mimic the collection conditions used for invasive cells (). MTLn3 cells were injected into the mammary fat pads of female Fischer 344 rats to form primary tumors. After tumor growth, tumor cell blood burden was determined as described previously (). For the measurement of spontaneous metastases, after blood removal and euthanization of the rat, the lungs were removed, and the visible metastatic tumors near the surface of the lungs were counted as described previously (). In a separate experiment, animals bearing MTLn3-LIMK1 (F), MTLn3-KS (KS), or MTLn3-GFP (GFP) tumors were used to observe the effect of LIMK1 on animal survival. At least nine animals were used in each group. The animals were dissected after the animal died, and the metastases were scored by checking the visible metastatic tumors on the lung surface. After 7 wk, all of the animals were killed and dissected to check the metastases. To study experimental lung metastasis, 7.5 × 10 cells were injected into the lateral tail veins of 7-wk-old female rats. 2.5 wk after injection, the rats were killed, and the lungs were removed for isolation of carcinoma cells using the method described below. To check the expression level of LIMK1 in the resident tumor and lung metastases, quantitative real-time PCR analysis was performed by using a sequence detection system (ABI PRISM 7900; Applied Biosystems) with sequence-specific primer pairs for LIMK1 as described previously (). To isolate the carcinoma cells from primary tumor or lung, a small piece of tumor or lung tissue was minced and filtered twice through a nylon filter to obtain a single cell suspension. FACS was performed on the resulting single cell suspensions based on their GFP expression in tumor cells using a cell sorter (FACSVantage; Becton Dickinson). SYBR Green was used for real-time monitoring of amplification. Results were evaluated with ABI PRISM SDS 2.0 software (Applied Biosystems). The normalization of all samples was performed using at least two housekeeping genes (β-actin and glyceraldehyde-3-phosphate dehydrogenase).