ATX-2, the C. elegans Ortholog of Human Ataxin-2, Regulates Centrosome Size and Microtubule Dynamics

Centrosomes are critical sites for orchestrating microtubule dynamics, and exhibit dynamic changes in size during the cell cycle. As cells progress to mitosis, centrosomes recruit more microtubules (MT) to form mitotic bipolar spindles that ensure proper chromosome segregation. We report a new role for ATX-2, a C. elegans ortholog of Human Ataxin-2, in regulating centrosome size and MT dynamics. ATX-2, an RNA-binding protein, forms a complex with SZY-20 in an RNA-independent fashion. Depleting ATX-2 results in embryonic lethality and cytokinesis failure, and restores centrosome duplication to zyg-1 mutants. In this pathway, SZY-20 promotes ATX-2 abundance, which inversely correlates with centrosome size. Centrosomes depleted of ATX-2 exhibit elevated levels of centrosome factors (ZYG-1, SPD-5, γ-Tubulin), increasing MT nucleating activity but impeding MT growth. We show that ATX-2 influences MT behavior through γ-Tubulin at the centrosome. Our data suggest that RNA-binding proteins play an active role in controlling MT dynamics and provide insight into the control of proper centrosome size and MT dynamics.


Author Summary
The microtubule (MT) cytoskeleton undergoes dynamic rearrangements during the cell cycle. As the primary microtubule-organizing center, centrosomes orchestrate MT dynamics and play a key role in establishing bipolar spindles in mitosis. Errors in centrosome assembly lead to missegregation of genomic content and aneuploidy. Thus, stringent regulation of centrosome assembly is of vital importance for the fidelity of cell division and survival. Using the nematode Caenorhabditis elegans (C. elegans) as a model, we study the role of the RNA-binding protein, ATX-2, a C. elegans homolog of Human Ataxin-2 in early cell division. A number of RNAs and RNA-binding proteins are shown to be associated with centrosomes and MTs, and influence the assembly of mitotic spindles. In C. elegans, the RNA-binding role of SZY-20 is implicated in regulating centrosome size. We show that ATX-2 functions together with SZY-20 in centrosome size and MT behavior. SZY-20 promotes ATX-2 protein levels, and the amount of ATX-2 influences centrosome

Introduction
As the primary microtubule-organizing centers, centrosomes are vital for the maintenance of genomic integrity in animal cells [1]. The centrosome consists of two barrel-shaped centrioles surrounded by a network of proteins termed pericentriolar materials (PCM). To maintain the fidelity of cell division, each cell must duplicate a pair of centrioles precisely once per cell cycle, one daughter per mother centriole. Mishaps in centrosome assembly result in chromosome missegregation and other cell cycle defects. Thus, stringent regulation of centrosome assembly is imperative for proper cell division and survival.
Studies in C. elegans have discovered five evolutionarily conserved proteins (ZYG-1, SPD-2, SAS-4, SAS-5 and SAS-6) that are required for centrosome assembly [2][3][4][5]. Many other factors, including protein phosphatase 2A, also regulate the production, activity, or turnover of core regulators, and are equally important in regulating centrosome assembly [6,7]. Like other biological processes, centrosome assembly is regulated by a combined action among negative and positive regulators [8]. While the kinase ZYG-1 promotes centriole duplication in C. elegans, szy-20 acts as a genetic suppressor of zyg-1 [9,10]. The szy-20 gene encodes a centrosome-associated RNA-binding protein that negatively regulates centrosome assembly by opposing ZYG-1. Centrosomes in szy-20 mutants exhibit elevated levels of centrosomal proteins, resulting in defective microtubule (MT) behavior and embryonic lethality. SZY-20 contains putative RNAbinding domains (SUZ, SUZ-C). Mutating these domains has been shown to perturb in vitro RNA-binding of SZY-20 and its capacity to regulate centrosome size in vivo [10]. Other studies have shown that a number of RNAs and RNA-binding proteins are associated with centrosomes and MTs, and influence proper mitotic spindles and other aspects of cell division. In mammalian cells, several RNA-binding proteins (e.g., RBM8A, Hu antigen R, and the Ewing sarcoma protein) associate with centrosomes and play a role in regulating centrosome assembly during cell division [11][12][13][14][15]. In yeast, spindle pole body duplication is linked to translational control via the action of RNA-binding proteins [16]. In Xenopus, MT-guided localization of transcripts followed by spatially enriched translation is important for proper MT behavior and cell division, suggesting the importance of local translational control [17][18][19].
Despite the finding that SZY-20 negatively regulates ZYG-1, no direct interaction between the two proteins has been found. Thus, identifying additional factors that function between SZY-20 and ZYG-1 should provide further insights into the molecular mechanism by which the putative RNA-binding protein, SZY-20, influences centrosome assembly. Toward this end, we report here our identification of an RNA-binding protein ATX-2 that physically associates with SZY-20. ATX-2 is the C. elegans ortholog of human Ataxin-2 that is implicated in human neurodegenerative disease [20]. Specifically, human spinocerebellar ataxia type 2 is shown to be associated with an extended poly-glutamine (Q) tract in Ataxin-2 [20][21][22]. Ataxin-2 is an evolutionarily conserved protein that contains an RNA-binding motif (LSm: Sm-like domain) and a PAM domain for binding the poly-(A) binding protein (PABP1) [23][24][25]. It has been shown that Ataxin-2 binds directly to the 3'UTR of mRNAs and stabilize target transcripts, and that poly-Q expansion blocks the RNA-binding by Ataxin-2 in vitro [26]. Ataxin-2 homologs have been implicated in a wide range of RNA metabolism-dependent processes including translational control of circadian rhythm [27,28]. While ATX-2 in C. elegans is known to be responsible for embryonic development and translational control in germline development [29][30][31][32], the action of this RNA-binding protein in centrosome assembly and cell division has not been fully explored.
In this study, we investigate the role of C. elegans ATX-2 in early cell cycles and how ATX-2 acts together with SZY-20 and ZYG-1 in controlling centrosome size and MT behavior. We show that ATX-2 negatively regulates the key centriole factor ZYG-1. In the centrosome assembly pathway, SZY-20 acts upstream of ATX-2 and positively regulates embryonic levels of ATX-2; proper levels of ATX-2 contributes in turn to normal centrosome size and subsequent MT dynamics.
To confirm the physical interaction between SZY-20 and ATX-2, we used anti-SZY-20 to pull down SZY-20 and its associated proteins from embryonic lysates and examined co-precipitates by western blot (Fig 1A). Consistent with our mass spectrometry data, we detected ATX-2 and SZY-20 in the SZY-20-immunoprecipitates from wild-type embryonic extracts. Given that this protein complex consists of RNA-binding proteins, we asked if the physical association is mediated through RNA. To test RNA dependence, we repeated IP to pull down SZY-20 interacting proteins in the presence of RNaseA or RNase inhibitor and found that they co-precipitated in either condition (Fig 1A, S1B Fig), suggesting ATX-2 and SZY-20 physically interact in an RNA-independent manner although we cannot exclude the possibility that RNA bound by the ATX-2-SZY-20 complex could have been protected from RNase treatment.
The szy-20(bs52) mutation results in a truncated protein, deleting the C-terminal 197 aa residues including the SUZ-C domain, one of the putative RNA-binding domains in SZY-20 [10]. We utilized szy-20(bs52) embryos for IP analysis to determine if the C-terminal truncation of SZY-20 affected physical association of SZY-20 with ATX-2 (S1B Fig). Whereas ATX-2 coprecipitated with SZY-20 in wild-type extracts, ATX-2 was undetectable in co-precipitates from szy-20(bs52) extracts, suggesting that the C-terminus of SZY-20 influences physical interaction with ATX-2. IP assay using embryos expressing SZY-20-GFP-3xFLAG yielded a similar result: ATX-2 is undetectable in co-precipitates of SZY-20 tagged with GFP-3xFLAG at the Cterminus (S1E Fig), supporting that proper folding of the C-terminal domain is critical for SZY-20 to interact with ATX-2. Further, the C-terminal deletion in either ATX-2 or SZY-20 appears to alleviate its interaction with SZY-20 (S1C and S1D Fig). By additional IP assays with anti-GFP using embryos expressing various GFP-tagged proteins, we further confirmed that both ATX-2 and PAB-1 physically interact with SZY-20 or with each other (S1E Fig). Together, our data suggest SZY-20 forms a complex with known RNA-binding proteins ATX-2 and PAB-1 in vivo, via direct or indirect interaction.
Embryonic lethality by loss of atx-2 might result from defective cell division. To examine what role ATX-2 plays in cell division, we immunostained embryos for microtubules (MTs), centrosomes and DNA (Fig 1D-1H). Confocal microscopy of immunofluorescence (IF) revealed that knocking down atx-2 by RNAi results in multiple cell division defects including polar body extrusion failure (22%; S3 Movie), abnormal spindle positioning (3%), chromosome missegregation (10%) and cytokinesis failure (36%; n = 114). We observed similar cell division phenotypes, but with higher penetrance in temperature sensitive (ts) atx-2(ne4297) mutants. By 4D timelapse confocal microscopy, we observed that incomplete cytokinesis following successful centrosome duplication results in tetrapolar spindles in one-cell embryo (S1 and S2 Movies). In these embryos, the cytokinetic furrow initiates but cytokinesis fails to complete, resulting in a multinucleated cell with four centrosomes after the second mitosis. All of these cell division phenotypes resemble cell cycle defects observed previously in szy-20(bs52) embryos [10], suggesting that ATX-2 functions closely with SZY-20 in cell division. In contrast, pab-1(RNAi) produced only minor cell cycle defects such as atypical spindle positioning (S2A Fig).
Given the positive genetic interactions among szy-20, atx-2 and pab-1, we further asked if co-depleting these factors could enhance the suppression of zyg-1 (Fig 2D and 2E). At 24°C, co-depleting either atx-2 or pab-1 with szy-20 restored nearly 100% of centrosome duplication to zyg-1(it25) embryos, while single depletion produced~60% duplication in zyg-1(it25). Despite the restoration in centrosome duplication at 24°C, none of these embryos hatched, owing to other cell cycle defects such as cytokinesis failure described above. However, we were able to show partial restoration of embryonic viability in zyg-1(it25) by co-depleting either atx-2 or pab-1 with szy-20 at semi-restrictive temperature 23°C, whereas single depletion of atx-2 or pab-1 showed no effect on embryonic viability of zyg-1(it25). Our data indicate that like szy-20, atx-2 and pab-1 act as genetic suppressors of zyg-1. Thus, these RNA-binding proteins in a complex function together to negatively regulate centrosome assembly.

Loss of atx-2 leads to increased levels of centrosomal ZYG-1
Because atx-2 acts as a genetic suppressor of zyg-1, we reasoned that inhibiting ATX-2 might enhance ZYG-1 activity, thereby restoring centrosome duplication and embryonic viability to zyg-1(it25) embryos. By staining embryos for ZYG-1 and microtubules (Fig 4A), we quantified the fluorescence intensity of ZYG-1 at first metaphase centrosomes, finding that atx-2 mutant centrosomes possess twice as much ZYG-1 levels as those in control embryos (p <0.001, Fig  4B). Using the CRISPR-Cas9 method [39,40], we also generated a strain expressing HA-tagged ZYG-1 at endogenous levels from the native genomic locus (S4A- S4C Fig, S2 Table). Given that ZYG-1 localizes to centrosomes in a cell cycle-dependent manner, the observed increase in centrosomal ZYG-1 at first metaphase in atx-2 mutants could result from a shift in the cell cycle due to loss of atx-2. To examine cell cycle dependence of ZYG-1 localization to centrosomes, we utilized a strain expressing GFP-ZYG-1-C-term that contains a C-terminal portion (217-706 aa) lacking most of the kinase domain, but including the Cryptic Polo Box (CPB) that is sufficient for centrosomal targeting [5,41]. To observe dynamics of centrosome-associated ZYG-1 over time, we acquired 4D time-lapse movies of early embryos starting from pronuclear meeting up to separation of the centriole pair at first anaphase (Fig 4C and  4D, S5 and S6 Movies). Using these recordings, we first quantified the fluorescence intensity of the intensely labeled sub-centrosomal GFP signal, which presumably reflects the centriolar structure. Then, we measured pericentriolar GFP signal that likely represents PCM. Throughout the cell cycle, we observe a nearly two-fold increase in both centriolar and PCM-associated GFP signal in atx-2(RNAi) embryos compared to controls. It could be that the elevated levels of centrosomal ZYG-1 in atx-2 mutants reflect a global increase in ZYG-1 throughout the cells. To test this possibility we compared overall ZYG-1 levels by measuring cytoplasmic GFP signals, but found no increase in the cytoplasmic levels of atx-2(RNAi) embryos. In fact, we noticed a small decrease (p = 0.2) in the cytoplasmic GFP signal in atx-2(RNAi) embryos compared to controls, suggesting that increased centrosomal ZYG-1 levels are unlikely due to an increase in overall ZYG-1 expression. Similar results were also observed in atx-2(ne4297) mutants (S6 Movie). Together, our data show that inhibiting ATX-2 results in elevated levels of ZYG-1 at centrosomes without affecting overall ZYG-1 levels. Thus, we speculate that elevated levels of centrosomal ZYG-1 in atx-2 depleted embryos might partially compensate for the reduced activity of mutant ZYG-1 (P442L) in zyg-1(it25) [8], restoring centrosome duplication in zyg-1 mutants.

ATX-2 limits the number of microtubules emanating from the centrosome
The PCM factors, SPD-2, SPD-5 and γ-Tubulin, play a critical role in positively regulating the MT nucleating capacity of the centrosome [42,[45][46][47]. As atx-2 mutant embryos exhibit enlarged centrosomes with increased centrosomal SPD-5 and γ-Tubulin, we examined if atx-2 mutant centrosomes affected MT nucleating capacity. To investigate MT nucleation, we used a The atx-2 mutant centrosome exhibits more intense ZYG-1 focus than N2. (B) Relative fluorescence intensity of centrosomal ZYG-1 at the first metaphase (**p<0.001). (C) Still images from time-lapse movies of embryos expressing GFP-ZYG-1-C-term [41] at selected time points. Time (min) is relative to the first metaphase. These embryos exhibit centriolar as well as pericentriolar GFP-ZYG-1, with increased levels in the atx-2(RNAi) embryo. In the atx-2(RNAi) embryo, the centrosome on the right (arrow) is out of the focal plane as two centrosomes in atx-2(RNAi) embryos often do not align on the same focal plane (see S5 and S6 Movies). (D) Measurements of fluorescence intensity of centriolar, pericentriolar and cytoplasmic GFP-ZYG-1 using time-lapse recordings (n = 5-7). ZYG-1 levels steadily increase in control and atx-2(RNAi) embryos as cell cycle progresses. Note a slightly lower cytoplasmic levels of ZYG-1 in atx-2 (RNAi) embryos. Error bars are SD (centriolar and pericentriolar ZYG-1; p<0.0001; cytoplasmic ZYG-1; p>0.005). Bar, 5 μm. strain expressing EBP-2-GFP to mark the plus-ends of growing MTs [47] and acquired a series of 500 msec-interval snap shots at the center plane of first metaphase centrosomes (Fig 6, S7  Movie). First, we found that atx-2 mutant embryos exhibit a three-fold increase (p<0.0001) in centrosomal EBP-2-GFP signal compared to wild-type controls (Fig 6A and 6B). To assess the level of MT nucleation by the centrosome, we measured the fluorescent intensity of EBP-2-GFP in line regions proximal (25 pixels) to the centrosome, finding that atx-2 mutant centrosomes exhibit a two-fold increase (p<0.001) in MT nucleation (Fig 6C). Consistent with increased MT nucleation, kymograph analysis at the proximal region to the centrosome revealed a drastic increase (p = 0.002) in the number of astral MTs emanating from atx-2 mutant centrosomes over a 5-sec period, compared to controls (Fig 6F). Further, line scans of the kymograph in a 355°arc around the centrosome show that mutant centrosomes emanate increased numbers of spindle fibers as well as astral MTs (Fig 6D). Thus, atx-2 mutant centrosomes possessing elevated levels of PCM factors nucleate the increased number of astral and spindle MTs.

Excessive MT nucleation leads to aberrant MT growth in atx-2 mutants
Next we examined how these MTs nucleated by mutant centrosomes continue to grow out toward the cortex (Fig 6E, 6G, 6H, S7 Movie). To measure the number of MTs reaching the cortex, we used kymograph analysis over a 5-sec period and counted the number of EBP-2-GFP dots crossing an arc drawn 1.5 μm (10 pixels) inside the cortex. In mutant embryos, fewer MTs grew out to reach the cortex than in controls (p = 0.001; Fig 6H). Consistently, only a small portion of MTs nucleated by mutant centrosomes grew beyond the midpoint between the centrosome and cortex compared to controls (p = 0.029; Fig 6G). In fact, astral MT growth rates are significantly slower in mutant embryos (0.70 μm/sec ± 0.007) than in controls (0.88 μm/sec ± 0.013, p = 0.0014) (S5E Fig). Thus, MT growth appears to be impeded in the mutant embryo, presumably due to an excess of MT nucleation.
In C. elegans embryos, it has been shown that fast MT growth is subject to the MT stabilizing complex (ZYG-9/TAC-1) and the amount of free tubulin [47]. atx-2 mutant embryos, however, exhibit a significant increase in both centrosomal and overall TAC-1 levels (S5F and S5G  Fig), indicating that defective MT growth in atx-2 mutants is unlikely due to insufficient MT stabilization. Studies in other systems also showed that the amount of free tubulin influences the rate of MT polymerization in vitro [49,50] and such mechanism appears to be pertinent in C. elegans embryos [10,47,51]. We thus hypothesized that atx-2 mutant centrosomes possessing increased PCM nucleate more MTs, reducing the supply of free tubulin available for MT polymerization. Subsequently MT growth is interfered, resulting in shorter than normal MTs and cytokinesis failure. To test this, we generated atx-2 mutants overexpressing GFP-Tubulin to see if increasing Tubulin levels in the atx-2 mutant could partially rescue cytokinesis defects. As predicted, the incidence of cytokinesis failure is significantly (p<0.001) reduced in GFP-Tubulin overexpressing atx-2 embryos compared to control mutant embryos (Fig 7A). This partial rescue of cytokinesis failure further led to a significant decrease (p<0.001) in embryonic lethality (Fig  7B), suggesting that MT growth in atx-2 embryos is partly affected by limited supply of free tubulins, likely resulting from excessive MT nucleation. Next, we examined atx-2 mutant embryos overexpressing GFP-γ-Tubulin, a PCM factor, to see if further enhancing MT nucleation by increasing γ-Tubulin levels could exacerbate cytokinesis defects through worsened MT growth. Consistent with our hypothesis, cytokinesis defects in mutant embryos were significantly increased by GFP-γ-Tubulin overexpression compared to control mutants, leading to an increase in embryonic lethality (Fig 7A and 7B). Together, our data suggest that excessive MT nucleation and subsequent MT growth defects result in cytokinesis defects in atx-2 mutants.
Together, our results suggest a model in which loss of ATX-2 leads to elevated PCM levels at centrosomes and excessive MT nucleation, resulting in MT growth defects and subsequent cytokinesis failure (Fig 7H). Consistently, embryos depleted of ATX-2 exhibit a prominent cytokinesis failure phenotype, accompanied by enlarged centrosomes that cause excessive MT nucleation and aberrant MT growth, which results in cytokinesis failure and embryonic lethality. Thus, the proper level of centrosomal PCM factors is critical for normal MT nucleating activity to support normal MT dynamics. In this pathway, ATX-2 acts upstream to establish the proper centrosome size and MT nucleating activity.

ATX-2 forms a complex with SZY-20 and acts as a negative regulator of centrosome assembly in C. elegans embryos
In this study, we identified that ATX-2, together with PAB-1, physically associates with SZY-20 in vivo. While RNA-binding proteins ATX-2 and SZY-20, each containing unique RNA-failure is significantly reduced in atx-2 embryos overexpressing GFP-Tubulin (29%, n = 280). In contrast, cytokinesis failure is enhanced in atx-2 embryos overexpressing GFP-γ-Tubulin (68%, n = 172), while wild-type embryos (n = 63-109) overexpressing either GFP-protein exhibit no cytokinesis defects. (B) Embryonic lethality was scored at semirestrictive (20˚C) and restrictive (22˚C) conditions for atx-2 mutants. Overexpression of GFP-Tubulin in the mutant leads to a small but significant (***p< 0.0001) decrease in embryonic lethality: 78% (n = 3852) and 98.6% (n = 1949) in atx-2 vs 69% (n = 2776) and 95% (n = 1829) in atx-2; GFP-Tubulin at 20 and 22˚C, respectively. However, overexpressing GFP-γ-Tubulin leads to a small increase (p>0.05) in embryonic lethality: 81% (n = 3087) and 99% (n = 1149) at 20 and 22˚C, respectively, compared to the atx-2 mutant alone. (C) At 20˚C, tbg-1(RNAi) in atx-2 mutant embryos (10.4%, n = 235) decreased cytokinesis defects compared to atx-2 mutants (19.8%, n = 270), while tbg-1(RNAi) produced cytokinesis failure (67%, n = 331) in wild-type embryos. At 22˚C, the same trends were observed: 38 binding motifs (LSm, SUZ and SUZ-C, respectively), form a complex, these two proteins appear to physically interact independently of RNA. While it remains unknown what RNA molecules associate with these RNA-binding proteins, each RNA-binding protein in a ribonucleoprotein (RNP) complex might recruit a specific group of RNA through each own RNAbinding motif. In fact, Ataxin-2 has been shown to bind directly to mRNAs through its LSm domain and promote the stability of transcripts independently of its binding partner, poly (A)binding protein (PABP) in flies and humans [25,26]. Our prior study in C. elegans embryos showed that SUZ and SUZ-C RNA-binding motifs in SZY-20 exhibit RNA-binding capacity in vitro and that mutating these domains perturbs in vitro RNA-binding of SZY-20 and its capacity to regulate centrosome size in vivo [10]. However, there is no evidence that RNA-binding role of ATX-2 is directly involved in centrosome regulation, while C. elegans ATX-2 is shown to function in translational regulation during germline development [29].
In a multi-protein complex, ATX-2 and SZY-20 function closely to regulate cell division and centrosome assembly. While knocking down ATX-2 phenocopies a loss of function szy-20 (bs52) mutation, the different degree of phenotypic penetrance in atx-2(RNAi), a strong loss of function atx-2(ne4297) mutation, and a hypomorphic szy-20(bs52) mutation suggests a dosedependent regulation of ATX-2. Double knockdown by combining atx-2(RNAi) and szy-20 (bs52) mutation further enhances embryonic lethality, cytokinesis failure, the restoration of centrosome duplication to zyg-1(it25) embryos and the levels of centrosome-associated factors (ZYG-1, SPD-5, γ-Tubulin). Furthermore, the atx-2(ne4297) mutation produces a similar effect to that of atx-2(RNAi) combined with the szy-20(bs52) mutation. Our data thus suggest that atx-2 exhibits a positive genetic interaction with szy-20 in regulating cell cycle and centrosome assembly. In this pathway, SZY-20 acts upstream of ATX-2 to promote ATX-2 levels, with both acting upstream of zyg-1. We propose that SZY-20 influences centrosome size and MT dynamics indirectly through ATX-2 [10]. While it remains unclear whether C. elegans ATX-2 directly acts on ZYG-1, it has been shown that Plx4 (a Xenopus homolog of Plk4/ZYG-1) forms a complex with Atxn-2 (Xenopus homolog of ATX-2) [52]. Although a direct role for Atxn-2 in centrosome assembly has not been demonstrated, a physical connection between Xenopus Atxn-2 and Plx4 suggests a possible role of Xenopus Atxn-2 in centrosome assembly, via a mechanism that seems likely to be conserved between nematodes and vertebrates.

ATX-2 negatively regulates centrosome size in a dose-dependent manner
Our data indicate that ATX-2 acts as a negative regulator of centrosome size. A priori, increased PCM levels at atx-2 mutant centrosomes could be achieved by several mechanisms. First, centrosome factors might be overexpressed in atx-2 mutant cytoplasm via translational control, leading to increased recruitment of these factors to the centrosome by equilibrium, as shown by Decker et al., [53]. Second, atx-2 mutants might enhance the recruitment of factors to centrosomes post-translationally, without affecting overall levels of these factors. Third, loss of ATX-2 might promote local translation near centrosomes, leading to locally enriched centrosome factors. The first scenario is unlikely because our quantitative analyses reveal no significant changes in overall levels of centrosome factors (SPD-2, SAS-6, γ-Tubulin) or cytoplasmic levels of ZYG-1 by loss of ATX-2. Our current data do not differentiate between the second and third scenarios, but certain observations in C. elegans and other systems are consistent with the latter. It has been shown that Ataxin-2 assembles with polyribosomes and that ribosomes are associated with the MT cytoskeleton [25,54]. Furthermore, neuronal RNA granules are shown to contain translational machinery, allowing local translation upon arrival of transcripts at the right location [55]. We also observed ribosomal protein S6 spatially associated with C. elegans centrosomes (S3D Fig), suggesting the possibility of local translation around the centrosome. Indeed mRNA localization linked to the cytoskeleton and the following local translation is shown to be an efficient means of concentrating proteins at the functional site [19,56,57]. For example, spatial control of β-actin translation is executed by localizing its transcripts to actin-rich protrusions, which facilitates neuronal outgrowth [56]. Also, RNA-binding protein (CPEB/maskin) mediated localization of cyclin B1 transcripts to the mitotic apparatus leads to locally enriched Cyclin B proteins in the vicinity of spindles and centrosomes, supporting cell cycle progression [19]. Together such mechanism has been demonstrated to provide a tight control for temporal and spatial translation. How then, might ATX-2 regulate translation? For negative translational control, it has been proposed that Ataxin-2 binding to PABP inhibits translation by blocking the interaction between PABP and translational machinery or by directly blocking translation of mRNA targets at the initiation stage [58,59]. Alternatively, ATX-2 may facilitate the interaction between a microRNA and its mRNA target, leading to translational inhibition [60]. ATX-2 might also mediate translational control via its RNA-binding role as a component of the RNP complex comprising SZY-20/ATX-2/PAB-1. In fact, C. elegans PAB-1 is shown to associate with stress granules and processing bodies (P-bodies) that have been implicated in translational repression [38]. It has been observed that RNP complexes including P-bodies are involved in subcellular targeting of RNAs and precise timing of local translation at the final subcellular destination [58,61]. Further identification of specific RNAs that bind ATX-2 and/or SZY-20 will help to understand how RNA-binding role of an ATX-2 associated RNP complex plays a role in defining proper centrosome size.

ATX-2 contributes to proper MT nucleating activity of centrosomes and MT growth through γ-Tubulin
We have shown here that ATX-2 plays an essential role in cell division. atx-2 mutant embryos with enlarged centrosomes exhibit multiple cell division defects including cytokinesis failure and aberrant spindle positioning. In animal cells, contact between astral MT and the cortex is critical for the initiation of the cleavage furrow that is required for proper cytokinesis [62,63]. Thus cell cycle defects in atx-2(ne4297) embryos might associate with aberrant MT behavior, likely due to enlarged centrosomes. Recent work showed an additional role of ZYG-1 in regulating centrosome size, independently of its role in centriole duplication [10]. PCM factors SPD-2, SPD-5 and γ-Tubulin are known to positively regulate the MT nucleating activity of the centrosome [42,45,46,64]. atx-2 mutant centrosomes possessing increased (2-5 folds) levels of centrosome factors nucleate the drastically increased number of MTs. Such increase in MT nucleation by mutant centrosomes may lead to a substantial reduction in cytoplasmic free tubulins available for timely MT polymerization. Given the rapid cell divisions (~20 min/cell cycle) and high MT growth rate (0.88 μm/sec) in early C. elegans embryos, timely and sufficient supply of free tubulins in the cytoplasm must be immensely critical for proper cell cycle progression. In support of this, either overexpressing Tubulin or reducing positive regulators (ZYG-1 or γ-Tubulin) of MT nucleation partially restores normal cytokinesis, MT growth and embryonic viability to atx-2 mutants. Partial restoration of normal MT-dependent processes correlates with proper levels of centrosomal γ-Tubulin. Therefore, enlarged centrosomes are likely to be a primary cause of embryonic lethality in atx-2 mutants. Mutant centrosomes possess significantly increased levels of γ-Tubulin, and reducing its levels at mutant centrosomes partially reinstalls free tubulins available for MT growth, which in turn restores normal cell divisions and embryonic viability in atx-2 mutants. Our data suggest that ATX-2 contributes to MT nucleation and MT dynamics, in part, through γ-Tubulin at centrosomes, although it remains unknown how RNA-binding protein, ATX-2, regulates centrosomal levels of γ-Tubulin. Given our observation that overall levels of γ-Tubulin are not altered in atx-2 mutant embryos, it is curious how γ-Tubulin levels are locally enriched at the centrosome, perhaps via the regulation of recruitment to the centrosome or locally enriched translational control. Thus, it will be interesting to see if tbg-1 transcripts are elevated at the proximity of centrosomes by in situ hybridization. A recent study in Xenopus egg extracts reported that RNase treatment leads to defective spindle assembly due to hyperactive MT destabilizer MCAK (the ortholog of C. elegans KLP-7) by depleting RNA, suggesting a direct involvement of RNA in regulating MT organization [65]. Interestingly, we also found that atx-2 mutants show increased levels of KLP-7 at centrosomes and cytoplasm (S5H Fig), and that atx-2 and klp-7 exhibit a synergistic effect on MT dynamics and embryonic lethality (S6 Fig).
Another centriole factor, SAS-4, positively regulates centrosome size in C. elegans [66]. Interestingly, while we identify ATX-2 as a negative regulator of centrosome size, depleting ATX-2 had no effect on centriolar SAS-4 levels (S4G Fig), suggesting that ATX-2 regulates centrosome size independently of SAS-4, perhaps through a separate pathway. Thus, a balance between positive (e.g., SAS-4) and negative (e.g., ATX-2) regulators may contribute to establish proper centrosome size and MT nucleating activity, which in turn influences MT dynamics during the cell cycle. Improper levels of ATX-2 disrupt this balance, resulting in deregulated MT dynamics and subsequently abnormal cell divisions.
In summary, our work uncovers a role for ATX-2, the C. elegans ortholog of human Ataxin-2 in regulating centrosome size and MT cytoskeleton. In this pathway, SZY-20 positively regulates levels of ATX-2, which contributes to defining the proper centrosome size, leading to proper levels of MT nucleation and subsequent MT growth. While human Ataxin-2 and its poly Q stretch have been implicated in spinocerebellar ataxia [20], it remains largely unknown how this RNA-binding protein is linked to human diseases. Many RNA-binding proteins in neuronal RNA granules have been implicated in neuronal functions and growth via controlled local translation [55,58], suggesting a link between RNA-binding role and neuronal activity. In this regard, our work provides insights into a mechanistic link between the RNA-binding role of Ataxin-2 and the MT cytoskeleton.

Strains and genetics
All C. elegans strains were grown on MYOB plates seeded with E. coli OP50, and were derived from the wild-type Bristol N2 strain [67]. All strains were maintained at 18 or 20°C unless otherwise indicated. Transgenic strains were generated by standard particle bombardment transformation [68] and the CRISPR/Cas-9 method (S2 Table) [39,40]. The list of worm strains used in this study is shown in S1 Table. Standard genetics were used for strain construction and genetic manipulations [69]. RNAi feeding was performed as described and L4440 clone with the empty dsRNA-feeding vector served as a negative control [70].
Immunofluorescence and confocal microscopy were performed as described in [10]. For confocal microscopy, MetaMorph software (Molecular Devices) was used to acquire images from a Nikon Eclipse Ti-U microscope equipped with a Plan Apo 60 X 1.4 NA lens, a Spinning Disk Confocal (CSU X1) and a Photometrics Evolve 512 camera. Fluorescence intensity measurements were made using the MetaMorph software, and image processing with Photoshop CS6.
For quantification of centrosomal signals, the average intensity within a 25-pixel (pixel = .151 μm) diameter region was measured throughout a region centered on the centrosome. For each region, a focal plane with the highest average intensity was recorded. The same sized regions were drawn in the cytoplasm for cytoplasmic signal, and outside the embryo for background subtraction unless otherwise indicated. For centriolar signals, analysis was done the same way, but a 5-pixel diameter region was used.
For EBP-2-GFP movies, pre-mitotic embryos were selected and monitored under DIC until before metaphase I. Upon entrance into metaphase, 61 images were captured every 500 msec for a 30 sec period. For spindle MTs, a 355°circular region was drawn around each centrosome, and kymographs were generated for the first 5 seconds of recordings. Linescan measurements of the kymographs were taken to obtain average EBP-2 signal along every point of the circular region. Values for each linescan were averaged and plotted on a graph. To measure MT nucleation, the fluorescence intensity was quantified at a line region drawn in a semi-circle (25 pixel radius) around the centrosome. Linescan (5 pixel wide) measurements were taken in 5-sec time projections from recordings. Pixel intensities were generated for each point along the line and averaged for comparison. For background subtraction, cytoplasmic intensity in the same size region was used. To quantify the number of growing astral microtubules proximal to the centrosome, line regions were drawn in a semi-circle (25 pixel radius) around the centrosome. For midpoint measurements, line regions were drawn at a radius of 60 pixels from the centrosome. Kymographs were generated for the first 5 seconds of each movie. Individual pixel measurements from the kymographs were obtained, and pixels with GFP intensity over 40 were counted. Cortex measurements were obtained by generating 5 sec kymographs of a region drawn at 1.5 μm inside of the cell cortex. The number of EBP-2 signal was counted manually. For MT polymerization rate, 4.5 μm-long lines were drawn along the individual EBP-2-GFP tracks outward to the cortex. Kymographs were generated, and rate was calculated by distance over time (μm/sec).
Immunoprecipitation and quantitative western analysis 20-25 μl of Dynabeads Sheep-anti-Rabbit, Protein A magnetic beads (Invitrogen) or mouseanti-GFP magnetic beads (MBL) were used per IP reaction. Beads were washed 2x 15 min in 1 ml PBS with 0.1% Triton-X (PBST). For SZY-20 IP, beads were incubated overnight at 4°C in a ratio of 1mg/ml (α-SZY-20 or α-HA/beads volume) and 3 mg/ml (α-IgG/beads volume). Embryos were collected by bleaching worms grown in liquid culture, snap-frozen in liquid nitrogen, and stored at -80°C until use. Embryo pellets were ground up in microcentrifuge tubes in equal volumes of worm lysis buffer (50 mM HEPES; pH 7.4, 1 mM EDTA, 1 mM MgCl2, 200 mM KCl, 10% glycerol) [71] containing complete protease inhibitor cocktail (Roche) and MG132 (Tocris). For RNase A or RNase Inhibitor treatment, RNase (10 μg/ml, Roche) or RNase Inhibitor (200U/ml, Roche) was added to lysis buffer prior to grinding. Embryos were ground for 5 min, and sonicated for 3 min. The lysate was then centrifuged at 4°C for 2x 20 min at 15K rpm in a desktop centrifuge, and the supernatant were collected. For enzyme reaction, lysates were incubated for 15 min at room temperature (RT). Protein concentration was determined and adjusted before IP. Beads were washed for 2x 15 min with PBST + 0.5% BSA at RT, followed by 1x 15 min wash with 1X worm lysis buffer. Beads were resuspended in 1X lysis buffer, mixed with embryonic lysates and incubated at 4°C for 1 hour.

Supporting Information
(C) zyg-1(RNAi) almost abolishes HA-ZYG-1 signals. Compared to controls, atx-2(RNAi) embryos exhibit increased levels (~2-fold, n = 10) of centrosome-associated HA-ZYG-1 signal at late mitosis. (D) The specificity of γ-Tubulin antibody: Immunostaining and immunoblot reveal that anti-γ-Tubulin specifically detects endogenous γ-Tubulin, as tbg-1(RNAi) leads to a significant reduction in γ-Tubulin signals. (E) The specificity of ZYG-1 antibody: ZYG-1 is enriched at centrioles. zyg-1(RNAi) results in a drastic reduction in both centriolar and cytoplasmic ZYG-1 signals. zyg-1(RNAi) leads to monopolar mitotic spindles at the second mitosis (bottom right). (F) SPD-5 is not required for ZYG-1 localization at centrosomes. ZYG-1 localizes to centrosomes in spd-5(RNAi) embryos that exhibit cell cycle arrest at the first mitosis (Hamill et al., 2002). Quantification of ZYG-1 levels at centrosomes reveals no significant change in ZYG-1 levels by loss of spd-5: Mitotic arrest in spd-5(RNAi) embryos is likely to contribute to more frequent detection of higher fluorescence intensity as ZYG-1 levels at centrosomes peak at late mitosis. (G) SAS-4 levels are not affected by loss of atx-2: SAS-4 stained embryos display SAS-4 localization at early cell cycle stages. Note cell cycle defects (e.g., DNA segregation, spindle positioning, cytokinesis) in atx-2 mutants. Quantification of SAS-4 signals shows that atx-2 mutant embryos exhibit similar levels of SAS-4 at centrosomes (CE) and cytoplasm (Cyto) to those of wild-type embryos at metaphase. Insets are magnified 3-fold. Bar, 5μm. embryos. (C) atx-2(RNAi) embryos display a nearly 3-fold increase in centrosomal GFP-γ-Tubulin but no change in cytoplasmic levels. (D) Line scans of the kymographs on embryos immunostained for MTs show that atx-2 mutant embryos exhibit fewer MT-signals at the midpoint (9.6 μm away from centrosomes) and near the cortex (1.5 μm inside cortex) compared to wild-type controls, which is consistent with the analyses performed using EBP-2-GFP in Fig 6. Each dot on the graph represents an embryo. Horizontal bars indicate average values, ÃÃÃ p<0.001. (E) Polymerization velocity: Kymographs created along the growing MT. Lines (4.5 μm) were drawn along the individual EBP-2-GFP track extending outward to the cortex. Astral MT growth rates are significantly lower in mutant embryos (0.70 μm/sec ± 0.007) than in controls (0.88 μm/sec ± 0.013, ÃÃ p = 0.0014). (F) atx-2 mutant embryos exhibit increased levels of TAC-1. Quantitative immunoblot analysis using embryonic extracts of wild-type and atx-2(ne4297) mutants expressing GFP-TAC-1 [73] with α-GFP and α-TAC-1 [72] shows that mutant embryos possess increased levels of both GFP-TAC-1 and endogenous TAC-1. Additional analysis using wild-type and atx-2 embryos confirms that atx-2 embryos contain elevated levels of TAC-1 (n = 5). (A) 5-sec time projections of 500 msec interval live imaging of embryos expressing EBP-2-GFP at first metaphase illustrate the effects on MT behavior in RNAi-treated atx-2 mutant, compared to wild-type embryos. tbg-1(RNAi) slightly reduces EBP-2 signals at atx-2 mutant centrosomes whereas tbg-1(RNAi) drastically decreases EBP-2 signals at wild-type centrosomes (n>30), illustrating the effects on MT nucleation by tbg-1(RNAi). In contrast, klp-7(RNAi) increases centrosomal EBP-2 signals in both wild-type and mutant embryos. KLP-7 knockdown also affects cytoplasmic MT growth patterns. Magnified regions illustrate that while MTs emanate from the centrosome and reach the cortex in the wild-type embryo, EBP-2 tracks in mutant embryos appear randomly oriented. (B) Kymograph analysis in klp-7(RNAi) vs control treated wild-type and atx-2 mutants: Levels of EBP-2-GFP signals relative to control RNAi are presented for wild-type and atx-2 mutants, respectively. While klp-7(RNAi) increases MT nucleation by over 2-fold in wild-type ( ÃÃ p = 0.004), relative increase in MT nucleation by klp-7(RNAi) is much lower (1.2-fold) in control atx-2 mutants (p = 0.6). This difference in MT nucleation by klp-7(RNAi) suggests a limited supply of free tubulins available for MT nucleation by mutant centrosomes. The same trends are observed for the midpoint measurement. Interestingly, relative increase in the number of EBP-2 foci near the cortex is higher (p = 0.1) in atx-2 mutants, unlikely due to lengthened MTs growing out from centrosomes. (C) Embryos expressing GFP-Tubulin reveal that in klp-7(RNAi) in the wild-type embryo, MT nucleation is increased and MTs appear to grow longer, originating from centrosomes. However, while more MTs (based on the GFP-Tubulin signal) are found in the cytoplasm distant from the centrosome in klp-7(RNAi); atx-2 embryos compared to L4440; atx-2 embryos, these MTs seem to be disorganized and appear disconnected from the centrosome, which might explain the increase we observe in EBP-2 signal by cortex kymograph analysis, as opposed to an increase in MT growth. These abnormal patterns of astral MTs can be reminiscent of the result previously observed by Srayko et al., [47], in which freely moving MTs are detached from the centrosome owing to an excessive MT nucleation by KLP-7 depletion. Embryos expressing GFP-Tubulin reveal that tbg-1(RNAi) significantly decreases the number of MTs growing out from wild-type centrosomes, while tbg-1(RNAi)-treated mutant centrosomes still exhibit a decent level of MT nucleation although centrosomal MT signal is reduced compared to control. (A, C) Boxes are magnified 2-fold. (D) klp-7(RNAi) in atx-2 mutants leads to a synergistic increase in embryonic lethality. klp-7(RNAi); atx-2(ne4297) animals produced~90% of embryonic lethality compared to 30% in atx-2(ne4297) and only 6% in klp-7(RNAi) animals, suggesting that a small increase in the already aggravated MT nucleation and freely moving detached MTs by KLP-7 depletion negatively impacted embryogenesis. This increase in embryonic lethality might be due to enhanced cytokinesis failure resulting from exacerbated MT nucleation. ÃÃÃ p<0.001 (E) Knocking down both KLP-7 and ATX-2, indeed, leads to a significant increase in cytokinesis failure compared to single depletion. ÃÃ p<0.01 Each dot on the graph represents a centrosome. Horizontal bars indicate average values. Error bars are SD. Bar, 5 μm. (TIF) S1 Movie. ATX-2 is required for proper cell division. Embryos expressing GFP-α-Tubulin, mCherry-γ-Tubulin and mCherry-H2B. The atx-2(RNAi) embryo contains extra DNA, presumably resulting from polar body extrusion failure during meiosis (see S3 Movie). Before pronuclei meet, the spindle starts to form only around the sperm pronucleus, followed by later incorporation of the maternal DNA into the spindle. The atx-2(RNAi) embryo displays abnormal spindle positioning, improperly positioned axis of the metaphase plate, and lagging DNA at first anaphase. The atx-2(RNAi) embryo duplicates centrosomes properly but fails to complete cytokinesis (see S2 Movie), forming a tetrapolar spindle at the second mitosis. Note that in the atx-2(RNAi) embryo, centrosomes labeled with mCherry-γ-Tubulin is significantly larger than in the control. Also note shorter metaphase spindle and cell cycle delay in atx-2 (RNAi) embryo compared to the control. Each frame is equal to 1 min of elapsed time (t = 0 at first metaphase; 5 fps). Bar, 5 μm. (MOV) S2 Movie. ATX-2 is required for normal cytokinesis. Embryos expressing GFP-MOE marking Actin. Arrows indicate initial position of cytokinetic cleavage furrow. In the control embryo, the cleavage furrow starts to form on both dorsal and ventral (DV) sides of the embryo symmetrically, and extend toward the center of the embryo, then meet at the center to complete cytokinesis. In the atx-2(RNAi) embryo, the cleavage furrow only forms on one side of the embryo, and progresses to the center of the cell but is stalled at two-thirds of DV axis, resulting in cytokinesis failure. In the atx-2(RNAi) embryo, cortical distribution of actin appears disorganized and scattered. Note that in the atx-2(RNAi) embryo, position of the cleavage furrow is not sustained throughout cytokinesis. Movies are Z-projections of the middle 2-3 μm of the embryo. Each time frame is equal to 30 sec of elapsed time. (t = 0 at first indication of cleavage furrow formation; 5 fps). Bar, 5 μm. (MOV) S3 Movie. The atx-2 mutant embryo fails to extrude polar bodies, resulting in extra DNA. Control and atx-2(RNAi) embryos expressing GFP-α-Tubulin, mCherry-γ-Tubulin and mCherry-H2B. While the control embryo displays successful polar body extrusion, in the atx-2 (RNAi) embryo the extrusion of the second polar body fails (arrow at t = 12), leading to extra DNA associated with the maternal pronucleus. Note cell cycle delay and the orientation of meiotic spindle in the atx-2(RNAi) embryo-the meiotic spindle aligns 90°to the cell cortex throughout the meiosis II. However, in the control embryo the meiotic spindle aligns parallel to the cortex until anaphase and rotates 90°just before polar extrusion [74]. Each frame is equal to 1 min of elapsed time (t = 0 at metaphase of meiosis II; 5 fps). Bar, 5 μm. (MOV) S4 Movie. ATX-2-GFP localizes to cytoplasm and punctate cytoplasmic foci. A transgenic strain expressing atx-2-gfp-3x flag [75]. ATX-2-GFP localization is consistent with endogenous ATX-2 detected by anti-ATX-2. Each frame is equal to 2 min of elapsed time; 3 fps. Bar, 5 μm. (MOV) S5 Movie. atx-2(RNAi) results in increased levels of ZYG-1 at the centrosome. Embryos expressing GFP-ZYG-1-C-terminus [41]. Throughout the cell cycle from pronuclear migration to first anaphase, centriolar and PCM-like signals of GFP-ZYG-1 are much more intense in the atx-2(RNAi) embryo. Each frame is equal to 1 min of elapsed time (t = 0 at first metaphase; 5 fps. Bar, 5 μm. (MOV) S6 Movie. atx-2(ne4297) embryos exhibit elevated levels of centrosomal ZYG-1. Embryos expressing GFP-ZYG-1-C-terminus [41]. The atx-2 mutant embryo shows the similar pattern to atx-2(RNAi) embryos: atx-2 mutant embryos display increased levels of centriolar and PCM-like GFP-ZYG-1 throughout the first cell cycle. Each frame is equal to 1 min of elapsed time (t = 0 at first metaphase; 5 fps). Bar, 5 μm. (MOV) S7 Movie. atx-2 mutant embryos exhibit abnormal microtubule dynamics. Embryos expressing EBP-2-GFP that marks microtubule plus ends. More microtubules emanate from atx-2 mutant centrosomes, compared to the wild-type. However, only few of the microtubules that grow out from the mutant centrosome continue to grow out and reach the cortex. Note that the atx-2 embryo exhibits increased levels of centrosomal EBP-2 signal and shorter mitotic spindles, compared to the wild-type. Movies are acquired at 500 msec intervals over a 10 secperiod; 6 fps Bar, 5 μm. (MOV) S1