Sperm-specific glycogen synthase kinase 3 is required for sperm motility and the post-fertilization signal for female meiosis II in Caenorhabditis elegans

Nearly all sexually reproducing animals reproduce via a highly regulated form of cell-cell fusion, whereby gametic cells combine to form the zygote. This fusion can occur between sperm and egg or between sperm and oocyte, depending on the organism and the stage of female meiosis that is completed prior to fertilization. Regardless of the specific mechanisms used, meiotic maturation, activation and fertilization must be precisely coordinated with signals from the sperm to trigger female meiotic cell divisions and the post-fertilization developmental program. The molecular mechanisms that regulate the transition from an unfertilized cell to a developing zygote are largely unknown as a result of difficulties in observing oocyte maturation, fertilization and subsequent cell divisions in a living organism. Caenorhabditis elegans has been used as a genetic model to reveal signaling pathways required for the oocyte-to-embryo transition (Stitzel and Seydoux, 2007; Müller-Reichert et al., 2010; Robertson and Lin, 2013; Huelgas-Morales and Greenstein, 2018; Kipreos and van den Heuvel, 2019).

In most animals, oocyte development arrests at two points to allow growth, differentiation, coordination with fertilization and completion of female meiosis (Masui, 2001; Von Stetina and Orr-Weaver, 2011). In C. elegans, the first arrest occurs during prophase of meiosis I (MI), which is characterized by an increase in cell volume and synthesis of maternal components to support the development of the future embryo. Exit from this primary arrest and progression through the MI division is termed ‘oocyte maturation’. In many animals, a secondary arrest occurs in metaphase of meiosis II (MII). At this stage, the oocyte awaits sperm entry before completing the second meiotic cell division. In C. elegans, oocyte maturation and fertilization occur almost simultaneously, because a sperm typically enters the oocyte a few minutes after the onset of maturation (Marcello et al., 2013). Sperm provide a diffusible signal called major sperm protein (MSP), which triggers oocyte maturation prior to fertilization (Miller et al., 2001). Upon fertilization, sperm also provide an unknown second signal that is required for proper completion of female meiosis and transition to embryonic mitosis. In the absence of this second signal, the oocyte aborts anaphase I, skips MII and cycles mitotically (Ward and Carrel, 1979; Miller et al., 2001; McNally and McNally, 2005).

We previously reported the role of the novel meiosis-to-mitosis (memi) gene family in specifying the female MII cell division program in response to sperm entry (Ataeian et al., 2016). memi-1, memi-2 and memi-3 are maternally expressed and simultaneous loss of all three paralogs via RNAi results in a ‘skipped MII’ phenotype. In contrast, a gain-of-function hypermorph mutation, memi-1(sb41), results in normal entry into MII, but failure to exit female MII properly. MEMI proteins are normally degraded soon after female MI, but memi-1(sb41) results in persistence of MEMI-1 into mitosis, indicating that its timely removal is required for the MII-mitosis transition (Ataeian et al., 2016). The CUL-2 cullin-RING ubiquitin-ligase and its substrate adaptor, ZYG-11, are also maternal factors that are required for the MII-to-mitosis transition (Feng et al., 1999; DeRenzo et al., 2003; Liu et al., 2004; Sonneville and Gönczy, 2004; Vasudevan et al., 2007). Loss of ZYG-11 results in failure to degrade MEMI-1/2/3, and zyg-11(RNAi) cellular phenotypes resemble memi-1(sb41) mutants (Ataeian et al., 2016).

In a previous genome-wide RNAi screen for suppression of memi-1(sb41) lethality, we identified two highly similar functionally redundant genes, gsp-3 and gsp-4, the wild-type products of which act as genetic activators of the MEMI pathway. Intriguingly, these genes encode sperm-specific catalytic subunits of PP1 phosphatase. The prototype member, yeast GLC7, is involved in glycogen metabolism, meiosis and mitosis (Feng et al., 1991; Peggie et al., 2002; Tan et al., 2003; Bharucha et al., 2008). Loss of either C. elegans gsp-3 or gsp-4 results in no obvious phenotypes other than suppression of memi-1(sb41) (Ataeian et al., 2016; R.P.B., unpublished). However, complete loss of both gsp-3 and gsp-4 function causes defects in male meiosis, sperm activation and sperm motility (Chu et al., 2006; Wu et al., 2012). Interestingly, mouse PP1 phosphatases are also necessary for sperm development and fertility (Varmuza et al., 1999; Oppedisano et al., 2002; Chakrabarti et al., 2007), indicating that this part of the MEMI pathway could be widely conserved.

During C. elegans spermiogenesis, immotile amoeboid spermatids transform into motile spermatozoa that crawl by cytoskeletal treadmilling from the pseudopod tip toward the cell body (Chu and Shakes, 2013; Marcello et al., 2013). The cytoskeletal subunit used for pseudopod formation and sperm motility is MSP; thus, this protein serves as both a secreted signaling molecule for oocytes and the cytoskeleton for sperm (Huelgas-Morales and Greenstein, 2018; Ma et al., 2012). Similar to actin in other eukaryotic cells, MSP dynamics generates membrane tension in the direction of nematode sperm movement. Prior to the meiotic divisions, MSP coalesces into symmetrically elongating fibers called fibrous bodies (FBs), which are closely associated with Golgi-derived membranous organelles (MOs). After the meiotic divisions produce four inactive spermatids, the FB-MO complexes eventually disassociate, whereupon the MOs move to the plasma membrane and the FBs disassemble into MSP dimers, which diffuse throughout the cytoplasm (Roberts et al., 1986). Upon sperm activation, the MSP is redistributed to the pseudopod, where it is thought to undergo dynamic polymerization and depolymerization (Chu and Shakes, 2013), likely via addition of MSP subunits at the leading edge of the cell and MSP subunit removal near the base of the pseudopod (Wu et al., 2012). Based on data from the nematode Ascaris, phosphorylation of accessory proteins is involved in regulating MSP assembly and disassembly (Ma et al., 2012). It is unknown whether C. elegans MSP is similarly regulated, although the PP1 phosphatases GSP-3/4 have been predicted to facilitate MSP disassembly at the base of the pseudopod (Wu et al., 2012). Although the casein kinase-I homolog, SPE-6, has been shown to regulate the assembly of MSP into FBs and pseudopod formation (Varkey et al., 1993; Muhlrad and Ward, 2002; Price et al., 2021), it is still not clear how MSP assembles at the leading edge of the pseudopod.

To complement our previous RNAi screen for memi-1(sb41) suppressors, we used a forward genetic screen to find additional suppressors of memi-1(sb41). We identified four alleles of R03D7.5 (Herzog, 2018). R03D7.5 encodes a homolog of glycogen synthase kinase, type 3 (GSK3), a conserved serine-threonine protein kinase found throughout eukaryotes; hence, we named this gene GSK3-like-1 (gskl-1). Herein, we show that gskl-1 is functionally redundant with another GSK3 encoded by gskl-2 (formerly known as gska-3). Together, these genes are required for male meiotic sister-chromatid segregation and spermatogenesis and pseudopod treadmilling in sperm. We also found that gskl-1 and gskl-2 kinase genes cooperate with, rather than act in opposition to, gsp-3/4 phosphatases for post-fertilization functions. Furthermore, all four genes genetically interact with the MEMI pathway, consistent with a role in the fertilized oocyte. Our results suggest a new candidate kinase for MSP regulation and sperm motility, as well as a potential mechanism for triggering female MII after fertilization.

To understand the function of the memi-1(sb41) suppressor gene, gskl-1, and how it modulates the MEMI pathway, we first tested whether a deletion [(Δ)] mutant, tm4146, would also suppress the temperature-sensitive embryonic lethality of memi-1(sb41). Indeed, embryonic viability increased from 0% in control memi-1(sb41) to 35% in gskl-1(Δ) memi-1(sb41) double-mutant hermaphrodites at the nonpermissive temperature of 25°C (Fig. 1A; P<0.0001).

Although gskl-1(Δ) is predicted to abrogate gene function (see Materials and Methods), worms with this mutation did not exhibit any obvious phenotypes. gskl-1 is one of seven members of the GSK3 family of protein kinases in C. elegans: Y18D10A.5 (gsk-3), R03D7.5 (gskl-1), C36B1.10 (gskl-2), Y106G6D.4, C44H4.6, Y10G6GE.1 and F21F3.2 (Manning, 2005; see Materials and Methods; Table S1). Of these, only gsk-3 is an essential gene; it is required for endoderm specification and mitotic spindle orientation in the EMS cell of the four-cell embryo (Schlesinger et al., 1999). The viable single-deletion strains of all but gsk-3 appeared healthy, although all strains produced fewer total progeny (brood size) compared with wild type (Fig. 1B; Table S2). Suspecting that gskl-1 function could be masked by redundancy, we established double-deletion strains to test for synthetic phenotypes. From the five double-deletion strains, only gskl-1(tm4146); gskl-2(ok970), hereafter referred to as gskl-1(Δ); gskl-2(Δ), resulted in synthetic phenotypes. gskl-1(Δ); gskl-2(Δ) worms exhibited reduced brood size and 56% embryonic viability from self-fertilization (P<0.0001), suggesting that these genes have some overlapping function(s) affecting fecundity and embryogenesis (Fig. 1B; Table S2).

To characterize whether the genes acted in the oocyte or sperm, gskl-2(Δ); gskl-1(Δ) hermaphrodites were mated to wild-type males. Embryonic lethality was completely rescued, consistent with either zygotic or paternal rescue (P<0.0001; Fig. 1C). If the GSK3 double mutant was paternally rescued, then we would also expect all sperm from a heterozygous male to rescue a homozygous mutant hermaphrodite; in other words, the phenotype would be dependent on the genotype of the male parent and not on the zygote or individual sperm haplotype. Therefore, we genotyped individual progeny after mating gskl-2(Δ)/+; gskl-1(Δ) males to gskl-2(Δ); gskl-1(Δ) hermaphrodites. We observed a 1:1 ratio of the gskl-2 (+) and (Δ) alleles in surviving progeny (n=42; see Materials and Methods), indicating that homozygous mutant progeny were rescued by the presence of a wild-type product contributed by the male parent (Wood et al., 1980).

To further test the idea that embryonic lethality resulted from defective sperm, gskl-2(Δ); gskl-1(Δ) males were mated to fog-2 feminized hermaphrodites that do not make their own sperm and can only produce progeny if fertilized by males. In this case, the double-mutant males were expected to cause the same level of embryonic lethality as self-fertilized hermaphrodites. Surprisingly, we observed 91% embryonic viability in the cross, compared with the 56% viability in gskl-2(Δ); gskl-1(Δ) hermaphrodite self-progeny (P<0.005; Fig. 1D; Table S3).

We next considered that incomplete penetrance (56% versus 91% embryonic viability for mutant hermaphrodite sperm versus male sperm) might result from differences in individual gskl-2(Δ); gskl-1(Δ) sperm. Sperm competition might be more intense for male-derived sperm because they must first crawl from vulva to spermatheca to reach an oocyte, whereas hermaphrodite sperm are pre-stored within the spermatheca, which is adjacent to the site of their generation. If the fittest outcross sperm are selected for fertilization, then gskl-2(Δ); gskl-1(Δ) male mating should also partially rescue gskl-2(Δ); gskl-1(Δ) hermaphrodites. In this case, we observed that mated gskl-2(Δ); gskl-1(Δ) exhibited 88% viability, compared with 56% in non-mated controls (Fig. 1E, P<0.0005). Therefore, the increase in viability we observed upon mating mutant males to fog-2 hermaphrodites did not result from a maternal wild-type allele, but was likely the result of selection for healthier mutant male sperm. To further test this, single young adult males were mated to fog-2 hermaphrodites and the males were transferred every 8 h to mate with new hermaphrodites. In this way, fewer male sperm were expected to be deposited in the hermaphrodites compared with en masse matings. We reasoned that this might increase the chance for fertilization by less-competitive sperm. In this case, fog-2 hermaphrodites mated to single gskl-2(Δ); gskl-1(Δ) males displayed 59% embryonic viability compared with 100% in wild-type male-mated controls (P<0.0005; Fig. 1F; Table S4). These results indicated a correlation between the viability of embryos fertilized by gskl-2(Δ); gskl-1(Δ) sperm and the quantity and/or quality of sperm available for competition prior to fertilization.

Although the male-mating experiment confirmed that mutant male sperm do confer paternal-effect lethality, the effect was less severe than in hermaphrodites (59% versus 56% viability, respectively). Thus, we also considered the possibility that the GSK3 genes have hermaphrodite-specific sperm functions. Mutations of genes in the ‘spe-8 group’ (spe-8, spe-12, spe-19, spe-27 or spe-29) result in hermaphrodite-specific spermiogenesis defects that cause sterility (Shakes and Ward, 1989; Minniti et al., 1996; Nance et al., 1999, 2000; Geldziler et al., 2005). The spe-8 group likely negatively regulates the casein kinase, SPE-6, because a loss-of-function allele, spe-6(hc163), suppresses all known mutations within the group (Muhlrad and Ward, 2002). Therefore, in order to test whether the GSK3s might be part of the spe-8 group, we constructed a gskl-2(Δ); gskl-1(Δ); spe-6(hc163) triple mutant. spe-6(hc163) did not suppress gskl-2(Δ); gskl-1(Δ) embryonic lethality in unmated hermaphrodites (Fig. S1), suggesting that these two GSK3 genes do not contribute to the negative regulation of SPE-6.

Together, these data suggest that gskl-1 and gskl-2 are functionally redundant and exhibit incompletely penetrant paternal-effect embryonic lethality. Our data also revealed that selection among male-derived gskl-2; gskl-1 sperm can contribute to higher embryonic viability, especially when multiple males are allowed to mate with hermaphrodites.

In light of the above results, we hypothesized that gskl-1 and gskl-2 might be important for sperm motility or guidance. Therefore, we tracked the location of male sperm in mated hermaphrodites, using male worms that express the chromatin marker mCherry-histone. After mating transgenic wild-type males to unlabeled wild-type hermaphrodites for 24 h, fluorescent punctae corresponding to the chromatin masses of wild-type male sperm were detected only in the spermathecae, adjacent to the uterus (Fig. 2). In contrast, in all hermaphrodites mated to gskl-2(Δ); gskl-1(Δ) males, fluorescent punctae were scattered between the vulva and the spermatheca (Fig. 2). This suggested that gskl-2(Δ); gskl-1(Δ) sperm can be transferred from the male to the hermaphrodite during mating, but many sperm are inefficient in migrating to the spermatheca, representing a possible mechanism to select for fitter sperm. This result also provided a plausible explanation for the reduced penetrance of embryonic lethality upon male mating; sperm that reach the spermatheca could be healthier with respect to both motility and post-fertilization functions.

We next sought to determine whether gskl-1 and gskl-2 play a direct role in sperm motility. In C. elegans, activated sperm move by treadmilling of the cytoskeletal polymer MSP from the pseudopod tip toward the cell body (Roberts and Ward, 1982). Treadmilling rates, which directly correlate with sperm crawling velocity (Nelson et al., 1982), can be observed indirectly by tracking vesicles within the cytoplasm (Fig. 3A). In order to measure treadmilling, we used the mutant swm-1. swm-1(+) normally inhibits male sperm activation until after transfer to the hermaphrodite; thus, swm-1 mutant spermatids are constitutively activated to crawl (Stanfield and Villeneuve, 2006). The average pseudopod treadmilling rate for swm-1 him-5 male sperm was 19.6±2.4 µm/min, which is similar to previously reported data (Wu et al., 2012). In gskl-2(Δ); gskl-1(Δ); swm-1 him-5 mutants, the average treadmilling rate was significantly slower (8.8±2.5 µm/min; P<0.005; Fig. 3B). A histogram showing the distribution of individual pseudopod treadmilling events is shown in Fig. S2.

To visualize pseudopod development, sperm were treated with Pronase-A, which allows synchronous activation of male sperm in vitro (Ellis and Stanfield, 2014). Quantification showed that 100% of him-8 control male sperm activated within 12 min (Fig. 3C), as evidenced by the appearance of normal pseudopods (Fig. 3D). In contrast, only 3% of gskl-2(Δ); gskl-1(Δ); him-8 sperm activated within 10 min, and only 20% within 30 min (Fig. 3C). The inactivated sperm did not form any obvious protrusions (Fig. 3D). These data indicated that gskl-1 and gskl-2 were involved in sperm pseudopod formation and function, and provided an explanation for why the double-mutant male sperm were inefficient in migrating to the spermatheca.

Our genetics results and phenotypes suggested that gskl-1 and gskl-2 were expressed in sperm. We examined the expression pattern by tagging the respective genes and performing western blots on whole-worm lysates. We created 3×FLAG-tagged gskl-1 and Ollas-tagged gskl-2 strains using CRISPR-Cas9 (Paix et al., 2017). To test for sex-specific expression, we used temperature-sensitive mutants that produce only sperm (fem-3) or only oocytes (fem-1). Western blotting with anti-FLAG and anti-Ollas revealed that GSKL-1^FLAG and GSKL-2^Ollas were detected in sperm-only fem-3 worms but not in oocyte-only fem-1 females (Fig. 4A).

Consistent with the western blot experiments, immunostaining of endogenous GSKL-1^FLAG and GSKL-2^Ollas worms revealed strong fluorescence at multiple stages of spermatogenesis. In males, immunofluorescence was brightest in the proximal gonad, which is where spermatogenesis occurs (Fig. 4B). Immunofluorescence was also observed in the cytoplasm of primary and secondary spermatocytes, in a punctate pattern for both tagged proteins (Fig. 4B), but was not detected in similar immunostaining experiments using wild-type untagged control worms (Fig. S3). GSKL-1^FLAG and GSKL-2^Ollas immunofluorescence was also detected specifically in the hermaphrodite sperm (Fig. S4).

As sperm become activated to crawl, the location of GSP-3/4 changes from a distributed punctate pattern associated with fibrous bodies to a restricted band at the base of the pseudopod. During this transition, MSP becomes concentrated throughout the pseudopod, overlapping with GSP-3/4 at the base (Wu et al., 2012).

In inactive sperm from males, GSKL-1^FLAG was distributed throughout the cytoplasm, displaying a punctate pattern similar to that of GSP-4. However, GSP-4 only partially colocalized with GSKL-1^FLAG and, in many cases, fluorescent foci were located adjacent to GSP-4 foci (Fig. 5A). Similarly, in inactive sperm, GSKL-2^Ollas was distributed throughout the cytoplasm and partially colocalized with MSP (Fig. 5B). In contrast to the isotropic distribution of the tagged proteins in inactive sperm, the immunostaining revealed an asymmetric distribution for both proteins in activated sperm. In activated sperm, GSKL-2^Ollas and GSKL-1^FLAG immunofluorescence was strongest throughout the pseudopod, similar to MSP (Fig. 5A,B). Furthermore, co-immunostaining for GSP-4 revealed colocalization with GSKL-1^FLAG and GSKL-2^Ollas at the base of the pseudopod (Fig. 5; Fig. S5). GSKL-1^FLAG was also detected in the cell body of active sperm, whereas GSKL-2^Ollas was more restricted to the pseudopod and, in many cases, was located predominantly at the leading edge. The enrichment of GSKL-1^FLAG and GSKL-2^Ollas in the pseudopod of active sperm is consistent with a role for this kinase in regulating MSP dynamics and, hence, sperm motility.

Given that sperm activation also involves fusion of MOs with the membrane and a redistribution of MSP prior to pseudopod formation (Chu and Shakes, 2013; Marcello et al., 2013), we examined these structures in active sperm (see Materials and Methods). We observed a similar distribution pattern of MOs and MSP in swm-1 him-5 male sperm and in gskl-2(Δ); gskl-1(Δ); swm-1 him-5 male sperm (Fig. 5C). This suggests that, during sperm activation, GSKL-1 and GSKL-2 are not required for proper distribution of MOs and MSP.

From the above experiments, we noticed a variation in sperm size and that many sperm appeared to lack nuclei, as determined by differential interference contrast (DIC) imaging (Fig. S6). Therefore, we investigated male meiosis by performing live imaging of wild-type and gskl-2(Δ); gskl-1(Δ) male spermatocytes expressing GFP tubulin (microtubules) and mCherry-histone (chromatin). We found that many double-mutant spermatocytes exhibited defects. In wild-type males (n=16), homologous chromosomes separated in MI, and sister chromatids separated in MII to generate a four-bud structure (Fig. 6; Movie 1). Each bud in wild type eventually separated from the residual body to become an individual spermatid (n=16). In gskl-2(Δ); gskl-1(Δ) males (n=15), homologous chromosomes separated successfully to form secondary spermatocytes; however, one pair of sister chromatids did not separate in 3/15 spermatocytes (Fig. 6; Movie 2). Instead, these cells displayed an aberrant three-bud structure that would be expected to produce one aneuploid sperm. In 12/15 gskl-2(Δ); gskl-1(Δ) male spermatocytes, both homologous chromosomes and sister chromatids separated successfully to form a four-bud structure, similar to controls. However, in some of these, one or more sets of segregated chromosomes moved back into the residual body, resulting in the formation of anucleate sperm (5/12; Fig. 6; Movie 3). Immunostaining of gskl-2(Δ); gskl-1(Δ) revealed that MSP was distributed throughout the residual body of the aberrant three-bud structures (Fig. S7).

Given that gskl-2(Δ); gskl-1(Δ) double mutants exhibited paternal-effect embryonic lethality, we reasoned that these sperm components could also function in the fertilized oocyte. To look for defects during the meiosis-to-mitosis transition, we used gskl-2(Δ); gskl-1(Δ) embryos that express GFP tubulin (microtubules) and mCherry-histone (chromatin). Aneuploid sperm produce defective embryos (Sadler and Shakes, 2000); thus, we specifically looked at embryos fertilized by nucleated sperm (Movie 4). Time-lapse in utero imaging of gskl-1(Δ); gskl-2(Δ) embryos showed that MI spindle formation, homologous chromosome segregation and extrusion of the first polar body were similar to wild-type fertilized oocytes. However, we observed significant delays and defects in subsequent stages in a subset of mutant embryos (Fig. 7). For example, in gskl-2(Δ); gskl-1(Δ) embryos fertilized with a nucleated sperm, 7/13 cells exhibited defects after completion of MI. Five of these samples could be timed accurately, and displayed a severe delay in the appearance of the MII spindle after anaphase I onset (mean of 17 min 52 s) compared with controls (mean of 4 min 14 s; P<0.0005; Fig. 7A). In these cases, polar body II extrusion did not occur. In all seven meiosis-defective embryos, sperm-derived centrosomal microtubules were visible even though the female MII spindle persisted; normally, these microtubules do not appear prior to dissolution of the meiotic spindle. Eventually, the MII spindle disappeared and pronuclei formed and migrated toward each other (Fig. 7B; Movie 5). We also observed a separate defect in 7/20 samples that occurred as the oocyte traversed the spermatheca, whereby a portion of the cell formed an oblong extension containing maternal DNA, which eventually cleaved off from the rest of the oocyte (Movie 6).

gsp-3/4(RNAi) or gskl-1(Δ) individually suppress memi-1(sb41) (Ataeian et al., 2016; and results presented above). We also tested gsp-3(Δ); memi-1(sb41) and gskl-2(Δ); memi-1(sb41) double mutants and observed significant suppression of memi-1(sb41) (Fig. S8). Therefore, our data indicated that gskl-1 and gskl-2 act similarly to gsp-3/4 with respect to sperm function. We next tested whether gskl-1 and gsp-4 enhance each other genetically. Indeed, the double kinase/phosphatase deletion strain gsp-4(Δ); gskl-1(Δ) exhibited 68% embryonic viability compared with 97% in the individual mutant strains (Fig. 8A; Table S5). This suggests that the phosphatases and kinases together contribute to an essential post-fertilization function.

Next, we tested whether the phosphatase and kinase genes suppressed memi-1(sb41) through the same or different pathways. gsp-4(Δ) and gskl-1(Δ) independently suppressed the maternal-effect lethality of memi-1(sb41), up to 8.5% and 35% embryonic viability, respectively, compared with 0% for memi-1(sb41) controls at 25°C (Fig. 8B). If the two genes synthetically interact and suppress memi-1(sb41) beyond an additive effect, we should observe more than 40.5% viability in the triple mutant. The gsp-4(Δ); gskl-1(Δ); memi-1(sb41) triple mutant exhibited 29% embryonic viability (Fig. 8B). However, one complication is that the gsp-4(Δ) gskl-1(Δ) double mutant is synthetically lethal on its own, resulting in only 68% viability (Fig. 8A). Correcting for this maximum expected survival, the triple mutant resulted in 43% viability, which is only slightly higher than additive suppression from the two individual mutations (40.5%). These data are consistent with the idea that each of the GSK3 kinases and PP1 phosphatases contributes to a common function, and the attenuation of this function suppresses memi-1(sb41); however, it is not clear whether suppression of memi-1(sb41) occurs via a single or two parallel pathways.

If gskl-1, gskl-2 and gsp-4 activate memi(+) after fertilization, loss of these genes should cause a phenotype similar to the memi-1/2/3(RNAi) phenotype. We found that gsp-4(Δ) gskl-2(Δ); gskl-1(Δ) displayed decreased embryonic viability compared with gskl-2(Δ); gskl-1(Δ) (32% versus 68%; Fig. 8A; Table S5). We developed a strain that expressed GFP tubulin (microtubules) and mCherry-histone (chromatin) to characterize the phenotype of defective embryos in utero (Movie 7). In gsp-4(Δ) gskl-2(Δ); gskl-1(Δ) embryos, 11/22 fertilized oocytes displayed defects during the completion of female meiosis. Four of these embryos exhibited a skipped-MII phenotype, characteristic of memi-1/2/3 loss of function (Fig. 8C; Movie 8). Seven out of 11 defective embryos progressed through metaphase I but did not complete anaphase I or release the first polar body, and also exhibited delays in forming the MII spindle. These embryos exhibited abnormal anaphase II and the maternal DNA formed multiple maternal nuclei. Centrosomal microtubules appeared before dissolution of the MII spindle and pronuclear migration initiated soon after the MII spindle disappeared (Fig. 8C; Movie 9). As noted in our analysis of the double-mutant above, we also observed a separate defect in the triple mutant, occurring as the oocyte traversed the spermatheca. From a slightly larger dataset that included three samples in which the oocyte was not evidently fertilized, we observed 3/25 cases in which the oocyte fragmented as it moved through the spermatheca, resulting in a small chromatin-containing cell that blebbed off from the oocyte (Movie 10).

In summary, gskl-1 and gskl-2 encode paternal factors required for male MII, sperm activation and motility, as well as female meiotic divisions after fertilization. The embryonic phenotypes of the gskl-2; gskl-1 double mutant, as well as the triple mutant with gsp-4, occurred during the transition from female MI to MII. This is consistent with the idea that these genes are required for the sperm-derived post-fertilization signal that uses the memi genes for completion of female meiosis.

In C. elegans, the sperm send two signals, a diffusible MSP-based signal that triggers the oocyte to exit MI prophase (oocyte maturation; Miller et al., 2001) and, upon fertilization, a second signal to complete anaphase I and enter MII (Ward and Carrel, 1979). Evidence for the latter signal comes from analyses of fertilization-defective sperm that produce MSP but are unable to fertilize the oocyte. In the absence of fertilization, such MSP-stimulated ‘embryos’ enter anaphase I but fail to extrude the first polar body, skipping MII and proceeding to mitosis (McNally and McNally, 2005).

To date, the only genes known to be directly involved in the fertilization-dependent signal are the memi genes (Ataeian et al., 2016). However, by screening for suppressors of the gain-of-function mutation, memi-1(sb41), we have begun to identify more components of this pathway, such as the sperm-specific GSP-3/4 PP1 protein phosphatases (Ataeian et al., 2016) and the two GSK3 kinase genes described herein. Although originally identified as regulators of glycogen metabolism, GSK3 kinases have well-established roles in signaling pathways crucial for development (Patel and Woodgett, 2017). The gskl-2(Δ); gskl-1(Δ) double-deletion mutant revealed new phenotypes in sperm motility and spermatogenesis as well as incompletely penetrant embryonic paternal-effect lethality. The reason for the variability in phenotypic severity among individual sperm is still unclear, but there could be some further redundancy for GSK3 activity within the sperm and/or fertilized oocyte. Despite this, the phenotypes revealed an intriguing correlation between the motility of individual sperm and a post-fertilization function that is essential for development.

The data presented here suggest that GSKL-1/GSKL-2 kinases and GSP-3/4 phosphatases, despite having presumably opposite biochemical activities in these cells, likely perform at least some similar functions. For example, reduction of either GSP-3/4 phosphatases or GSKL-1/GSKL-2 kinases results in meiotic defects during male MII, delayed activation of spermatids, as well as reduced sperm motility and slow pseudopod treadmilling. In addition, all individual components act as suppressors of the maternal-effect mutation memi-1(sb41). Furthermore, the gsp-4(Δ); gskl-1(Δ) kinase/phosphatase double mutant suppressed memi-1(sb41), indicating that reducing both kinase and phosphatase activity did not cancel out individual suppression; instead, the suppression increased slightly. Finally, the gsp-4(Δ) gskl-2(Δ); gskl-1(Δ) triple mutant exhibited more severe embryonic lethality than did the double mutants alone, and some of those defective embryos displayed a skipped-MII phenotype, similar to loss of memi-1/2/3 function. Thus, all data indicate that, perhaps unexpectedly, reducing the activity of either GSKL-1/GSKL-2 kinases or GSP-3/4 phosphatases results in very similar phenotypes.

Current models of sperm activation and motility in the Ascaris nematode involve assembly of MSP fibers at the leading edge of the pseudopod and disassembly at the base. Given that MSP filaments lack polarity (Bullock et al., 1998; Baker et al., 2002), accessory proteins are required to facilitate unidirectional growth. At the leading edge, membrane-bound MSP polymerization organizing protein (MPOP) recruits a serine/threonine kinase MSP polymerization-activating kinase (MPAK) to phosphorylate MSP fiber protein 2 (MFP2) at a threonine residue (LeClaire et al., 2003; Yi et al., 2007). Upon its phosphorylation, MFP2 associates with the MSP fiber (Yi et al., 2007). Another accessory protein, MFP3, is important for stabilizing MSP fibers, with PP2A-induced dephosphorylation at the base of the pseudopod resulting in MFP3 release and fiber disassembly (Yi et al., 2009). Finally, MFP1 appears to be a negative regulator of assembly, whereby excess MFP1 reduces the rate of fiber elongation and inhibition of MFP1 increases the polymerization rate (Buttery et al., 2003; Roberts and Stewart, 2012). Thus, the data from Ascaris indicate that phosphorylation and dephosphorylation of accessory proteins is likely key to in vivo regulation of MSP fiber dynamics. It was shown recently that C. elegans homologs of MFP1 and MFP2 colocalize with MSP in the pseudopod (Morrison et al., 2021). However, it is still unclear whether any of the accessory proteins play a direct role in C. elegans sperm motility.

The phenotypes associated with gskl-2(Δ); gskl-1(Δ) double mutants, and the subcellular location of GSKL-1/GSKL-2 in the sperm pseudopod, suggests that these kinases could regulate MSP dynamics. Although the targets of GSK3 are unknown, a few MSP fiber components contain putative GSK3 target sites. For example, the MSP monomer itself contains a Ser/Thr-X-X-X-Ser/Thr consensus sequence, in which the first Ser/Thr is a target and the last Ser/Thr is a site for priming phosphorylation (ter Haar et al., 2001). In some members, the priming site increases the efficiency of GSK3 action (Thomas et al., 1999). Interestingly, X-ray crystallography of MSP-142 revealed that the GSK3 consensus sequence is within a conserved region required for MSP assembly in vitro (Baker et al., 2002; del Castillo-Olivares and Smith, 2008). Therefore, GSKL-1/GSKL-2 could theoretically regulate MSP directly. Alternatively, GSKL-1/GSKL-2 could act on accessory proteins to regulate MSP dynamics. For example, the accessory protein MFP2 (encoded by nsph-2 and nsph-3.2) also contains putative GSK3 target sites. In this case, GSKL-1/GSKL-2 could act similarly to the serine/threonine kinase MPAK in Ascaris, to promote MSP assembly at the leading edge of the pseudopod.

To date, the PP1 phosphatases GSP-3/4 represent good candidate enzymes to regulate MSP dynamics and treadmilling in the C. elegans sperm pseudopod (Wu et al., 2012). It is tempting to speculate that the GSK3 kinases and PP1 phosphatases would facilitate opposite effects with respect to the assembly and disassembly of MSP fibers, similar to the Ascaris situation involving MPAK and PP2A, respectively. However, our data indicated that reduction of function of either enzyme type resulted in similar phenotypes. One simple explanation for this observation is that MSP dynamics requires a precise balance of assembly and disassembly, and any shift in this equilibrium disrupts the entire process. Another idea is that the kinases and phosphatases could be part of a single pathway. In other systems, GSK3s engage in complex regulatory networks that involve PP1 and/or PP2A phosphatases, or their respective regulatory subunits. For example, bovine PP2A and PP1 activity in immature sperm has been correlated with increased activity of GSK3 (Vijayaraghavan et al., 1996). GSK3 can also phosphorylate inhibitor 2, a PP1 regulatory subunit (Lin et al., 2003), and this phosphorylation enhances PP1 (PPP1CC) catalytic activity in the motile caudal sperm in mice (Goswami et al., 2018). Although we have not identified a regulatory subunit for GSP-3/4, it is possible that a similar form of regulation occurs in C. elegans sperm. Alternatively, GSP-3/4 phosphatases could be a direct target of GSKL-1/GSKL-2 kinases, or vice versa. Having this type of interdependent regulation could allow more-precise spatial regulation of MSP dynamics within the relatively small pseudopod region.

Loss of GSP-3/4 or GSKL-1/GSKL-2 also results in chromosome segregation defects during male meiosis, suggesting an additional function of this system. In mouse and rat, GSK3B is expressed in type B spermatogonia and preleptotene spermatocytes, suggesting that vertebrate counterparts also play a role in spermatogenesis (Guo et al., 2003; Bhattacharjee et al., 2015). The precise role for vertebrate GSK3 in spermatogenesis is still unclear, but, in C. elegans, the phenotypes could result from changes in MSP regulation. It has been suggested that MSP must be sequestered or inactivated during male meiotic cell divisions in order to prevent the cytoskeletal polymer from interfering with microtubule- and actin-related functions required for meiosis and spermatid budding (Varkey et al., 1993; Price et al., 2021). We observed MSP within the residual body of defective three-bud structures, suggesting that the dysregulation of MSP at these stages might contribute to chromosome segregation and spermatocyte budding phenotypes.

As summarized in Fig. 9A, the localization data suggest a connection between GSKL-1/GSKL-2 and GSP-3/4 for sperm motility, but it is difficult to explain how a loss of either kinase or phosphatase relates to the post-fertilization function for this system, as reflected by their suppression of memi-1(sb41). The memi-(sb41) mutation results in a P74S change at a putative proline-directed phosphorylation site and it has been shown that MEMI-1 degradation depends on this motif (Ataeian et al., 2016). The suppression data appear to be inconsistent with a simple mechanism involving direct phosphorylation or dephosphorylation of MEMI by GSKL-1/GSKL-2 and GSP-3/4, respectively. Instead, we suggest that the common link between these genes and the MEMI pathway might involve treadmilling-competent MSP (Fig. 9B). MSP has unusual properties; it is both a cytoskeletal polymer within the sperm cytoplasm and a diffusible hormone when secreted by sperm to trigger oocyte maturation (Miller et al., 2001). Thus, it is possible that MSP could have a third function, as part of a sperm-derived signal that activates the maternal MEMI proteins to trigger female MII. This signal would function independently of the secreted MSP-based oocyte maturation signal, and be delivered to the oocyte upon fertilization. In this context, treadmilling-competent MSP includes all the necessary regulatory proteins; therefore, the post-fertilization signal for the oocyte-to-embryo transition could involve one or more of these components.

If GSKL-1/GSKL-2 kinases and GSP-3/4 phosphatases (and, by extension, treadmilling-competent MSP) are essential components of a sperm-derived signal that triggers entry into female MII after fertilization, one would expect that the loss of such components should result in a skipped-MII phenotype, similar to a complete loss of MEMI function. Using a combination of double and triple mutants from these redundant pathways, we observed the skipped MII phenotype in many fertilized oocytes, consistent with the idea that GSKL-1/GSKL-2 and GSP-3/4 are required for the transition to MII in the fertilized oocyte. It is possible that some female meiosis defects could result from aneuploidy or mis-segregation of other sperm components during male meiosis. However, we think that the defects are specific for this pathway, because other mutations that produce anucleate or aneuploid sperm give rise to different phenotypes. For example, the paternal-effect mutants, such as emb-27 (Sadler and Shakes, 2000) and top-2 (Jaramillo-Lambert et al., 2016), form anucleate sperm that are capable of fertilization. In these mutants, the early female meiotic divisions appear normal, indicating that the post-fertilization phenotypes of gskl-2(Δ); gskl-1(Δ) double mutants are not simply a result of chromosome segregation defects during male meiosis.

Paternal spe-11 has also been reported to be involved in post-fertilization events, whereby mutants exhibit defects in MI and MII polar body extrusion (McNally and McNally, 2005), eggshell formation, mitotic spindle positioning and cytokinesis (Hill et al., 1989; Browning and Strome, 1996). However, unlike some of the fertilized gsp-4(Δ) gskl-2(Δ); gskl-1(Δ) triple mutant oocytes, the spe-11 mutants transition out of MI and MII and have not been reported to skip MII. In addition, spe-11 mutant embryos are permeable to DAPI as a result of defects in eggshell formation, whereas gskl-2(Δ); gskl-1(Δ) embryos are not (R.P.B., unpublished). Although it is still possible that a connection exists between spe-11 and the MEMI pathway, the defects associated with loss of gskl-2, gskl-1 or gsp-4 appear distinct. Furthermore, we found that embryonic lethality of memi-1(sb41) at 25°C was not suppressed by loss-of-function spe-11(hc77) (R.P.B., unpublished). Therefore, although spe-11 might be involved in some overlapping functions that involve gskl-1/2 and gsp-3/4, it does not appear to be a genetic activator of the MEMI pathway.

In summary, this work reveals a new role for two homologs of the GSK3 family, gskl-1 and gskl-2, in sperm meiosis and sperm motility in C. elegans. Our data also indicate that the GSK3 kinases could act with GSP-3/4 PP1 phosphatases to activate the MEMI pathway for accurate completion of female MII after fertilization. The molecular link between the MEMI pathway and these sperm components remains unclear, but at least one intriguing possibility is MSP. Future work will assess this and other possibilities.

C. elegans strains were maintained at 20°C following standard protocols, except for memi-1(sb41)-containing strains, which were maintained at 15°C (Brenner, 1974). him-8(e1489) facilitated production of males (Hodgkin et al., 1979; Phillips et al., 2005) and swm-1(me66) was used to observe constitutively active spermatozoa (Stanfield and Villeneuve, 2006). gskl-1(tm4146) 439 bp deletion removes part of the 3rd exon, and exons 4-6, and is not expected to produce a functional kinase. All strains used in this study are listed in Table S1.

Single larval-stage (L) 4 hermaphrodites were allowed to lay eggs on individual Nematode Growth Medium (NGM) agar plates with fresh Escherichia coli OP50 at the indicated temperature, with transfer to fresh plates every 24 h until the worms stopped laying eggs. Eggs that were not hatched after 48 h at 20°C or 36 h at 25°C were scored as dead. Embryonic viability refers to the number of hatched larvae divided by the total progeny. To test suppression of memi-1(sb41), worms were transferred to 25°C as young L4 larvae.

Paternal-effect embryonic lethality was determined by four different methods. First, him-8(e1489) males were mated to L4 hermaphrodites of their respective genotypes to score embryonic viability from the mated hermaphrodites, as indicated by the presence of 50% outcross male progeny. Second, males of their respective genotypes were mated to fog-2 females, which can only produce progeny after mating with males, followed by scoring total embryonic viability. Third, gskl-2/+; gskl-1 males were mated to gskl-2; gskl-1 hermaphrodites to determine whether both gskl-2/+; gskl-1 and gskl-2; gskl-1 progeny were viable in a 1:1 ratio. The genotypes of 42 outcross progeny were confirmed through PCR. Out of the 42 worms, 20 worms were gskl-2; gskl-1 and 22 worms were gskl-2/+; gskl-1. Fourth, individual male matings were performed as follows: six males of the specified genotype were mated individually to six fog-2 feminized hermaphrodites for 8 h. The six males were then removed and allowed to mate individually with six new hermaphrodites for another 8 h. The total progeny from each hermaphrodite was scored.

Endogenous 3xFLAG::gskl-1 and Ollas::gskl-2 N-terminal fusion tags were generated using a CRISPR-Cas9 method with some modifications (Paix et al., 2017). Homology arms in the repair template were 50 bp. A 20 µl injection mixture contained 5 μl of 10 µg/µl Cas9 (Alt-R S.p. Cas9 Nuclease V3; Integrated DNA Technologies), freshly prepared 1 μl 1 M KCl (Invitrogen), freshly prepared 0.75 μl 200 mM HEPES (pH 7.4; Invitrogen), 5 µl 0.17 mM tracrRNA (Alt-R CRISPR-Cas9 tracrRNA, 5 nmol; Integrated DNA Technologies ), 0.4 µl 0.6 mM dpy-10 crRNA (Alt-R CRISPR-Cas9 crRNA, 2 nmol; Integrated DNA Technologies), 1 µl 0.6 mM targeted gene crRNA (Alt-R CRISPR-Cas9 crRNA, 2 nmol; Integrated DNA Technologies), 0.55 µl 16 µM dpy-10 single-stranded oligo DNA nucleotides (ssODN; Integrated DNA Technologies), 10 pmolar ssODN repair (Integrated DNA Technologies) and nuclease free water (Invitrogen) to a final volume of 20 µl. L4 worms were allowed to mature into young adults at 15°C for 14-16 h before injection and injected worms were kept at 25°C. F1-generation Dpy/Rol worms were screened in pools of three by PCR. FLAG::GSKL-1 segregated with the co-CRISPR marker dpy-10 in the F2 generation. dpy-10 was later removed by outcrossing. Non-Dpy Ollas::gskl-2 worms were selected directly from the F2 generation. All putative insertions were confirmed by DNA sequencing. Short guide (sg)RNA and repair template sequences used in this study are listed in Table S6. Deletion mutants for gskl-1 and gskl-2 have no obvious phenotype on their own. Thus, in order to confirm that the tagged genes were functional, homozygous fusion-tagged males were crossed to memi-1(sb41). In each case, no suppression was observed, indicating that the fusion tags likely did not disrupt gene function.

Male worms expressing mCherry-histone and GFP-tubulin were dissected in 3.5 µl sperm media (SM) buffer that contained 50 mM HEPES (pH 7), 50 mM NaCl, 25 mM KCl, 1 mM MgSO₄, 5 mM CaCl₂ and 1 mg/ml bovine serum albumin (BSA). An 18×18 mm coverslip was placed on top and gentle pressure was applied to release spermatocytes from the gonad. The coverslip was sealed with a thin ring of Dow Corning high-vacuum grease to prevent evaporation. Time-lapse images were captured with a Hamamatsu Orca R2 camera on an inverted Olympus IX81 microscope using a 60× oil objective (NA 1.42), equipped with a Yokogawa CSU-10 spinning disc confocal head modified with a condenser lens in the optical path (Quorum Technologies). Z-stacks of three planes with 1.2 µm spacing were captured at 15 s-intervals, controlled by MetaMorph software. Image files were analyzed using MetaMorph software.

Relevant mutant and control strains that expressed mCherry-his-58 were constructed to allow visualization of sperm chromatin (Table S1). In total, 35-40 males were mated to five L4 hermaphrodites at 20°C for 24 h, followed by isolation of the hermaphrodites for 1 h. Then, hermaphrodites were anesthetized with 200 µM levamisole in M9 buffer (22 mM KH₂PO₄, 42 mM Na₂HPO₄, 85 mM NaCl and 1 mM MgSO₄) and mounted onto a 2% agarose gel pad for confocal fluorescence imaging as described above. Fluorescent images of 14 planes, spaced 1.5 µm apart, were taken for each worm (n=7 for each set). The fluorescent image stacks were projected as a maximum before merging with DIC in Adobe Photoshop.

L4 virgin swm-1(me66) him-5(e1490) and gskl-2(Δ); gskl-1(Δ); swm-1(me66) him-5(e1490) males were transferred to fresh NGM plates for 18 h at 20°C. Next, males were dissected in 50 µl SM buffer on a polylysine-coated coverslip to release the activated sperm (Nelson and Ward, 1980). Pseudopod treadmilling was captured by time-lapse recordings with a 1 s interval, using an Olympus DIC microscope with a 60× oil objective (NA 1.42) and a Hamamatsu Orca R2 camera controlled by MetaMorph software. Vesicles within the pseudopod were tracked for 6 s to determine treadmilling rates (Wu et al., 2012). Pseudopod treadmilling rates were determined from 2 sperm from each of 7 individual worms (14 in total) of the respective genotype.

L4 virgin him-8 and gskl-2; gskl-1 males were transferred to fresh NGM plates for 18 h at 20°C to prevent spermatid activation. Two males were dissected in activation solution that contained 200 µg/ml Pronase in SM buffer with 1 mg/ml BSA. Dow Corning high-vacuum grease was used to surround the buffer and an 18×18 mm coverslip was placed on top to prevent evaporation. Time-lapse DIC images were captured at 20 s intervals using a 60× oil objective (NA 1.42) on an Olympus microscope as described above.

L4 males of their respective genotypes were harvested after overnight incubation at 20°C in the absence of hermaphrodites to observe inactive spermatids or active spermatozoa. In total, 20-25 males were dissected in 5 µl SM media on a polylysine-coated slide with an 18×18 mm coverslip placed on top. Gentle pressure was applied to release spermatids from the gonad. To observe embryos, approximately 35 young adult hermaphrodites were picked into 5 µl egg buffer (188 mM NaCl, 48 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂ and 25 mM HEPES, pH 7.3) on a polylysine-coated slide and gentle pressure was applied to squeeze the embryos out of the worms. Slides were immediately frozen in liquid nitrogen. After freeze-cracking by removal of the coverslip, slides were fixed in methanol for 15 min at −20°C. After washing in PBS for 20 min, slides were incubated with 25% normal goat serum for one hour. Slides were incubated in primary antibody solution (5% normal goat serum, 0.01% Triton-X100 and 1 µg/µl primary antibodies) for 1 h at room temperature after washing in PBS for 10 min. 3xFLAG was detected by mouse anti-FLAG antibody (F1804, 1/500, Sigma-Aldrich), Ollas was detected by rat anti-Ollas antibody (NBP1-06713, 1/250, Novus Biologicals), MOs were observed by AlexaFluor 488-conjugated wheat-germ agglutinin (W11261, 1/1000 dilution, Invitrogen) (Kelleher et al., 2000), MSP was visualized by mouse anti-MSP 4A5 antibody (4A5; 1/1000; Developmental Studies Hybridoma Bank) and rabbit anti-GSP-3/4 antibody (1/1000 dilution; Ataeian et al., 2016). Slides were washed in PBS three times for 10 min each before incubation with secondary antibody solution (5% normal goat serum, 0.01% Triton-X100 and 1 µg/µl secondary antibodies) for 1 h at room temperature. Alexa488 goat anti-mouse (A11001; 1/100 dilution; Invitrogen), Alexa546 goat anti-mouse (A11003; 1/100 dilution; Invitrogen), or Alexa488 goat anti-rat secondary antibodies (A48262; 1/100 dilution; Invitrogen) were used. Slides were then washed in PBS three times for 10 min each and mounted in Fluoroshield with DAPI (F6057, Sigma-Aldrich). Images were captured using a 60× oil objective (NA 1.42) with an Olympus IX81 spinning disc confocal inverted microscope as described above.

Western blotting was performed on whole-worm lysates as previously described (Gusnowski and Srayko, 2011). Briefly, worms were pelleted at 1000 g for 1 min and washed three times in 400 ml H₂O. An equal volume of 2× Laemmli buffer (Sigma-Aldrich) was added and the sample was boiled at 95°C for 5 min. Samples were then loaded onto a 10% SDS-PAGE resolving gel. A prestained protein marker (Page ruler pre-stained protein ladder 26616, Thermo Fisher Scientific) was used to estimate the relative mass of the proteins. Proteins were transferred to a nitrocellulose membrane (Hybond-N, GE Healthcare) at 100 V for 2 h. The membrane was blocked in 8% skim milk in Tris-buffered saline with Tween 20 (TBST; 20 mM Tris-HCl, pH 7.4, 500 mM NaCl and 0.05% Tween 20) for 1 h. The following primary antibodies were used: mouse anti-FLAG (F1804, 1/500, Sigma-Aldrich), rat anti-Ollas (NBP1-06713, 1/250, Novus Biologicals), anti-tubulin (DM1A, 1:500, Sigma-Aldrich) and anti-MSP 4A5 antibodies (4A5, 1/1000, Developmental Studies Hybridoma Bank). All antibodies were diluted in TBST+4% skim milk and incubated for 1 h at room temperature. Goat anti-rat (31470; 1/5000; Invitrogen) and goat anti-mouse (1705047; 1/5000; BioRad) HRP-bound secondary antibodies in TBST+4% skim milk were incubated with the membrane for 1 h at room temperature. The secondary antibodies were detected via Super Signal West Pico ECL (Thermo Fisher Scientific) and imaged on a ChemiDoc MP Imaging system (Bio-Rad).

Worms were immobilized in 7 µl egg buffer containing 5 mM tetramisole hydrochloride before imaging. When the worms stopped moving, the cover slip was inverted onto a 2% agarose pad. Fluorescent images of three planes, spaced 1.5 µm apart, were taken at 30 s intervals using a 60× oil objective (NA 1.42) on an Olympus IX81 spinning disc confocal microscope, as described above.

Statistical analyses were performed using GraphPad software (USA). All Student's t-tests were unpaired, two-tailed, unless otherwise stated. Results were significant at P<0.05.

We thank the Caenorhabditis Genetics Center [funded by National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440)], Dr Shohei Mitani [National Bioresource Project (Tokyo)] and Dr David Pilgrim (University of Alberta, Canada) for providing strains, and Dr Shelagh Campbell (University of Alberta, Canada) for providing Alexa488 goat anti-rat antibodies. We thank Drs Paul Mains and Shelagh Campbell for insightful comments on the manuscript, and members of the Srayko laboratory for helpful discussions throughout the course of this work.