The centralspindlin complex regulates cytokinesis and morphogenesis in the C. elegans spermatheca Free

Organ morphogenesis is a complicated dynamic process that requires the right number of cells with the right size and shape positioned in the right place with the right orientation at the right time. Great progress has been made in the elucidation of the genes and mechanisms involved in morphogenesis of a variety of organs in model organisms (Carron and Shi, 2016; Langdon and Mullins, 2011; Niehrs, 2004; Stuckenholz et al., 2005; Wieschaus, 2016). For example, a morphogenesis screen in the Drosophila tracheal system has identified over 70 patterning and morphogenesis genes on the third chromosome that contribute to the development of this organ (Ghabrial et al., 2011). Despite the successful identification of genes involved in organ morphogenesis, examination of the cellular mechanisms (for example, cell division, cell movement and rearrangement) of the building process itself remains technically challenging owing to the complexity of organs even in lower model organisms, such as Drosophila. The challenges are even greater in vertebrates, such as zebrafish and mouse. The exquisitely simple organization and invariant lineage of Caenorhabditis elegans (C. elegans) offer unique technical advantages for overcoming this difficulty.

C. elegans is a transparent roundworm with digestive, excretory and reproductive organs that consist of simple tubes of single-layered epithelia (Shaye and Soto, 2021). In our previous study, we used animals engineered with green fluorescent protein (GFP)-labeled lumenal membranes and examined the requirement for single genes for tubulogenesis of the intestine and excretory canal by a genome-wide morphological RNA interference (RNAi) analysis (Zhang et al., 2011, 2012). Owing to a spectrum of loss-of-function phenotypes induced by RNAi, our screen, as well as others, enabled identification of a subset of upstream regulators and ‘housekeeping’ genes and their specific function in tubulogenesis (Winter et al., 2012; Zhang et al., 2011, 2012). However, the functions of many essential genes in organ/tube morphogenesis remain elusive because knockdown of these genes using a standard protocol leads to lethality at an early embryonic stage when the intestine and excretory canal have not started or only just started to form, and conditional post-embryonic RNAi results in no phenotype in these two tissues, probably due to the completion of their morphogenesis before RNAi initiation.

To understand further the roles of these genes in organ morphogenesis, we specifically looked for the genes involved in regulating the formation, size and shape of a different tubular organ – the spermatheca. The hermaphrodite spermatheca is composed of 24 epithelial cells organized into an accordion-like stretchable tube that acts in sperm storage, ovulation and fertilization (Aono et al., 2004). The spermatheca develops post-embryonically starting from late L3/early L4 stages, and is therefore ideal for a tube morphogenesis screen with conditional post-embryonic RNAi. Such a screen will bypass the essential early functions of embryonic- and larval-lethal genes, allowing those genes to be identified if they play a role in spermathecal morphogenesis.

In our conditional RNAi screen for genes that affect spermathecal architecture using the ERM-1::GFP strain, we found that post-embryonic depletion of cyk-4 led to a highly penetrant (over 90%) spermathecal morphology defect. C. elegans CYK-4 encodes an ortholog of human RACGAP1 (Rac GTPase activating protein 1; previously known as MgcRacGAP1) and exhibits GTPase activator activity that generally promotes GTPase activity, thereby terminating signaling events (Jantsch-Plunger et al., 2000). CYK-4 forms the centralspindlin complex with the kinesin KIF23 (MKLP1)/ZEN-4 to specify contractile ring assembly and is required for cytokinesis in nearly all cell types (Glotzer, 2005). In the early C. elegans embryo, CYK-4 functions to activate RhoA through the RhoGEF ECT-2, thereby promoting the assembly of linear actin for contractile ring formation via the formin CYK-1 (Loria et al., 2012; Simon et al., 2008; Zhang and Glotzer, 2015). CYK-4 also inhibits the RAC proteins CED-10 or RAC-2, which in turn inhibits the Arp2/3 component ARX-2-dependent assembly of branched F-actin formation and stimulates furrow ingression (Canman et al., 2008; Pavicic-Kaltenbrunner et al., 2007; Pintard and Bowerman, 2019). Centralspindlin complex formation depends on upstream factors, such as Aurora B kinase encoded by air-2 in C. elegans. AIR-2 is part of the chromosome passenger complex and acts to promote cytokinesis and regulate abscission timing by phosphorylating ZEN-4 and facilitating centralspindlin oligomerization (Basant et al., 2015; Guse et al., 2005). Although centralspindlin complex components act together in most somatic cells, a recent study showed that CYK-4 functions independently of its centralspindlin partner ZEN-4 to activate RHO-1, which in turn regulates cytokinesis-like oocyte cellularization in germline syncytia (Green et al., 2011; Lee et al., 2018). CYK-4 and ZEN-4 are also required for foregut tubulogenesis independent of their general activities in cytokinesis (Portereiko et al., 2004).

Intuitively, cytokinesis is linked with morphogenesis by ensuring that the correct number of cells undergo morphogenesis. For example, Racgap1 (Ogre)-mediated cytokinesis is important for zebrafish nervous system development by ensuring a sufficient number of differentiated neurons (Warga et al., 2016). Recent studies have uncovered specific features of cytokinesis during morphogenesis of various epithelial tissues (Bai et al., 2020). Thus, there is compelling evidence for a function of cytokinesis in morphogenesis beyond its role in cell division. Consistent with this view, Rathbun et al. (2020) showed that in the zebrafish Kupffer's vesicle, mitotic cells strategically place their cytokinetic bridge at the rosette center, which in turn functions as a platform for Rab11-mediated targeting of Cftr and other apical protein components to facilitate de novo lumen formation. Here, we characterized the function of several cytokinesis regulators in spermathecal morphogenesis. We revealed that the centralspindlin complex components CYK-4 and ZEN-4 coordinate with their downstream effectors to direct spermathecal morphogenesis by regulating cytokinesis of spermathecal cells.

To characterize the phenotypes upon cyk-4 knockdown in detail, post-embryonic RNAi was carried out by bleaching 50-60 adult worms to allow the eggs to hatch on individual plates with bacteria expressing double-stranded RNA transcribed from part of the cyk-4 cDNA or bacteria containing empty vector. By 72 h after bleaching at 20°C, the majority of the control worms had reached adulthood and generated progeny. In contrast, cyk-4 RNAi resulted in a sterile phenotype with altered organization of the reproductive system and severely reduced brood size (almost no eggs laid; Fig. 1A,B Fig. S1A). About 42% of cyk-4(RNAi) animals did not produce embryos in the uterus and showed severely altered morphology of the spermatheca (as well as the uterus and vulva; Fig. 1Ai), whereas the rest, with mild morphological defects, had one or two embryos in the uterus (Fig. 1Aii). We also observed a protruding vulva phenotype in about 32% of cyk-4(RNAi) worms (Fig. 1Ai). A small percentage of worms (<10%) also had an endomitotic oocyte (Emo) phenotype, evidenced by the appearance of endomitotic nuclei under Nomarski imaging or large dense areas of nuclear staining indicative of condensed chromosomes (Fig. 1Aiii, Fig. S1B; white arrows).

To evaluate spermathecal morphogenesis specifically, we examined the spermatheca of worms carrying transgenes that label apical (ERM-1, Ezrin/Radixin/Moesin), junctional (AJM-1, adherens junction molecule) and basolateral membranes (LET-413, scribble planar cell polarity protein). A comparison of cyk-4(RNAi) and control worms revealed markedly altered localization patterns of these membrane markers in spermatheca (Fig. 1C). In L4-stage control animals, the membrane markers outlined the regular cell boundaries of two rows of cells in the spermatheca, especially in the distal portion. In adult spermatheca containing an oocyte, the cells that form the sac were expanded and stretched and all the membrane markers showed a consistent pattern. However, in L4 and adult cyk-4 RNAi animals, all the markers revealed distortion of the spermatheca, and the GFP signals at the cell boundaries were largely missing, indicating the absence of cellular partitions (Fig. 1C). These results suggested that CYK-4 is required for the proper development of the spermatheca.

A previous study revealed that CYK-4 is required for oocyte cellularization in germline syncytia (Lee et al., 2018). To rule out the possibility that the altered spermathecal morphology arises as a secondary consequence of germline defects, we carried out the cyk-4 RNAi in rrf-1(ok589) mutants. In these mutants, RNAi acts normally in the germline but is defective in many somatic tissues. In contrast to an almost 100% sterility of cyk-4(RNAi) worms (Fig. 1A), rrf-1(ok589);cky-4(RNAi) worms had a partial sterility phenotype, with a ∼60% decrease of brood size compared with control animals, and 100% embryonic lethality of the eggs laid. However, almost all rrf-1(ok589);cky-4(RNAi) spermathecae examined exhibited intact morphology (Fig. 1D, Fig. S2, Table S1). These results suggest that germline-specific RNAi of cyk-4 is not sufficient to produce the spermathecal morphology defects. Thus, the spermathecal morphology defects observed upon global cyk-4 RNAi are unlikely to be secondary to the effects on the germline. To confirm the role of CYK-4 in spermathecal development, we examined the spermathecal morphology in cyk-4(t1689ts) and cyk-4(or749ts) mutants. The cyk-4(t1689ts) mutant carries an S15K mutation in the N-terminal coiled-coil domain of CYK-4 that disrupts CYK-4/ZEN-4 interaction (Jantsch-Plunger et al., 2000), whereas the cyk-4(or749ts) mutant carries an E448K mutation that compromises the function of the GAP and C1 domains (Zhang et al., 2015) at the non-permissive temperature (25°C). To assess the effect of CYK-4 mutations on spermathecal development, we upshifted mutants to 25°C at various time points and for various periods. We found that upshifting cyk-4(t1689ts) mutants at L3 stage for 8 h already affected spermathecal morphogenesis, and the earlier and the longer the time of the upshift, the more severe the effect was. cyk-4(t1689ts) spermathecae exhibited a spectrum of morphological defects, ranging from a reduced number of cellular partitions with multiple nuclei to spermathecal sacs lacking cell boundaries, similar to cyk-4 RNAi animals (Fig. 1E, Fig. S3A). However, upshifting cyk-4(or749ts) mutants at L3, L2 and L1 for 24-40 h did not perturb spermathecal development (Fig. 1E), unless the shift occurred immediately after egg-laying, which produced <10% of spermathecal morphology defects, probably as a result of non-specific confounding factors. This result is consistent with the previous report that the CYK-4 GAP and C1 domains are dispensable for germline formation during the L1-L4 stages, whereas the coiled-coil domain, which mediates the CYK-4/ZEN-4 interaction, is required during the L1-L4 stage (Lee et al., 2018). These data suggest that CYK-4 coiled-coil domain-dependent CYK-4/ZEN-4 interaction, but not CYK-4 GAP activity, is important for spermathecal morphogenesis and function.

Lack of cell boundaries in cyk-4 RNAi animals suggests a reduction of cell numbers. We therefore constructed transgenic lines that specifically mark the nuclei of spermathecal cells by using a spermathecal promoter F55B11.3 to drive expression of GFP fused with a nuclear localization signal and histone 2B (PF55B11.3::GFP::H2B). The expression of F55B11.3 can be observed around mid-L4 stage. Around this stage, all the spermathecal cell divisions are just completed, yielding 24 cells organized into two regional groups: distally, eight cells are aligned in two rows that form a narrow corridor or neck; proximally, 16 cells form a wider bag-like chamber. Indeed, we observed 22-24 nuclei in control animals carrying this transgene. In contrast, cyk-4 RNAi adult animals showed a significantly reduced average of 12.6 (n=29) spermathecal nuclei (Fig. 2A,B). Moreover, the spermathecal nuclei tended to gather in clusters (Fig. 2A, arrows). We suggest that this phenotype is caused by continuous nuclear divisions that occur without cell divisions as a result of unsuccessful cytokinesis. Similar results were obtained by counting the number of nuclei in the spermathecal region after staining nuclear DNA with 4′,6-diamidino-2-phenylindole (DAPI) (Fig. 2C, brackets), although the nuclear staining in other tissues obscured the spermathecal nuclei to some extent.

To corroborate these findings, we examined the expression of two markers of spermathecal differentiation: GFP fused with the promoter of let-502, encoding the homolog of Rho-associated kinase, and mel-11, encoding a myosin phosphatase regulatory subunit (Wissmann et al., 1999). The let-502-driven GFP was expressed in 30 spermathecal cells, including six spermathecal-uterine valve cells, with the highest levels in the proximal and distal cells. mel-11-driven GFP was expressed mainly in 16 spermathecal sac cells. Consistent with the reduced number of spermathecal cells, both markers exhibited altered expression patterns in cyk-4 RNAi animals. A high level of let-502::GFP expression in both proximal and distal regions was observed in control animals, whereas let-502::GFP was only detected in the proximal region in cyk-4 RNAi animals. In control animals, mel-11::GFP signals were evenly distributed throughout the whole spermathecal sac, whereas in cyk-4 RNAi animals, the mel-11::GFP expression pattern was disorganized and clustered in part of the spermatheca (Fig. 2D). The altered expression patterns of both markers are consistent with a cell division defect, resulting in a failure in generation of distal spermathecal cells. Together, these data indicate that CYK-4 is essential for spermathecal cell division.

The significantly reduced cell/nucleus number and clustering and mispositioning of nuclei observed in cyk-4 RNAi spermathecal cells suggested a failure in cytokinesis in these animals. In order to evaluate cytokinesis during spermathecal cell division, we visualized the spermathecal cell membranes by expressing GFP fused with the pleckstrin homology (PH) domain from phospholipase C (Audhya et al., 2005; Klompstra et al., 2015) under control of the erm-1 promoter. We examined spermathecal development from mid-L3 until young adult stages, with a focus on the period when the spermathecal cell divisions take place. To describe spermathecal development conveniently, we adapted the ten sub-L4 stages that were used for distinguishing L4 vulva morphogenesis (Cohen et al., 2020). In wild-type animals expressing the membrane marker driven by the erm-1 promoter, GFP was barely detected in spermathecal precursor cells [SPCs; Fig. 3Ai, Fig. S4i (with brightfield for sub-L4 staging according to vulval morphology)] from mid-L3 to early L4 (roughly L4.0) stage, although GFP clearly labeled the membranes of uterine cells. Gradually, three or four spermathecal cells became visible by L4.1 stage and five to seven by L4.2 stage (Fig. 3Aii-iii, Fig. S4ii,iii), corresponding to one round of cell division from two SPCs generated in mid-L3 stage plus two or three spermathecal cells derived from the uterine lineage. At L4.3 stage, the uterine-derived spermathecal cells finished one round of division, yielding 10-11 visible cells (Fig. 3Aiv). By L4.4 stage, after another round of division of the distal SPC cells, 12 spermathecal cells were visibly aligned in pairs in the middle plane (Fig. 3Av, Fig. S4iv). At L4.5,L4.6 stages, while the uterine lumen was forming and expanding, two more rounds of SPC-derived cell divisions were completed, resulting in 24 cells in total with 18 SPC-derived cells and six cells derived from the uterus lineage. In the middle plane of the spermatheca at L4.5-L4.6 stages, five or six uterine-derived and 10-12 SPC-derived spermathecal cells could be observed (Fig. 3Avi,vii, Fig. S4v,vi), depending on the spermathecal alignment. It is worth noting that from L4.4 to L4.6 stages, spermathecal cells undergo rearrangement to some extent, resulting in significant organ shape change. The length along the proximal-distal axis is reduced and the boundary between uterus and spermatheca is more distinct. The spermathecal lumen starts forming around L4.6-L4.7 (Fig. 3Aviii, Fig. S4vii) after the uterine lumen has almost fully formed and during the last spermathecal cell division (Kimble and Hirsh, 1979). Shortly afterwards, the spermathecal cell membranes, especially those on the apical and lateral surfaces, become folded and convoluted (Fig. 3Aix, Fig. S4viii). At the young adult stage, the spermatheca-uterine valve becomes more apparent (Fig. S4ix) owing to the full extension of the core cell body away from the uterus (Praslicka and Gissendanner, 2015). Spermathecal cell membranes expand during oocyte passage and then collapse after oocyte exit is complete (Fig. S4x-xii).

Upon RNAi depletion of CYK-4, we found that cytokinesis of the spermathecal lineage cells was initiated in most animals, but the dividing membrane was often distorted or failed to elongate properly. In most animals, the first cellular cleavage managed to reach completion, but the daughter cells contained more than one nucleus with variable sizes owing to arrest of the subsequent cell cycles at the nuclear division stage (Fig. 3B, top left and top middle). In some animals, regardless of the failure of the first cleavage, the spermathecal precursor cells continued through multiple cell cycles, accumulating multiple cortical ruffles and partially formed cleavage furrows (Fig. 3B, top right). As a result, these spermatheca displayed morphological defects with characteristic formation of multinucleated cells and the presence of seemingly different numbers of cellular partitions. We considered it as two partitions when the furrow between two nuclei reached at least half of the distance from the plasma membrane to the presumptive location of the central spindle. We quantified the cell division attempts by counting the number of complete or incomplete partitions. Compared with control animals with multiple completely divided cells in early L4 spermatheca (Fig. 3A), cyk-4 RNAi animals at equivalent stages exhibited two to nine spermathecal cell partitions (Fig. 3B). As development proceeded, abnormal membrane formation became more evident, including membrane branching, multiple breaks and the presence of various membranous shapes and folds alongside the presumed cleavage furrow, without altering the cell partition numbers (Fig. 3B, Late L4 and ‘Adult’ panels). Similar membrane remodeling defects have been observed in early-stage embryos with cell division defects (Green et al., 2011; Kniazeva et al., 2012).

To gain a greater insight into this process, we monitored the mitotic progression of spermathecal cells subjected to CYK-4 depletion. Time-lapse imaging was performed on wild-type and cyk-4 RNAi animals at early- to mid-L4 stages. In wild-type larvae, several cell division events were recorded within 20-30 min. Although the individual cleavage furrow ingressions were not easily discerned owing to different division planes, we did observe full membrane extension around the newly generated cells, which suggests that cytokinesis was complete (Fig. 3C, top, white arrowheads; Movie 1). However, in some CYK-4-depleted spermathecal cells, cytokinesis was initiated, and cleavage furrows formed and ingressed asymmetrically to some extent towards the presumptive apical membrane of the spermatheca before stalling (Fig. 3C, middle, blue arrowheads; Movie 2). Some other cells exhibited partial furrow ingression followed by regression (Fig. 3C, middle, yellow arrowheads; Movie 2). We also observed attempted furrow ingression characterized by restricted membrane budding (Fig. 3C, bottom, red arrowheads; Movie 3). In addition, we detected successful ingressions, albeit with distorted membranes (Fig. 3C, bottom, white arrowheads; Movie 3). Thus, CYK-4 is required for efficient completion of cytokinesis in spermathecal cells.

RacGAP and Rho-specific guanine nucleotide exchange factor activity at the central spindle is thought to provide a signal for RhoA-mediated F-actin reorganization and contractile ring assembly at the furrow, subsequently stimulating ingression and membrane delivery (D'Avino et al., 2005). We therefore sought to examine whether F-actin assembly is affected in spermathecal cells of cyk-4 RNAi animals.

To visualize actin dynamics during spermathecal development, we used F55B11.3 to drive expression of the fluorescent actin reporter ABD::GFP, which consists of the F-actin-binding domain of the spectraplakin VAB-10 fused to GFP. ABD::GFP has previously been used to monitor F-actin in other C. elegans cells (Bosher et al., 2003). In wild-type L4 to young adult stage spermathecal cells, the actin cytoskeleton was predominantly assembled as an actin filament network in the cytoplasm, with a tortuous, branching and randomly oriented, interconnected pattern (Fig. 4A,B, blue arrowheads). As reported previously (Wirshing and Cram, 2017), after oocyte entry in adults, the cytoplasmic actin network is remodeled into circumferential actin bundles oriented along the long axis of each cell (Fig. 4C, yellow arrowheads). During cell division, F-actin is targeted to the cytokinetic furrow through the redistribution of cortical actin filaments or vesicle-based delivery to form the actomyosin contractile ring, which in turn may be remodeled and incorporated into cortical actin at the cell–cell contact after cell division is completed (Albertson et al., 2008). Cortical actin (Fig. 4A,B, white arrowheads) associated with the plasma membrane and actin puncta (Fig. 4A,B, red arrowheads) presumptively associated with vesicles were also detected. In cyk-4 RNAi animals, the cytoplasmic F-actin network was similar to that in wild-type spermathecal cells, although the overall actin pattern and the alignment of the filaments sometimes appeared to be messy (Fig. 4A,B, blue arrowheads), which may be secondary to the cell division defect. However, serial sections revealed that actin was completely absent from the presumptive cell equator between the dividing or divided nuclei (Fig. 4B, dotted white lines and white arrowheads). These results corroborate the cytokinesis defects observed using membrane markers (Fig. 3B) and are consistent with the hypothesis that the recruitment of actin to the contractile ring is perturbed in cyk-4 RNAi animals. Interestingly, a seemingly normal pattern of circumferential actin bundles at least in some spermathecal cells was observed in all the spermathecae examined (n=17) once there was an oocyte or an oocyte-like structure trapped in the spermatheca, even though only part of the spermathecal tube was formed owing to a lack of cells (Fig. 4C, yellow arrowheads). This is consistent with the notion that the cue for remodeling of circumferential actin bundles is independent of morphogenesis, and comes instead from oocytes (Kelley and Cram, 2019; Stephens et al., 2017).

We next examined whether CYK-4 is expressed in spermatheca by imaging an in situ-tagged GFP::CYK-4 (Wang et al., 2021). CYK-4 is expressed in spermathecal cells in a stage-dependent manner. From L4 to young adult stages, CYK-4 appeared as scattered spots at the interface of the two rows of spermathecal cells (Fig. 5A). However, the number of spots decreased at later stages, and the spots completely vanished after oocytes passed through the spermatheca (Fig. S5A). Both the spotty pattern and dynamics of CYK-4 in spermatheca are consistent with the previous observation of CYK-4 persisting at the division remnants for several cell cycles in early embryos (Jantsch-Plunger et al., 2000). To delineate further the localization of CYK-4, we fused the CYK-4 coding sequence with Tomato under the control of the erm-1 promoter and confirmed that CYK-4 localization in the transgenic lines is consistent with its endogenous pattern (Fig. S5B). We then introduced CYK-4::Tomato into Perm-1::PH::GFP and PF55B11.3::ABD::GFP strains. In the double transgenic lines of CYK-4::Tomato and PH::GFP, most of the CYK-4-positive spots did indeed reside on the spermathecal cell membranes with predominant localization on the apical/lumenal surface of the spermatheca (Fig. 5B, Fig. S5C, yellow dots in enlarged view). We also observed localization of CYK-4 in the cytoplasm adjacent to the apical membrane of some spermathecal cells (Fig. 5B, Fig. S5C, white arrow in enlarged view). This suggests that CYK-4 is retained on internalized midbody remnants. Of note, the predominantly apical localization implies that the midbody remnants migrate to apical subdomains after cytokinesis, in line with the observation in other epithelia (Bai et al., 2020).

Given that the expression of F55B11.3 cannot be clearly detected until L4.5-L4.6 stages, a time point when spermathecal cell divisions are almost completed, we could not examine the relative location of actin and CYK-4 in dividing cells. However, we did occasionally detect colocalization of CYK-4 with actin in post-mitotic spermathecal cells (Fig. S5D, yellow arrows), which may represent the actin that was associated with the cell membrane or transiently assembled on the remnants of previous cell divisions (Waddle et al., 1994).

To understand further the midbody dynamics during spermathecal morphogenesis, we monitored the CYK-4::Tomato pattern in cells at early stages of spermathecal development. Upon examination of spermathecal cells undergoing mitosis, we noted that the CYK-4-positive midbodies are positioned where the future lumen forms early around L4.1-L4.2 stages (Movie 4). We also observed new cell membranes budding from the lumen formation midline, accompanied by CYK-4 assembly at the midline at L4.3-L4.5 stages (Movie 5). These observations suggest that completion of cytokinesis is a prerequisite for deposition of midbody at the lumen formation site, and also implicate a strategic role of midbody deposition in polarization of spermathecal cells throughout spermathecal development.

To examine whether cytokinesis is a prerequisite for CYK-4 localization at apical membrane, we created a cytokinesis-deficient condition by using zen-4 knockdown. In C. elegans early-stage embryos, ZEN-4 and CYK-4, the two components of the centralspindlin complex, function interdependently to ensure cytokinesis completion (Fig. 6A) (Jantsch-Plunger et al., 2000). We therefore expect that ZEN-4 is required for cytokinesis in spermathecal cells. Indeed, zen-4 RNAi initiated from L1 stage leads to fully penetrant sterility in all the strains tested. Over 95% of zen-4 RNAi animals displayed spermathecal morphology defects, which are not secondary to abnormal germline development, similar to those observed in cyk-4 RNAi animals (Fig. 6B,C, Fig. S2, Table S1). The temperature-sensitive zen-4(or153) mutants phenocopied zen-4 RNAi animals after being upshifted to the restrictive temperature (25°C) from L2 stage (Fig. 6D, Fig. S3A).

We next investigated the effect of zen-4 RNAi on CYK-4 localization. In both the endogenously tagged CYK-4::GFP strain and the Perm-1::PH::GFP/Perm-1::CYK-4::Tomato transgenic strain, CYK-4 puncta were completely abolished in zen-4 RNAi animals (Fig. 5C). Time-lapse video imaging revealed that CYK-4 signals appeared transiently at least in some cells; however, CYK-4 was unable to localize normally because of an unrecognizable midline after failed cytokinesis and/or lack of ZEN-4 in zen-4 RNAi animals (Fig. 5D, Movie 6). Thus, these data imply that CYK-4 recruitment to the midline or presumptive apical membrane depends on ZEN-4-mediated cytokinesis in spermathecal cells.

Localization of CYK-4 to the apical membrane suggests a potential function of midbody remnants as a polarization cue that directs lumen formation. Indeed, the severe morphological defects observed in the spermatheca of cyk-4 RNAi and mutant animals, including unrecognizable or distorted midlines evidenced by PH::GFP-labeled plasma membrane and ERM-1::GFP-tagged apical membrane, and misorientated actin filaments, provide support for this hypothesis. In line with this hypothesis, we further found that, in control spermathecae at L4.3-L4.4 stages ERM-1 was localized predominantly at the apical membrane, whereas in cyk-4 RNAi animals ERM-1 exhibited an abnormal apical localization with significant expansion toward the lateral membrane domain and branching in different directions (Fig. S3B). This observation suggests a polarity defect prior to the profound morphology defect upon cyk-4 RNAi. Phenotypically, this polarity defect is different from the cytokinesis-independent polarization defects observed in the C. elegans foregut, where cytokinesis and cell numbers are normal but the apical and junctional markers fail to be targeted to the apical domains together with disorganized actin and microtubules (Portereiko et al., 2004). In contrast, the spermathecal cells in cyk-4 RNAi animals are able to polarize and to form junctions (see the staining patterns in Figs 1C, 2C and 6C) despite reduced cell numbers. Thus, the apically placed midbody remnants may contribute to epithelial polarization, and this function is linked to the conventional role of CYK-4 in cytokinesis.

Next, we investigated whether CYK-4 affects spermathecal development in the context of other cytokinesis regulators deduced from the studies in embryos (Fig. 6A). We depleted relevant components, such as AIR-2 (Aurora B kinase) and CYK-1 (formin), as well as the two GTPases RHO-1 and CDC-42, by post-embryonic RNAi and analyzed the spermathecal morphology using transgenic worms carrying Perm-1::PH::GFP. Similar to cyk-4 and zen-4 RNAi, over 95% of rho-1 RNAi animals displayed spermathecal morphology defects, in a germline development-independent manner (Fig. 6B, Fig. S2, Table S1).

In striking contrast, AIR-2 depletion led to a low penetrance (<20%), whereas CYK-1 and CDC-42 depletion had very little effect (<5%) on spermathecal morphology (Fig. 6B, Table S2). We confirmed RNAi efficiency by observing that standard cyk-1 RNAi led to embryonic lethality, whereas air-2 RNAi caused sterility with reduced brood size and 100% embryonic lethality. Based on this observation, it is tempting to conclude that cytokinesis of spermathecal cells might not require CYK-1, although it is essential for embryonic cytokinesis. However, the cyk-1 and air-2 RNAi results may also be explained by a phenotypic threshold effect whereby these two proteins are required at very low levels in spermatheca (Gjuvsland et al., 2007; Rossignol et al., 2003).

CYK-4 activates GTP hydrolysis via RhoA and Cdc42 in vitro (Jantsch-Plunger et al., 2000). We therefore tried to explore the genetic interaction between CYK-4 and RHO-1 or CDC-42. The effects of RHO-1 or CDC-42 inhibition on spermathecal morphology were validated using multiple membrane and nucleus markers (Fig. 6C). We detected a reduced number of junctional cell boundary signals tagged by AJM-1, irregular and circular patterns of ERM-1 signals, and a reduced number of H2B-marked nuclei in rho-1, but not cdc-42, RNAi animals (Fig. 6C). These observations suggest that RHO-1, similar to CYK-4, contributes to spermathecal morphogenesis likely by regulating cytokinesis. Considering that single knockdown of cyk-4 and rho-1 generates severe phenotypes, we titrated and optimized the knockdown level by diluting the RNAi bacteria with empty vector or unc-22 bacteria to generate a mild phenotype (Fig. 7). The final bacterial ratios were cyk-4:rho-1:empty vector (or unc-22)=1:2:1 for double RNAi, cyk-4:empty vector (or unc-22)=1:3 for cyk-4 single RNAi and rho-1::empty vector (or unc-22)=1:1 for rho-1 single RNAi. This ensured that the amount of cyk-4 and rho-1 bacteria under single and double RNAi conditions remained the same. We found that cyk-4 and rho-1 double RNAi animals generally showed stronger and/or more penetrant phenotypes than the single RNAi animals in various marker backgrounds (Fig. 7A,B, Fig. S6). These results corroborate the role of CYK-4/RHO-1 in cytokinesis in early embryos, and also suggest that RHO-1 and CYK-4 regulate the cytokinesis of spermathecal cells through the same or a parallel pathway.

Knockdown of CDC-42 resulted in neither significant change in spermathecal morphology nor reduced cell numbers (Figs 6C and 7C,D), even though it led to reduced brood size and embryonic lethality (Table S2). Interestingly, double knockdown of cyk-4 and cdc-42 at a ratio of 1:3 led to a partial suppression of spermathecal cell cytokinesis defects, compared with single knockdown of cyk-4 [cyk-4:empty vector (or unc-22) at 1:3; Fig. 7C,D]. This result is consistent with the previously suggested inhibitory effect of CYK-4 on CDC-42 activity (Zhuravlev et al., 2017). However, the question of whether CDC-42 is a direct substrate of CYK-4 awaits further investigation with more detailed biochemical and cell biological analyses.

Although the regulators and mechanisms controlling cytokinesis have mostly been revealed, little is known about the effects of cytokinesis on morphogenesis. In this study of the interaction between spermathecal cell cytokinesis and morphogenesis, we provide several insights into the function of the centralspindlin complex and its effectors in controlling cytokinesis and spermathecal morphogenesis. First, our work extends the previous study in multiple epithelia describing the function of the Aurora B kinase AIR-2 during cytokinesis and its contribution to organ morphogenesis (Bai et al., 2020), and reveals a link between centralspindlin-mediated cell division and spermathecal morphogenesis. We cannot formally rule out the possibility that the spermathecal morphology defects may be derived from other defects, such as cell fusion or mispatterning prior to differentiation of the spermathecal lineage. However, given the known role of the centralspindlin complex in cytokinesis of early-stage C. elegans embryos, together with the results shown in this study, it is very likely that the centralspindlin complex as well as RHO-1 regulates spermathecal development by controlling cytokinesis. It is also possible that CYK-4, ZEN-4 and RHO-1 may play a role in spermathecal morphogenesis independently of their role in cytokinesis. This possibility should be further investigated in future studies but does not weaken our conclusion regarding the role of the centralspindlin complex in spermathecal cell cytokinesis and morphogenesis. Second, retention of CYK-4 at the apical midline suggests that the CYK-4-positive midbody remnants function as a polarizing cue during apicobasal polarity establishment in the C. elegans spermatheca. Considering that CYK-4 is specifically enriched along the apical surface, we propose that it may also recruit cell-polarity proteins and signaling complexes into the lumen formation site to transmit cell polarity. In this model, it is likely that CYK-4 fills the same instructive role in spermatheca that AIR-2 plays in other tissues (Bai et al., 2020). However, we note that the role of CYK-4 in spermathecal polarization is mechanistically different from its cytokinesis-independent function in polarizing the foregut epithelium. Third, this study supports a model in which the central spindle components contribute to furrow ingression and spermathecal morphogenesis through RHO-1-mediated recruitment of F-actin. Loss of CYK-4 results in dramatic depletion of F-actin from the equatorial cell cortex. This result suggests a role of the central spindle in coordinating F-actin with cytokinesis. Support for this model comes from previously published evidence that actin and vesicles are targeted to the cleavage furrow via microtubules and the central spindle in Drosophila melanogaster embryos in a Rho-specific guanine nucleotide exchange factor-dependent manner (Albertson et al., 2008). Fourth, this study indicates that two Rho family members, RhoA (RHO-1) and CDC-42, are involved in spermathecal morphogenesis by participating in cell division. Loss of RHO-1 phenocopies and further enhances the spermathecal morphogenesis phenotypes in cyk-4 RNAi worms. These data are consistent with the established role of RHO-1 as an effector of CYK-4 and ZEN-4 in the cytokinesis of C. elegans embryos. Knockdown of CDC-42 does not affect spermathecal cell cytokinesis on its own, yet leads to substantial rescue of the cyk-4 RNAi-induced spermathecal cell cytokinesis phenotype. It is seemingly paradoxical that GTPase activity may not be required for spermathecal development, evidenced by the lack of a spermathecal phenotype in cyk-4(or759ts) mutants, in which GAP/C1 function is disrupted. However, we envision that RHO-1 and CDC-42 are probably not the direct substrates of CYK-4 at least under the scenario of spermathecal morphogenesis.

In early-stage embryos, CYK-4, together with its centralspindlin complex partner, the kinesin-6 ZEN-4, specifies contractile ring assembly by activating RhoA and inhibiting Rac1 (Glotzer, 2005; Pintard and Bowerman, 2019). Interestingly, CYK-4, ZEN-4 and RHO-1 play an analogous role during spermathecal morphogenesis by ensuring the completion of cytokinesis and subsequent patterning of cells for correct organ development. Surprisingly, under our experimental setting, knockdown of cyk-1 or air-2 did not lead to the same level of defects in spermathecal morphogenesis as those observed in cyk-4, zen-4 or rho-1 RNAi animals. Given the nature of RNAi knockdown, it is likely that the residual function of CYK-1 and AIR-2 is enough to drive the completion of cytokinesis in the majority of spermathecal cells. Another possibility is that not all cell division components are required at similar levels in the same set of cells. Recently, it has been shown that cytokinetic regulation in individual cell types has more variation than previously realized. A cell type-specific variation in the cytokinetic requirement for a robust formin CYK-1-dependent filamentous-actin (F-actin) cytoskeleton has been identified (Davies et al., 2018). Similarly, there might be other kinases and/or actin modulators that exert this function in spermathecal cells.

Our results also mimic the phenotypes observed in the gonad of cyk-4(RNAi)/rho-1(RNAi) or cyk-4(or749ts) mutants (Green et al., 2011; Lee et al., 2018). Knocking down CYK-4 or RHO-1 in the gonad leads to severe multinucleation with absent partitions throughout the gonad, which demonstrates a role of CYK-4 in gonad cellularization. In the germline, oocyte cellularization does not require the ZEN-4 subunit of centralspindlin. However, we show here that CYK-4 and ZEN-4 are both required for cytokinesis in spermathecal cells, in line with the proposal that cytokinesis in somatic tissues and cellularization in the germline are two independent processes although they share some regulatory components. Although CYK-4 and ZEN-4 have tissue-specific functions, they are expected to function together in the centralspindlin complex in regulating cytokinesis in most, if not all, cell types.

In human, the two centralspindlin complex components RACGAP1 and the kinesin family member KIF23 act together with RHOA GTPase (Konstantinidis et al., 2015; Romero-Cortadellas et al., 2021; Wontakal et al., 2022) to regulate erythroblast cytokinesis. Recently, mutations of RACGAP1 and KIF23 were linked with congenital dyserythropoietic anemia III (CDAIII), a rare disorder characterized by multinucleated erythroblasts that are reminiscent of cells undergoing endoreplication owing to cytokinesis failure (Romero-Cortadellas et al., 2021). Specifically, in individuals with CDAIII, variants of CYK4 (p.L396Q and p.P432S) and KIF23 (p.P916R) cause erythropoiesis-specific cytokinesis defects as a result of altered substrate specificities of the GAP activity of RACGAP1 and impaired centralspindlin clustering, which results in ineffective erythropoiesis. Further mechanistic studies of the centralspindlin complex and its cell type-specific downstream effectors in cytokinesis will provide insights into the common underlying basis for morphological and physiological defects in CDAIII and other disorders.

The C. elegans strains were maintained, cultured and crossed as described previously (Brenner, 1974). Most of the strains were grown at 20°C, with the exception of the temperature-sensitive mutant strains and their derivatives, which were grown at 16°C. Transgenic strains were derived from wild-type Bristol N2 worms. All the strains used in this study are listed in Table S3.

To construct PF55B11.3::GFP::H2B for labeling spermathecal cell nuclei, a 2 kb F55B11.3 promoter fragment was amplified from C. elegans genomic DNA using Phusion High-Fidelity DNA Polymerase (New England Biolabs) and cloned into pJM370 vector (gift from J. D. McGhee University of Calgary, Canada) to substitute the elt-2 promoter by Gibson assembly (New England Biolabs). The cDNA fragment corresponding to codons 1-290 of vab-10a, which encode the predicted actin-binding domain (ABD) of spectraplakin (Bosher et al., 2003), was cloned into plasmid pPD95.75 (gift from A. Fire, Addgene plasmid #1494) between the BamHI and KpnI sites upstream of GFP to make the ABD::GFP plasmid. PF55B11.3::ABD::GFP was generated by stitching the F55B11.3 promoter with ABD::GFP by fusion PCR (Hobert, 2002). The fragment encoding the PH domain was amplified from a plasmid containing mCherry-PH_PLC1∂1 (Klompstra et al., 2015), and cloned into plasmid pPD95.75 between the BamHI and KpnI sites upstream of GFP. Perm-1::PH::GFP was constructed by joining the erm-1 promoter, obtained by PCR from the Perm-1::erm-1::GFP::PP7::erm-1 3′UTR plasmid (Li et al., 2021), to the PH::GFP coding sequence through homologous recombination. Perm-1::CYK-4::Tomato plasmid was generated by SunyBiotech. DNAs were prepared from multiple independent isolates, then verified by restriction digestion and sequencing. A mixture was used for germline transformation of animals by microinjection. Constructs were normally injected at 50-100 ng ml⁻¹, along with rol-6 selection markers. The primer sequences are listed in Table S4. At least three independent transgenic lines were generated for each construct.

RNAi induction and feeding were performed as previously described (Timmons et al., 2001). HT115 (DE3) bacteria harboring the ‘empty’ KS+-based vector pL4440 (containing two T7 promoters flanking a poly linker) were used as a control for RNAi feeding experiments. Most of the RNAi experiments were carried out using a mild RNAi condition initiated from L1-stage larvae (Zhang et al., 2015). Briefly, eggs were isolated from gravid hermaphrodites by a standard bleaching protocol and allowed to hatch directly on the RNAi plates seeded with RNAi bacteria. The worms were observed and scored for phenotypes when they grew into L4 and/or adult stages (around 48-96 h after seeding eggs). Bacterial clones were picked from the Ahringer genome-wide RNAi feeding library (J. Ahringer, Wellcome Trust/Cancer Research UK Gurdon Institute, Cambridge, UK) and sequenced before using.

To find an RNAi dose that results in an intermediate cyk-4 phenotype that can be used as a baseline for detecting genetic interaction, knockdown levels were titrated and optimized by diluting the RNAi bacteria with empty pL4440 vector. For genetic interaction assays of cyk-4/rho-1 and cyk-4/cdc-42, in which two genes were simultaneously knocked down by RNAi, bacterial cultures of Escherichia coli expressing the appropriate dsRNA were mixed together with empty vector-containing bacteria or unc-22 RNAi bacteria in certain ratios seeded onto RNAi plates as described above. To examine the genetic interaction between CYK-4 and RHO-1, the final ratios were cyk-4:rho-1:empty vector (or unc-22)=25:50:25 for double RNAi, cyk-4:empty vector (or unc-22)=25:75 for cyk-4 single RNAi, and rho-1::empty vector (or unc-22)=50:50 for rho-1 single RNAi. To examine the genetic interaction between CYK-4 and CDC-42, the final ratios were cyk-4:cdc-42 =25:75 for double RNAi, cyk-4:empty vector (or unc-22)=25:75 for cyk-4 single RNAi and cdc-42::empty vector (or unc-22)=75:25 for cdc-42 single RNAi.

To count spermathecal nuclei, spermathecae of control and RNAi animals at late L4 stage were scanned using a confocal microscope. The GFP-positive nuclei in the projection images were counted. Thirty spermathecae were observed simultaneously and the experiments were repeated three times.

Young adults with no visible eggs in the uterus were singled and transferred to new RNAi plates every 24 h, and the number of eggs and larvae were counted (Praslicka and Gissendanner, 2015). This was repeated for 4 days until the parent worms stopped laying eggs. Each day, the progeny production was recorded and compared with the wild-type controls. Thirty worms were observed simultaneously and the experiments were repeated three times.

Hermaphrodites carrying AJM-1::GFP were collected after 48 h of growth on bacteria producing cyk-4 dsRNA or control bacteria and subjected to DAPI staining. Worms in M9 buffer containing 0.25 mM levamisole were dissected using a pair of 25 G syringe needles to achieve gonad extrusion (Gervaise and Arur, 2016; Strome, 1986). Extruded gonads were fixed in 3.7% formaldehyde in PBS for 15-45 min at room temperature. For immunofluorescence staining, the samples were sequentially incubated in blocking buffer (PBS buffer containing 5% BSA and 0.5% Triton X-100) for 15 min, anti-AJM-1 primary antibody (1:20; MH27, Developmental Studies Hybridoma Bank, University of Iowa) for 1 h, and TRITC-conjugated goat anti-mouse IgG secondary antibody (1:100; A5278, Sigma-Aldrich) for 1 h. For DAPI staining, the fixed gonads were incubated in PBS containing 1 mg/ml DAPI for 30 min, as described for larvae (Li et al., 2021). After two short washes, the gonads were mounted in Antifade Prolong Gold mounting medium (Life Technologies Corporation) on poly-lysine-coated glass slides for imaging.

To evaluate the brood size and sterile phenotype, live animals were directly observed on plates under a Nikon SMZ800 microscope. To observe spermathecal morphology in detail, live worms were mounted on glass slides and anesthetized using 10 mM sodium azide (Sigma-Aldrich) and visualized using a Zeiss LSM 710 confocal microscope (Carl Zeiss). Single-plane images were taken as 6-20 sections along the z-axis at 0.2 µm intervals. z-stacks were collected and then merged into a single projection image with maximum intensity using Zen software (Carl Zeiss) for additional analysis. DAPI, GFP and Tomato/Cherry were visualized using 405 nm, 488 nm and 561 nm excitation laser lines, respectively. Multi-channel images were taken after adjusting individual channels to eliminate bleed-through and sequentially scanned. Images were taken at minimal laser settings unless indicated otherwise. Identical laser and confocal settings were used when comparing experimental animals with controls. Images were arranged using Adobe Photoshop with occasional small adjustments for contrast and brightness.

For time-lapse imaging, live worms were mounted on 3-5% (wt/vol) agarose pads with 0.1% tetramisole and sealed with Vaseline. Imaging was performed on an inverted spinning disk confocal microscope (Olympus SpinSR10) using a Yokogawa CSU W1 scanner system, equipped with a 60×/1.4 NA objective and two Hamamatsu ORCA Flash sCMOS cameras. All movies were acquired under the control of cellSens Dimension software (Olympus) and analyzed using ImageJ.

All data were analyzed using GraphPad Prism software (GraphPad Software) and are presented as mean+s.d. unless otherwise stated. Statistical analyses were performed using two-tailed Student's t-test for between-group comparisons. All data are representative of those of at least three independent experiments unless otherwise indicated. Statistical significance levels are presented as: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

We thank V. Gobel, R. Legouis and Y. Tse for strains and plasmids, and SunyBiotech for generating the CYK-4::Tomato plasmid used in this study. Some strains were provided by the CGC, which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440; University of Minnesota, USA). We also thank Isabel Hanson for critical reading of the manuscript, and the Bioimaging and Stem Cell Core in the Faculty of Health Sciences for their excellent technical support for the microscope systems.