Actin organization and endocytic trafficking are controlled by a network linking NIMA-related kinases to the CDC-42-SID-3/ACK1 pathway

Molting is an essential process in the nematode Caenorhabditis elegans during which the epidermal apical extracellular matrix, termed the cuticle, is detached and replaced at each larval stage. The conserved NIMA-related kinases NEKL-2/NEK8/NEK9 and NEKL-3/NEK6/NEK7, together with their ankyrin repeat partners, MLT-2/ANKS6, MLT-3/ANKS3, and MLT-4/INVS, are essential for normal molting. In nekl and mlt mutants, the old larval cuticle fails to be completely shed, leading to entrapment and growth arrest. To better understand the molecular and cellular functions of NEKLs during molting, we isolated genetic suppressors of nekl molting-defective mutants. Using two independent approaches, we identified CDC-42, a conserved Rho-family GTPase, and its effector protein kinase, SID-3/ACK1. Notably, CDC42 and ACK1 regulate actin dynamics in mammals, and actin reorganization within the worm epidermis has been proposed to be important for the molting process. Inhibition of NEKL–MLT activities led to strong defects in the distribution of actin and failure to form molting-specific apical actin bundles. Importantly, this phenotype was reverted following cdc-42 or sid-3 inhibition. In addition, repression of CDC-42 or SID-3 also suppressed nekl-associated defects in trafficking, a process that requires actin assembly and disassembly. Expression analyses indicated that components of the NEKL–MLT network colocalize with both actin and CDC-42 in specific regions of the epidermis. Moreover, NEKL–MLT components were required for the normal subcellular localization of CDC-42 in the epidermis as well as wild-type levels of CDC-42 activation. Taken together, our findings indicate that the NEKL–MLT network regulates actin through CDC-42 and its effector SID-3. Interestingly, we also observed that downregulation of CDC-42 in a wild-type background leads to molting defects, suggesting that there is a fine balance between NEKL–MLT and CDC-42–SID-3 activities in the epidermis.

The genome of the nematode Caenorhabditis elegans encodes four NEK-like proteins, termed NEKL-1-NEKL-4 [25]. NEKL-3 is a close ortholog of mammalian NEK6 and NEK7, whereas NEKL-2 is most similar to NEK8 and NEK9 [25,26]. We have previously shown that NEKL-2 and NEKL-3, together with three conserved ankyrin repeat partners, MLT-2/ANKS6, MLT-3/ANKS3, and MLT-4/INVS, are required for the completion of molting [25][26][27]. During molting, larvae release their old apical extracellular matrix (ECM), termed the cuticle, and synthesize a new one underneath, thereby allowing for further growth and development [28]. Whereas null mutations in nekls or mlts lead to a complete failure to shed the old cuticle during the first larval molt, hypomorphic alleles typically arrest during the second molt and exhibit a partial release of the old cuticle [25][26][27]. Our previous studies have suggested that the role of NEKL-MLT proteins in molting may be linked to their regulation of endocytic trafficking, which is critical for the normal molting process [25][26][27]. Whether the perturbation of normal trafficking is primarily responsible for molting defects in nekl and mlt mutations, however, is currently unresolved.
In this study, we further clarify the mechanism through which the NEKL-MLT pathway is involved in molting. More specifically, by using two independent screens, we have linked the NEKL-MLT pathway with the Rho-like GTPase, CDC-42/CDC42, and its conserved effector protein, SID-3/ACK1. CDC42 was initially identified as an essential cell polarity and budding regulator in the yeast Saccharomyces cerevisiae and its activity is in large part controlled through its association with GTP (in the active state) and GDP (in the inactive state) [29,30]. Studies in mammals have shown that CDC42 becomes activated in response to different cdc-42 suppression of NIMA kinases intracellular and extracellular signals and, depending on the type of the stimulus, can interact with a wide variety of effectors to control diverse cellular functions. These include actin reorganization and polarization, intracellular trafficking, polarized growth, microtubule polarization, and septin organization [31][32][33][34][35][36][37][38][39][40][41]. CDC42 is thus considered to be a key regulator in the establishment and maintenance of cellular polarity, a function that is essential for normal proliferation, differentiation, and morphogenesis [34]. ACK1 is a non-receptor tyrosine kinase and serine/threonine protein kinase that is implicated in cell morphology, migration, survival, growth, and proliferation [42][43][44]. More specifically, ACK1 functions in trafficking and clathrin-mediated endocytosis through binding to the epidermal growth factor receptor and clathrin [45][46][47], and also regulates the formation of actin filaments by phosphorylating both the Wiskott-Aldrich syndrome protein (WASP) and cortactin [48,49]. In addition, ACK1 has been reported to promote the activation state of CDC42 [50]. In C. elegans, mutations in sid-3 affect the transfer of dsRNA between cells, suggesting a possible connection to endocytosis and RNA import [51]. Nevertheless, the functions of SID-3 are not well characterized in C. elegans, and previous studies have not linked SID-3 functions to CDC-42 in this species.
Our study for the first time functionally connects CDC-42/CDC-42 and SID-3/ACK1 with the NIMA-related kinase network. Moreover, we demonstrate that these pathways are critical for the regulation of actin-dependent ECM remodeling and vesicular trafficking in the polarized epithelium of C. elegans.

Inhibition of CDC-42 suppresses molting defects in nekl-3 mutants
We have previously shown that mutations in components of the NEKL-MLT network lead to a spectrum of molting defects during larval development. In the case of nekl-3(sv3) hypomorphic mutants, larvae typically arrest at the second (L2/L3) molt and exhibit a stereotypical corset phenotype, whereby the unshed L2 cuticle constricts a significant portion of the central body region [25][26][27]. To gain insights into the underlying mechanisms controlling molting, as well as targets and components of the NEKL-MLT network, we carried out a screen for RNAi suppressors of nekl-3(sv3) molting defects. Among the genes identified, an RNAi clone targeting cdc-42 led to penetrant suppression of nekl-3(sv3), such that most treated animals progressed through L3 and L4 stages as compared with control animals that arrested at the L2/L3 transition (Fig 1A and 1B). We also tested the ability of cdc-42(RNAi) feeding to suppress defects in nekl-3(gk506) null mutants, which arrest during the L1/L2 molt and exhibit complete encasement within both the L1 and L2 cuticles [25]. In contrast to the nekl-3(sv3) hypomorphic allele, suppression of nekl-3(gk506) by cdc-42(RNAi) was relatively weak but did lead to an increase in the average width of treated animals (S1A Fig).

Repression of CDC-42 suppresses defects in multiple nekl-mlt mutant backgrounds
To test if a reduction in CDC-42 can suppress loss-of-function phenotypes in other members of the NEKL-MLT network, we examined the effects of cdc-42(RNAi) feeding on the hypomorphic alleles of nekl-2(fd91) and mlt-4(sv9). In addition, we tested cdc-42(RNAi) on nekl-2 (fd81); nekl-3(gk894345) double mutants. Notably, single mutants of nekl-2(fd81) and nekl-3 (gk894345) are aphenotypic, whereas~99% of double mutants arrest as larvae with molting defects [26]. On control RNAi plates, all three strains displayed near uniform L2/L3 arrest ( Fig  1C-1E), consistent with previous reports [25,26]. RNAi feeding of cdc-42 led to partial suppression of molting defects in all three strains, whereby animals developed through L3 and L4 stages and in some cases progressed to adulthood (Fig 1C-1E). Suppressed adult animals, however, appeared to be sterile, which is likely due to a requirement for cdc-42 in the germline [52,53]. In addition, we tested for cdc-42(RNAi) suppression of the nekl-2(gk839) null allele, which leads to arrest with a complete encasement phenotype at the L1/L2 molt [25][26][27]. Similar to findings for nekl-3(gk506), cdc-42(RNAi) led to a significant increase in larval width (S1B Fig), although suppression was quite weak. The inability of cdc-42(RNAi) to robustly suppress null alleles of nekl-2 and nekl-3 indicates that reduction of cdc-42 cannot fully compensate for the complete loss of NEKL activity. Because hypomorphic alleles are not available for the two additional characterized members of the nekl-mlt network, mlt-2 and mlt-3, we used a combined RNAi-feeding approach to assay for cdc-42 suppression. Animals were fed mixtures of bacteria expressing RNAi against either mlt-2 or mlt-3, together with either cdc-42 or control RNAi. Although suppression under these conditions was weaker than that observed for the hypomorphic alleles described above, cdc-42 (RNAi) led to a significant increase in body width (S2 Fig), which may be due to partial release of the unshed cuticle. Taken together, these results demonstrate that depletion of CDC-42 can alleviate molting defects in all tested members of the conserved nekl-mlt network.

CDC-42 function is required for molting in C. elegans
Given our observation that partial loss of cdc-42 activity can lead to suppression of molting defects in nekl-mlt mutants, we were interested to determine if perturbation of CDC-42 levels might have an effect on molting in an otherwise wild-type background. Injection of cdc-42 dsRNA, which induces a strong reduction in CDC-42 activity in the germline, leads to severe defects in the early embryonic development of F1 progeny because of disruptions in cell polarity and spindle orientation [52,54]. In contrast, cdc-42(gk388) null homozygous mutants derived from heterozygous mothers complete embryogenesis but arrest later in development, likely because maternal CDC-42 provided by the germline is sufficient for progression through early stages of development [52][53][54][55].
Interestingly, we found that 32% (n = 79) of cdc-42(gk388) homozygous progeny derived from heterozygous mothers displayed clear molting defects. Most commonly, cuticle detachments were observed in the head and tail regions, but in some cases shedding defects were visible in other locations (Fig 2). Based on P-cell divisions, the majority of cdc-42 homozygotes arrested at the L3 stage, although 14% of examined animals progressed to L4 or adult stages (n = 79). Injection of genomic DNA encoding wild-type cdc-42 into a balanced cdc-42(gk388) strain resulted in normal molting in the F1 generation of transformed larvae (n>20), demonstrating that the molting defects in this strain are specifically attributable to a lack of zygotic cdc-42 expression. Taken together, our findings indicate that whereas weak inhibition of CDC-42 can partially compensate for a reduction in nekl-mlt activities, CDC-42 is nevertheless required for normal molting.

CDC-42 is expressed in the epidermis and colocalizes with NEKL-MLT components
Because the epidermis is the primary tissue that modulates molting in C. elegans [25][26][27]56], we next examined the pattern of CDC-42 expression in the larval epidermis. A functional GFP::CDC-42 reporter (S1 Table) was expressed in epidermal syncytia (hyp1-hyp11) as well as in lateral specialized epithelial seam cells (Fig 3A). GFP::CDC-42 accumulated in rounded, rod-like, or irregularly shaped structures throughout the cytosol that were enriched in the apical region of hyp7. GFP::CDC-42 was also observed in ring-shaped structures, suggesting that CDC-42 is associated with the membranes of vesicles ( Fig 3A). Furthermore, smaller CDC-42 puncta accumulated at the boundary between the main epidermal syncytium hyp7 and seam cells (Fig 3A). We failed to observe any obvious differences in the pattern of GFP::CDC-42 expression in molting versus intermolt animals (Fig 3B), suggesting that CDC-42 localization may not be grossly altered during the molting program. We did, however, detect a statistically significant increase in mean levels of GFP::CDC-42 expression in molting versus intermolt animals, although these changes were relatively modest ( Fig 3C).

The NEKL-MLT network controls CDC-42 localization in the epidermis
Given the observed functional and spatial connections between CDC-42 and NEKL-MLT components, we next determined if the localization of CDC-42 is dependent on the NEKL-MLT network. Notably, GFP::CDC-42 epidermal localization was altered in 89% (n = 53) of nekl-2(fd91) and 79% (n = 53) of nekl-3(sv3) molting-defective larvae (Fig 4A-4C). In particular, apical expression of CDC-42::GFP was altered either throughout the entire epidermis or within limited regions. Changes included smearing (diffuse expression) of the GFP::CDC-42 marker as well as increased aggregation. Although GFP::CDC-42 was highly enriched at seam cell boundaries in nekl-2(fd91) and nekl-3(sv3) mutants, consistent with wild type, seam cells in these mutants were abnormally small and round relative to wild-type controls (Fig 4A-4C), suggesting a non-autonomous role for NEKL-2 and NEKL-3 in seam cell processes. In

CDC-42 activity is altered in nekl-2 and nekl-3 mutants
In addition to gross changes in CDC-42 expression, we were interested to determine if CDC-42 activation was affected by loss of nekl function. As a dynamic switch, CDC-42 cycles between GTP-bound active and GDP-bound inactive forms. To assay CDC-42 activity, we made use of two independent reporters of CDC-42 activation. We first examined a GST::GFP:: WSP-1(GBD) reporter that associates with GTP-bound active CDC-42 and is expressed under the control of the cdc-42 promoter [57]. In the case of wild type larvae we observed a punctate pattern of expression near the apical surface, which was very similar to our observations for the GFP::CDC-42 reporter ( Fig 4E). Notably, we observed both qualitative and quantitative differences in patterns of GST::GFP::WSP-1(GBD) in wild-type larvae and nekl-2(fd91) and nekl-3(sv3) single mutants. GST::GFP::WSP-1(GBD) tended to form aggregations in the mutants (Fig 4F and 4G). In addition, apical expression levels of GST::GFP::WSP-1(GBD) in mutants were nearly twice the levels observed in wild type larvae (Fig 4H), suggesting increased CDC-42 activity in the mutants.
To confirm these results, we used another reporter for CDC-42 activity, WSP-1(CRIB):: mCherry, which is expressed exclusively in the epidermis under the control of the lin-26 promoter [58]. In wild-type larvae, WSP-1(CRIB)::mCherry was primarily observed in small puncta and filament-like structures throughout the epidermis of molting and intermolt animals, with occasional localization within slightly larger aggregates (S5A Fig). The different expression pattern of this reporter in comparison to both GFP reporters for active and total CDC-42 may be due to difference in the fluorescence tag and its C terminal localization. Although we observed some variability in the pattern of WSP-1(CRIB)::mCherry between individual larvae, no consistent differences were detected between molting (n = 25) and intermolt (n = 31) animals. In contrast, we observed a strong increase in the size (S5 Fig) of WSP-1 (CRIB)::mCherry puncta in 93% of nekl-2(fd91) (n = 28) and 96% of nekl-3(sv3) (n = 28) larvae, as well as an increase in the number of puncta in some animals. This included the formation of very large accumulations that were never observed in wild type (S5 Fig). In contrast to nekl mutants, we observed a wild-type pattern for WSP-1(CRIB)::mCherry in 91% of fbn-1(tm290) molting-defective mutants (n = 21) (S5D Fig), and the size of mCherry puncta in fbn-1(290) mutants was much closer to that of wild type (S5 Fig), suggesting that the observed effects on CDC-42 activity are likely specific to perturbation of the NEKL-MLT network. Collectively, these findings indicate that NEKL-MLT components affect both CDC-42 subcellular localization as well as its activity levels.
We also tested the effect of expressing two predicted CDC-42 hyperactive variants (G12V and Q61L) [54,[59][60][61] in the epidermis using the dpy-7 promoter, which drives specific expression within epidermal cells [62]. Whereas expression of wild-type CDC-42 from transgenic arrays was well tolerated and gave rise to transmitting lines, both the G12V and Q61L CDC-42 variants were highly toxic, leading to severe morphological defects and accompanying embryonic and L1 larval lethality in most GFP-marked F1 progeny (n>1000). In one representative experiment using cdc-42(G12V) with the co-injection marker sur-5::GFP, we observed 57% embryonic lethality and 27% early L1 lethality (n = 165) of GFP-positive worms whereas cdc-42(Q61L) led to 55% embryonic lethality and 22% L1 lethality (n = 176). This level of lethality was observed using a range of DNA concentrations and is likely due to the deleterious effects of hyperactive CDC-42 on the epidermis of embryos undergoing morphogenesis. Approximately 20% of F1s did not arrest as embryos or L1 larvae but failed to produce stable transmitting lines. F1s that escaped early arrest were typically highly mosaic for expression of the GFP-containing array with a majority displaying some form of morphogenetic defect including molting deficits (S6 Fig). In some cases, we observed complete entrapment within cdc-42 suppression of NIMA kinases the old cuticle at the L2/L3 stage, whereas other worms failed to shed only a portion of the cuticle or became entrapped at later stages (S6 Fig). These findings are consistent with our above data and indicate that epidermal-expressed hyperactive CDC-42 can cause defects in molting. One caveat to these studies, however, is that the observed effects on molting could be indirectly due to defects in epidermal cells induced by the CDC-42 gain-of-function mutants earlier in development. Nevertheless, molting phenotypes were observed in some animals that did not display obvious morphological defects (S6 Fig), suggesting that these effects are separable.
Mutations in the C. elegans activated Cdc42-associated kinase 1 ortholog, SID-3, suppress molting defects in nekl-2 and nekl-3 mutants In a parallel forward genetic screen [63], we identified a mutation, fd139, that strongly suppressed molting defects in nekl-2(fd81); nekl-3(gk894345) double mutants (Fig 5A and 5B). Molecular identification of fd139 revealed it to affect sid-3, the ortholog of activated Cdc42-associated kinase 1 (ACK1). ACK1 was initially described as a positive regulator of CDC42 activity, although later studies have shown that it also functions as a CDC42 effector and has roles in actin regulation and intracellular trafficking [45][46][47][48][49][50]. In C. elegans, SID-3 regulates systemic RNAi and is expressed in the epidermis in a pattern that is similar to what we observed for CDC-42 [51].
fd139 is a 5-bp deletion in sid-3 exon 13, leading to a frameshift following Ser 989 of the SID-3 protein (B0302.1a). Using CRISPR/Cas9 methods, we generated five additional alleles of sid-3, all of which induced strong suppression of the corset phenotype in nekl-2(fd81); nekl-3 (gk894345) animals ( Fig 5A and S2 Table). This included two C-terminal (fd213 and fd214) and three N-terminal (fd218, fd219, fd221) alleles, of which the latter group caused frameshifts within 60 bp downstream of the start codon. Notably, none of these alleles has any obvious phenotype on their own, consistent with a previous report [51]. The ability of the N-terminal sid-3 alleles to suppress nekl-2(fd81); nekl-3(gk894345) mutants indicates that complete inactivation of SID-3 can suppress molting defects in the double mutants. These results further strengthen the functional connection between NEKL-MLT components and CDC-42 and also suggest that SID-3 cooperates with CDC-42 in the molting network. The viability of null mutations in sid-3, however, indicates that unlike CDC-42, SID-3 is not required for normal molting [51].

NEKL-MLT components are required for normal actin organization in the epidermis
In mammals, CDC42 and Ack1 have important roles in the regulation of the actin cytoskeleton. Importantly, epidermal actin is reorganized during molts to form circumferentially oriented arrays, which have been proposed, though not formally demonstrated, to be critical for molting [64]. The observation that inhibition of both cdc-42 and sid-3 can suppress defects in nekl-mlt mutants further suggested that the NEKL-MLT network may control actin localization during molting.
We first sought to recapitulate the observations of Costa et al. [64] using phalloidin staining of actin in fixed wild-type worms. We observed that intermolt L2 larvae generally showed very weak actin staining in the epidermis (Fig 6A), suggesting that actin in these animals may be largely diffuse. Occasionally, intermolt animals contained dispersed epidermal actin puncta as well as some accumulation at the seam cell boundary (Fig 6A). In contrast, phalloidin staining in molting animals was much stronger, and actin was clearly organized into parallel apical bundles, consistent with Costa et al. (Fig 6B) [64].
To test if epidermal actin organization is regulated by the NEKL-MLT network, we first examined actin staining in nekl-2(fd91) (n = 29) and nekl-3(sv3) (n = 18) hypomorphic animals. We failed to observe normal actin bundles in any of the examined animals from both genetic backgrounds. In most larvae, the pattern of actin differed dramatically from wild type in that puncta were more numerous and tended to form aggregations (Fig 6C and 6D). Moreover, although mutants occasionally contained short filamentous-like structures, they failed to complete the formation of parallel bundles observed in wild-type molting animals (S7 Fig). In 24% of nekl-2(fd91) and 11% of nekl-3(sv3) mutants we observed some actin filaments or rows of actin puncta in the dorsal and ventral areas of hyp7, usually above the body wall muscles (S7B and S7C Fig). Notably, patches with actin filaments were more frequently observed in body regions constricted by the old cuticle; it is possible that actin filaments formed and rapidly disassembled in the non-constricted regions. In addition, actin accumulations at the hyp7-seam-cell boundary were frequently undetectable.
Collectively, our findings suggest that formation of actin bundles can be initiated but not completed in nekl hypomorphs. Furthermore, filament formation is more severely disrupted in stronger loss-of-function nekl mutants. The strong loss of function allele nekl-3(sv3) shows very subtle and rare formation of actin filaments, whereas actin bundle formation was more extensive when two weak loss of function alleles, nekl-2(fd81) and nekl-3(gk894345), were combined. From these studies we conclude that NEKL-2 and NEKL-3 play an important role in actin organization during molting.

Depletion of CDC-42 and SID-3 reverse actin organization defects in nekl mutants
Given that cdc-42 or sid-3 depletion can suppress molting defects in nekl-2 and nekl-3 mutants, we next determined if epidermal actin organization was restored in suppressed larvae. We observed that cdc-42(RNAi) had a strong effect on actin organization in both nekl-2(fd91) and nekl-3(sv3) backgrounds, both by reducing the occurrence of actin accumulations and by promoting the formation of actin filaments (Fig 6G and 6H). In the case of nekl-2(fd91); cdc-42 cdc-42 suppression of NIMA kinases (RNAi), actin filaments were observed in 96% (n = 24) of animals, however, these filaments were frequently disorganized. 92% (n = 12) of nekl-3(sv3); cdc-42(RNAi) suppressed animals also contained numerous actin filaments, but these did not form parallel circumferential bundles and in some cases were in an anterioposterior orientation (Fig 6H). The observed differences between nekl-2 and nekl-3 strains may be due to differences in their specific allele strengths or because cdc-42(RNAi) can more efficiently suppress reduced NEKL-2 functions.
To further analyze effects of CDC-42 downregulation on actin organization in the apical epidermis, we carried out actin staining on cdc-42(gk388) molting defective mutants. Although many of these animals were mostly impermeable to phalloidin, we observed a range of abnormal actin patterns. Importantly, in 60% of mutants (n = 25) actin puncta were organized in parallel rows, reminiscent of molting-specific actin bundles in wild type (S8C and S8D Fig). Nevertheless, filaments connecting these puncta were often missing. Furthermore, the boundary with the seam cells in cdc-42(gk388) mutants was not always enriched in actin filaments. In the remainder of cdc-42(gk388) molting defective animals (40%), actin puncta and rare filaments showed abnormal actin organization patterns and did not form parallel rows (S8E and S8F Fig). Thus, loss of cdc-42 activity in an otherwise wild-type background leads to alterations in the normal pattern of actin staining, although this phenotype was generally less severe than that observed in nekl-2 and nekl-3 mutants.
We also examined actin localization in suppressed and nekl-2(fd81); nekl-3(gk894345) sid-3 (fd139) animals and observed either a normal molting-specific actin pattern in animals that contained filaments (n = 6) ( Fig 6I) or a normal punctate intermolt actin pattern (n = 13). Taken together, our data imply that NEKL-2 and NEKL-3 influence the formation of epidermal apical actin structures during development and that this function is tightly balanced by the activities of CDC-42 and SID-3.

Apical actin colocalizes with NEKL-MLT components
Because NEKL-MLT components partially colocalize with CDC-42 and because nekl depletion led to a strong perturbation of the actin cytoskeleton, we next asked whether members of the NEKL-MLT network colocalize with actin. For these studies, we used a strain carrying an integrated vab-10 actin-binding domain (ABD) mCherry fusion reporter, which can serve as a proxy for actin expression in live animals [65]. A CRISPR-generated reporter for NEKL-2:: NeonGreen (S1 Table) strongly colocalized with the VAB-10(ABD)::mCherry reporter in 100% of worms (n = 15) at subapical regions of the epidermal-seam cell boundary, with little colocalization observed in other regions (Fig 7A and S9A Fig). Likewise, a CRISPR-generated reporter for MLT-4 (S1 Table), which forms a putative complex with NEKL-2 [26], showed a similar degree of overlap with VAB-10(ABD)::mCherry in the same region (n = 16) (Fig 7B  and S9B Fig). Lastly, a CRISPR-generated reporter for NEKL-3 (S1 Table) was also found to strongly colocalize with VAB-10(ABD)::mCherry in hyp7 in 100% of examined animals (n = 11), however, in contrast to NEKL-2 and MLT-4, this overlap occurred throughout the cytoplasm and rarely at the seam cell boundary, where NEKL-3 is largely absent (Fig 7C and  S9C Fig). Collectively, our data place members of the NEKL-MLT complex in close proximity to both actin and CDC-42, consistent with a role for NEKL-MLT proteins in regulating the actin cytoskeleton through the CDC-42-SID-3 pathway.

Discussion
We have shown that inhibition of the small Rho GTPase CDC-42 leads to the suppression of molting defects in nekl-mlt hypomorphic backgrounds. Likewise, we independently found that mutations in sid-3, the C. elegans ortholog of activated Cdc42-associated kinase 1, also suppress nekl mutants, indicating that CDC-42-SID-3 functions are closely connected to NEKL-MLT signaling in the epidermis. Furthermore, we have observed that cdc-42 downregulation in wild type can lead to molting defects and developmental arrest, suggesting that there is a functional balance between NEKL-MLT and CDC-42 activities in the epidermis. This balance becomes particularly important during molting, when inhibition of either component impairs normal ECM remodeling. Moreover, our observation that inhibition of CDC-42 and SID-3 can suppress molting defects in nekl-mlt mutants indicates that NEKL kinases are most likely negative regulators of CDC-42-SID-3. This control could be at the level of CDC-42-SID-3 intrinsic activity, through effects on subcellular localization, or both.
Consistent with a strong functional link, we observed that CDC-42 partially colocalizes with components of both NEKL-MLT complexes, NEKL-2-MLT-2-MLT-4 and NEKL-3-MLT-3. Furthermore, normal GFP::CDC-42 localization and activity depend on NEKL-2 and NEKL-3, suggesting that there is both a spatial and functional interaction between CDC-42 and the NEKL-MLT protein network. Interestingly, independent proteomic studies to identify binding partners of mammalian NEK6 and NEK7 both identified CDC42 [68,69]. These findings suggest that CDC42 interacts directly or indirectly with NEK6 and NEK7, the highly conserved mammalian orthologs of NEKL-3. Furthermore, a CDC42 modifier, Rho-GAP SNX26 (ARHGAP33), which is implicated in regulation of vesicular trafficking, has also been identified as interacting with NEK6 [68]. Because the phosphorylation targets of NEK kinases are largely unknown, we hypothesize that CDC42, its regulators, or its effectors could be controlled by NEK phosphorylation (Fig 9). This is consistent with findings that the phosphorylation status of CDC42 and its GEFs is important for their localization and activity [70][71][72].
CDC42 has important functions in the establishment of cell polarity, cytoskeletal reorganization and the regulation of endocytosis [34,59,73,74]. In this study we analyzed the effects of nekl-mlt depletion on two CDC-42-dependent processes, actin remodeling and endocytosis, which are known to be important for molting in C. elegans [27,64,75]. It has been previously proposed that apical actin bundles reorganize at each molt to facilitate proper ECM formation [64]. In support of this model, we have shown that loss of NEKL-2 and NEKL-3 activity leads to defects in the pattern of apical actin, a failure to form molting-specific arrays of actin bundles, and defective molting (Fig 9). These apical actin bundles are proposed to contract as new cuticle is synthesized between the actin filaments, thus allowing for the formation of a new cuticle that is larger than the previous one [64]. Our finding that NEKL-2 and NEKL-3 depletion strongly perturb apical actin organization suggests a mechanism by which both detachment of the old cuticle and new cuticle synthesis might be impaired in nekl-mlt mutants. Namely, disorganization of the actin cytoskeleton may alter the distribution and timing of new cuticle synthesis and create an abnormal juxtaposition of the old and new cuticles, resulting in a failure to fully separate. Our observation that nekl mutants often contain incompletely formed molting-type actin bundles within constricted body regions supports this idea. In addition, colocalization data from this study indicate that components of the NEKL-MLT network and actin are partially coexpressed in the same subcellular compartments, which, together with CDC-42 colocalization data, further support a model for the specific control of actin organization by NEKL kinases through CDC-42. Consistent with this model, inhibition of CDC-42 and SID-3 can suppress actin localization defects in nekl mutants, suggesting that NEKL-MLT components are required to spatially and/or temporally repress CDC-42-SID-3 activity in the epidermis.
Interestingly, studies in mammalian cells have implicated MLT-4/INVS in apical actin regulation as INVS knockout mouse fibroblasts form excessive actin-rich filopodia during interphase as well as in mitosis [76]. Notably, filopodia formation is CDC42 dependent, and filopodia overproduction by mammalian cells is a characteristic of mutations that lead to higher CDC42 activity [77,78]. Moreover, CDC42 activity increases in mammalian cells following depletion of INVS, consistent with INVS acting as a negative regulator of CDC42 [79].
Our study suggests that this function for MLT-4/INVS may be conserved in C. elegans and demonstrates for the first time that other components of the NEKL-MLT network control CDC42 activity and actin organization.
Both CDC42 and the actin cytoskeleton have been previously implicated in multiple aspects of vesicular trafficking. For example, actin may control the spatial organization of endocytic machinery components, promote the formation of membrane invaginations, create forces required for vesicle fission and internalization, and provide a barrier function that negatively regulates the rate of endocytosis [80][81][82]. Actin is essential for clathrin-mediated endocytosis in yeast, although its role in mammalian endocytosis may be more limited. Interestingly, actin is required for mammalian endocytosis at cell surfaces that are enriched for actin filaments [81]. Notably, such cortical enhancement of actin filaments is found at the apical surface of epidermal syncytium in C. elegans [64], providing a potential connection between actin and the regulation of vesicular trafficking during molting.
Importantly, endocytosis is a critical process for ECM remodeling in C. elegans, and inactivation of core endocytic components, including CHC-1, leads to molting defects [25-27, 75, 83-86]. One possible role of endocytosis in ECM remodeling may be in resorption and recycling of old cuticle components [27]. Furthermore, epidermal endocytosis has been implicated in the process of sterol uptake from the surrounding environment, and sterols have a proposed role in the synthesis of hormonal cues that initiate molting in nematodes [27,56]. In addition, endocytosis is closely coupled to exocytosis [87], a process critical for cuticle synthesis [27]. Namely, the rate of exocytosis must be balanced with a commensurate level of endocytosis to allow for the recycling of membrane components back to intracellular compartments and to maintain a relatively constant cell surface area [87].
Previously, we have shown that the NEKL-MLT protein network regulates endocytosis in the epidermis [25,26]. In this study, we found that highly penetrant trafficking defects in nekl-2(fd81); nekl-3(gk894345) double mutants can be partially or completely suppressed by CDC-42 and SID-3 downregulation. Nevertheless, we observed that suppression of molting defects was not always well correlated with the restoration of normal clathrin localization patterns. This could be due to different threshold requirements for the suppression of molting defects versus the normalization of clathrin-mediated endocytosis. For example, a partial correction of vesicular trafficking may be sufficient to prevent overt molting defects. Alternatively, impaired endocytosis may not be a primary cause of molting defects in nekl-mlt mutants, consistent with our earlier finding that trafficking alterations are not detected in all molting-defective animals depleted for components of the NEKL-MLT network [25,26]. Instead, the observed endocytosis defects could be a secondary consequence of actin disorganization in nekl-mlt mutants, given that apical actin assembly and disassembly are critical for normal endocytosis [81,82,[88][89][90][91][92][93]. In addition, it is possible that perturbation of both endocytosis and apical actin organization contribute to the observed molting defects in nekl-mlt mutants and that suppression can be achieved through compensatory effects on either process.
In summary, our data implicate the C. elegans orthologs of NIMA-related kinases NEK6, NEK7, NEK8, and NEK9, along with their conserved ankyrin-repeat binding partners, in the regulation of actin remodeling and intracellular trafficking through functional interactions with CDC42 and its effector ACK1. We propose that, by modulating CDC42-ACK1 activity, NEK kinases may control a wide variety of cellular processes including effects on actin morphology and endocytosis (our study), along with potential downstream effects on cell cyclerelated processes and ciliogenesis. Future studies will be required to determine the generality of our observations along with the specific nature of this regulatory connection.

Strains
C. elegans strains were maintained according to standard protocols [94] and were propagated at 21˚C. Strains used in this study include FT1459

RNA interference
RNAi was performed using bacterial strains from Geneservice using standard protocols [95]. In the study that identified cdc-42, we screened all available clones on LGII (~3000 clones) for clones that allowed nekl-3(sv3) mutants to bypass L2/L3 arrest. Control RNAi experiments were carried out using the bacterial strain HT115 carrying either a vector plasmid (pPD129.36) or gfp (RNAi). Marker localization studies involving mlt-3(RNAi) were carried out using strains that were previously grown for several generations on lin-35(RNAi) plates, which increases RNAi sensitivity [96]. Combined RNAi treatments in S2 Fig were performed using RNAi-hypersensitive rrf-3(pk1426) mutants. Bacteria expressing mlt-2, mlt-3, cdc-42, and gfp(RNAi) constructs were grown to the same density and combined at specific ratios. A bacterial culture expressing mlt-2(RNAi) was combined at a 2:1 ratio with bacteria expressing cdc-42(RNAi) or gfp(RNAi), whereas the mlt-3(RNAi) bacterial strain was mixed with the above RNAi clones at a 1:2 ratio.

Genetic identification and analysis of sid-3
sid-3(fd139) was identified as described in Joseph et al. [63]. CRISPR/Cas9 technology was used to generate additional loss-of-function alleles of sid-3 in the nekl-2 (fd81); nekl-3(gk894 345) background. Specifically, genome engineering was performed using CRISPR/Cas9 ribonucleoproteins in combination with the dpy-10 co-CRISPR method [97][98][99]. crRNAs were cdc-42 suppression of NIMA kinases designed to target sequences in the N terminus (5'-AGTGCTCCGCAAAG CACAGT-3') or C terminus (5'-TGAGCCGATTCTCTCGTCTG-3') of sid-3. Absence of repair templates allowed for non-homologous repair mechanisms to generate random frameshifts and premature stop codons in the desired regions. Six loss-of-function alleles were isolated, including fd213, a 16-bp insertion and one point mutation in exon 13, causing a frameshift after L987 and premature stop at amino acid position 1034; fd214, a 7-bp deletion in exon 13, causing a frameshift after L987 and premature stop at amino acid position 1006; fd218, a 5-bp deletion in exon 1, causing a frameshift after K19 and premature stop at amino acid position 25; fd219, a 1-bp deletion in exon 1, causing a frameshift after A20 and premature stop at amino acid position 101; and fd221, a 31-bp insertion and 1-bp deletion in exon 1, causing a frameshift after A20 and premature stop at amino acid position 24 (also see S2 Table). Suppressed nekl mutants were analyzed at the last molt (Fig 6) or as L4 or adult worms (Figs 5  and 8).

Actin staining
Actin staining was performed following a modified procedure from Costa et al. [64]. Animals were washed in M9 solution, and fixation was carried out by incubating animals in fixative (2% paraformaldehyde in 0.05 M Na 3 PO 4, pH 7.4) for 2 h at room temperature. Worms were washed three times in PBS and stored at 4˚C. Phalloidin-TRITC (0.1 mg/ml, dissolved in methanol) was diluted in permeabilization solution (0.05% Triton X-100, 0.1 M NaCl, 3.7% sucrose) at a ratio of 1:500. Equal amounts of liquid containing fixed animals and the phalloidin-containing solution were mixed and incubated in a dark chamber for at least 1 h. To avoid possible post mortem artifacts in our analyses, we scored only the animals that did not show any alterations in the morphology of their internal organs.

Microscopy
Fluorescent images were acquired using an Olympus IX81 inverted microscope with a spinning-disc confocal head (CSUX1 Yokogawa). Confocal illumination was provided by an ILE-4 laser launch (SpectralAppliedResearch). MetaMorph7.7 software was used for image acquisition. DIC images were acquired using a Nikon Eclipse epifluorescence microscope and Open Lab software. This setup was also used for measuring body length and width. Body width was measured in the central region of animals; 50 animals were scored for each category. Animals were immobilized using a 0.1 M solution of levamisole. Distinguishing molting larvae from intermolt larvae was achieved using 0.5-μm red beads mixed with food (Sigma L3280), following the protocol of Nika et al. [100]. For Fig 6, wild-type molting and intermolt animals were categorized using a mlt-10p::gfp-pest reporter [101,102].
Fluorescence intensity of GFP::CDC-42 in Fig 3 was measured using the FIJI program. For each animal, average fluorescence intensity was measured in 10 equal, randomly selected regions of hyp7 and average values were determined. GFP::CDC-42 was analyzed in 27 molting and 30 intermolt wild-type animals. Fluorescence intensity of GST-GFP::WSP-1(GBD) in Fig 4 was measured using ImageJ program. Mean intensities in selected areas were calculated for 11 animals or more and averaged. Background mean intensities were subtracted. Average area of puncta in S5 Fig and number of puncta in S10 Fig were calculated using FIJI program, using ten animals for each genetic background. Colocalization analyses were performed on L4-stage larvae to allow for better visualization of fluorescent structures. Line scanes in S3 and S9 Figs were made using Metamorph. Colocalization with VAB-10(ABD)::mCherry marker was performed in apical regions of epidermis that are slightly more medial than the plane containing actin bundles, as this region showed the greatest extent of colocalization. Transverse cdc-42 suppression of NIMA kinases plane reconstructions of GFP::CHC-1 morphology were created using the FIJI program 3D projection function in selected 1000-pixel × 7-pixel regions of 24-frame z-stacks. Distances between neighboring planes in each z-stack were 0.2 μm. Composite sagittal plane projections were created from the same z-stacks in FIJI, using the Temporal-Color Code function.