Cell signaling facilitates apical constriction by basolaterally recruiting Arp2/3 via Rac and WAVE

Cell shape changes are fundamental to shaping animals during development. One of the most common cell shape changes is apical constriction—the shrinkage of the apical surface of a cell (Martin and Goldstein, 2014). Apical constriction promotes tissue remodeling in various homeostatic and developmental contexts, such as gastrulation in many organisms (Martin and Goldstein, 2014). Neural tube formation in vertebrates also depends on apical constriction, and defective neural tube formation is one of the most common classes of human birth defects (Nikolopoulou et al., 2017). Understanding the mechanisms underlying apical constriction is essential to understanding normal development and disease states.

Apical constriction is driven by the contraction of cortical actomyosin networks and the linkage of these networks to adherens junctions at the plasma membrane (Clarke and Martin, 2021; Martin and Goldstein, 2014). While there is some understanding of how spatially localized actomyosin regulators drive cytoskeletal rearrangement during processes like cell crawling and division, we know far less about how actomyosin networks are regulated during apical constriction. We predict that various actin regulators may have important roles in various sites in apically constricting cells, much as actin regulators do in crawling cells (Blanchoin et al., 2014). To address this knowledge gap, here, we focused on the local regulation of an important actin regulator, the Arp2/3 complex, using Caenorhabditis elegans gastrulation as our model.

Gastrulation in C. elegans begins when the embryo is at the 26-cell stage. Two endodermal precursor cells (EPCs), E anterior and E posterior (Ea/p), internalize from the surface of the embryo by apical constriction (Fig. 1 A) (Lee and Goldstein, 2003; Nance et al., 2003).

Loss-of-function studies in diverse organisms have revealed the importance of the Arp2/3 complex in various morphogenetic events, including apical constriction in C. elegans—100% of Arp2/3-depleted embryos form membrane blebs and exhibit a gastrulation-defective (Gad) phenotype, with EPCs failing to become fully covered by surrounding cells before the EPCs divide (Roh-Johnson and Goldstein, 2009; Severson et al., 2002). The Arp2/3 complex is an actin nucleator that binds existing actin filaments and initiates the formation of new filaments at a characteristic ∼70° angle (Mullins et al., 1998; Svitkina and Borisy, 1999). Arp2/3-generated branched actin networks are present in the cell cortex and near the surface of many membranous organelles (Svitkina, 2018). These networks generate pushing forces during processes such as cell migration and vesicular trafficking (Galletta et al., 2008; Suraneni et al., 2012; Wu et al., 2012).

Arp2/3 is activated by nucleation-promoting factors (NPFs) (Bieling and Rottner, 2023). Moreover, different NPFs fulfill specialized functions at distinct subcellular locations (Molinie and Gautreau, 2018). For instance, the Wiskott-Aldrich syndrome protein and verprolin homologue (WAVE) complex at the leading edge of lamellipodia facilitates cell migration, particularly persistent directional migration (Steffen et al., 2004). Neural Wiskott-Aldrich syndrome protein (N-WASP) plays a crucial role in endocytosis at the plasma membrane (Benesch et al., 2005; Merrifield et al., 2004). The Wiskott-Aldrich syndrome protein and SCAR homologue (WASH) complex activates Arp2/3 on the surface of endosomes, regulating endosomal sorting and trafficking (Gomez and Billadeau, 2009; MacDonald et al., 2018). We were curious whether this division of roles also exists in C. elegans gastrulation, and if so, whether we could take advantage of this to further dissect which subcellular population of Arp2/3 is relevant for apical constriction, and how this population is locally regulated.

Whereas vertebrates typically have several NPF complexes, C. elegans has just three known NPF complexes—a WAVE complex, a WASP complex, and a WASH complex, with each complex member encoded by a single gene (Fig. S1 A) (Kollmar et al., 2012; Smolyn, 2020). This provided us with an opportunity to examine Arp2/3 regulation in apical constriction by studying a complete set of NPFs and avoiding potential function redundancy between different isoforms (Tang et al., 2020). To answer how NPFs might control Arp2/3 activity in apical constriction, we visualized the localization of Arp2/3 along with WAVE, WASP, and WASH via live-cell imaging of tagged endogenous proteins. We observed that WAVE, WASP, and WASH had distinct subcellular locations. Further, these NPFs colocalized with Arp2/3 and controlled Arp2/3 localization at each of these subcellular locations. We then used the NPFs as tools to study the contributions of different populations of Arp2/3 in apical constriction. We confirmed that WAVE is required for gastrulation (Sullivan-Brown et al., 2016) and found that WASP plays a minor role that is redundant with WAVE, whereas depleting WASH had no detectable effect on gastrulation. Depleting WAVE led to a reduction of F-actin levels at regions of cell-cell contact. To determine which signaling pathway(s) might activate WAVE and Arp2/3 at a subset of cell–cell contacts, we targeted Rac genes expressed in the early embryo since Rac small GTPases bind and activate WAVE in diverse systems (Rottner et al., 2021). We identified a Rac that is required for WAVE and Arp2/3 localization to cell–cell contacts. These results lead to the hypothesis that cell–cell contacts serve as a symmetry-breaking cue that localizes WAVE and WAVE-activated Arp2/3 to sites on the basolateral but not apical cell cortex. We confirmed this model by creating ectopic cell-cell contacts and finding that the ectopic contacts were sufficient to recruit WAVE and Arp2/3. Our results suggest that cell-cell signaling via Rac and WAVE positions Arp2/3 basolaterally, where it makes an important contribution to apical constriction by mechanisms that are not yet fully understood.

To elucidate which subcellular population(s) of Arp2/3 contribute to apical constriction, we first examined its localization. Our lab previously generated an antibody against one of the Arp2/3 subunits, ARX-5, to examine Arp2/3 distribution in fixed samples. We observed Arp2/3 enriched near plasma membranes in 26- to 28-cell stage embryos (Roh-Johnson and Goldstein, 2009). To examine Arp2/3 localization over time using live-cell imaging, we filmed embryos with ARX-2 tagged with TagRFP at its C-terminus (Wu et al., 2017). The ARX-2::TagRFP strain was fully viable (Table S2), despite ARX-2 being an essential protein (Severson et al., 2002), indicating that C-terminally tagged ARX-2 is functional. With endogenously tagged ARX-2, we now show that the Arp2/3 complex enriched near specific cell–cell contacts as well as vesicle-like structures in the cytoplasm (Fig. 1 B, white arrows point to vesicle-like structures in the cytoplasm). We observed a brighter signal where the internalizing endodermal precursor Ep and the germline precursor P₄ contact each other, compared with other cell–cell contacts. To quantitatively analyze the distribution of ARX-2 over time, we collected time-lapse movies of multiple embryos and temporally aligned the movies using the birth of the neighboring MSxx cells (i.e., the four granddaughters of the MS cell) as a reference point. EPCs complete apical constriction ∼18 min after MSxx division, so we quantified the ARX-2 signal intensities at regular intervals before then: at 0, 6, and 12 min. We focused on three regions in Ea/p and their neighboring cells: the Ep–P₄ contact, the contacts between Ea and Ep with each other and their neighboring cells excluding P₄, and the cytoplasm (Fig. 1 C). We found that at all the time points analyzed, ARX-2 was most enriched at Ep-P₄ contacts, with a signal intensity approximately five times that of the cytoplasmic signal; ARX-2 was also enriched at other cell–cell contacts, although to a lesser extent; ARX-2 signal showed a punctate distribution at the non-Ep-P₄ cell–cell contacts, and the average signal intensity along these cell–cell contacts, which included the punctae, was ∼1.3 times that of the cytoplasmic signal (Fig. 1 D and Fig. S1 B). Because the overall localization patterns were similar at all time points, we showed 6 min as a representative time point in Fig. 1 D, and the rest of the time points are included in Fig. S1 B.

Next, we aimed to determine how the Arp2/3 complex became localized at the aforementioned locations. Because different NPFs have been shown to fulfill specialized functions at distinct subcellular locations in migrating cells (Molinie and Gautreau, 2018), we hypothesized that NPFs play critical roles during apical constriction by modulating Arp2/3 activities at specific sites. To test this hypothesis, we examined the localization of the three known NPFs in C. elegans—the WAVE, WASP, and WASH complexes—in early embryos. Specifically, we visualized tagged endogenous components of these complexes (Fig. S1 A). Using the same quantification pipeline as we used for the Arp2/3 complex, we analyzed the distribution of each NPF.

To examine the localization of the WAVE complex, we filmed embryos expressing GEX-3 tagged with GFP at its N terminus (Heppert et al., 2016). The tagged strain had an embryo hatching rate of 99%, indicating that the tagged GEX-3 was functional (Table S2). For simplicity, we refer to the gex-3 gene and GEX-3 protein as WAVE hereafter. WAVE localized to most cell–cell contacts with an average signal intensity along the contacts ∼1.25 times that of the cytoplasmic signal, except at the Ep-P₄ contact, where the signal was indistinguishable from the cytoplasmic signal (Fig. 1, B and D; and Fig. S1 B). This localization to cell–cell contacts during gastrulation is consistent with WAVE’s basolateral localization at later stages in epithelial cells of the epidermis (Bernadskaya et al., 2012; Patel et al., 2008).

The wsp-1 locus is predicted to encode two protein isoforms, WSP-1A and WSP-1B, using two different transcriptional start sites. We used a strain expressing WSP-1A tagged with GFP at its N terminus (Wu et al., 2017) and tagged WSP-1B with mNeonGreen at its N terminus using CRISPR/Cas9-dependent genome editing. Both strains were fully viable (Table S2). WSP-1A was primarily localized to the Ep-P₄ contact, with a signal intensity about four times that of the cytoplasmic signal. WSP-1A was also enriched at other cell–cell contacts, with an average signal intensity along contacts ∼1.25 times that of the cytoplasmic signal (Fig. 1, B and D; and Fig. S1 B). WSP-1B was at levels below the detectable threshold on our imaging system, appearing similar to unlabeled, wild-type N2 control embryos at the time of Ea/Ep internalization (Fig. S1, C–E). For simplicity, we refer to the wsp-1 locus and WSP-1A protein as WASP.

To examine the localization of the WASH complex, we used CRISPR/Cas9-dependent genome editing to tag wshc-5 with mNeonGreen at its C terminus. The tagged strain is fully viable, indicating it was functional (Table S2). For simplicity, we refer to the wshc-5 gene and WSHC-5 protein as WASH hereafter. WASH was not enriched at any cell–cell contacts but localized to vesicle-like structures in the cytoplasm (Fig. 1 B). Since WASH localization had not been documented previously in C. elegans, we compared our findings with observations from other systems. WASH is known to interact with F-actin capping proteins (Hernandez-Valladares et al., 2010; Jia et al., 2010), so we performed colocalization analyses between WASH and the capping protein CAP-1 with a dual-labeled strain (Fig. S1 F). Our analysis focused on the previously defined regions: the Ep–P₄ contact, the other cell–cell contacts, and the cytoplasm (Fig. 1 C). We measured and compared the Pearson correlation coefficient (PCC) at each of these locations for WASH and CAP-1; WASH and CAP-1 colocalized most strongly at punctate structures in the cytoplasm. This is consistent with previous work in other systems that found WASH localized to endocytic vesicles (MacDonald et al., 2018).

We proceeded to quantify WASH localization. The total signal intensities at all measured locations showed no difference, likely because the contribution from the vesicle-like structures is too weak to significantly affect the total cytoplasmic intensity measurements (Fig. 1 D and Fig. S1 B). Therefore, to quantify the signal distribution within the cytoplasm for WASH and other proteins, we performed line scans across the cytoplasm of Ea and Ep (Fig. 1 C). Because the overall localization patterns were similar at all time points based on our previous results (Fig. 1 D and Fig. S1 B), we used 6 min as a representative time point for this and the rest of the quantification in this study except where noted. We found that the signals for both WASH and Arp2/3 showed peaks along the line scan corresponding to the sites of the vesicle-like structures, whereas WAVE and WASP exhibited a more uniform distribution (Fig. 1 E). We used the difference between the maximum and minimum intensity along the line scan to quantify the degree of punctate signal distribution in the cytoplasm. We found that WASH and Arp2/3 had a consistently larger intensity difference compared with WAVE and WASP (Fig. 1 F).

We concluded that three NPFs in C. elegans localize to mostly distinct subcellular locations, with some overlap between WAVE and WASP at cell–cell contacts.

To investigate whether the three NPFs colocalize with Arp2/3 at their respective subcellular locations, we performed live-cell imaging of dual-labeled strains for each NPF and Arp2/3 (Fig. 2 A). We conducted colocalization analyses using the quantification pipeline that we established for WASH and CAP-1. Each of the NPFs colocalized with Arp2/3. WAVE and Arp2/3 colocalized most strongly at the non-Ep–P₄ cell–cell contacts (Fig. 2 B). WASP and Arp2/3 exhibited the strongest colocalization at the Ep–P₄ contact, with a weaker but still significant colocalization at the other cell–cell contacts, and no obvious colocalization in the cytoplasm (Fig. 2 B). WASH and Arp2/3 colocalized most strongly in the cytoplasm (Fig. 2 B).

Our results so far led us to hypothesize that NPFs modulate Arp2/3 activities at distinct subcellular locations, with potential functional overlap between WAVE and WASP at cell–cell contacts. To test this hypothesis, we observed and quantified Arp2/3 localization after depleting the NPFs by RNAi. We maximized RNAi efficiency by synthesizing double-stranded RNAs (dsRNAs) using C. elegans cDNA as templates and performing RNAi by injection. Then, we quantitatively assessed the extent of protein knockdown by imaging and quantifying the total fluorescence from three sets of embryos side by side: unlabeled (wild-type) embryos, embryos expressing the fluorescent tags, and embryos with tagged components and targeted by dsRNA. In all cases, the total fluorescence in the tagged, knockdown embryos was indistinguishable from that in unlabeled embryos, indicating that RNAi was effective at depleting these proteins to undetectable levels (Fig. S2, A and B). Although the wsp-1 RNAi was designed to target both wsp-1a and wsp-1b, WSP-1B was not detectable in untreated early embryos (Fig. S1, C–E), so we attributed phenotypes observed upon wsp-1 RNAi to loss of the sole isoform detectable in early embryos, WSP-1A.

We then depleted each NPF in Arp2/3-labeled embryos and quantified the change in Arp2/3 localization by imaging and comparison of RNAi-treated and control embryos side by side. In Fig. 1 and Fig. 2, we imaged embryos from an en-face ventral view to capture the details of protein localization. While lateral view imaging provides less detail overall, it offers better visualization of protein distribution along the apical–basal axis. Since our results suggest that the NPFs and Arp2/3 localize to cell–cell contacts and within the cytoplasm, we switched to the lateral view for subsequent imaging (Fig. 3 A). WAVE(RNAi) reduced the Arp2/3 signal at the non-Ep-P₄ cell–cell contacts to the cytoplasmic level while leaving the signal at the Ep–P₄ contact and in the cytoplasm mostly unchanged (Fig. 3, A and B). WASP(RNAi) reduced the Arp2/3 signal at the Ep-P₄ contact to the cytoplasmic level. Arp2/3 at the non-Ep-P₄ cell–cell contacts was not affected by WASP(RNAi), and the signal in the cytoplasm increased, likely due to the failure to recruit Arp2/3 to the Ep–P₄ contact (Fig. 3, A and B). WASH RNAi eliminated most of the cytoplasmic, vesicle-like Arp2/3 signal as observed through live imaging (Fig. 3 A). Our quantification method detected no changes at any measured locations (Fig. 3 B), consistent with our previous observations (Fig. 1 D and Fig. S1 B). Therefore, to quantify the effect of WASH(RNAi), we again implemented the line scan method to capture the change in signal intensity in the cytoplasm. WASH(RNAi) reduced the difference between maximum and minimum Arp2/3 signal intensity along the line scan, whereas WAVE(RNAi) and WASP(RNAi) led to no obvious change (Fig. 3 C).

We conclude that discreet NPFs build Arp2/3-enriched structures at distinct cellular and subcellular locations: WAVE is mainly controlling Arp2/3 at the non-Ep-P₄ cell–cell contacts, WASP is responsible for the signal at the Ep-P₄ contact, and WASH is responsible for Arp2/3 at the vesicle-like structures.

Our RNAi results positioned us to study the contributions of different subcellular populations of Arp2/3 in apical constriction. We investigated how NPF deletion might affect C. elegans gastrulation using RNAi against each NPF alone or in combination. Embryos from injected worms were filmed using differential interference contrast (DIC) microscopy and examined for gastrulation defects, defined as the failure of the two EPCs to fully internalize, leaving part of at least one EPC uncovered by the neighboring cells before division.

Consistent with our previous findings (Sullivan-Brown et al., 2016), WAVE RNAi alone resulted in a highly penetrant Gad phenotype in 40 out of 46 (87%) embryos (Fig. 4, A and B). Depleting WASP and WASH individually or in combination did not disrupt gastrulation (Fig. 4, A and B). However, combining WASP(RNAi) with WAVE(RNAi) significantly enhanced the WAVE(RNAi) phenotype, with 63 out of 64 (98%) embryos showing gastrulation defects (Fig. 4, A and B), indicating that WASP is redundant with WAVE in apical constriction. Deleting all NPFs simultaneously resulted in 100% gastrulation defects in all 44 embryos examined, mirroring the 100% defect observed in Arp2/3 RNAi embryos (Fig. 4, A and B). Since the premature division of the EPCs can prevent their internalization (Lee et al., 2006), we examined the timing of the Ea/p division by measuring the time between the MS division and the Ea/p division (Table S5). None of the RNAi treatments led to a significant change in division timing, indicating that Ea and Ep did not divide prematurely in NPF-depleted embryos. Taken together, our results suggest that the subpopulation of Arp2/3 activated by WAVE at the non-Ep-P₄ cell–cell contacts is the main contributor to apical constriction (Fig. 4 B). We next aimed to understand (1) does WAVE locally regulate F-actin and cadherin levels, and (2) how does WAVE become activated at these cell–cell contacts.

WAVE-activated Arp2/3 drives branched F-actin formation at various subcellular locations, including the leading edge of migrating cells, endocytic vesicles, and cell–cell contacts of epithelial and endothelial cells (Papalazarou and Machesky, 2021). Previously, our lab demonstrated that three of the six neighbors of Ea and Ep cells form F-actin-rich processes during apical constriction (Roh-Johnson and Goldstein, 2009). F-actin is transiently enriched at the front of these MSxx neighbors (MSxx–Ea contact) compared with the back (MSxx–ABxxxx contact), and this enrichment depends on Arp2/3 (Roh-Johnson and Goldstein, 2009). To investigate whether WAVE-activated Arp2/3 regulates F-actin at non-Ep-P₄ cell–cell contacts, including the F-actin–rich processes, we depleted WAVE by RNAi in embryos with a single-copy transgene expressing an F-actin marker, Utrophin::mScarlet-I (Fig. 5 A). We then quantified and compared F-actin levels at multiple subcellular locations in control versus RNAi embryos: (1) the front of the extension-forming MSxx (Ea-MSxx cell contact regions), (2) the back of the extension-forming MSxx (MSxx-ABxxxx), (3) the contact between two Ea/Ep cells (Ea-Ep), (4) the contact between Ea/Ep cells and non-extension-forming ABxxxx cells (Ea/p-ABxxxx), and (5) the Ep–P₄ contact (Fig. 5 A).

We found that WAVE RNAi caused varying degrees of F-actin reduction at all examined cell–cell contacts except for the Ep–P₄ contact (Fig. 5 B), in line with WAVE’s role in activating Arp2/3 at non-Ep–P₄ cell–cell contacts. Consistent with previous observations, the front of MSxx exhibited ∼1.2 times higher F-actin levels compared with the back in control embryos, but this enrichment was lost in the WAVE RNAi-treated embryos (Fig. 5 C). These results suggest that WAVE is critical for maintaining F-actin levels at non-Ep-P₄ cell–cell contacts and that the formation of F-actin–rich extensions in MSxx cells depends on WAVE.

Because branched F-actin stabilizes adherens junctions in other systems (Efimova and Svitkina, 2018; Li et al., 2020; Verma et al., 2012), we investigated whether depleting WAVE affects E-cadherin levels at cell–cell contacts. Using a strain with endogenously tagged E-cadherin (HMR-1::GFP), we examined E-cadherin levels at cell–cell contacts upon WAVE RNAi (Fig. S3 A). Since F-actin reduction was observed at all non-Ep–P₄ contacts, we initially quantified E-cadherin levels across all non-Ep–P₄ contacts as a group and the Ep–P₄ contacts but did not detect significant changes in both (Fig. S3 B). We then specifically examined E-cadherin levels at the extension-forming Ea–MSxx contacts. We found that WAVE depletion did not reduce E-cadherin levels at any of the locations examined (Fig. S3 B).

Together, we found that depleting WAVE-activated Arp2/3 reduced F-actin levels but did not affect E-cadherin levels at non-Ep-P₄ cell–cell contacts. This suggests that WAVE-activated Arp2/3 regulates F-actin levels during apical constriction but does not modulate junctional cadherin levels.

In migrating cells and epithelial cells, WAVE is known to be activated by Rac family GTPases at the leading edge (Bernadskaya et al., 2012; Innocenti et al., 2004; Miki et al., 1998). To determine if Rac also regulates WAVE during apical constriction, we first checked the expression of Rac genes in early embryos and the germline using published RNA sequencing databases from our lab and others. We found that two of the three putative Rac genes, ced-10 and mig-2, are expressed in early embryos and the germline, whereas rac-2 showed no detectable expression in these stages (Diag et al., 2018; Tintori et al., 2016). We then RNAi-depleted ced-10 or mig-2 in WAVE-labeled embryos and quantified WAVE localization. Depletion of CED-10 reduced the WAVE signal at cell–cell contacts to the level of cytoplasmic WAVE (Fig. 6, A and B). mig-2(RNAi) did not alter WAVE localization (Fig. S4, A and B).

These findings indicate that CED-10 is crucial for recruiting WAVE to cell–cell contacts in the early embryo. We then explored if CED-10 controls Arp2/3 localization and found that ced-10 RNAi reduced the Arp2/3 signal at non-Ep-P₄ cell–cell contacts to the cytoplasmic level, while the signal at Ep–P₄ contact and in the cytoplasm remained mostly unchanged (Fig. 6, A and B). Since depletion of CED-10 or WAVE resulted in similar reductions in Arp2/3 levels (i.e., down to cytoplasmic levels), we conclude that CED-10 is key in activating Arp2/3 at specific cell–cell contacts. To examine the localization of CED-10, we tagged it with mNeonGreen at the N-terminus. This stable new strain has a hatching rate of 91% (Table S2). Because ced-10 null worms are not viable (Lundquist et al., 2001), we concluded that the tagged protein is partially functional. Live imaging of tagged endogenous CED-10 showed that it localizes to all cell–cell contacts, including the Ep–P₄ contact (Fig. S4 C). Using this tagged endogenous CED-10, we verified that ced-10(RNAi) was highly effective at depleting CED-10 protein to undetectable levels (Fig. S4, D and E).

Because of the role of CED-10 in recruiting WAVE and Arp2/3 to a subset of cell–cell contacts, we hypothesized that CED-10 function contributes to C. elegans gastrulation. To test this, we examined ced-10 RNAi embryos by DIC imaging and found that 31 out of 39 targeted embryos exhibited gastrulation defects (Fig. 6, C and D). Taken together, our results demonstrate that the CED-10/Rac GTPase recruits WAVE and Arp2/3 to specific cell–cell contacts and plays a crucial role in gastrulation.

Our results suggested that the Rac–WAVE–Arp2/3 signaling axis is activated at a subpopulation of cell–cell contacts. We hypothesized that cell–cell contacts serve as a symmetry-breaking cue to localize WAVE and Arp2/3 specifically to basolateral sites. To test this, we removed the eggshell and envelope from developing early embryos and created new cell–cell contacts by placing pairs of embryos in contact (Fig. 7 A). Both WAVE and Arp2/3 localized to the ectopic contacts (Fig. 7 A). The signal intensity at the ectopic contacts was significantly higher than the normally contact-free cell apex and was comparable with the endogenous contacts (Fig. 7 B). We concluded that cell–cell contacts are sufficient to recruit WAVE and Arp2/3 to sites of cell–cell contact. Because cell contacts normally define basolateral sites and we could recruit WAVE and Arp2/3 to previously contact-free apical sites by ectopically providing cell contacts there, we concluded that cell–cell contact acts as an apico-basal symmetry-breaking cue that recruits WAVE and Arp2/3 specifically to basolateral sites.

The actin cytoskeleton is crucial for generating and transmitting forces to neighboring cells in apical constriction (Lee and Goldstein, 2003; Martin et al., 2009; Slabodnick et al., 2023). However, we still have a limited understanding of the degree to which actin regulation in various subcellular locations contributes to apical constriction. Here, we focused on one of the major nucleators of the actin cytoskeleton, the Arp2/3 complex, to gain insights into its function and regulation in C. elegans gastrulation. We discovered that three NPFs, the WAVE, WASP, and WASH complexes, activate distinct subpopulations of Arp2/3 at specific cellular and subcellular locations. WAVE RNAi, but not WASP or WASH RNAi, recapitulated the gastrulation defect caused by Arp2/3 RNAi, suggesting that the Arp2/3 complex activated by WAVE at a subset of cell–cell contacts is key in apical constriction. We identified that a Rac protein, CED-10, is responsible for localizing WAVE and Arp2/3 at these cell–cell contacts. Further, we found that ectopic cell–cell contacts between embryos could recruit WAVE and Arp2/3, revealing cell–cell contacts as an apico-basal symmetry-breaking cue for the localization of these proteins. In summary, our work reveals that cell–cell signaling makes an essential contribution to apical constriction by basolaterally recruiting Arp2/3 via Rac and WAVE, as depicted in the model in Fig. 8, A and B.

Although our work reveals a subcellular population of Arp2/3 that contributes to gastrulation and reveals cellular and molecular mechanisms by which this specific population is specified, we do not yet fully understand how basolateral Arp2/3 contributes mechanistically to apical constriction. The reduction of F-actin levels at non-Ep–P₄ cell–cell contacts suggests that WAVE-activated Arp2/3 may contribute to apical constriction via regulating F-actin. Moreover, Arp2/3-mediated actin polymerization and myosin II-dependent contractility have been shown to play opposing roles in regulating cortical tension: active myosin II at the cortex slides actin filaments with respect to one another and generates cortical tension, whereas activation of the Rac–WAVE–Arp2/3 signaling axis decreases cortical tension (Bergert et al., 2012; Chugh et al., 2017; Tinevez et al., 2009). Elevated myosin II activity and cortical tension can lead to bleb formation, characterized by local detachment or rupture of the cortex from the plasma membrane (Tinevez et al., 2009). CED-10, WAVE, and Arp2/3-depleted embryos form membrane blebs, which may indicate elevated membrane tension and/or decreased membrane–cortex linkage (Lamb et al., 2025; Roh-Johnson and Goldstein, 2009; Sullivan-Brown et al., 2016). To test if the blebbing phenotype is linked to the gastrulation defects seen in WAVE-depleted embryos, we attempted to rescue the gastrulation defects by tuning myosin levels with partial RNAi. Using a strong loss-of-function strain of GEX-3/WAVE, WM43, we found that reducing myosin levels with partial RNAi could restore membrane morphology and partially rescue gastrulation defects in WAVE-depleted embryos (Fig. S5, A and B). Although the membrane blebs hindered our ability to segment the cell–cell boundary, measure myosin speed near the edges of the apical domain, and assess its linkage to the membrane, our partial RNAi data indicated that the defects caused by WAVE RNAi can be rescued by reducing myosin levels. This observation suggests to us that the gastrulation defects caused by WAVE depletion may be related to the blebbing phenotype and membrane–cortex linkage issues rather than reducing contractility. We speculated that active Rac-WAVE-Arp2/3 signaling at cell–cell contacts is crucial for maintaining proper membrane–cortex linkage—perhaps by biasing the actin network toward the branched form versus a more contractile linear actin network (Kadzik et al., 2020; Muresan et al., 2022)—and that strong linkage at sites of cell–cell contact is important for successful apical constriction.

Components of the WAVE complex including Nckap1 and Abi1, as well as Rac1, have been implicated in neural tube closure in mice, with disruptions in these genes leading to neural tube defects (Dubielecka et al., 2011; Migeotte et al., 2011; Rakeman and Anderson, 2006; Rolo et al., 2016). Neural tube closure is a complex morphogenetic process that requires the coordination of many cellular events, including apical constriction to bend the tissue and create the neural folds and the fusion and zippering of the neural folds to create a tube-shaped tissue (Hashimoto et al., 2015; Nikolopoulou et al., 2017). Rac-1 has been shown to drive protrusion formation in non-neural (surface) ectoderm, and such protrusions are important for neural fold fusion in the mouse spinal region (Rolo et al., 2016). The MSxx neighbors of the Ea/Ep cells also formed WAVE-dependent, actin-enriched extensions during apical constriction. Because partial myosin RNAi did not rescue 100% of the gastrulation defects, it is possible that the extensions from the MSxx neighbors also contribute to the completion of E-cell internalization. These findings suggest that C. elegans gastrulation is potentially a valuable model for further investigating multiple cell biological mechanisms underlying neural tube formation. It may be of value to use cell-specific manipulations of protein levels and/or localization to further dissect the function of the RAC–WAVE–Arp2/3 signaling axis in Ea/Ep cells and their neighbors.

Our genetic analysis indicated that WASP makes a minor, redundant contribution to apical constriction. This may reflect the importance of strong Arp2/3 enrichment at the Ep-P₄ contact. Alternatively, WASP might recruit a small amount of Arp2/3 at other cell–cell contacts, an amount that was below the detection limit of our current quantification methods. A functional coordination between WAVE and WASP has been shown in C. elegans neuroblast migration—WAVE mutations impair migration, while WASP mutations alone do not affect migration but exacerbate defects in WAVE-deficient cells (Zhu et al., 2016). Our findings indicate that WAVE and WASP may be coordinating with each other at a cellular and/or subcellular level in apical constriction. In the future, we are interested in understanding the striking enrichment of WASP and the exclusion of WAVE at the Ep-P₄ contact and its potential role in apical constriction and other processes, such as asymmetric division. Additionally, this may also help explain the previously observed enrichment of capping protein at the endoderm/germline precursor cells’ contacts (Goldstein, 2000; Zhang et al., 2023).

WASH depletion did not result in gastrulation defects. However, it has been shown to play a role in C. elegans embryonic development, perhaps by affecting processes that happen at later stages (Smolyn, 2020). We found that WASH colocalizes with the capping protein CAP-1, as shown in other systems. CAP-1 has been observed surrounding the germ cell nuclei (Ray et al., 2023), and the authors of this paper speculated that this perinuclear localization is due to the capping protein being a component of the dynactin complex. Our findings suggest an alternative explanation for this localization: CAP-1 may localize to these sites as a member of the WASH complex. These possibilities are not mutually exclusive: WASH has been reported to assemble a super complex with dynactin where the capping protein is exchanged from dynactin to the WASH complex (Fokin and Gautreau, 2021). The distinct perinuclear location in early embryos may offer an opportunity to further dissect the function and organization of these complexes.

The quantification pipeline we developed in this paper will be valuable for analyzing and understanding the function and regulation of other actin regulators during apical constriction, particularly those active at similar cellular locations, including the capping proteins. The regulation and interplay between different NPFs play crucial roles in processes such as directional cell migration, membrane remodeling, and vesicular trafficking. Our results highlight their distinct contributions to apical constriction by regulating the Arp2/3 complex. The system we present here offers a valuable tool for scientists interested in NPF-dependent processes to dissect their regulation and activation, potentially revealing organizing principles beyond the context of apical constriction.

The Rac–WAVE–Arp2/3 signaling axis plays crucial roles in various fundamental processes, including lamellipodia formation, junction establishment and maintenance, and regulation of cortical tension (Bergert et al., 2012; Sasidharan et al., 2018; Steffen et al., 2004; Yamazaki et al., 2007). In migrating cells, Rac activates Arp2/3 at the leading edge to produce branched-actin networks and generate pushing force. Myosin II, on the other hand, typically gets activated at the rear end of migrating cells and contributes to trailing edge retraction and steering the direction of migration (Allen et al., 2020; Ridley et al., 2003; Vicente-Manzanares et al., 2007). Intriguingly, in apically constricting cells in C. elegans embryos, Rac signaling and a myosin-activating kinase are also active at distinct locations: WAVE and Arp2/3 localize basolaterally via cell–cell contacts, while myosin is enriched at the contact-free apices of Ea and Ep cells (Fig. 8 C). Because lamellipodia-based cell crawling is found more broadly than in metazoans and therefore is likely to have evolved earlier than apical constriction (Brunet, 2023), the similarity between the two systems could feasibly reflect an ancient evolutionary co-opting of cell crawling mechanisms for bending cell sheets. Consistent with this possibility, Rac and Rho function antagonistically in both crawling cells and some apically constricting cells (Burridge and Wennerberg, 2004; Chauhan et al., 2011; Nobes and Hall, 1995). Alternatively, lamellipodia-based cell crawling and apical constriction might have evolved independently. Due to limitations of resolution and imaging depth, we were not able to resolve whether WAVE and Arp2/3 contribute to treadmilling structures extended from the basolateral sides of cells. In the future, newer techniques may clarify whether active migration at the basolateral side of cells contributes to apical constriction.

Table S1 lists the strains used in this study. Some strains were provided by the Caenorhabditis Genetics Center (CGC), which is funded by the National Institutes of Health (NIH) Office of Research Infrastructure Programs (P40 OD010440). Worms were cultured on Nematode Growth Medium plates at 20–22°C following standard protocols (Brenner, 1974; Sternberg et al., 2024).

The new endogenously tagged strains described in this work were created using two previously established CRISPR/Cas9-mediated genome editing protocols: a Cas9 protein-based method (Ghanta and Mello, 2020) and a Cas9-expressing plasmid method (Dickinson et al., 2015). Cas9 targeting sequences for each gene were selected using the CRISPR design tool CRISPOR (Concordet and Haeussler, 2018). In the protein-based method, trans-activating CRISPR RNA (tracrRNA) and CRISPR RNA (crRNA) were mixed with Cas9 protein (all reagents from IDT) and incubated at 37°C for 15 min. Then, a double-stranded DNA donor, containing codon-optimized mTurquoise2 (amplified from pDD315, RRID:Addgene_73343), mNeonGreen (amplified from pDD268, RRID:Addgene_132523), or mScarlet-I (amplified from pMS050, RRID:Addgene_91826) with 35 bp homology arms, was melted, cooled, and added to the mixture. Additionally, pRF4 (rol-6(su1006)) plasmid was included as a coinjection marker. The injection mix was then centrifuged, transferred to a fresh tube, and kept on ice during the injection. Heterozygous mutants were selected by genotyping F1 non-rollers or examining F1 under a Zeiss Axiozoom V16 stereo microscope. In the plasmid-based method, repair templates were constructed by integrating homology arm PCR products, which were amplified from worm genomic DNA, into vectors containing a fluorescent protein and a selection cassette using Gibson Assembly (New England Biolabs), as thoroughly detailed by Dickinson et al. (2015). Cas9 guide sequences were inserted into the Cas9-sgRNA expression vector pDD162, RRID:Addgene_47549, and coinjected into the adult germlines along with the repair template vector and array markers. The selection of the edited worms was performed using previously established techniques (Dickinson et al., 2015). Homozygotes made with both methods were sequenced to confirm the edits. sgRNAs and homology arms sequences can be found in Table S3.

Primers were designed to amplify ∼1,000 bp of the protein-coding sequence of each target gene (Table S4). cDNA from wild-type N2 (WBSTRAIN:WBStrain00000001) worms was prepared using SuperScript III First-Strand Synthesis SuperMix according to the manufacturer’s instructions (Invitrogen). Each primer included 15 bases of the T7 promoter sequence at the 5′ end for use in a two-step PCR with the wild-type cDNA. The PCR product was purified using a Zymoclean Gel DNA Recovery Kit (Zymo Research) and used as a template for a second PCR with primers containing the full-length T7 promoter sequence. After another round of purification with the Zymoclean Gel DNA Recovery Kit, the PCR product served as a template in a T7 RiboMAX Express RNAi System (Promega) following the manufacturer’s instructions. Animals were randomly selected to receive RNAi treatment or be used as control. Purified dsRNA, at a concentration of 300–1,000 ng/μl, was injected into L4 or young adult hermaphrodites using a Narishige injection apparatus, a Parker Instruments Picospritzer II, and a Nikon Eclipse TE300 microscope with DIC optics. Excess dsRNA was aliquoted to avoid repeated freeze–thaw cycles and stored at −80°C. For complete knockdown, injected worms were allowed to recover on a seeded NGM plate for at least 28 h at room temperature before harvesting embryos for imaging. For partial RNAi, worms were collected for imaging after 6–8 h.

Embryos were dissected from gravid adults and mounted on poly-L-lysine–coated coverslips in C. elegans egg buffer, then sealed with VALAP (a 1:1:1 mixture of vaseline, lanolin, and paraffin). For ventral mounts, glass beads (∼23 µm in diameter, MS0023; Whitehouse Scientific Monodisperse Standards) and clay feet were used to create separation between the slide and coverslip. For lateral mounts, 2.5% agar pads were used. Fluorescence imaging was carried out using a Hamamatsu ORCA QUEST qCMOS camera mounted on a Nikon Eclipse Ti inverted confocal microscope equipped with a Yokogawa CSU-X1 spinning-disk scan head. Images were captured at room temperature using a 60× 1.4 NA oil immersion lens and lasers at 445-, 488-, 514-, and 561-nm wavelengths. Some images were taken with a 1.5× magnifier. Z-stacks with a step size of 0.5–1 μm were collected every 1–3 min. Four-dimensional DIC video microscopy was performed on a Nikon Eclipse 800 microscope using a Diagnostic Instruments SPOT2 camera, capturing images with a 60× 1.4 NA oil objective every minute at 1-μm optical sections. Images were acquired and analyzed using MetaMorph software (RRID:SCR_002368).

Signal intensity at different subcellular regions was collected in ImageJ/FIJI (Schindelin et al., 2012). For contacts between cells, the freehand line tool was used with a line width of five pixels (∼0.44 μm) to select the region of interest and measure the mean gray value. The Ep-P₄ contact refers to the region where Ep and P₄ cells touch. The other cell contacts’ measurement reflects contacts between Ea and Ep with each other and their neighboring cells excluding P₄. For cytoplasmic signal intensity in Ea and Ep, a region of interest within each cell was drawn, excluding the cell boundaries and nucleus. Background subtraction was performed post hoc by manually subtracting the mean gray value of a region outside of the embryo to account for camera noise. Measurements were collected at 0, 6, and 12 min following division of the neighboring mesoderm precursor cells (MSx). All measurements were normalized to the average cytoplasmic signal.

The extent of colocalization between two proteins was determined using ImageJ/FIJI (Schindelin et al., 2012) and the Coloc 2 plugin (https://imagej.net/plugins/coloc-2). First, images were processed with the rolling ball background subtraction algorithm (radius: 50 pixels) and a Gaussian blur filter (radius: two pixels) to account for noise. Regions of interest were selected as described in the previous section; however, a thicker line width of 20 pixels (∼1.78 μm) was used to ensure that we captured areas of both high and low signal intensity for comparison. The Coloc 2 plugin was performed using the Costes threshold regression method, and the PCC was collected. Measurements were collected at 6 min following the division of the neighboring MSx.

To quantify the uniformity of protein localization across the cytoplasm, we collected line scans in ImageJ/FIJI (Schindelin et al., 2012). First, images were processed with a Gaussian blur filter (radius: two pixels) to account for noise. Straight lines with a line width of 10 pixels (∼0.89 μm) were drawn across the cytoplasm, excluding cell–cell boundaries and the nucleus. Maximum and minimum gray values within the region of interest were measured, and the differences between them were calculated and plotted. Representative line scan plots were also provided. Measurements were collected at 6 min following the division of the neighboring MSx.

To quantitatively evaluate the efficacy of the RNAi treatments, we mounted stage-matched wild-type, knock-in, and RNAi-treated embryos side-by-side. Relative intensity measurements were obtained from the sum intensity Z-projections of the embryos in ImageJ/FIJI (Schindelin et al., 2012). Intensity was measured by drawing a region of interest around the embryo and measuring the mean gray value, and then manually subtracting the mean gray value of a background region to account for camera noise.

Control and RNAi-treated embryos were mounted side by side to facilitate quantification. To quantify changes in signal intensity due to RNAi treatment, measurements were taken in the previously described regions of interest using a line width of five pixels (∼0.44 μm) in ImageJ/FIJI (Schindelin et al., 2012). Background subtraction was performed manually by subtracting the mean gray value of a region outside the embryo to account for camera noise. Measurements were collected 6 min after the division of neighboring MSx. Measurements at each region were normalized to the average signal intensity of the control cytoplasm.

Ventral mounts of individual control and RNAi-treated embryos were used to allow for maximum exposure of multiple cell–cell contacts. To quantify changes in signal intensity due to RNAi treatment, measurements were taken in the regions of interest using a line width of five pixels (∼0.44 μm) in ImageJ/FIJI (Schindelin et al., 2012). Background subtraction was performed manually by subtracting the mean gray value of a region outside the embryo to account for camera noise. Measurements were collected 9 min after the division of neighboring MSx. The 9-min time point was chosen as a representative time point because our lab previously measured MSxx extension formation from 6 to 12 min after MSx division (Roh-Johnson and Goldstein, 2009). Measurements at each region were normalized to the average signal intensity of the whole embryo.

Control and RNAi-treated embryos were mounted side by side to facilitate quantification. To quantify changes in signal intensity due to RNAi treatment, measurements were taken in the regions of interest using a line width of five pixels (∼0.44 μm) in ImageJ/FIJI (Schindelin et al., 2012). Background subtraction was performed manually by subtracting the mean gray value of a region outside the embryo to account for camera noise. Measurements were collected 6 min after the division of neighboring MSx.

The eggshell and vitelline envelope of the embryos were removed as described previously (Edgar and Goldstein, 2012). Briefly, a solution containing chitinase (5 U/ml, C6137; Sigma-Aldrich) and chymotrypsin (10 mg/ml, C4129; Sigma-Aldrich) was used to digest the eggshell. The vitelline envelope was then mechanically stripped off. Embryos, now free of the eggshell and vitelline envelope, were pushed together to facilitate the formation of ectopic cell–cell contacts. These newly formed chimeras were mounted on glass coverslips with glass beads and clay feet in Shelton’s Medium and then sealed with VALAP (a 1:1:1 mixture of vaseline, lanolin, and paraffin) for imaging.

To compare the amount of protein localized to ectopic contacts, endogenous contacts, and contact-free cell apexes, measurements were taken in these regions using a line width of five pixels (∼0.44 μm) in ImageJ/FIJI (Schindelin et al., 2012). Endogenous contacts and contact-free cell apexes were selected near the ectopic contacts to minimize potential variation from uneven illumination. For chimeras with multiple ectopic contacts, an average measurement for each region within the same embryo was calculated and used for statistical analysis. Background subtraction was performed manually by subtracting the mean gray value of a region outside the embryo to account for camera noise. Measurements at each region were normalized to the average signal intensity of the contact-free cell apexes.

All statistical analyses were performed using GraphPad Prism 10 (RRID:SCR_002798). Statistical tests for each experiment with numerical data were chosen based on the normality and variance of the data and described in the figure legends. Only two-sided tests were used. Data were tested for normality using a Shapiro–Wilk test. For all parametric tests, variance equality was checked with an F-test (for two groups) or a Brown–Forsythe test (for multiple groups). Categorical data (comparative analyses of Gad phenotypes across different groups) were conducted using Fisher’s exact test and described in the figure legends.

Fig. S1 shows the characterization of WAVE, WASP, and WASH during apical constriction. Fig. S2 verifies the effectiveness of WAVE, WASP, and WASH RNAi through quantification of fluorescence in wild-type/knock-in/RNAi-treated embryos. Fig. S3 shows that WAVE RNAi did not affect E-cadherin levels at cell–cell contacts. Fig. S4 shows the characterization of Rac proteins during apical constriction. Fig. S5 shows that Myosin II partial RNAi rescued gastrulation defects caused by WAVE depletion. Table S1 lists the strains used in this study. Table S2 reports embryo hatching rates. Table S3 provides the sequences used for CRISPR/Cas9 genome editing. Table S4 provides the sequences used for generating dsRNA. Table S5 summarizes the time between MSx and Ea/p division in all conditions examined.

The data are available from the corresponding author upon reasonable request.

We thank members of the Goldstein lab, Dan Dickinson, Ed Munro, Dave Reiner, Martha Soto, and Jessica Sullivan-Brown for helpful feedback, discussion, and/or critical reading of the manuscript.

This work was supported by a Maximizing Investigators’ Research Award (R35GM134838 to B. Goldstein) from the NIH. T.N. Medwig-Kinney was supported by postdoctoral fellowships from the American Cancer Society (A24-0591-001) and the L’Oréal For Women in Science Program. E.A. Breiner was supported by a Summer Undergraduate Research Fellowship from the University of North Carolina Chapel Hill Office for Undergraduate Research. Some strains were provided by the CGC, which was funded by the NIH Office of Research Infrastructure Programs (P40 OD010440).

Author contributions: P. Zhang: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing, T.N. Medwig-Kinney: Formal analysis, Methodology, Supervision, Visualization, Writing - review & editing, E.A. Breiner: Formal analysis, Investigation, J.M. Perez: Formal analysis, A.N. Song: Formal analysis, Investigation, B. Goldstein: Funding acquisition, Supervision, Writing - review & editing.