Tissue-specific RNA-seq defines genes governing male tail tip morphogenesis in C. elegans

Cellular morphogenesis, a process that is essential for animal development, involves precise temporal, spatial and often sex-specific coordination of various cytological events, such as cell fusion, change in cell shape and migration. It has been studied mostly in embryonic processes such as tubulogenesis, gastrulation, neural crest migration and the epithelial-to-mesenchymal transition (EMT) (Debnath et al., 2022; Gheisari et al., 2020; Goldstein and Nance, 2020; Hashimoto and Munro, 2019; Iruela-Arispe and Beitel, 2013; Shaye and Soto, 2021; Szabó and Mayor, 2018). However, it is also important for postembryonic structures, such as the excretory system of C. elegans, which is used as a model for tubulogenesis (Shaye and Soto, 2021) and for sexual dimorphisms manifested during the juvenile-to-adult transition in animals (Kopp, 2011; Mason et al., 2008; Ohde et al., 2018). In C. elegans, sexually dimorphic morphogenesis forms the vulva and uterus of hermaphrodites (Gupta et al., 2012), and the male tail (Mason et al., 2008; Nelson et al., 2011; Nguyen et al., 1999). In nearly all cases, a major deficiency in our understanding is how the cell biological events of morphogenesis are regulated transcriptionally in the gene regulatory network (GRN).

To investigate the transcriptional regulation of sexually dimorphic morphogenesis, we focus here on the male tail of C. elegans. In larvae, the male tail is morphologically similar to that of adult hermaphrodites and has a pointed tip (Fig. 1). In adult males, however, the tail tip is short and round. The morphogenetic process, called male tail tip morphogenesis (TTM), leading to the short male tail tip happens during the 4th larval stage (L4) at the juvenile-to-adult transition (Kiontke et al., 2024). During this time, other male-specific structures form in the tail: a pair of spicules and nine pairs of sensory processes (rays) that are embedded in a cuticular fan.

Neurons in the tail also undergo male-specific morphogenesis during L4: the two PHC neurons in the tail tip differentiate into hub neurons that are required for male mating behavior (Serrano-Saiz et al., 2017). A second pair of male-specific neurons, PHDL and PHDR, develop de novo through transdifferentiation of the phasmid socket cells (Molina-García et al., 2020).

Previous research (Mason et al., 2008) has demonstrated that TTM is controlled by the DM-domain transcription factor (TF) DMD-3, which is a homolog of Drosophila Doublesex and mammalian Dmrt1. DMD-3 is expressed specifically in males, and is both necessary and sufficient for TTM. Loss-of-function (lf) mutations in the dmd-3 gene result in partial failure of TTM, and misexpression of DMD-3 in hermaphrodites leads to ectopic TTM. TTM, and indeed all male tail morphogenesis, fails completely in double loss-of-function mutants of dmd-3 and its paralog mab-3 (Mason et al., 2008). mab-3(lf) mutants have low-penetrance TTM phenotypes (Shen and Hodgkin, 1988), suggesting that mab-3 contributes to robustness of TTM but that dmd-3 has the major role as a ‘master regulator’ of the process. dmd-3 is also necessary and sufficient for the differentiation of the PHC neurons into male hub neurons (Serrano-Saiz et al., 2017).

A genome-wide RNAi screen for TTM defects found 210 genes that contribute to TTM (Nelson et al., 2011). Many of these genes play a regulatory role and belong to pathways that determine, for example, when (via the heterochronic pathway), where (via HOX TFs) and in which sex (via the sex determination pathway) TTM takes place. Other genes are involved in the processes that execute TTM, e.g. fusion, vesicular transport and rearrangement of the cytoskeleton. The current hypothesis for the gene network for TTM is that it has a bow-tie architecture with DMD-3 and MAB-3 at the core (Nelson et al., 2011). Regulatory pathways determine the expression of DMD-3, which then coordinates the cell biological processes during morphogenesis. How DMD-3 is linked to these cell biological processes is insufficiently known. In fact, how transcriptional regulators communicate with the cell machinery to drive morphogenesis is largely unknown for most systems (Bernadskaya and Christiaen, 2016; Debnath et al., 2022; Gildor et al., 2021; Kenny-Ganzert and Sherwood, 2024; Shaye and Soto, 2021; Yamakawa et al., 2023).

We know that morphogenesis involves universal and pleiotropically acting ‘cellular machines’ that are specialized for particular aspects of cell behavior (e.g. apicobasal polarity, actomyosin, endocytosis, cell-cell and cell-ECM adhesion) and comprise groups of interacting genes, which we here call ‘universal modules’. How the activity of these machines is integrated in a particular morphogenetic event is dependent on cues from the cellular environment, which are relayed by tissue-specific TFs. However, TFs are unlikely to affect the universal modules directly, but instead regulate components of signaling pathways, such as kinases and G-proteins or their GEF or GAP switches (Bernadskaya and Christiaen, 2016), which we refer to as ‘effectors’. In addition, we posit that morphogenesis employs other groups of proteins that act together in various contexts but are not universally present in every cell. For example, C. elegans vulva morphogenesis includes the dynamic expression of a collection of apical extracellular matrix proteins, some of which are also used in eggshell formation or in morphogenesis of the excretory duct (Cohen and Sundaram, 2020). We call such groups of proteins ‘conditional modules’. These conditional modules may also be regulated by effectors rather than by TFs. Under these assumptions, one model for the TTM system architecture could be quite simple, with few direct DMD-3 targets, all of which are effectors (Fig. 2A). This network model is similar to that for Ciona cardiac progenitor migration where the TFs Mesp and FoxF control cellular processes through receptors for signaling pathways, Rho GTPase and aPKC (Bernadskaya and Christiaen, 2016). An alternative model is that DMD-3 targets parallel cascades of TFs, which – likely also through effectors – regulate TTM (Fig. 2B). Vertebrate neural crest formation is an example of such a regulatory architecture (Bernadskaya and Christiaen, 2016). Other more-complex models are also possible.

In a first step towards distinguishing between these possible network models for TTM, we sought to identify genes that are specifically expressed in the tail tip during TTM. Our strategy was to compare the transcriptomes of tail tips that undergo TTM [wild-type males and dmd-3(gf) hermaphrodites ectopically expressing DMD-3] with the transcriptomes of tail tips that do not [wild-type hermaphrodites and dmd-3(lf) males]. The tail tip of L4 animals can easily be isolated using laser capture microdissection and the tissue processed for RNA-seq (Woronik et al., 2022). A differential expression (DE) analysis of the transcriptomes obtained from pooled tail tips of animals with and without TTM yielded 564 genes that are directly or indirectly controlled by DMD-3 and are candidates for being involved in TTM.

Using gene ontology enrichment analysis and information from the literature, we found groups of DE genes that are acting in TTM, are potentially co-regulated, and constitute universal and conditional modules in TTM. We find evidence that, for example, cytoskeleton genes, ABC transporters, the UPR (unfolded protein response) pathway, chondroitin proteoglycan synthesis and the hypertonic stress response are activated during TTM, and that cuticle synthesis or maintenance are repressed. There are also 40 TFs among the DE genes and several effectors, i.e. GTPases, kinases and phosphatases. We use this information to propose an updated model for the type of regulatory network that might act downstream of DMD-3 in TTM.

To determine which genes are important for morphogenesis of the male tail tip, we sequenced the transcriptome of laser micro-dissected tail tips of animals at the beginning of the L4 stage when DMD-3 is expressed but TTM had not started yet (Kiontke et al., 2024). We sampled pools of 230-400 tail tips from four different conditions: wild-type males and dmd-3(gf) hermaphrodites that undergo TTM (referred to as TTM+), and wild-type hermaphrodites and dmd-3(-) males that do not undergo complete TTM (referred to as TTM−). The CEL-Seq2 method (Hashimshony et al., 2016) was used to prepare libraries from extracted RNA of each pool. With this method, we obtained on average ∼653,000 UMI (unique molecular identifiers) per sample (Tables S1 and S2).

To conduct sample level quality control and to explore the strong patterns driving variation in our dataset, we performed a principal component analysis. Visualization of the principal component analysis revealed that all samples, except those from dmd-3(gf) hermaphrodites, clustered by tail tip phenotype (Fig. 3A). dmd-3(-) males and wild-type hermaphrodites that are TTM− cluster together, whereas samples of wild-type male tails, which undergo DMD-3 directed TTM, cluster separately. Wild-type male samples do not cluster as tightly as the dmd-3(-) male and wild-type hermaphrodite samples. This pattern likely arises because dynamic gene expression during TTM exacerbated slight variations in developmental timing across samples. The dmd-3(gf) hermaphrodite bioreplicates did not cluster together in the PCA plot, nor did they cluster with wild-type male samples, as might be expected for TTM+ tail tips. These large differences between dmd-3(gf) hermaphrodite samples might be due to heterochronic shifts in the development of these mutants: some animals initiate TTM precociously in L3 (Mason et al., 2008).

Next, we conducted a differential expression analysis on three comparisons in our dataset: wild-type males versus dmd-3(-) males, wild-type males versus wild-type hermaphrodites, and dmd-3(gf) hermaphrodites versus wild-type hermaphrodites. Using a cutoff for the adjusted P-value of 0.01, we found 760 genes were differentially expressed (DE) between wild-type and dmd-3(-) males, 846 genes were DE between wild-type males and hermaphrodites, and 202 genes were DE between dmd-3(gf) hermaphrodites and wild-type hermaphrodites. Like the PCA analysis, the DE analysis with dmd-3(gf) hermaphrodite samples raised concerns. The genes that were DE between dmd-3(gf) hermaphrodites and wild-type hermaphrodites exhibited little overlap with DE genes from other TTM+ versus TTM− comparisons. Of the 202 DE genes identified in the dmd-3(gf) hermaphrodites versus wild-type hermaphrodites comparison, only 24 were also DE in the dmd-3(-) males versus wild-type males comparison, and only 39 were DE in the wild-type males versus wild-type hermaphrodites analysis. In contrast, 564 DE genes were common between the comparisons of wild-type males versus dmd-3(-) males and wild-type males versus wild-type hermaphrodites. The overlap of all three comparisons was 18 genes. Taken together with the PCA results, we believe the dmd-3(gf) hermaphrodite samples do not capture the biological signal at a comparable stage of TTM, likely due to documented heterochronic shifts in these mutants (Mason et al., 2008). We therefore did not have confidence in our ability to draw biological conclusions from these samples and excluded them from downstream analyses.

A comparison of the log₂-fold change of the 564 candidate genes that overlap in the wild-type males versus dmd-3(-) males and the wild-type males versus wild-type hermaphrodites DE analyses showed that not only was the direction in which expression differed the same [e.g. genes have higher expression in wild-type males than in hermaphrodites and in dmd-3(-) males], but also the degree to which these genes were DE (Fig. 3C). The overlap of these gene sets is visualized in Fig. 3D. The 282 DE genes in male versus hermaphrodite comparison but not in the wild-type male versus dmd-3(-) male comparison (blue sector) are sex specific but not regulated by DMD-3, and thus not likely to be important for TTM. The 196 genes that are DE in wtM versus lfM but not DE in wtM versus wtH (pink sector) are interpreted to be DMD-3 targets but might not be important for TTM. Thus, we focused primarily on the 564 genes in the intersection (purple); we call these our high quality (HQ) candidates. Of the 564 HQ candidate genes, 202 showed higher expression in wild-type males (Table S3). Because DMD-3 is expressed only in wild-type male tail tips, we can conclude that these 202 genes are directly or indirectly activated by DMD-3. The remaining 362 genes are repressed by DMD-3.

To validate the candidate TTM genes obtained with our RNA-seq data, we made GFP reporters for the proximal promoter regions of a selection of DE genes (Table 1, Fig. 4). Reporters were built to include the region upstream of the gene between the transcription start site and the next gene. We also included the first intron, which often contains regulatory sites (Fuxman Bass et al., 2014). We focused on genes that were activated by DMD-3 and showed a large log₂-fold change between TTM+ and TTM− animals. We used three criteria to evaluate the expression of the reporters: (1) Is the reporter expressed in the male tail tip during TTM? (2) Is the reporter more brightly expressed in male tail tips – or in more cells – than in hermaphrodites? (3) When crossed into dmd-3(-) mutants, is the reporter less brightly expressed than in wild-type males? Of 49 reporters, 26 were expressed in male tail tips more strongly than in hermaphrodite tail tips. Seventeen of these reporters were investigated in the dmd-3(-) genetic background; 16 showed reduced tail tip expression in males lacking functional DMD-3. Ten additional reporters were expressed in tail tips but were equally bright in males and hermaphrodites. It is expected that more subtle differences in the regulation of the genes between sexes or conditions are not always captured with multi-copy extrachromosomal arrays. Thus, reporters for which we did not see a difference between sexes would not invalidate the DE results. In summary, for many genes tested, transcriptional reporter expression mirrored the RNA-seq results, providing confidence that our RNA-seq dataset reflects bona fide DMD-3-regulated genes that are involved in TTM.

As a second method for validation, we compared the 564 HQ candidate genes with the 210 genes that yielded a TTM phenotype in a previous genome-wide, post-embryonic RNAi feeding screen (Nelson et al., 2011). We found an overlap of only 11 genes (Table 2). RNAi of these genes led to a Lep phenotype (a protruding tail tip in adult males resulting from defective TTM). Seven are activated and four are repressed by DMD-3. Five DE genes were not evaluated by Nelson et al. (2011) because RNAi led to larval lethality. For one of these genes, lin-41, Del Rio Albrechtsen et al. (2006) found an over-retracted RNAi phenotype. Because the overlap of the two lists was small, we also examined the male tail phenotype of loss-of-function mutants for a subset of genes whose transcriptional reporters showed clear male-biased expression in the tail tip: fos-1, grl-26, chw-1, C06A6.4, crml-1, zip-5, bli-1, mrp-2, cdh-7, wrt-5, elp-1 and srw-85. None of these mutations caused TTM defects, consistent with the previous negative RNAi results.

The small overlap between RNAi-positive and DE genes is not surprising for several reasons. First, the RNAi screen identified genes that were both upstream and downstream of DMD-3, whereas our DE analysis is expected to only identify genes downstream of DMD-3. Second, we know of some genes that are important for TTM, but are likely regulated post-transcriptionally, e.g. cdc-42 and genes for components of the cytoskeleton, where activation or localization rather than changes in expression are required for TTM. Such genes may show an RNAi phenotype without being DE. Third, we expect a large amount of redundancy in pathways involved in morphogenesis (Sawyer et al., 2011; Wieschaus, 1997). For that reason, many genes that are DE may not show any phenotype when knocked down by RNAi. Fourth, our RNA-seq dataset only represents a single time point during TTM, whereas the RNAi screen identified genes important for TTM throughout development. In addition, a gene required for TTM may be expressed in a tissue adjacent to the tail tip and regulate TTM non-cell-autonomously, as was shown for egl-18 (Nelson et al., 2011). Fifth, RNAi is more likely to identify genes whose function is required for TTM than those that need to be turned off. Thus, many of the genes that are repressed by DMD-3 might not show an RNAi phenotype. In this context, it is surprising that RNAi of some of the repressed genes caused Lep tails. Here, one might expect to observe the opposite of failed TTM, i.e. the over-retracted phenotype. Genes that nevertheless show a Lep RNAi phenotype may be in a negative-feedback loop with DMD-3, as was demonstrated for nhr-25: NHR-25 is required for the onset of DMD-3 expression, but its expression is then repressed by DMD-3 (Nelson et al., 2011). Finally, not all DE genes in the tail tip need to function in TTM per se; their mutation or knockdown would thus not cause a TTM defect.

To investigate whether the tail tip transcriptome bears information about the structure of the TTM gene network, we first searched our list of HQ DE genes for TFs and effectors. TFs were identified through comparison with those listed by Ma et al. (2021), the database cisBP (Weirauch et al., 2014) and putative TFs described in WormBase (Davis et al., 2022). We found 40 genes likely involved in transcriptional control (Table 3), 12 of them activated and 28 repressed by DMD-3. Among the latter is nhr-25, for which repression by DMD-3 was previously shown (Nelson et al., 2011). Another known TTM TF is TLP-1 (Zhao et al., 2002). The gene encoding it is repressed by DMD-3, even though its mutant phenotype is Lep. It is possible that TLP-1, like NHR-25, acts in a negative-feedback loop with DMD-3 (Nelson et al., 2011).

To investigate possible connections between the TFs downstream of DMD-3, we turned to information about TF targets in the database TFlink (Liska et al., 2022) and the TF yeast-2-hybrid data by Reece-Hoyes et al. (2013) (Fig. S1). This analysis showed that many, but not all, of these TFs are targets of the well-studied TFs JUN-1 and FOS-1 (activated by DMD-3) and/or NHR-25 (repressed by DMD-3). NHR-25 is a DNA-binding scaffold protein that requires binding of other factors for specific gene regulatory capacity (Asahina et al., 2020 preprint). Most of the protein interactions among the TFs downstream of DMD-3 are with EGL-13.

As defined above, post-translational modifiers such as kinases, phosphatases and GTPases are referred to here as effectors. Several effector genes are DE in the tail tip (Table 3). Among the kinases repressed by DMD-3 is kin-20, one of the few DE genes that also shows an RNAi phenotype. drl-1 and wnk-1, which play a role in the hypertonic stress response (see below), as activated. Among the DE GTPases is chw-1, which has a male-specific expression pattern. CHW-1 is known from its involvement in polarity establishment during vulva development, where it boosts signaling through the WNT receptor LIN-17/Frizzled (Kidd et al., 2015).

By scrutinizing the list of DE genes further, searching for known connections between them, we looked for groups of genes that may act as universal modules, i.e. molecular machines present in every cell but activated or repressed during TTM, or genes that constitute conditional modules, i.e. that act together in different contexts, here in TTM. We also used a gene ontology enrichment (GO) analysis to assist with this search (WormBase enrichment tool, q value threshold=0.03). For GO enrichment analyses, we found that more informative results were obtained when we analyzed DMD-3-repressed and -activated genes separately (Table S4).

This approach identified several universal modules, some of which were previously proposed to be involved in TTM (Nelson et al., 2011), i.e. cytoskeleton and polarity, cell-matrix adhesion, vesicular trafficking, ABC-type transporters, cell fusion and unfolded protein response in the ER (Table S5). We also found conditional modules, i.e. DMD-3-repressed cuticle synthesis/maintenance, DMD-3-activated collagen synthesis, chondroitin proteoglycan synthesis and hypertonic stress response.

Eleven genes that are part of the cytoskeleton or its regulation are activated, and seven are repressed, by DMD-3. The enrichment analysis identified the GO term ‘spindle localization’ (GO:0051653) as over-represented. This term is associated with genes involved cell migration (hmr-1, ced-1 and ced-6) and par-3 involved in polarity establishment, processes known to be important for directing morphogenetic events such as TTM (Nance and Priess, 2002; Naturale et al., 2023; Vuong-Brender et al., 2016). Cell-matrix adhesion has been previously shown to be crucial for morphogenesis (Bernadskaya et al., 2019; Labouesse, 2012; Wu et al., 2023). In the tail tip, the laminin LAM-3 and the integrin INA-1, both associated with adhesion between cells and basement membrane, are activated by DMD-3. A strong connection of the tail tip cells to the basement membrane is maintained during TTM and may involve these proteins (Kiontke et al., 2024). HMR-1, a component of apical junctions, is also activated by DMD-3, although apical junctions between the tail tip cells dissociate during TTM. It is possible that hmr-1 is needed at the junction between the tail tip cells and the adjacent tail cells. In addition, hmr-1 may take on a non-junctional role of hmr-1 as in the zygote (Padmanabhan et al., 2017). Four genes related to vesicular trafficking were identified as important for TTM in the RNAi screen (Nelson et al., 2011); here, three different genes in this category are DE and activated by DMD-3. ‘ABC-type transporter activity’ was one of the GO terms enriched for the genes activated by DMD-3 (Table S4). abcx-1, which showed an RNAi phenotype (Nelson et al., 2011), is DE in the comparison of wtM versus lfM. The function of these transporters in TTM is unclear. Cell fusion is one of the processes involved in TTM (Kiontke et al., 2024; Mason et al., 2008; Nguyen et al., 1999). The fusogen EFF-1 has previously been shown to be controlled transcriptionally by DMD-3 (Mason et al., 2008). Our data confirm this finding. Communication across gap junctions before fusion was hypothesized to be important for morphogenesis (Hall, 2017). The gap junction genes inx-12 and inx-13 showed an RNAi TTM phenotype, and inx-13 is DE and activated in the comparison of wtM versus lfM. We thus posit that innexins and eff-1 act in cell fusion during TTM and together form a module.

The GO enrichment analysis for genes activated by DMD-3 identified another universal module involved in TTM. The most highly enriched GO terms are ‘IRE1-mediated unfolded protein response’ (GO:0036498) and ‘response to topologically incorrect protein’ (GO:0035966) (Table S4). These GO terms refer to the UPR (unfolded protein response), which senses and responds to an overabundance of unfolded proteins in the ER. Such proteins are found during normal development in tissues with high protein secretory load (Mitra and Ryoo, 2019). The UPR is activated in the anchor cell before invasion, where ribosome biogenesis is also upregulated to accommodate increased translation of transmembrane and secreted proteins (Costa et al., 2023). It is possible that the UPR is also activated by increased translation in the tail tip, although we do not find enrichment of ribosome biogenesis genes. However, the UPR also has other noncanonical roles in differentiation and morphogenesis (Hetz et al., 2020).

The most highly enriched GO categories for genes repressed by DMD-3 are ‘structural constituent of cuticle’ (GO:0042302) with 25 genes for collagens and the cuticlin CUT-2, ‘collagen trimer’ (GO:0005581) and ‘molting cycle’ (GO:0042303) (Table S4). Among the DE genes in the latter category are several for astacin-like proteases (also in the enriched GO category ‘metalloendopeptidase activity’ GO:0004222) and nine genes in the Hedgehog-like signaling pathway (ptc-3 and eight Ptr genes). In addition to these putative Hedgehog-like receptors, 13 putative ligands are DE and repressed by DMD-3 (Table S6). The Hedgehog pathway has been implicated in: vulva and male tail development (Zugasti et al., 2005); importantly, in molting (Lažetić and Fay, 2017); and in the organization of the precuticular ECM (Cohen et al., 2021; Serra et al., 2024). Thus, cuticle and molting genes along with their regulators form a conditional module that is repressed by DMD-3 in the male tail tip cells. These cells dissociate from the L4 cuticle early during TTM and are largely internalized by the time TTM is completed. Thus, they are unlikely to contribute to the maintenance of the L4 cuticle or the production of the adult body cuticle and the precuticlar ECM that precedes it. This may explain why transcription of collagens and other cuticle and molting genes is repressed in the male tail tip.

Network analysis suggests that several DMD-3-repressed genes with unknown function may also be part of this module responsible for cuticle maintenance or synthesis (Fig. S2, Table S7). These genes are among those with the greatest negative log2-fold change in our DE analyses. Some are highly expressed in TTM− tail tips but either not expressed in TTM+ tails, or expressed by an order of magnitude less. We used interaction information in WormBase to generate a network for all DMD-3-repressed genes and found interaction clusters for 117 (Fig. S2, Table S7). Most of the genes with unknown function cluster together with genes known to be involved in cuticle formation (cluster 1 in Fig. S2): the cuticulin cut-2 (one of the most highly expressed genes in the TTM− tail tips), the zinc carboxypeptidase suro-1 (Kim et al., 2011), and the two patched-related receptors ptr-4 and ptr-16 (Cohen et al., 2021). This cluster is linked to two other clusters that contain primarily collagens.

Given that most collagens that are DE in the tail tip are repressed by DMD-3, we turned our attention to the four collagens whose expression is activated by DMD-3: BLI-1, COL-20, COL-46 and COL-89. Several enzymes that act in collagen biogenesis (Winter et al., 2007) are also activated (the disulfide isomerases pdi-1, pdi-2, pdi-6 and dnj-27, and the catalytic alpha subunit of collagen prolyl 4-hydroxylase phy-2), indicating that collagen synthesis plays a role in TTM. To better understand the involvement of the activated collagens in TTM, we tagged endogenous BLI-1, COL-20 and COL-89 N-terminally with fluorescent proteins and studied their expression. Unexpectedly, BLI-1 and COL-20 were localized in the cytoplasm of the tail tip cells (Fig. 5 and Kiontke et al., 2024). Expression of BLI-1 in hyp8-11 and hyp13 starts immediately after the L3/L4 molt and continues into adulthood. Beginning at stage L4.4 (for stage assignment, see Kiontke et al., 2024), C-terminally tagged BLI-1::GFP is also expressed in the body epidermis of hermaphrodites and males, where it is later localized to the struts of the adult cuticle, as previously described (Johnson et al., 2023; Kiontke et al., 2024; Lints and Hall, 2005). The N-terminally tagged BLI-1 is expressed in the hypodermis in mid-L4, after which hypodermal expression disappears, probably because GFP is cleaved off at the RXXR furin cleavage site along with the N-terminal region of the protein (Johnson et al., 2023). Tail tip cell expression, however, is still seen in young adult males.

The COL-20 protein begins to be expressed in the seam of both sexes late during the L4/adult lethargus phase. In early adults, the GFP::COL-20 reporter is seen in hyp7 (Fig. 5D). This reporter is not expressed in the hermaphrodite tail tip cells or in the vulva. In young adult males, GFP::COL-20 is expressed in the cytoplasm of the ventral tail hyp (a syncytium that includes the tail tip cells; Kiontke et al., 2024). It is also seen in cells of the proctodeum, hyp7 and possibly hyp11.

GFP::COL-89 shows very bright expression in cells of the male proctodeum, which largely conceals the much fainter tail tip expression when observed with a wide-field microscope. Only one cell in the tail (likely R9.p) occasionally expresses COL-89 more brightly (Fig. 5E). In hermaphrodites, GFP::COL-89 is seen in cells of the proctodeum.

It appears that all three collagens are expressed in the tail tip cells of males but not hermaphrodites. As a caveat, we do not know where the mature COL-20 and COL-89 are located as our reporters were N-terminally tagged and the fluorescent protein was cleaved off during collagen processing. However, unprocessed BLI-1 and COL-20 remain in the cytoplasm of the tail tip cells throughout TTM into adulthood. The function of these collagen monomers in the tail tip cells remains unclear. Adult bli-1(RNAi) males do not show TTM defects, although the Bli phenotype is 100% penetrant. However, it should be noted that all three collagens (bli-1, col-20 and col-89) were found in RNAi screens (Chandler et al., 2023; Rohlfing et al., 2011) for genes that affect the expression of gpdh-1, a gene in the hypertonic stress response module (see below).

Several genes in the chondroitin synthesis pathway are upregulated in the tail tips of males undergoing TTM. Genes in this pathway were initially identified in a screen for mutations in vulva morphogenesis that resulted in the ‘Squashed vulva’ (Sqv) phenotype (Herman et al., 1999). Subsequently, the Sqv genes were shown to encode transporters and enzymes that are required for the production of chondroitin proteoglycans (CPGs), proteins to which a polysaccharide chain is attached (Berninsone, 2006). The chondroitin synthesis pathway comprises eight SQV proteins and the chondroitin polymerizing factor MIG-22 (Suzuki et al., 2006). Genes for SQV-4, SQV-5, SQV-7 and MIG-22 are DE in tail tips with higher expression in wild-type males. An additional nucleotide sugar transporter gene, nstp-4, is also DE, as well as the uncharacterized gene C29H12.2, with a putative sugar phosphate transporter domain. We found that transcriptional reporters for sqv-4, sqv-5 and sqv-7 are expressed in tail tips (Fig. 4, Fig. 6A). Except for sqv-4, the reporters were expressed in tails of males and hermaphrodites. RNAi-knockdowns of sqv-5 and mig-22 were shown to cause Lep tail tip phenotypes (Nelson et al., 2011), highlighting the importance of this pathway for TTM.

To investigate whether knockdown or mutation of other Sqv genes would also show a tail tip phenotype, we performed RNAi by feeding or injection and observed male tails of mutants. We found that mutation or RNAi of sqv-2, sqv-4, sqv-5 and sqv-7 led to similar defects in the morphology of the adult male tail (Fig. 6C) with a slightly unretracted tail tip (and defects in further anterior tail structures). We then tagged endogenous SQV-4, SQV-5, SQV-7, MIG-22 and NSTP-4 with fluorescent proteins (Fig. 6D). SQV-4::mKate is expressed cytoplasmically as expected (Berninsone, 2006). All other proteins are seen in puncta, consistent with their localization to the Golgi apparatus (Fig. 6D). In males, SQV-4::mKate expression begins after the L3/L4 molt in all tail tip cells and hyp13, and shows a marked increase as TTM progresses (Fig. S3). SQV-4::mKate expression is faint in hermaphrodite tail tips and does not increase during L4. When crossed into the dmd-3(-) background, brightening of the reporter is delayed (Fig. S3). SQV-5::GFP expression is seen in male tail tips in bright puncta (Fig. 6D). Expression is less bright in hermaphrodite and dmd-3(-) tails, and the puncta are smaller (Fig. S3). SQV-7::GFP is expressed in the seam in both sexes and in males in the tail tip and the Rn.p cells (Fig. 6D), with increasing brightness as development progresses through L4. A GFP::NSTP-4 reporter is expressed in the seam and hyp 7 in both sexes. In males, it is seen in the tail tips and in other cells in the developing tails, where it is concentrated at the membranes between the cells in puncta and diffuse in the cytoplasm (Fig. 6D). MIG-22::GFP is seen in hyp7 and the tail tip of males and hermaphrodites. However, in males, expression in the tail tip is brighter than in other tissues. The presence of these proteins in the tail tip is consistent with a role in TTM.

Next, we tried to find evidence for a chondroitin core protein in the tail tip. Using mass spectrometry, Noborn et al. (2018) identified many CPGs; currently 23 are known and three more are predicted. We detected mRNA for 12 of these proteins in tail tips. fbn-1 and the predicted CPG B0365.9 are DE in the comparison of wild-type males versus wild-type hermaphrodites, and another predicted CPG, F52E1.5, is on our list of HQ DE genes. A translational reporter for fbn-1 is expressed in the tail of L4 males, at first only in the proctodeum and the area around the anus, in the extracellular space between tail tissue and L4 cuticle. Beginning with stage L4.4, FBN-1::mCherry is also seen in the extracellular space behind the tail tip (Fig. 6D), where expression gets brighter as TTM continues until late during L4, when it becomes restricted to the area of the developing fan. FBN-1 also plays a role in vulva development, where this protein is laid down around a central core, which is proposed to also contain CPGs (Cohen et al., 2020). FBN-1 appears in the vulva beginning at stage L4.2 with peak expression during stages L4.5 and L4.6. Given that FBN-1::mCherry expression appeared relatively late during TTM, it is likely that other CPGs are being secreted at an earlier stage, similar to those forming the central core in the developing vulva (Cohen et al., 2020).

When C. elegans encounters high salinity, it elicits the HTSP, which leads to the intracellular accumulation of the osmolyte glycerol. Glycerol synthesis is upregulated by transcriptional activation of the glycerol-3-phosphate-dehydrogenase GPDH-1 (Urso and Lamitina, 2021), in part through the kinases DRL-1 and WNK-1. gpdh-1, drl-1 and wnk-1 are DE and activated by DMD-3 in the tail tip, suggesting that here, too, glycerol synthesis is upregulated. In HTSP, gpdh-1 expression is also affected by signaling through the patched-related transmembrane protein PTR-23 (Igual Gil et al., 2017; Rohlfing et al., 2011); RNAi against its mRNA leads to male tail morphogenesis and molting defects (Zugasti et al., 2005). In the tail tip, DMD-3 represses ptr-23.

Previous RNAi screens for genes affecting the expression of gpdh-1 transcriptional reporters (Chandler et al., 2023; Rohlfing et al., 2011) found several genes that are also DE in the tail tip: atx-2, bbln-1, bli-1, col-20, col-89, crt-1, F01G10.9, ndg-4, srw-85 (all activated by DMD-3), col-77, ell-1, F01G10.10, fath-1, gdh-1, H03E18.1, hum-5, moe-3, pgph-3 and ptr-4 (repressed by DMD-3). In addition, 18 of the genes with a gpdh1 RNAi phenotype were positive in the RNAi screen for TTM defects (Nelson et al., 2011). Among these is the gene for the tail tip-expressed WNT LIN-44 and the gene for the GEF VAV-1, which was previously shown to be regulated by DMD-3 (Nelson et al., 2011). Finally, two collagens, dpy-7 and dpy-9, which are both involved in HTSR signaling (Dodd et al., 2018), also showed a TTM RNAi phenotype. This overlap of RNAi-positives for HTSR and TTM also suggests an overlap of the function of these genes. Another DE gene activated by DMD-3 and involved in the HTSP is that for the aquaporin AQP-8. Aquaporins are channel-forming proteins that allow the passage of water and small molecules through cell membranes. aqp-8 mRNA levels increase approximately eightfold during hypertonic stress (Huang et al., 2007). Finally, lea-1, a gene that is upregulated during desiccation stress (Hibshman and Goldstein, 2021) is DE in the comparison of wtM versus lfM. Taking all of this together, we hypothesize that the HTSP genes aqp-8 and possibly lea-1 constitute a module in TTM.

We propose that the module for chondroitin proteoglycan synthesis and the HTSR module function together in TTM (Fig. 7). For vulva development, it was proposed that the hygroscopic chondroitin sulfate draws water into the vulva lumen, thereby expanding it and creating pressure that prevents the vulva from collapsing (Hwang et al., 2003). It is now clear that CPGs act with other aECM proteins in a complex luminal scaffold to shape the vulva lumen during morphogenesis. However, the initial lumen appears to be created by CPGs (Cohen et al., 2020). We propose that, likewise, in TTM, the extracellular space is filled with a hygroscopic matrix as the tail tip tissue dissociates from the L4 cuticle and rounds up. We further hypothesize that CPGs draw water from the tail tip tissue into the extracellular space, leading to hypertonic stress in the tail tip cells. To counteract this stress, gpdh-1 transcription and osmolyte production are activated. Consistent with this model, the water channel protein AQP-8 is also activated by DMD-3. In HTSR, aqp-8 transcription and translation are elevated upon exposure to high osmolarity in the environment (Huang et al., 2007; Igual Gil et al., 2017). Testing this model and evaluating how genes in other modules function in TTM and how they are regulated by DMD-3 will be a subject of future research.

Through tail-tip-specific RNA-seq, we have identified 564 genes that are differentially expressed in wild-type male tail tips that express DMD-3 and undergo TTM versus tail tips that do not express DMD-3 and do not undergo TTM. We verified selected candidates by male tail-specific transcriptional reporter expression.

We found support for the involvement of several universal processes in TTM that we categorized as universal modules: genes acting in cytoskeleton and polarity establishment, cell-matrix adhesion, vesicular trafficking, ABC transporters, and the unfolded protein response. In addition, we identified conditional modules that play a role in this process: downregulation of cuticle maintenance, production of male tail tip-specific collagens, chondroitin proteoglycan synthesis and the hypertonic stress response.

We also found 40 TFs DE in the tail tip during TTM, as well as several effectors (GTPases, their regulators, kinases and phosphatases). Our data allow us to distinguish between the two network models for the TTM GRN proposed above (Fig. 2). The large number of DE TFs suggest that transcriptional control plays a bigger role in TTM than, for example in Ciona, cardiac progenitor cell migration. At the same time, the TTM transcriptome also contains genes for effectors. Therefore, our current working hypothesis for the TTM network model is one where DMD-3 regulates different modules through a variety of TFs in addition to regulating effectors that facilitate the appropriate activity of universal pleiotropic cellular machines (Fig. 8). This is still a very preliminary hypothesis, which does not identify which regulators are directly downstream of DMD-3 and whether there are TF cascades upstream of the modules. Determining the direct targets of DMD-3 is a subject of future research. Likewise, our RNA-seq analysis captured only a snapshot of the dynamic process of TTM at one time point in early L4. How TTM genes are regulated dynamically in the tail tip will be investigated in the future.

C. elegans strains were grown on a lawn of E. coli (OP50-1) according to standard methods (Lewis and Fleming, 1995; Stiernagle, 2006).

To obtain synchronized populations of larvae, two methods were used: L1 arrest and ‘hatch-off’. For L1 arrest (Lewis and Fleming, 1995), gravid hermaphrodites were treated with alkaline hypochlorite to release embryos. Embryos were then washed with M9 buffer and left shaking in M9 buffer at 25°C for 18-24 h. Arrested L1 larvae were placed on plates seeded with E. coli and allowed to develop at 25°C until the desired developmental time. For hatch-off, the method described by Woronik et al. (2022) was used. Briefly, gravid hermaphrodites were allowed to lay eggs overnight. Hermaphrodites and all hatched larvae were then removed by washes with M9 buffer. The remaining embryos were incubated at 25°C. After 1 h, all L1 that had hatched during this time were collected in M9, placed on food and allowed to develop to the L3 stage (∼20-22 h post hatch at 25°C), at which time males could be distinguished from hermaphrodites and picked onto separate plates. The strains used are listed in Table S8.

This experiment was performed with four types of samples: tail tips of L4 wild-type hermaphrodites (N2), wild-type males (CB4088), dmd-3(-) males (UR156) and dmd-3(gf) hermaphrodites (UR257).

Worm development was synchronized by L1 arrest. Gravid CB4088 hermaphrodites were exposed to an alkaline hypochlorite solution (Porta-de-la-Riva et al., 2012) to release embryos. Embryos were then washed with M9 buffer and left shaking in M9 buffer at 25°C for 18-24 h. Arrested L1s were placed on plates seeded with E. coli and allowed to develop at 25°C for 22 h, after which time L3 males could be distinguished from hermaphrodites. Males were picked onto a new plate under a dissection microscope at room temperature. To collect hermaphrodites, the N2 strain was used, but L4 animals were still picked to account for imperfect synchronization. After 2 h of picking, worms were collected into a tube with M9 buffer and washed twice with ice-cold M9 buffer supplemented with a small amount of detergent (Tween20). After washing, the worms were fixed in ice-cold freshly made 70% methanol, washed twice in 70% methanol and stored in this solution for at least 1 h or, at most, 24 h until further processing.

Drops of worms fixed in 70% methanol were pipetted onto a Leica PEN membrane slide (76×26 mm P/N 11505158). The slide was then placed on a slide warmer at 50°C to evaporate the methanol. Once dried, the slide was mounted onto a Leica laser dissection microscope. Costar Thermowell tubes (0.5 ml with flat cap, 6530) were loaded into the collection holder and 65 µl of lysis buffer (Ambion RNAqueous-Micro Kit, ThermoFisher Scientific, AM1931) was pipetted into the tube cap. Using the 40× objective, 50-100 tail tips were dissected into each cap. Before unloading the tube from the collection holder, an additional 35 µl of lysis buffer was added to the tube cap. Tubes were spun down, flash frozen in liquid nitrogen or on dry ice and stored at −80°C until enough tissue was collected for one RNA extraction (240-440 tail tips). RNA was extracted using the Ambion RNAqueous-Micro kit following the manufacturer's protocol and eluted twice into 8 µl elution solution. The quality and quantity of the extracted RNA was evaluated with Agilent high sensitivity RNA screen tapes (5067-5579).

To obtain enough DNA for sequencing, the single cell RNA-Seq method CEL-Seq2 (Hashimshony et al., 2016) was used for library construction. Instead of single cells, purified RNA was used as starting material for the protocol, which was otherwise followed as published. For the CEL-Seq2 method, RNA of each sample is indexed at the first reverse transcription step and UMIs are introduced. Subsequently, several samples are combined into the same library. Here, two libraries were made with the same sample RNAs at two starting concentrations: 300 pg/sample and 118 pg/sample. Each library contained four replicate samples for dmd-3(-) males, three for wild-type males and five for wild-type hermaphrodites (Table 1). The CEL-Seq2 method includes a linear and a PCR amplification step, and preserves strandedness: briefly, RNA is reverse transcribed with oligo-dT primers. After second strand synthesis, the DNA is in vitro transcribed and thereby amplified. The amplified RNA is quantified, fragmented and reverse transcribed with random hexamers. cDNA that contains adapter sequences introduced during the first cDNA synthesis step are subsequently PCR amplified (11 cycles). During this step, Illumina sequencing adapters are added. This method results in 3′ reads that can be easily mapped back to the C. elegans genome. Non-polyadenylated RNAs and splice variants are not captured with this method. Libraries were sequenced on two lanes of an Illumina HiSeq 2500 in Rapid Mode (2×50, 15×36 bp) by the Genomics Core at New York University's Center for Genomics and Systems Biology.

Raw reads were adapter trimmed using CutAdapt (v1.12) (Kechin et al., 2017). Fastqc (v. 0.11.8) (https://qubeshub.org/resources/fastqc) was used to evaluate read quality. Trimmed reads were mapped to the C. elegans genome (WB235) using Bowtie2 (v. 2.3.2) (Langmead and Salzberg, 2012) via the CelSeq2 analysis pipeline (v. 0.2.5; FASTQ_QUAL_MIN_OF_BC: 10, CUT_LENGTH: 35, ALN_QUAL_MIN: 0) (Hashimshony et al., 2016). The read counts matrix generated by the CelSeq2 pipeline was input for an analysis in DESeq2 (v. 1.26.0) (Love et al., 2014). The dmd-3(gf) hermaphrodite A replicate was removed from the matrix because it had too few read counts (the B and C replicates had 20- and 30-fold greater read counts than A). Genes with no mapped reads were removed from the matrix, then differential expression analyses (alpha=0.01, P adjusted<0.01) were conducted for wild-type versus dmd-3(-) males, wild-type males versus wild-type hermaphrodites, dmd-3(-) males versus wild-type hermaphrodites and wild-type hermaphrodites versus dmd-3(gf) hermaphrodites. We then identified genes that were differentially expressed in TTM(+) versus TTM(−) comparisons (i.e. wild-type versus dmd-3(-) males, wild-type males versus wild-type hermaphrodites, and dmd-3(gf) hermaphrodites versus wild-type hermaphrodites), but not in the TTM(−) versus TTM(−) comparison [i.e. dmd-3(-) males versus wild-type hermaphrodites] to generate a list of high-quality candidate genes involved in TTM.

To gain biological insights for our high-quality candidates, we performed enrichment analyses using the Enrichment Analysis Tool (q value threshold=0.1) on Wormbase (v. WS290) (Angeles-Albores et al., 2016, 2018). The Enrichment Analysis Tool uses a hypergeometric statistical model to identify gene, anatomy and worm phenotype ontology terms that are over-represented in a list of genes.

Transcriptional reporters were made either by overlap extension PCR (Nelson and Fitch, 2011) or by Gateway cloning, restriction enzyme cloning or Gibson assembly (Gibson et al., 2009). Generally, a region extending from the gene upstream of the transcription start site of a candidate gene plus the first exon and intron of that gene was fused to the sequence of GFP in plasmid pPD95.75 (Fire Vector kit; Addgene, #1000000001), which also contains the unc-54 3′UTR. For PCR, Q5 high fidelity DNA polymerase (NEB), PrimeStar Max (Takara) or PrimeStar GXL (Takara) were used. The PCR products were purified with the Promega Wizard SV Gel and PCR clean-up system or the Qiaquick PCR purification kit (Qiagen). For restriction enzyme cloning, the regulatory region of each gene was amplified and ligated into SphI/AgeI- or SphI/KpnI-digested pPD95.75. For Gateway cloning, the regulatory region of sqv-5 was amplified, inserted into a pENTRY vector and transferred into pENTRY-SrfI-mCherry using the Multisite Gateway System (ThermoScientific). For Gibson Assembly, the regulatory region of each gene was inserted into SmaI-digested pUC19 together with GFP amplified from pPD95.75 using NEBuilder HiFi DNA Assembly Master Mix (NEB - E2621). PCR products or plasmids were micro-injected into CB4088 [him-5(e1490)] hermaphrodites with myo-2::mCherry from plasmid pCFJ90 (Addgene, #44472) as co-injection marker, or into EM574 [pha-1(e2123); him-5(e1490)] with plasmid pBX as co-injection marker, which rescues the temperature-sensitive lethal phenotype of the pha-1(e2123) mutation. Microinjections were performed as previously described (Evans, 2006). Several lines were maintained and examined for GFP expression in males and hermaphrodites with the emphasis on expression in the tail tip. Some reporters were then crossed into strains containing the dmd-3(-) allele. To evaluate the expression pattern of these reporters, we examined and imaged between 10 and over 50 animals per gene and sex. Reagents for making transcriptional reporters are provided in Tables S9 and S10.

For some candidates, we made endogenously tagged translational reporters via CRISPR Cas9 genome editing. To identify PAM sites near the 3′ or 5′ end of a gene, we used the algorithms implemented in Benchling (https://www.benchling.com). If a suitable PAM site was within 0-3 nucleotides of the insertion site (before translation start or stop), we made homology repair templates (HRTs) with 35 nucleotide long homology arms. For PAM sites further away from the insertion site, HRTs were designed with 200 nucleotide homology arms. Oligonucleotides with these overhangs and a region complementary to the sequence of GFPnovo2 (in plasmid pSM; Addgene, #116943) or mKate (in plasmid pCFJ350; Addgene, #34866) were used to PCR amplify the HRT (Takara PrimeStar Max). After purification (Promega Wizard SV Gel and PCR clean-up system), the concentration of the HRT was adjusted to 100 ng/µl as measured with Nanodrop (ThermoFisher Scientific). Guide RNAs were purchased from IDT either as sgRNA or as crRNA, and resuspended in IDT Duplex buffer to a concentration of 100 µM or 200 µM, respectively. crRNAs were annealed to the IDT tracerRNA (200 µM) at 95°C for 5 min. The concentration of the guide RNA was adjusted with water to 62 µM. 0.5 µl of IDT Cas9 protein was mixed with 0.5 µl of guide RNA and incubated for 5 min at room temperature. The HRT was heated to 95°C for 5 min and immediately placed on ice. An injection mix was made by adding 1 µl RNP complex 1 µl 1×TE buffer (pH 8), 2 µl HRT and 6 µl water. The mixture was loaded into needles and injected into both gonads of 10 young hermaphrodites. No coCRISPR reagents were used. 16-32 L4 stage progeny of the injected hermaphrodites were picked onto individual plates and kept at 25°C until they had laid several eggs. Then, each F1 hermaphrodite was picked into 10 µl proteinase K lysis buffer (Fay and Bender, 2006), frozen in liquid nitrogen, thawed and refrozen three times, and incubated for 90 min at 60°C and 15 min at 95°C. This worm lysis was used for PCR with primers flanking the insertion site. L4 hermaphrodite progeny of worms with edits were picked onto individual plates and allowed to lay eggs. Lysis and PCR were performed, and this process repeated until a homozygous animal was identified. Reagents for introducing fluorescent reporters and mutations with CRISPR Cas9 genome editing are provided in Table S11.

To introduce the him-5(e1490) mutation into some strains, we turned to CRISPR Cas9 genome editing, which was performed as described above with a HRT that was ordered as ssDNA from IDT. Individual F1 hermaphrodites were isolated as L4 and allowed to reproduce. From plates that contained males, L4 hermaphrodites were picked to individual plates to homozygose the mutation. No PCR test for an edit was performed.

To create a dmd-3(lf) mutation, we deleted the entire coding region of the gene with two CRISPR Cas9 cuts and repaired it with a HRT that bridged the gap. Homozygous mutants were isolated as for him-5 above by observing the Lep phenotype in the F1 progeny. Reagents for introducing fluorescent reporters and mutations with CRISPR Cas9 genome editing are provided in Table S11.

For microscopy, animals were either picked from mixed staged plates or from bleach synchronized populations. Worms were paralyzed in 20 mM sodium azide and mounted on a pad of 4% Noble Agar in 50% M9 buffer supplemented with 20 mM sodium azide. They were examined with a Zeiss AxioImager equipped with Colibri LED illumination and an Apotome. Generally, image stacks were recorded with 0.5 µm distance between slices. If the fluorescent signal was too dim for Apotome imaging, the deconvolution algorithm implemented in the Zeiss ZenBlue software v. 2.6 was used (‘good/fast iterative’ setting) to improve image quality. Some DIC and fluorescent images were generated using a Nikon Ti-S compound inverted microscope equipped with a Hamamatsu C11440-10C Flash 2.8 camera. The NIS Elements Version 4.2 software was used for image acquisition, and digital images were processed using the Adobe Photoshop Elements software.

RNA interference was performed by injection or feeding. For sqv-4(RNAi) by injection, the sqv-4-coding sequence (+23 to +1397) with flanking T7 promoters was amplified from C. elegans RNA using SuperScript One-Step RT-PCR with Platinum Taq (Invitrogen). This PCR product was used as a transcription template to generate in vitro synthesized double-stranded RNA using the MEGAscript RNAi Kit (Ambion). Purified double-stranded RNA (100 ng/μl) was injected into the gonads of rrf-3(pk1426) II; qIs56 [lag-2::GFP] him-5(e1490) V adult hermaphrodites and F₁ adult males were examined for tail tip phenotypes. sqv-5 and par-3 RNAi feeding was carried out as previously described (Kamath et al., 2001) using the pPD129.36 nhr-67 RNAi construct from the Ahringer Library (Geneservice). Briefly, HT115(DE3) bacterial cells were freshly transformed with either the pPD129.36 sqv-5 or par-3 RNAi plasmid. Plasmid-containing bacteria were induced to express double-stranded RNA using IPTG and plated on M9 lactose plates. rrf-3(pk1426) II; qIs56 [lag-2::GFP] him-5(e1490) V worms were bleach synchronized, and L1-arrested worms were plated onto RNAi-expressing bacteria. The tail phenotype of adult males was assessed ∼3 days later. To generate males in strains with wild-type alleles of him-5, worms were grown on bacteria expressing double stranded RNA for him-8/klp-16 using the pLT 651 plasmid (Addgene, 59998) (Timmons et al., 2014).

Interaction data of repressed genes was mined from WormBase (WS290) using their REST API (http://rest.wormbase.org/rest/field/gene/{WORMBASE_GENE_ID}/interactions). We did with respect to discriminate with respect to the type of interaction, i.e. predicted, physical, regulatory or genetic. Interactions between DMD-3 repressed genes and non-repressed genes were removed. As the final network is undirected, interactions that differed only in their directionality were collapsed into one. The AutoAnnotate app (v. 1.4.1) (Kucera et al., 2016) in Cytoscape was used to calculate clusters using the MCL clustering algorithm (Van Dongen, 2008).

We thank the ∼50 undergraduate Siena College and six NYU students who helped generate transcriptional reporters for DMD-3 regulated genes. We specifically thank independent research students Liam Peterson, Chelsea LeBlanc, George Bushey, Athena Bennani and Lillian Gardner, and the students in Adam Mason's 2020, 2022 and 2023 Developmental Biology Course. We further thank Shehana Gunasekara for help with the CELSeq2 protocol and Raza Mahmood for help with data analysis. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440).