Effectors of anterior morphogenesis in C. elegans embryos

The development of embryos requires the tight coordination of cell proliferation, cell differentiation, cell movements, and tissue morphogenesis. Whereas many of the genes and mechanisms underlying proliferation, differentiation, and migration have been characterized in considerable detail, the morphogenesis of tissues, organs, and whole organisms is less well understood. This is likely due to the complexity of the morphogenetic process, constraints associated with genetic approaches including genetic redundancy, challenges with carrying out in vivo imaging and manipulations, and technical hurdles or limitations inherent to ex vivo studies.

The nematode Caenorhabditis elegans is a useful model in which to study various aspects of development including the morphogenesis of embryos and organs. Whereas certain aspects of C. elegans embryonic morphogenesis – such as gastrulation (Goldstein and Nance, 2020), embryonic elongation (Vuong-Brender et al., 2016), and dorsoventral enclosure (Chisholm and Hardin, 2005; Chisholm and Hsiao, 2012) – have been reasonably well described, the development and morphogenesis of the anterior embryo, in particular the epidermis and its integration with the oral cavity, have just begun to be characterized in detail (Grimbert et al., 2021; Portereiko and Mango, 2001).

Work from our lab and others has shown that the C. elegans SYM (synthetic lethal with mec-8) proteins, SYM-3 and SYM-4, act with the MEC (mechanosensory abnormality) protein, MEC-8, to prevent deformation of the anterior epidermis and internalization of the buccal cavity (mouth) in response to pharyngeal pulling forces during embryogenesis (Davies et al., 1999; Kelley et al., 2015; Yochem et al., 2004). Specifically, mec-8; sym-3 and mec-8; sym-4 double mutants display a pharynx ingressed (Pin) phenotype and consequent larval lethality due to an inability to feed. In contrast, mec-8, sym-3, and sym-4 single mutants are largely viable, as are sym-3 sym-4 double mutants. These genetic results suggest that MEC-8 may act in a pathway or process that functions in parallel to the SYMs, whereas SYM-3 and SYM-4 may function together in a common pathway or complex.

Additional findings indicate that MEC-8, an ortholog of the mammalian RBPMS2 splicing factor (Lundquist et al., 1996), controls the alternative splicing of multiple RNA targets including fbn-1 (Kelley et al., 2015; Spike et al., 2002), a protein related to vertebrate fibrillins (Frand et al., 2005). Moreover, misregulation of fbn-1 is a key factor in causing the Pin phenotype of mec-8; sym double mutants (Kelley et al., 2015). FBN-1 functions as a structural component of the embryonic sheath – a specialized embryonic pre-cuticle – and in the larval cuticle, in which it is required for normal molting (Cohen et al., 2020b; Frand et al., 2005; Katz et al., 2022; Kelley et al., 2015). Both the sheath and cuticle are a category of apical extracellular matrix (aECM) that is secreted in large part by the epidermis (Page and Johnstone, 2007). Whereas the larval and adult cuticles serve as an exoskeleton and protect worms from the environment (Johnstone, 1994; Page and Johnstone, 2007), the sheath functions as a semi-rigid scaffold to prevent the embryonic epidermis from deformation by mechanical forces, including inward-pulling forces exerted by the elongating pharynx (Kelley et al., 2015; Priess and Hirsh, 1986; Vuong-Brender et al., 2017). Abnormal pharyngeal ingression has also been observed in dex-1 and dyf-7 mutants (Cohen and Sundaram, 2020), which, like fbn-1, encode components of the aECM (Heiman and Shaham, 2009). Moreover, additional roles for the aECM in various aspects of morphogenesis and cell shape control have come to light, emphasizing the importance of this relatively understudied extracellular component of developing tissues (Cohen and Sundaram, 2020; Labouesse, 2012; Lamkin and Heiman, 2017; Li Zheng et al., 2020; Sundaram and Cohen, 2017).

Whereas the roles of MEC-8 family members in splicing are well established, the molecular and cellular functions of SYM-3 and SYM-4 are less known. SYM-4 is a predicted β-propeller protein with seven WD-repeats, suggesting a role in coordinating protein interactions. The mammalian ortholog of SYM-4, WDR44/RAB11BP/Rabphilin-11, was identified independently by two groups as a binding partner and candidate effector of RAB11 (Mammoto et al., 1999; Zeng et al., 1999), a highly conserved GTPase involved in several aspects of cell trafficking and cell biology (Ferro et al., 2021; Guichard et al., 2014; Welz et al., 2014; Zhang et al., 2021). Although more-recent studies have expanded these initial findings (Chung et al., 2014; Lucken-Ardjomande Hasler et al., 2020; Thibodeau et al., 2023; Walia et al., 2019), the full range of WDR44 functions largely remains to be elucidated. Even less is known about SYM-3, the ortholog of FAM102A/EEIG1. FAM102A family proteins contain an N-terminal C2 (NT-C2) domain, a motif involved in lipid binding that is found in membrane-associated proteins including the conserved trafficking factors EHBP1 and Synaptotagmin-1 (Lynch et al., 2007; Shi et al., 2010; Zhang and Aravind, 2010). Whereas the specific molecular and cellular functions of FAM102A family proteins are unknown, FAM102A mutations have been implicated in the progression of glaucoma (Khor et al., 2016; Li et al., 2020; Nongpiur et al., 2018; Shi et al., 2021; Zhuang et al., 2018).

Here we provide evidence in support of the model that SYM-3/FAM102A and SYM-4/WDR44 family members act together in a pathway or complex and that these proteins may carry out functions connected to intracellular trafficking. Furthermore, using a candidate-based enhancer screening approach, we have implicated ∼30 additional proteins in aECM biology in the worm embryo. We also test the model that SYM-3 and SYM-4 promote the deposition of aECM proteins during embryogenesis.

Given the previously described genetic interactions between sym-3, sym-4, and mec-8 (Davies et al., 1999; Kelley et al., 2015; Yochem et al., 2004), we hypothesized that SYM-3 and SYM-4 may function together in a common pathway or complex or at a common subcellular location. To test this last prediction, we obtained strains expressing endogenously CRISPR-tagged SYM-3::GFP, SYM-3::mScarlet, SYM-4::GFP, and SYM-4::mScarlet and examined their expression and subcellular localization patterns. Both SYM-3 and SYM-4 were detected at low levels throughout development, with their most prominent expression observed at punctate structures along the basolateral boundary of the hyp7 epidermal syncytium and epidermal seam cells in larvae (Fig. 1A,B,E,F). Additional puncta were also detected within the cytoplasm of hyp7 and the seam cells, although these were less abundant. Notably, both color combinations of tagged markers revealed clear colocalization of SYM-3 and SYM-4 in the larval epidermis, consistent with these proteins acting together in cells (Fig. 1 and File S1).

Because SYM-3 and SYM-4 play a role in anterior morphogenesis (Davies et al., 1999; Kelley et al., 2015; Yochem et al., 2004), we examined these markers in early comma-stage embryos. Although expressed at low levels, SYM-3 and SYM-4 were detected at punctate structures throughout the embryo and, like their expression in larvae, were observed to overlap in their localization (Fig. S1 and File S1). As a complement to CRISPR-tagged markers, we examined the expression of sym-3::GFP and sym-4::GFP fosmid-based transgenes (Sarov et al., 2012) that were expressed from extrachromosomal arrays (Fig. S1). Both transgenes rescued the Pin phenotype of corresponding mec-8; sym-3 and mec-8; sym-4 mutants, indicating that the protein fusions are functional. Although more variable than the CRISPR reporters, analysis of the transgenes suggested the potentially broad expression of both proteins throughout development including localization to membranes, the cytoplasm, and punctate structures (Fig. S1 and File S1).

To identify candidate protein partners of SYM-3 and SYM-4 family proteins, we carried out affinity capture proteomics using C-terminally tagged FAM102A::GFP and WDR44::GFP “bait” proteins expressed in HeLa cells (Fig. 2A). Expression of both fusions was verified through western blotting and immunofluorescence, which suggested a punctate pattern in the cytoplasm of HeLa cells (Fig. 2A). Two independent affinity capture proteomics experiments were carried out for each bait protein in parallel with a GFP-only control (Files S2 and S3). To narrow down the list of potential interactors, we further filtered candidates by requiring a minimum of three spectral counts (SCs) for each candidate protein as well as a ratio of bait SCs to control SCs of ≥5 (File S2). Filtered results are summarized in Fig. 2B. In total we identified 333 proteins in one or both FAM102A pull-downs and 253 proteins in one or both WDR44 pull-downs.

FAM102A was the most abundant protein identified in both FAM102A replicates (201 and 259 sequence SCs), whereas WDR44 was the most abundant protein identified in both WDR44 replicates (208 and 457 sequence SCs), indicating that the GFP pulldowns worked as expected. Moreover, FAM102A was the second-most-abundant protein identified in one of the two WDR44 replicates (19 bait SCs versus 0 control SCs), and, although enriched in the other replicate (10 bait SCs versus 5 control SCs), it failed to make the ratio cut-off for our filtered list. Correspondingly, WDR44 was identified in one of the two FAM102A replicates (5 bait SCs versus 0 control SCs). These results are consistent with both the genetic and colocalization data from C. elegans and suggest that FAM102A and WDR44 family proteins may function together in a conserved protein complex. We note that we were unable to validate physical interactions between SYM-3 and SYM-4 CRISPR-tagged proteins from worm lysates using immunoprecipitation–western blotting approaches, which may be due to low levels of the proteins or because our worm lysis protocols disrupted SYM-3–SYM-4 interactions.

As indicated by the Venn diagram (Fig. 2B), our analysis identified proteins that were common to multiple datasets including six proteins present in both WDR44 datasets and 25 proteins present in both FAM102A datasets (File S2). Notably, RAB11A and RAB11B GTPases were identified in both datasets for WDR44 (Fig. 2C), consistent with findings in the literature demonstrating that these proteins physically interact and function together in membrane trafficking (Mammoto et al., 1999; Thibodeau et al., 2023; Zeng et al., 1999).

Interestingly, several proteins identified in the FAM102A pulldowns are also connected to membrane trafficking including the Endophilin-A1 family members SH3GL1 and SH3GL2 (Fig. 2C). Endophilins contain membrane-binding/bending BAR domains along with SH3 domains and have been implicated in several trafficking functions including clathrin-independent endocytosis (Casamento and Boucrot, 2020; Kjaerulff et al., 2011). In addition, interactions between FAM102A and two members of the retromer complex, VPS35 and VPS26, were also detected, as was an interaction with the retromer-associated protein SNX2 (Griffin et al., 2005; Gullapalli et al., 2004) (Fig. 2C). Retromer is a conserved complex implicated in several endocytic processes including endosome-to-Golgi retrieval and endosome-to-plasma-membrane recycling (Seaman, 2012; Seaman, 2021; Tu and Seaman, 2021). We also note a detected interaction between FAM102A and WDR5, which is connected to the retromer complex through protein interaction data (Fig. 2C). Collectively, these physical interactions suggest that WDR44 and FAM102A may perform functions connected to membrane trafficking.

Given a potential role for SYM-3 and SYM-4 in the membrane trafficking of proteins, together with prior genetic interaction data (Davies et al., 1999; Kelley et al., 2015), we hypothesized that loss of either or both proteins might alter the abundance and/or localization of the matrix component FBN-1 to the aECM during embryogenesis. We obtained an endogenously tagged FBN-1::sfGFP based on a previous design (Cohen et al., 2020b) with the fluorescent marker inserted after H2408 (fbn-1a exon 21), just N-terminal to the extracellular ZP domain (C2436–P2671). We first detected expression of FBN-1::sfGFP in wild-type embryos just prior to the onset of morphogenesis with weak localization observed at the embryonic apical surface. Expression of aECM-associated FBN-1::sfGFP increased during early stages of embryonic morphogenesis, with expression continuing for the remainder of embryogenesis (Fig. S2; Fig. 3A, and Movies 1–3).

A preliminary visual inspection of FBN-1::sfGFP in sym-3 and sym-4 single mutants failed to reveal obvious differences between the mutants and wild type during morphogenesis (Fig. 3A and Fig. S2B). To screen for more subtle effects, we quantified expression in early comma-stage embryos, ∼370 min post-fertilization (at 22°C) (Chisholm and Hardin, 2005), a time point that just precedes attachment of the primordial pharynx to the epidermis (Portereiko and Mango, 2001). With respect to whole-embryo expression levels, we failed to detect a significant difference in the mean intensity of FBN-1::sfGFP in wild-type, sym-3, and sym-4 embryos (Fig. 3A,B). In addition, we examined the level of apically localized FBN-1::sfGFP in the anterior region of embryos (referred to as anterior traces), where the aECM is subsequently required to withstand pharyngeal pulling forces. Here we observed a very slight decrease (∼10–15%) in the levels of FBN-1::sfGFP in sym mutants, although this trend was at most marginally significant (Fig. 3A,B).

To test for more subtle differences, we averaged the anterior traces for wild-type, sym-3, and sym-4 embryos, then quantified FBN-1::sfGFP levels in the region corresponding to the future buccal opening, as well as in the adjacent dorsal and ventral regions (Fig. S3A). This analysis indicated a ∼25% decrease in the mean FBN-1::sfGFP signal in the nascent buccal region (P<0.01) in sym-3 and sym-4 mutants. In contrast, no significant differences were detected in the adjacent dorsal and ventral regions. Although tempting to speculate that the effect on the central region could contribute to the Pin phenotype, the wide variation observed between individual embryos of each genotype along with their substantial overlap in values (Fig. S3A), strongly suggests that this averaged trend cannot by itself account for the phenotypic effects of sym loss on the Pin phenotype (also see Discussion).

As an additional test we examined FBN-1::sfGFP in sym-3 sym-4 double mutants. Again, no difference in mean intensities was observed for whole embryos, although we unexpectedly detected a slight increase (∼1.25-fold) in the anterior apical FBN-1::sfGFP signal in sym-3 sym-4 embryos (Fig. 3A,B). This latter finding was consistent with a trend towards slightly increased FBN-1::sfGFP within each subregion of the sym-3 sym-4 anterior traces (Fig. S3B). These small differences, if meaningful, would appear to run counter to the model that SYM-3 and SYM-4 promote FBN-1 trafficking and secretion to the aECM. As an additional test, we asked whether the distribution of apically localized FBN-1::sfGFP along the anterior aECM was generally more variable in sym mutants by determining the coefficient of variation for each anterior trace. By this measure, sym-3, sym-4, and sym-3 sym-4 mutants showed at most a modest increase (∼1.1-fold) in variability relative to wild type (Fig. 3E). Taken together, these results fail to support a clear role for SYM-3 or SYM-4 in the gross deposition of FBN-1 to the aECM during embryogenesis.

We next considered the possibility that SYM-3 and SYM-4 may act on components of the aECM other than FBN-1. For this purpose, we chose to test NOAH-1, a ZP domain–containing sheath protein that acts in a mechano-transducing and structural pathway that functions in parallel to the FBN-1 (Vuong-Brender et al., 2017). Similar to the findings for FBN-1::sfGFP, however, we failed to detect substantial differences between wild-type, sym-3, sym-4, and sym-3 sym-4 embryos, although apical anterior traces of NOAH-1::mCherry suggested a slight trend toward increased variability in the mutants (∼1.1- to 1.2-fold; Fig. 3C,D). In addition, our analysis of averaged anterior traces by subregion failed to detect any differences between the four genotypes (Fig. S3C). Whereas it remains possible that SYM-3 and SYM-4 have more substantial roles in the deposition of other aECM components, our current data do not support an obvious function for SYM-3 or SYM-4 in the deposition of aECM proteins to the embryonic sheath.

We previously showed that the RNA-binding protein MEC-8/RBPMS2 regulates the splicing of fbn-1 isoforms (Kelley et al., 2015). Specifically, mutations in mec-8 lead to the pronounced retention of sequences corresponding to introns 14–16 of fbn-1 (Fig. 4A). In testing for potential effects of mec-8 loss on FBN-1::sfGFP expression, we were surprised to find that FBN-1::sfGFP abundance and localization to the aECM were not strongly affected in mec-8 mutants (Fig. 4B,C). This result was unexpected in part because intron retention often leads to the introduction of premature stop codons, leading to the degradation of mRNAs by nonsense-mediated decay and a corresponding reduction in protein levels (Longman et al., 2008; Lykke-Andersen and Jensen, 2015). If anything, however, we observed a slight increase (∼1.25-fold) in the total abundance of FBN-1::sfGFP as well as a very slight increase (∼1.15-fold) in the level of FBN-1::sfGFP along the anterior perimeter of early comma-stage embryos (Fig. 4B,C). Given that the Pin phenotype of mec-8; sym-4 mutants can be largely rescued by the expression of a single fbn-1 cDNA isoform (fbn-1e) (Kelley et al., 2015), this suggests that the underlying defect of FBN-1 in mec-8 mutants is due to altered functional properties of the protein rather than a reduction in total levels of the protein or defects in its targeting to the aECM.

The above result may be partially explained by the absence of any in-frame stop codons in sequences corresponding to fbn-1 introns 14–16 (Fig. 4D; WormBase) (Stein et al., 2001). Moreover, RNA sequencing (RNAseq) data available on WormBase indicate that introns 14–16, although present at relatively low levels in wild-type worms, are retained at higher frequencies than fbn-1 intronic sequences that are not regulated by MEC-8 (Fig. S4A; WormBase). Together, these findings suggest that introns 14–16 may be more accurately categorized as alternate exons, which are retained to some extent even within wild-type (MEC-8 competent) worms. Consistent with this, fbn-1 was not identified as a target of endogenous nonsense mediated decay in C. elegans (Muir et al., 2018), suggesting that these elongated fbn-1 species may serve some physiological function.

As a follow up, we were interested to know if sequences corresponding to fbn-1 introns 14–16 are conserved. In examining the peptide sequences encoded by these regions, we failed to identify any known motifs, although one region within intron 16 is predicted to encode a 190-aa peptide segment associated with amidase domain–containing proteins (7.59e-05; based on an NCBI–CD Search). BLAST searches for related peptide sequences in other Caenorhabditis species failed to detect significant similarities. Furthermore, predicted gene structures of fbn-1 orthologs in other Caenorhabditis species did not indicate strong conservation of exon–intron boundaries (WormBase). Nevertheless, we did find one example of a potential alternative exon corresponding to intron 14 of Caenorhabditis briggsae fbn-1. This region consists of a 1095-bp predicted intronic sequence that, based on the available data, contains no in-frame stop codons and forms a continuous open reading frame with the adjacent upstream and downstream exons (Fig. S4B,C). These observations suggest that alternative exons in fbn-1 homologs – and possibly other genes – may have escaped previous detection due to gene prediction models and/or low representation of these sequences within wild-type mRNA pools.

Recent studies have implicated multiple proteins that contribute to the functions and mechanical properties of the aECM during development (Cohen et al., 2020a; Cohen et al., 2019; Cohen et al., 2020b; Cohen and Sundaram, 2020; Forman-Rubinsky et al., 2017; Heiman and Shaham, 2009; Kelley et al., 2015; Li Zheng et al., 2020; Vuong-Brender et al., 2017). These include FBN-1, DEX-1, and DYF-7, all three of which function in the resistance of the anterior epidermis to pharyngeal pulling forces (Cohen et al., 2019; Kelley et al., 2015). To identify novel aECM components and regulators, we carried out a directed RNAi screen focusing on gene categories often associated with aECM proteins (Cohen and Sundaram, 2020). In total we screened 425 genes whose products were indicated to be secreted (Suh and Hutter, 2012) or glycosylated (Fan et al., 2005; Kaji et al., 2007; Olsen and Mann, 2013) or that oscillate during larval molting cycles (Hendriks et al., 2014; Meeuse et al., 2020), with most tested genes matching at least two of these criteria (File S4). Specifically, we tested for enhancement of the larval-lethal Pin phenotype in lin-35; sym-4 mutants by RNAi-feeding methods, in which the lin-35 mutation served to increase the sensitivity to somatic RNAi (Lehner et al., 2006; Wang et al., 2005). For reference, whereas 8–11% of lin-35; sym-4 mutants grown in the presence of the control RNAi vector displayed the Pin phenotype, fbn-1(RNAi) led to expression of the Pin phenotype in ∼97% of larvae (Fig. 5 and File S4), consistent with our previous findings (Kelley et al., 2015).

Of the 425 genes tested, we identified 32 that reproducibly increased the Pin phenotype of lin-35; sym-4 mutants to levels that were at least 2-fold higher than that of the vector control (Fig. 5A and File S4). Not surprisingly, we identified a number of genes encoding proteins with known or proposed functions in aECM biology including sheath and/or cuticle structural components such as NID-1/nidogen 1 (Kang and Kramer, 2000); the ZP domain-containing protein CUT-3 (Sapio et al., 2005); and the cuticulin-like proteins CUTL-1, −8, −14, −18, and −23, about which little is known. In addition, the screen identified candidate aECM modifiers including the peroxidase collagen cross-linker MLT-7 (Thein et al., 2009). Consistent with a role for MLT-7 in the aECM, a fosmid-based MLT-7::GFP reporter localized to the larval cuticle (Fig. 5B). Other candidate aECM regulators included the receptor tyrosine kinase ROL-3 (Jones et al., 2013); the neprilysin-family peptidase NEP-1 (Spanier et al., 2005), an ortholog of the human endoplasmic reticulum peptidase F49B2.6/ERAP1; and a Patched-related protein, PTR-23, implicated in molting and membrane integrity (Choi et al., 2016; Zugasti et al., 2005). In addition, the screen identified several uncharacterized proteins as well as proteins with conserved motifs but with no previous connection to aECM biology.

We note that of the 32 RNAi clones that tested positive in our assay for Pin enhancement, only two had been previously shown to affect larval development in RNAi-feeding screens using RNAi-hypersensitive strains harboring either the lin-35 or rrf-3 mutations (File S4) (Ceron et al., 2007; Simmer et al., 2003). As such, the lin-35; sym-4 background may be useful for identifying and testing additional candidate aECM regulators and components.

As an alternative means of identifying genes involved in anterior morphogenesis and force resistance, we examined the literature for evidence of previously reported Pin-like phenotypes. Notably, deletion of all six members of the mir-51 family of microRNAs (miRNAs; mir-51–56) leads to abnormal pharyngeal ingression, although this phenotype differs from that of other aECM mutants in its morphology and because of the presence of detached cell-like bodies (referred to hereafter as ‘cell bodies’) proximal to the nascent buccal cavity (Fig. 6A–C) (Shaw et al., 2010).

To determine if the Pin phenotype of mir-51–56 mutants is due to an inability to withstand pharyngeal pulling forces, we generated mir-51–56 mutant strains in which pharyngeal attachment was prevented by loss of pha-1. Approximately 85% of pha-1 null mutants fail to establish a connection between the pharynx and adjacent anterior cells of the forming buccal cavity (Fay et al., 2004; Kuzmanov et al., 2014), and loss of this connection strongly suppresses pharyngeal ingression defects in mec-8; sym-3 double mutants (Kelley et al., 2015). Notably, whereas 21 of 22 mir-51–56 mutant embryos exhibited the Pin phenotype, only 5 of 38 pha-1; mir-51–56 mutants in which the pharynx failed to attach (the Pun phenotype) showed anterior ingression defects (Fig. 6E). Moreover, restoring a pharyngeal connection in pha-1; mir-51–56 mutants by expression of a wild-type pha-1⁺ transgene (fdEx182) (Granato et al., 1994) led to ingression defects in 25 of 25 embryos (Fig. 6F).

To further quantify these effects, we measured ingression depths in pha-1; mir-51–56 embryos in which pharyngeal attachment occurred (fdEx182⁺– pha-1⁺) versus those in which it had failed (fdEx182^–– pha-1^–). Whereas 1.5- to 2-fold-stage embryos in which attachment failed had an average ingression depth of 0.31 µm, sibling pharynx-attached embryos had an average ingression depth of 4.1 µm, which increased to 5.3 µm at the 2- to 3-fold stage (Fig. 6G). Our data are consistent with pharyngeal pulling forces being an underlying cause of the Pin phenotype in mir-51–56 mutants. Moreover, the finding that mir-51 family members regulate cadherins (Shaw et al., 2010) suggests that cell–cell adhesion may contribute to the resistance of pharyngeal pulling forces by anterior epithelial cells.

We were also interested to determine if pharyngeal pulling might contribute to the cell-body detachment phenotype of mir-51–56 mutants. In this case we failed to observe any difference between the percentage of embryos containing detached cell bodies in pha-1; mir-51–56 embryos (98%; n=50) versus pha-1; mir-51–56; fdEx182 embryos that express the pha-1⁺ transgene (97%; n=79). In addition, cell detachment could be observed in some early comma-stage mir-51–56 embryos, a time point that precedes pharyngeal attachment (Fig. 6D). Thus, whereas the ingression phenotype of mir-51–56 mutants can be attributed to pharyngeal pulling forces, the cell detachment phenotype appears to occur independently of pharyngeal pulling. Collectively, these data implicate miRNAs and their targets in the regulation of anterior morphogenesis and the resistance to pharyngeal pulling forces.

Genetic, in vivo expression, and proteomic data are consistent with the model that SYM-3/FAM102A and SYM-4/WDR44 function together in a shared pathway or complex. This includes previous genetic interaction data in C. elegans (Davies et al., 1999; Kelley et al., 2015; Yochem et al., 2004) along with new data showing that SYM-3 and SYM-4 colocalize to cytoplasmic and membrane-associated puncta (Fig. 1 and Fig. S1). Although the identity of these puncta are unknown, they are reminiscent of vesicles or compartments associated with membrane trafficking. In addition, our proteomic data in mammalian cells (Fig. 2; Files S2 and S3) are consistent with WDR44 and FAM102A family proteins functioning together in a complex and potentially acting with additional trafficking components such as RAB11, endophilins, and the retromer complex. Notably, retromer has been implicated in apical exocytosis in conjunction with retrograde trafficking, including the export of aECM proteins (Chou et al., 2016; Dong et al., 2013; Dong et al., 2014).

WDR44 was first identified as a binding partner and putative effector of RAB11 (Mammoto et al., 1999; Zeng et al., 1999). These studies implicated WDR44 in transferrin recycling and cell migration, consistent with the functions of RAB11 (Bhuin and Roy, 2015; Chung et al., 2014; Gibieza and Petrikaite, 2021; Ramel et al., 2013; Welz et al., 2014). WDR44 associates specifically with the activated GTP-bound form of RAB11 and is colocalized with RAB11 at perinuclear regions (Mammoto et al., 1999; Zeng et al., 1999). Subsequent work has shown that WDR44 and RAB11 also function together in the control of cilium growth and that their interaction is controlled by Akt phosphorylation of WDR44 (Goldenring, 2019; Walia et al., 2019). Although these studies may suggest a potential role for SYM-4 in conjunction with C. elegans RAB-11 proteins, a recent study characterizing the peptide domains responsible for WDR44–RAB11 binding identified a key region within WDR44 that is not conserved in SYM-4 (Thibodeau et al., 2023).

WDR44 also binds Sec13 (Mammoto et al., 2000), a component of COPII vesicles required for budding from the endoplasmic reticulum (Gillon et al., 2012; Lord et al., 2013; Roberg et al., 1997). More recently, WDR44 was found to promote the export of newly synthesized E-cadherin to the plasma membrane in mammalian cells through its interactions with GRAF1b/2 (Lucken-Ardjomande Hasler et al., 2020), raising the possibility that SYM-4 could contribute to anterior morphogenesis through the regulation of cadherins. Notably, loss of mir-51 family members leads to cadherin misregulation along with associated defects in anterior morphogenesis (Shaw et al., 2010), including the Pin phenotype (Fig. 6).

Relatively little is known about the molecular functions of mammalian FAM102A/EE1G1, although protein interaction data from our work and other high-throughput studies suggest possible roles in trafficking (Fig. 2; Files S2 and S3). SYM-3/FAM102A contains an NT-C2 membrane localization domain that is found in several trafficking proteins (Shi et al., 2010; Wang et al., 2016; Zhang and Aravind, 2010). Moreover, a study in worms and flies implicated sym-3 and its Drosophila melanogaster ortholog, CG8671, along with several endocytic regulators, in the uptake of dsRNA molecules during RNA silencing (Saleh et al., 2006). Finally, recent reports have implicated FAM102A variants in glaucoma (Khor et al., 2016; Li et al., 2020; Nongpiur et al., 2018; Shi et al., 2021; Zhuang et al., 2018), an eye disorder linked to altered ECM components and abnormal ECM histology (Hernandez and Ye, 1993; Nikhalashree et al., 2019; O'Callaghan et al., 2017; Tribble et al., 2018; Vranka et al., 2015; Wiggs, 2015). Thus, FAM102A family members may carry out conserved functions in trafficking and ECM biology.

Despite the above inferences, our results do not support the straightforward model that SYM-3 and SYM-4 promote the trafficking or deposition of sheath-associated aECM proteins during embryogenesis (Fig. 3; Figs S2, and S3). Namely, loss of sym-3 and/or sym-4 had at most a modest effect on the apical localization of FBN-1 and no effect on the localization of NOAH-1 in embryos. Nevertheless, we acknowledge that the SYMs could play more prominent roles in the deposition of other aECM proteins, such as those implicated by our RNAi screen. Moreover, the SYMs could potentially exert small effects on the deposition of multiple aECM proteins, which may collectively weaken the aECM and its ability to withstand mechanical forces. Lastly, it remains possible that the SYMs are specifically required for the trafficking of FBN-1 isoforms that contain peptide sequences normally removed by MEC-8 (i.e. coding sequences within annotated ‘introns’ 14–16). In this scenario, the SYMs may become required for FBN-1 trafficking in mec-8 mutants, which would constitute a form of induced essentiality (Tischler et al., 2008). At present, further studies in C. elegans and other systems will be necessary to obtain a clearer picture of the cellular and developmental functions of both FAM102A and WDR44 family proteins.

Nevertheless, our studies were successful in identifying new regulators of anterior morphogenesis (Fig. 5 and File S4), many of which likely affect the structure and functions of the embryonic aECM. In addition, we obtained somewhat unexpected results regarding the role of the conserved MEC-8/RBPMS splicing factor in the regulation of FBN-1; loss of MEC-8 likely leads to the increased incorporation of previously unannotated alternative exons (Fig. 4) (Kelley et al., 2015). Collectively, our findings help to lay the groundwork for future investigations into roles and regulation of the apical ECM during development.

Localization studies were carried out using strains PHX1529, PHX4755, PHX1532, PHX4831, PHX1726, which were designed by DSF and generated by SUNY Biotech with fluorescence tag insertions in sym-3 and sym-4 occurring before the stop codon; sfGFP was inserted after H2408 (fbn-1a exon 21). Colocalization studies were carried out using strains WY1972 and WY1973. ML2482 (noah-1::mCherry) was generated as described (Vuong-Brender et al., 2017). sym-3::GFP(fdEx244) [WY1012], sym-4::GFP(fdEx245) [WY1026], and mlt-7::GFP(fdEx310) [WY1313] extrachromosomal array lines were generated by injecting recombineered fosmids containing the sym-3 and sym-4 genomic loci with GFP fused to the C termini (Sarov et al., 2012), together with pRF6 (rol-6gf) (Mello et al., 1991; Sarov et al., 2012). For complete genotypes see File S5.

Confocal fluorescence images in Figs 1A–C, E–G, 3A,C, 4B; Figs S1A–C and S1F, and S2 and Movies 1–3 were obtained using an Olympus IX83 inverted microscope coupled with an Olympus SPINSR W1 SoRa Spinning Disk Confocal microscope. A 100×, 1.35 N.A. silicone oil objective was used to obtain the z-stack images. cellSens Dimension v3.1 built-in software (Olympus) was used for image acquisition with appropriate filters (GFP and RFP). Images in Figs 5B and 6A–F were obtained using a Cannon compound microscope (100×) with DIC and fluorescence capabilities. Images were pseudocolored for presentation.

A Wiener deconvolution algorithm in the cellSens Dimension v3.1 software was used to deconvolute the raw z-stack images. The required z-plane was extracted from both raw z-stack and deconvoluted z-stack images. The deconvoluted images were then filtered using the Gaussian filter option (radius=10 pixels), and the resulting filtered images were then subtracted using the image calculator function from the original deconvoluted image. Subsequently, filter-subtracted images were subjected to a thresholding function to obtain thresholded binary images to be used as masks. The thresholded binary masks were then combined with the background-subtracted raw images (rolling ball radius algorithm; radius=50 pixels) using the ‘AND’ Boolean operation. Finally, the polygon tool was used to identify the region of interest (e.g. seam cell region) in the combined images. The Coloc2 plugin was used to calculate the colocalization between both green and red channels based on the Mander’s coefficient. Apart from the deconvolution, the entire colocalization analysis was performed using NIH Fiji software (Schindelin et al., 2012).

The anti-GFP nanobody was purified as described (Topalidou et al., 2016). Briefly, the bacterial expression vector pLaG16 was transformed in Arctic Express (DE3) cells (Agilent). Bacteria were then induced with 0.1 mM IPTG for 16 h at 8°C and centrifuged at 5000 g for 10 min at 4°C. Cells were incubated for 1 h in TES buffer (0.2 M Tris-HCl, pH 8; 0.2 mM EDTA; 0.5 M sucrose) and spun down for 10 min at 6000 g at 4°C, and the pellet was resuspended in TE buffer (0.2 M Tris-HCl, pH 8; 0.5 mM EDTA). The resuspended cells were then incubated at 4°C for 45 min and centrifuged at 30,000 g at 4°C for 30 min. The supernatant (periplasmic fraction) was bound to PerfectPro Ni-NTA Agarose affinity resin (5 Prime Sciences) for 1 h at 4°C. The resin was then washed once with wash buffer I (20 mM sodium phosphate, pH 8.0; 0.9 M NaCl) and twice with wash buffer II (20 mM sodium phosphate, pH 8.0; 150 mM NaCl; 10 mM imidazole). Elution was carried out using His elution buffer (20 mM sodium phosphate, pH 8.0; 150 mM NaCl; 250 mM imidazole), and the eluent was dialyzed with PBS. Recombinant anti-GFP nanobody was conjugated to epoxy-activated magnetic Dynabeads M270 (Life Technologies) using 10 μg recombinant protein/1 mg Dynabeads. Conjugations were carried out for 18 h in 0.1 M sodium phosphate, pH 8.0 and 1 M ammonium sulfate at 30°C. Beads were washed once with 100 mM glycine, pH 2.5; once with 10 mM Tris, pH 8.8; four times with PBS; and twice with PBS plus 0.5% Triton X-100. Beads were stored at −20°C in PBS with 50% glycerol.

HeLa cells (American Type Culture Collection) were grown at 5% CO2 at 37°C in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). Cells were passaged every 2–3 days when confluent. Cells were transfected when needed with 15 μg plasmid using Lipofectamine 2000 (Thermo Fisher Scientific).

2×10⁵ HeLa cells per well were seeded onto UV sterilized cover slips placed in 12-well cell culture plates. 24–48 h later, cells were rinsed three times with ice-cold PBS and fixed with 4% paraformaldehyde in PBS for 20 min at room temperature. After fixation, cells were rinsed twice with PBS, permeabilized with 0.5% Triton X-100 in PBS for 5 min and washed again twice with PBS. Cells were then placed in 5% milk in PBS for 1 h, followed by staining with mouse monoclonal anti-GFP (1:300, Santa Cruz Biotechnology, #sc-9996) in 0.5% milk in PBS for 1 h at room temperature. The cells were then washed with PBS three times (5 min each) and incubated with Alexa Fluor 488 anti-mouse secondary antibody (1:1000, Jackson ImmunoResearch #115-545-146) at 4°C overnight. The cells were washed with PBS three times for 5 min each and examined by fluorescence microscopy (Nikon 80i wide-field compound microscope).

For protein extraction and immunoblotting cells were lysed as described below (Mass spectrometry). 40 µl of the lysed supernatant was resuspended in 6x Laemmli loading buffer. Samples were resolved on 10% SDS-polyacrylamide gels and blotted onto nitrocellulose membranes. Membranes were blocked using 3% milk in TBST (50 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Tween 20) for 1 h at room temperature and stained with mouse monoclonal anti-GFP (1:2000, Santa Cruz Biotechnology, #sc-9996) in 3% milk in TBST at 4°C overnight, followed by three 5-min washes in TBST. Membranes were then stained with Alexa Fluor 680–conjugated goat anti-mouse antibody (1:20,000; Jackson Laboratory; #115-625-166) in 3% milk in TBST at room temperature, followed by three 5-min washes in TBST. Images were developed using a LI-COR processor.

The WDR44::GFP construct was constructed by amplifying human WDR44 cDNA from a HeLa cDNA library using primers: 5′–gactcagatctcgagctcaagcttcgatggcgtcggaaagc–3′ and 5′–ctcaccatggtggcgaccggtggatcagatacattttttcttttattaacaaacac–3′. The FAM102A::GFP construct was constructed by amplifying human FAM102A cDNA from a HeLa cDNA library using primers: 5′–gactcagatctcgagctcaagcttcgatggctttcttgatgaagaaga–3′ and 5′–ctcaccatggtggcgaccggtggatcatggctttcaatcacaactgg–3′. The PCR products were inserted in pEGFP-N1 vectors digested with EcoRI/BamHI using Gibson cloning.

For immunoprecipitation, 4×10⁶ HeLa cells were plated onto 10-cm Petri dishes. Twenty-four hours later, cells were transfected with either FAM102A::GFP, WDR44::GFP, or GFP. After 24 h, cells were washed with PBS and harvested in 400 μL lysis buffer (50 mM Tris, pH 7.5; 150 mM NaCl; 1% NP-40; Protease inhibitor cocktail [Pierce]). Lysates were passed ten times through a 20-G needle and incubated for 30 min at 4°C. Lysates were then pre-cleared by centrifugation at 20,000 g for 15 min at 4°C, and the supernatant was incubated with 30 μg anti-GFP nanobody bound to magnetic beads for 2 h at 4°C. The beads were washed three times with 250 μL wash buffer (50 mM Tris, pH 7.5; 300 mM NaCl; 1% NP-40). Samples were eluted in two steps with (1) 0.2 M glycine (pH 2.5) and (2) 1 M NH₄OH. Eluents were neutralized with 0.33 M Tris-HCl (pH 7.5) and stored at −80°C.

Mass spectrometry was carried out at the University of Washington Proteomics Resource (Seattle, WA, USA). Peptide digests were analyzed with an Orbitrap Fusion Tribrid Mass Spectrometer (Thermo Fisher Scientific). Peptides were separated online by reversed-phase chromatography on a nanoACQUITY UPLC system (Waters, Milford, MA, USA) using a heated C18 column over a 120-min gradient. The mass spectrometer was operated in data-dependent acquisition mode. Full survey MS scans were acquired in the Orbitrap at a resolution of 120,000 at 400 m/z with a scan range of 375 to 1575 m/z. MS/MS scans were acquired on selected precursor ions which were fragmented using higher-energy collisional dissociation fragmentation with a 29% collision energy and a stepped collision energy of 5%, detected at a mass resolution of 15,000 at 400 m/z in the Orbitrap, with a maximum injection time of 60 ms and an automated gain control target of 30,000. The MS/MS spectra were searched against the UniProt human protein sequence database using the Comet search engine with the following settings: fully tryptic allowing up to two missed cleavages, 20 ppm precursor tolerance, oxidized methionine as a variable modification, and carbamidomethylated cysteine as a static modification. The peptide-spectrum matches were analyzed by the PeptideProphet tool, and those results uploaded to the Mass Spectrometry Data Analysis Platform (MSDaPl) for protein inference and subsequent analysis.

Data resulting from the proteomic studies, including filtered and peptide-read data, are contained in Files S2 and S3, respectively. Data presented in File S2 were analyzed using the Mass Spectrometry Data Platform (MSDaPl) (Sharma et al., 2012). Identified candidates were filtered by requiring a minimum of three spectral counts (SCs) for each candidate protein as well as a ratio of bait SCs to control SCs of ≥5.

Analyses of FBN-1::sfGFP and NOAH-1::mCherry in sym mutants (Fig. 3) were carried out using strains WY2047, WY2048, WY2063, WY1978, WY1979, and WY2062. Analyses of FBN-1::sfGFP in mec-8 mutants (Fig. 4) was carried out using WY2028. The Pin enhancer screen (Fig. 5) was carried out using WY964. The mir-51–56 analysis (Fig. 6) was carried using SX356 (Shaw et al., 2010) and WY884 For complete genotypes see File S5.

Both anterior perimeter and whole worm mean intensity analyses were performed using Fiji software (Schindelin et al., 2012). Desired planes were selected from the z-stack of raw embryo images. A segmented line tool was used to trace the anterior perimeter of each embryo from dorsal to ventral; the width of the traced line was 10 pixels. The length of the trace was set to be the same for all embryos (1000 pixels). Mean intensity data were acquired from anterior perimeter traces and plotted. Meanwhile, a coefficient of variation (CV) analysis was performed using the anterior perimeter mean intensity data with the following formula: CV=SD/mean (i.e. the SD of each embryo was divided by its corresponding mean). For Fig. S3, mean intensities for all measured traces were averaged and plotted, individual traces were smoothed using a gaussian filter with sigma=10. Bins were designated at 0–250, 250–650, and 650–1000 pixels for each trace and mean intensities were calculated across these intervals. Trace plots were created using seaborn (0.12.2), scipy (1.7.0) and matplotlib (3.6.2). Movies were created from the z-stack images using Fiji software (Movies 1, 2 and 3). For whole embryos the mean intensity of a small box outside the embryo where there was no fluorescence (considered as background) was subtracted from the whole worm intensity (region of interest; ROI) using the subtraction function to obtain a background-subtracted image. A polygon tool was used to trace the whole worm (ROI) using the default line width. The mean intensities of whole worms were determined, and background-subtracted image values were plotted.

Ahringer library RNAi clones (Geneservice Library) corresponding to the genes listed in File S4 were tested on strain WY964 [lin-35(n745); sym-4(mn619)] using standard methods (Kamath and Ahringer, 2003). The RNAi hypersensitive mutation lin-35(n745) was used to increase the detection of RNAi enhancers (Lehner et al., 2006; Wang et al., 2005). Control/Empty Vector RNAi-feeding assays were carried out using a bacterial strain carrying the RNAi vector pDF129.36, which produces an ∼200-bp dsRNA not homologous to any C. elegans gene (Timmons et al., 2001). Screening was carried out by placing ∼5–10 fourth larval stage (L4) worms (the P0 generation) of strain WY964 onto RNAi plates. The P0 s were allowed to develop and have progeny for ∼48 h, at which point they were removed from the plates. The resulting semi-synchronized F1 s were then screened after an additional ∼24 h using a dissecting microscope to identify plates with an increased frequency of larvae arrested at the first larval stage (L1) with an abnormal (misshapen or bulbous) anterior. After the initial screening, phenotypes for RNAi clones were confirmed a minimum of two times by quantifying the percentage of Pin F1 s using a compound microscope. All positive RNAi clones were also sequence confirmed.

Statistical tests were performed as outlined (Fay and Gerow, 2013) using GraphPad (Prism). Comparisons between the means of two samples were carried out using the (non-parametric) two-tailed Mann–Whitney test (Figs 3B,D,E, 4C). In addition, for comparisons of >2 samples, a simple one-way ANOVA was also carried out with results as follows: 3B–Whole embryo (WT, sym-3, sym-4; P=0.34); 3B–Anterior perimeter (WT, sym-3, sym-4; P=0.027); 3D–Whole embryo (WT, sym-3, sym-4, sym-3 sym-4; P=0.26); 3D–Anterior perimeter (WT, sym-3, sym-4, sym-3 sym-4; P=0.72); 3E–FBN-1 (WT, sym-3, sym-4, sym-3 sym-4; P=0.004); 3E–NOAH-1 (WT, sym-3, sym-4, sym-3 sym-4; P=0.001). For proportions (Fig. 5A) 95% CIs were calculated using the Wilson procedure with a correction for continuity (http://vassarstats.net/prop1.html). For comparisons between two proportions a Fisher's exact test was used.

We thank Tony Hyman for recombineered fosmid constructs, Meera Sundaram for strains, Pushph Khanal and John Yochem for help with mir-51–56 strain constructions, Jimmy Eng and Priska von Haller at the University of Washington Proteomics Resource for their kind help with acquiring datasets, and Amy Fluet for editing this manuscript.