Polarized Rac-dependent protrusions drive epithelial intercalation in the embryonic epidermis of C. elegans

Morphogenesis requires the concurrent rearrangement of many cell types during animal development. One such coordinated cell movement, mediolateral cell intercalation, is a ubiquitous process deployed in many developmental contexts to make a tissue longer and thinner (Walck-Shannon and Hardin, 2014). Cells derived from all three germ layers can undergo mediolateral intercalation, including epithelia. The intercalation of epithelial cells is especially complex because apicobasal polarity must be maintained while cells move along the mediolateral axis. Thus far, most research on epithelial cell intercalation has focused on RhoA-mediated shrinkage of apical junctions oriented along the axis of tissue shortening (Bertet et al., 2004; Blankenship et al., 2006; Chacon-Heszele et al., 2012; Karner et al., 2009; Lienkamp et al., 2012; Nishimura et al., 2012; Simões et al., 2014).

A less well understood mechanism of epithelial intercalation involves basolateral protrusive activity. Although basolateral protrusions have been described in several systems, including the sea urchin archenteron (Hardin, 1989), the ascidian notochordal primordium (Munro and Odell, 2002a,b) and the vertebrate neural tube (Williams et al., 2014), little information is available regarding how they form. Intriguingly, recent evidence suggests that a cooperative mechanism between apical junctional constriction and basal protrusions mediates cell intercalation in the mouse neuroepithelium (Williams et al., 2014). Together, these examples suggest that basolateral protrusive activity is an important but underappreciated mechanism of epithelial cell intercalation.

A promising context for exploring basolateral protrusive activity during epithelial cell rearrangement is the dorsal epidermis of the C. elegans embryo. Four hours after fertilization, 20 dorsal epidermal cells migrate past one another in the process of dorsal intercalation (Sulston et al., 1983). These intercalating cells are bona fide, apicobasally polarized epithelia, with intact apical junctions before, during and after intercalation (McMahon et al., 2001; Patel et al., 2008; Soto et al., 2002; Williams-Masson et al., 1998; E.W.-S. and J.H., unpublished). As cells become mediolaterally polarized they adopt a wedge shape; their medial tips extend contralaterally, while their lateral edges remain rounded. Eventually, the cell body and nucleus follow as an intercalating cell makes contact with contralateral seam epidermal cells, completing intercalation (Fig. 1A).

Dorsal intercalation is marked by the formation of medially directed protrusions that extend basal to the apical junction (Williams-Masson et al., 1998). These basolateral protrusions, and not apical junction rearrangement, appear to be the major driver of dorsal intercalation, as null mutants for cadherin complex components can intercalate successfully (E.W.-S. and J.H., unpublished; Costa et al., 1998). Four-dimensional microscopy indicates that basolateral protrusions in dorsal epidermal cells bear similarities to lamellipodia in other migrating cell types (Heid et al., 2001). Pharmacological studies suggest that actin filaments are required for intercalation (Williams-Masson et al., 1998). Beyond a basic role for actin, however, the molecular basis for medial tip extension is not well understood.

Actin networks can be formed through nucleation of existing filaments via the Arp2/3 complex, which must be activated by nucleation-promoting factors of the Wiskott-Aldrich syndrome protein (WASP) family, namely WASP and WAVE (Takenawa and Suetsugu, 2007). Rho family GTPases are the predominant regulators of such F-actin networks. In its GTP-bound state, Rac can activate the highly conserved, pentameric WAVE complex, while Cdc42 can activate WASP by relieving its autoinhibition (Kim et al., 2000; Miki et al., 1998; Patel et al., 2008). The RhoG ortholog mig-2 also genetically interacts with wsp-1 (Shakir et al., 2008), although it is unclear if the activation is direct. Rather surprisingly, the main Rac homolog expressed during early embryonic development in C. elegans, ced-10, is not reported to be required for intercalation (Patel et al., 2008; Soto et al., 2002). By contrast, mutations in components of the C. elegans WAVE complex, WVE-1/WAVE, GEX-2/Sra1 and GEX-3/Nap1, lack any significant epidermal motility, indicating that they are required for epidermal intercalation (Patel et al., 2008; Soto et al., 2002). Previous analyses of WSP-1/WASP function (Withee et al., 2004) have not identified a role for WSP-1 during dorsal intercalation, although no detailed analysis of motility was performed.

Here, we analyze in detail the formation of actin-rich dynamic protrusions in intercalating dorsal cells. We use a novel tissue-specific system to show that, rather than RhoA, the Rho family GTPases CED-10/Rac and MIG-2/RhoG redundantly control these protrusions upstream of the WAVE complex and WASP, respectively. Furthermore, we show that the guanine nucleotide exchange factor (GEF) UNC-73/Trio, which is known to activate both CED-10/Rac and MIG-2/RhoG, regulates intercalation. Additionally, we show that CRML-1/CARMIL normally polarizes protrusive activity upstream of UNC-73 by inhibiting motility at the lateral edges of intercalating cells. Our results represent the first detailed analysis of actin dynamics during dorsal intercalation, and identify a novel, phylogenetically conserved molecular cassette that regulates protrusion dynamics in intercalating cells that rely on basolateral protrusive activity.

To investigate F-actin dynamics in living embryos we utilized the actin-binding domain of a C. elegans spectraplakin, VAB-10, fused to GFP and specifically expressed in the epidermis (Gally et al., 2009) as a probe, focusing on embryos that had mosaically lost the transgene in one half of intercalating cells. We found numerous protrusions throughout intercalation (Fig. 1B). Initially, protrusions formed along the entire apicobasal axis of intercalating cells, but only persisted basolaterally (Fig. 1C). As intercalation proceeded, protrusions significantly decreased in number [12.3±0.7 early, 9.4±0.8 late (mean±s.e.m.); P<0.005, two-tailed Student's t-test] and became increasingly restricted to the medial edge (Fig. 1D). Together, these data indicate that dynamic F-actin-containing protrusions are directly correlated with migration of dorsal epidermal cells, and that these protrusions rapidly become polarized along the mediolateral axis.

Next, we investigated the functional importance of these protrusions during dorsal intercalation. Rac is the canonical Rho family GTPases responsible for tractive lamellipodia (Nobes and Hall, 1995). Given the lamellar appearance of the protrusions in dorsal epithelial cells, we reasoned that Rac would be an important regulator of their motility. The C. elegans genome encodes two Rac orthologs, CED-10 and RAC-2, and a distinct Rac- and Cdc42-like protein called MIG-2, an ortholog of mammalian RhoG and Drosophila Mtl (Lundquist, 2006). Given its predominant role in morphogenesis we focused first on CED-10/Rac.

To study CED-10 specifically in epidermal cells, we developed a novel tissue-specific, inducible, transgenic expression system. We expressed canonical dominant-negative (DN) T17N or constitutively active (CA) Q61L mutants of CED-10 (Bourne et al., 1991) driven by a variant of the lin-26 promoter that is expressed specifically in epidermal cells (see Materials and Methods) (Landmann et al., 2004). To regulate the expression of these transgenes, we exploited the endogenous nonsense-mediated mRNA decay (NMD) system, which identifies and degrades aberrant transcripts with a premature stop codon (Chang et al., 2007). The DNA encoding each fusion protein was fused to sequence encoding a long 3′UTR that contained an early stop codon, thus rendering the mRNA produced from these constructs sensitive to NMD. To make the system inducible, these tissue-specific, dominant, NMD-sensitive transgenes were put into a temperature-sensitive NMD background, smg-1(cc546ts) (Domeier et al., 2000). At the restrictive temperature (25°C) the transgenes are expressed, but at the permissive temperature (15°C) the transcripts are degraded, allowing inducible, tissue-specific protein expression (Fig. 2A, Fig. S1).

Conditional expression of CED-10(DN) in dorsal epidermal cells led to blunt medial edges in a subset of cells (29.4%, n=13), which were devoid of F-actin protrusions (Fig. 2D, Movie 1). This phenotype was also seen in the ced-10 null mutants tm597 (Shakir et al., 2006) and n3417 (Lundquist et al., 2001; Soto et al., 2002), but was harder to detect due to concurrent morphological defects (Fig. S2). Expression of CED-10(CA), on the other hand, led to rounded cells that were often unable to complete intercalation (Fig. 2D) and which exhibited excessive, unpolarized protrusions (Fig. 2B,C, Movie 2). These results suggest that CED-10 controls the intercalation of epidermal cells cell-autonomously. However, owing to the partial penetrance of these defects, we hypothesized that another Rho family GTPase functions redundantly with CED-10.

Embryonic RNAseq revealed no stage-appropriate expression of the other C. elegans Rac ortholog rac-2 (Celniker et al., 2009) (Fig. S3, supplementary materials and methods). Therefore, we focused on mig-2, which can act redundantly with ced-10 in other contexts (Lundquist et al., 2001). Examination of translational reporters of mig-2 and ced-10 confirmed that both proteins were present at the plasma membrane in intercalating cells (Fig. S3). Using an F-actin reporter in the dorsal epidermis, we observed that intercalating dorsal epidermal cells in mig-2(mu28) null mutants had significantly fewer protrusions than in wild type, whereas CA mig-2(gm103gf) mutants (Zipkin et al., 1997) had significantly more, unpolarized protrusions (Fig. 2B-D). These data indicate that, in addition to CED-10, MIG-2 also regulates protrusive activity during dorsal intercalation.

We next assessed whether mig-2 and ced-10 cooperatively regulate dorsal intercalation. Unfortunately, our attempts to generate mig-2; ced-10 loss-of-function animals carrying the F-actin reporter repeatedly failed. However, based on DIC microscopy, mig-2(mu28) mutants closely resembled wild type; however, mig-2(mu28) strongly enhanced intercalation defects in both an epidermal-specific CED-10(DN) (Fig. 3A) or a ced-10(n1993) reduced-function background (Fig. 3, Fig. S4). Moreover, these defects were more severe than in either mig-2 or ced-10 null mutants alone, suggesting that CED-10 and MIG-2 function in parallel to control intercalation. Taken together, our results indicate that the spatiotemporal regulation of F-actin protrusions is essential for proper dorsal intercalation, and that these protrusions are regulated by both CED-10/Rac and MIG-2/RhoG.

We posited that CED-10 and MIG-2 function during intercalation by interacting with effectors known to direct dendritic actin polymerization. To determine the relationship of WVE-1/WASP to CED-10/Rac and MIG-2/RhoG we performed pairwise epistasis tests. The strong wsp-1 loss-of-function allele gm324 enhanced intercalation defects in ced-10(e1993) embryos but not mig-2(mu28) embryos (Fig. S5), suggesting that wsp-1 functions in parallel to ced-10 but in the same pathway as mig-2 during intercalation. Consistent with this interpretation, wsp-1(gm324) was unable to suppress epidermal-specific CED-10(CA) (Fig. 4B), whereas it significantly suppressed intercalation defects of embryos heterozygous for mig-2(gm103gf) (Fig. 4A). During intercalation, mig-2(gm103gf) acts dominantly [heterozygotes are not significantly different from homozygotes but both are significantly different from wild type (P<0.0001, Student's t-test); Fig. 4A].

WAVE is encoded by wve-1 in C. elegans. There is strong evidence in C. elegans that null mutations in wve-1/WAVE, along with other components of the WAVE complex, namely gex-2/Sra1 and gex-3/Nap1, eliminate epithelial morphogenesis (Patel et al., 2008; Soto et al., 2002). Indeed, we find that dorsal epidermal cells in wve-1(ne350) or gex-2(ok1603) embryos show negligible motility (Fig. S6). To investigate the role of wve-1 and its genetic interactions during intercalation more informatively, we sought to achieve weak loss of wve-1 function. To do so, we performed RNAi, which as described previously yielded partially penetrant wve-1 knockdown (Patel et al., 2008) and intercalation failure (19.3%, n=31). The proportion of embryos that failed to intercalate was enhanced in the mig-2(mu28) background (Fig. S6), suggesting that MIG-2 and WVE-1 are in separate pathways. However, wve-1(RNAi) also enhanced defects in the ced-10(n1993) background. Based on previous studies in C. elegans growth cones (Shakir et al., 2008), we hypothesize that weak perturbation of two components in the same pathway leads to stronger loss of that pathway's function during intercalation. Providing further evidence for this interpretation, we found that wve-1(RNAi) could suppress epidermal-specific, CED-10(CA) (Fig. 4B) but not overactive mig-2(gm103gf) (Fig. 4A, Fig. S7).

In summary, these epistasis tests suggest that, as in neuronal growth cones in C. elegans (Shakir et al., 2008), the two Rho family members CED-10/Rac and MIG-2/RhoG function redundantly and act via different downstream effectors: MIG-2/RhoG acts upstream of WSP-1/WASP and CED-10/Rac acts upstream of WVE-1/WAVE (Fig. 4E). As expected given this ordering of gene function, we found that wsp-1(gm324) enhances both the intercalation failure (Fig. S6) and F-actin defects in wve-1(RNAi) (Fig. 4C,D).

To narrow down the candidate list of C. elegans GEFs that activate Rac/RhoG during dorsal intercalation, we mined the literature on GEFs for high-throughput expression data and GTPase exchange specificities (Fig. S8). We predicted that the C. elegans Trio ortholog, UNC-73, was likely to activate Rac/RhoG during dorsal intercalation. UNC-73 is part of a unique family of proteins that contain two separate GEF domains with different GTPase specificities. Each GEF domain is defined by sequential Dbl homology (DH) and Plekstrin homology (PH) domains. The first GEF domain, GEF1, activates CED-10/Rac and MIG-2/RhoG, whereas the second GEF domain, GEF2, activates RHO-1/RhoA (Fig. 5A) (Bellanger et al., 1998; Blangy et al., 2000; Debant et al., 1996; Kubiseski et al., 2003; Steven et al., 1998; Wu et al., 2002).

Domain-specific lesions in unc-73 allowed us to take a structure/function approach to ask which GEF domains in UNC-73 are required for intercalation. gm40 is a premature stop codon that results in a protein truncated before both GEF domains; however, alternative promoters within unc-73 allow multiple GEF2 isoforms to be expressed (Steven et al., 2005). rh40 is a missense mutation within the GEF1 domain that decreases its ability to activate CED-10 and MIG-2 (Kubiseski et al., 2003). ev802 is a deletion of the GEF2 domain, which can be rescued by unc-73 DNA with the GEF1 domain deleted, suggesting that it is a GEF2-specific mutant (Steven et al., 2005). As expected, mutations disrupting GEF1 (gm40 and rh40) resulted in significantly longer intercalation times than in wild type due to blunted medial edges, similar to CED-10(DN) (Fig. 5B,C). After normalizing for overall slower development, intercalation time in the GEF2 mutant ev802 was not significantly different than similarly adjusted wild-type controls. As predicted, dorsal cell protrusions were dramatically decreased in unc-73(gm40) mutants (Fig. 5D, Movie 3). The requirement for the GEF1 domain of UNC-73 indicates that the main function of UNC-73 is to act as an upstream regulator of Rac/RhoG-dependent protrusive activity during dorsal intercalation. Previous biochemical studies confirm that the GEF1 domain of UNC-73 can activate both MIG-2 and CED-10, consistent with this conclusion (Kubiseski et al., 2003).

We expected UNC-73B::GFP, a GEF1-specific UNC-73 fusion protein (Steven et al., 2005), to be enriched at protrusively active medial edges of intercalating cells. Instead, we found that during dorsal intercalation UNC-73B::GFP was localized along the entire dorsal cell membrane, even at edges that are not protrusively active (Fig. 5E). Because the protrusions themselves are obviously polarized (Fig. 1D) but the localization of an upstream regulator, UNC-73, is not, we sought to identify proteins that might contribute to the spatial restriction of UNC-73 activity at or near the leading edge. We took a candidate gene approach to determine if known regulators of UNC-73, Trio or the shorter vertebrate isoform Kalirin have intercalation phenotypes, and found that crml-1(n1962) mutants displayed significantly increased intercalation time (Fig. S9A). crml-1 is a member of the conserved CARMIL (capping Arp2/3 myosin I linker) gene family first discovered in Dictyostelium (Jung et al., 2001). CARMILs consist of a non-canonical plekstrin homology (PH) domain, leucine-rich repeats (LRR), a capping protein-binding domain (CPBD) and a proline-rich region (PRR) (Zwolak et al., 2013) (Fig. 6A). In vertebrates, CARMIL proteins can allosterically inhibit the ability of capping protein to bind the barbed ends of F-actin, thereby allowing more actin polymerization and hence protrusive activity (Fujiwara et al., 2014; Kim et al., 2012; Takeda et al., 2010; Yang et al., 2005). However, based on co-immunoprecipitation analysis, human CARMIL1 (LRRC16A) (Liang et al., 2009) and C. elegans CRML-1 (Vanderzalm et al., 2009) are also part of a complex with Trio and UNC-73, respectively. The significance of this interaction is largely uncharacterized. It has been suggested that CRML-1 functions upstream of UNC-73 to inhibit its function, as crml-1 mutations fail to suppress unc-73 mutations in neurons (Vanderzalm et al., 2009). Consistent with the hypothesis that CRML-1/CARMIL inhibits protrusions during intercalation, we find that nonsense mutations in crml-1, gm326 and n1962, or crml-1(RNAi) result in excessive, unpolarized F-actin protrusions in dorsal epidermal cells that are inversely related to intercalating cell velocity (Fig. 6B, Movie 4). Moreover, the extra, nonpolarized protrusions in crml-1(gm326) embryos are suppressed by the unc-73(rh40) mutation (Fig. 6). Together, these data suggest that UNC-73/Trio functions downstream of a negative regulator, CRML-1/CARMIL, to inhibit protrusive activity during dorsal intercalation.

We predicted that CRML-1 might spatially restrict UNC-73 activity at the lateral (rear) edges of dorsal epidermal cells, where protrusions are usually absent, thus confining UNC-73 activity to the medial, extending tip. To examine the spatial localization of CRML-1, we made an epidermal-specific CRML-1::GFP fusion protein (using the lbp-1 promoter; Plenefisch et al., 2000). This construct rescued the excessive F-actin protrusions in dorsal epidermal cells of crml-1(gm326) mutants (Fig. S9), indicating that the construct is functional and specifically required in the epidermis. Moreover, during tip extension we found that CRML-1::GFP consistently accumulated in lateral regions of dorsal epidermal cells (compared with broadly localized PH::mCherry controls) (Fig. 7B), where protrusions are normally absent. Later during intercalation we also observed transient expression of CRML-1::GFP in the advancing tips after cells reached the contralateral side (n=9/10; white arrowheads, Fig. 7A). This late accumulation after protrusions cease further supports the interpretation that CRML-1 inhibits protrusive activity. Together, these data suggest that polarized CRML-1 expression at the rear of intercalating cells spatially inhibits UNC-73 and dampens downstream protrusive activity.

Our work defines a new molecular network that promotes epithelial cell intercalation, driven by Rac/RhoG. CED-10/Rac and MIG-2/RhoG function redundantly downstream of activation by the GEF1 domain of UNC-73/Trio to drive dendritic actin polymerization. Our data suggest that CED-10/Rac signals through the WAVE complex, whereas MIG-2/RhoG signals through WSP-1. Moreover, protrusive activity is polarized through the negative UNC-73/Trio regulator CRML-1/CARMIL, which accumulates at the lateral edges of dorsal epidermal cells. Together, these data suggest that cell intercalation in the dorsal epidermis of C. elegans proceeds via protrusions promoted by a Rac/RhoG-dependent pathway that is tightly regulated, both temporally and spatially (Fig. 8). These protrusions provide the major driving force for intercalation. When they are nearly completely abrogated, as in wsp-1(gm324); wve-1(RNAi) or wve-1(ne350) embryos, intercalation fails (Fig. 4C,D, Fig. S6). Moreover, intercalation requires that these protrusions are polarized, as shown in CED-10(CA) (Fig. 2) or crml-1 mutants (Fig. 6), in which unpolarized protrusions result in dorsal intercalation failure (Fig. 4B) or marked delay (Fig. S9A), respectively. Together, these data suggest that polarized protrusions are crucial for the directional translocation of epithelial cells during dorsal intercalation in C. elegans.

Most of the existing work on epithelial intercalation has focused on the Rho-driven shrinkage of junctions oriented specifically in the axis of tissue shortening (Levayer et al., 2011; Nishimura et al., 2012; Simões et al., 2014; Warrington et al., 2013). In the C. elegans dorsal epidermis, by contrast, zygotic rho-1(ok2418) mutants or epidermal-specific RHO-1(T17N) embryos can still intercalate, even though cytokinesis has failed and the cells are binucleate in the latter (E.W.-S. and J.H., unpublished). This is, to our knowledge, the first study to investigate Rac and RhoG during epithelial intercalation. Previous reports suggest that basolateral protrusions exist in other epithelial tissues as well (Hardin, 1989; Munro and Odell, 2002b; Williams et al., 2014). Therefore, Rac- and/or RhoG-driven basolateral mechanisms of epithelial intercalation might be more widespread than previously appreciated.

Racs are known to act via downstream effectors that stimulate Arp2/3-dependent dendritic actin networks (Heasman and Ridley, 2008). Previously, there had been some ambiguity about whether actin polymerization was required for dorsal intercalation (Sawa, 2003; Soto et al., 2002). Here, we provide several lines of evidence that the WAVE complex and WASP are required for dorsal intercalation downstream of CED-10 and MIG-2, respectively. First, wsp-1(gm324) can both enhance ced-10(n1993) mutants and suppress mig-2(gm103gf) mutants. Second, wve-1(RNAi) can enhance mig-2(mu28) mutants and suppress CED-10(CA) specifically expressed in the epidermis. Third, wsp-1(gm324) can enhance the intercalation failure of wve-1(RNAi). However, given the relatively low penetrance of intercalation defects in ced-10/Rac null mutants compared with wve-1/WAVE null mutants, which display complete intercalation failure, additional upstream activators are likely to regulate the WAVE complex. Recent evidence suggests that within the WAVE complex there is a second, Rac-independent interaction site, which is conserved in C. elegans, that can be activated by a diverse set of receptors (Chen et al., 2014). Future studies focusing on whether this second interaction site is required for the WAVE complex during dorsal intercalation would likely prove informative.

Within intercalating cells, the asymmetric localization of molecules orthogonal to or along the direction of movement is a common strategy to polarize cells during intercalation (reviewed by Bertet and Lecuit, 2009; Gray et al., 2011; Walck-Shannon and Hardin, 2014; Wallingford, 2012; Zallen, 2007). During intercalation of the Drosophila germband, for example, actin, nonmuscle myosin II and its regulators accumulate at shrinking junctions, whereas cadherin is enriched at junctions that persist during intercalation (Blankenship et al., 2006; Levayer et al., 2011; Simões et al., 2010; Zallen and Wieschaus, 2004). Asymmetric localization of proteins in epithelial cells that use basolateral protrusive activity during rearrangement has not previously been examined. We found that a negative regulator of UNC-73/Trio, CRML-1/CARMIL, predominantly localizes to lateral edges, where protrusions are normally absent. We hypothesize that CRML-1 polarization contributes to the directional movement of intercalating cells by preventing non-directional (lateral) protrusions. At a genetic level, crml-1 regulates the activity of unc-73, rather than its expression in dorsal epidermal cells, because UNC-73::GFP localization in the crml-1(gm326) background is indistinguishable from that in wild type (Fig. S10). It remains unclear how CRML-1 becomes spatially restricted in its localization. Repeated attempts to disrupt dorsal intercalation by perturbing the function of conserved homologs of planar cell polarity components have failed to detect defects (King et al., 2009; E.W.-S. and J.H., data not shown); it is therefore likely that additional proteins are involved. CRML-1::GFP also accumulated at the tips of cells as intercalation concluded. This separate pool of CRML-1 might represent highly localized areas of UNC-73 inhibition at the migrating tips as they make contact with contralateral seam cells; alternatively, this pool could have a second, separate function that relates to capping protein at medial tips.

There are multiple isoforms of human CARMIL. Most effort has focused on CARMIL1, which can form a complex with Trio and promotes protrusive activity (Liang et al., 2009). The role of CARMILs in vertebrates has focused almost entirely on their inhibition of capping protein (Edwards et al., 2014). However, C. elegans CRML-1 seems to inhibit protrusive activity (this work; Vanderzalm et al., 2009) rather than promote it. The opposite effect would be expected if CRML-1 acts primarily via anti-capping protein activity. Here we show that CRML-1 has a crucial role in regulating F-actin dynamics through the inhibition of the RacGEF (GEF1) activity of UNC-73. In this regard, crml-1 loss of function more closely mimics CARMIL2 mutant phenotypes in vertebrate cells, which include loss of cell polarity and multiple lamellipodia (Liang et al., 2009). It is possible that C. elegans CRML-1 retains functions of both CARMIL1 (Trio binding/inhibition) and CARMIL2 (regulation of cell polarity/capping protein activity), and that these are each deployed in different contexts in C. elegans. CARMILs might have undergone subfunctionalization in vertebrates for separate, more specialized functions.

Our study unites previous observations (Patel et al., 2008; Williams-Masson et al., 1998) into a concise model for dorsal intercalation in C. elegans, in which Rac/RhoG control of dendritic actin polymerization is the major driver of medial protrusions, and hence cell rearrangement. This observation also raises the intriguing possibility that other epithelia use both mechanisms to varying degrees. Asymmetric Rho-mediated constriction of apical junctions has motivated most study of intercalation within the Drosophila germband, but embryos that carry mutations for eve and Toll-like receptors, which pattern these molecular asymmetries, still have germbands that extend 1.6× their original length, compared with 2× extension in wild type (Irvine and Wieschaus, 1994; Paré et al., 2014). Although oriented cell divisions may contribute to the remaining tissue extension (da Silva and Vincent, 2007), it is also possible that basolateral protrusions contribute to convergent extension in the Drosophila germband but have escaped detection thus far. Further evidence that the Rac/RhoG-dependent mechanism described here may constitute an additional general mechanism for driving intercalation in epithelia is suggested by studies of the mouse neural tube. In this neuroepithelium, the planar cell polarity pathway directs both apical junctional rearrangement and basolateral protrusions to achieve mediolateral intercalation (Williams et al., 2014). These results support the idea that apical and basolateral pathways during intercalation are not mutually exclusive. Instead, different epithelial cell types might utilize each pathway to varying extents. Understanding the extent to which various epithelial cells use one program or the other and the reasons why various epithelia favor one pathway over another is an important question for future research.

C. elegans were maintained on OP50 bacteria as previously described (Brenner, 1974). The wild-type strain was Bristol N2. Experiments were performed at 20°C unless otherwise stated. The following genetic lesions were utilized in this study. LG1: smg-1(cc546ts), unc-73(ev802), unc-73(gm40), unc-73(rh40), crml-1(gm326). LGIV: ced-10(n1993), ced-10(n3417), ced-10(tm597), wsp-1(gm324). LGX: mig-2(gm103gf), mig-2(mu28). The following transgenic arrays were made for, or utilized in, this study: mcEx227[P_lin-26::VAB-10(actin binding domain)::GFP, rol-6(su1006)] (Gally et al., 2009), jcEx222[P_lin-26::ced-10(T17N/DN)::smg sensitive 3′UTR, sur-5::mCherry], jcEx224[P_unc-73::UNC-73::GFP), rol-6(su1006)], reIs6[P_lin-26::ced-10(Q61L/CA)::smg sensitive 3′UTR, rol-6(su1006), P_myo-3::gfp)], ltIs44[P_pie-1::PH_PLC1∂1::mCherry, unc-119(+)] (Kachur et al., 2008). Additional strains are described in the supplementary materials and methods.

We imaged live embryos mosaically expressing a previously validated F-actin reporter, mcEx227[P_lin-26::VAB-10(actin binding domain)::GFP, rol-6(su1006)] (Gally et al., 2009), in the dorsal epidermis with a Nikon Eclipse E600 microscope connected to a Yokogawa CSU10 spinning disk scanhead. Images were gathered with a Hamamatsu ORCA-ER charge-coupled device (CCD) camera and Micromanager software (Edelstein et al., 2010, 2014) with a z-slice spacing of 0.4 μm. ImageJ (Schneider et al., 2012) was used to project 20-25 z-slices, which is shown in dorsal view images. Reslices were also performed in ImageJ. mcEx227 was crossed into all other genetic backgrounds to maintain similar expression levels.

Protrusions were quantified from maximum projections of z-stacks of F-actin reporter expression as described above. Relative protrusion area was calculated as the ratio between a cell's area in maximum projection images (which includes all protrusions) to a cell's lateral area (measured at the depth of the cell nucleus). Protrusion number was obtained by counting aggregations of F-actin reporter signal extending at least 0.5 μm from the cell body. Protrusion position was recorded using the ROI tool in Image J. Angles were obtained using trigonometric functions in Microsoft Excel to compare protrusion position relative to the cell centroid (also calculated in ImageJ). At least 15 cells from at least 10 embryos were analyzed per genotype. JMP software (SAS) was used for statistics and graphing of protrusion area and number; PAST was used for angular statistics and graphs (Hammer et al., 2001). We designated cells as ‘early’ if they were in the process of polarizing into a wedge shape and ‘late' if the cell tip or body had begun to extend. Cell velocity was calculated by measuring medial edge displacement over time. Linear regression and statistical analyses were carried out in JMP.

Four-dimensional (4D) DIC movies were gathered on either a Nikon Optiphot-2 connected to a QImaging camera or Olympus BX50 connected to a Scion camera. ImageJ plugins (available at http://worms.zoology.wisc.edu/research/4d/4d.html) were used to analyze movies. For this study, we defined intercalation time as the time between terminal epidermal (Cpaa.a/p) divisions and when contralateral nuclei met at the midline after cell migration was complete. For unc-73 mutants, intercalation time was measured relative to the time between the Cpaa and Cpaa(a/p) divisions. At least 20 embryos from at least 10 mounts were analyzed per genotype. JMP was used to create box plots and for statistical analysis.

The epidermal-specific, NMD-sensitive plasmid backbone was made as follows. (1) The hybrid eFGHi lin-26 promoter (including minimal myo-2 promoter) is expressed only in the epidermis and support cells (Landmann et al., 2004): P_lin-26(eFGHi) was amplified from pML433 with primers (5′-3′; lowercase letters are restriction sites) TTTTTTgcggccgcGGCGCGCCCGAGGTTAATATCTGAGCTCC and TTTTTTggatccTggccggccACTCATTTTTTCTGAGCTCGGTACCCTCC and cloned into pBluescript cut with NotI/BamHI (pCM1.1). (2) The NMD-sensitive 3′UTR sequence (inverted let-858 coding sequence without a sense polyadenylation sequence) was amplified from pPD118.44 (gift of A. Fire, Stanford University, CA, USA) using primers TTTTatcgatcctaaTCGTCGAGTCGGTCACAATCACC and TTTTgggcccGCTCATGTTTAGATTTGGATTG and cloned into pCM1.1 digested with ClaI/ApaI (pCM1.3). pCM1.1 plasmid for this step was grown in SCS110 bacteria because ClaI is Dam sensitive. The wild-type ced-10 sequence was amplified from pPR37 (Reddien and Horvitz, 2000) using primers AAAAAAAAggccggcctggcATGCAAGCGATCAAATGTGTCGTCG and TTTTTTatcgatTTAGAGCACCGTACACTTGCTCTTTTTGG and cloned into pCM1.3 using FseI/ClaI to generate pCM3.2. To make ced-10(Q61L/CA) (pCM3.3) and ced-10(T17N/DN) (pEWS27) variants, we used PCR site-directed mutagenesis, with pCM3.2 as template and the following primers (forward and reverse; underlining indicates bases mutated relative to wild type): ced-10(Q61L/CA), GGGATACAGCTGGACTGGAAGATTACGATCGAC and GTCGATCGTAATCTTCCAGTCCAGCTGTATCCC; ced-10(T17N/DN), ACTGTCTCCTGATATCCTACAC and TTTTACCGACGGCTCCGTCA.

Gonads of wild-type animals were microinjected with these constructs at 40 ng/μl and a co-injection marker (Mello and Fire, 1995). Resulting lines in the wild-type background were screened for defects using smg-1 feeding RNAi (Kamath et al., 2003). Representative lines were crossed into the smg-1(cc546ts) background, which can be detected by PCR using forward CAGTCGTGAGCTTTGGATGCGTGC and reverse TCGGGGATACGCAGATTCTTTCCC followed by digestion specifically of wild-type product using MslI. At least three lines were analyzed per construct. The resulting extrachromosomal array jcEx222[P_lin-26::ced-10(T17N/DN)::smg sensitive 3′UTR, sur-5::mCherry] was used for experiments. ced-10(Q61L/CA) was integrated into the genome using UV irradiation, resulting in reIs6[P_lin-26::ced-10(Q61L/CA)::smg sensitive 3′UTR, rol-6(su1006), Pmyo-3::gfp)]. Lines were maintained at 15°C and heat shocked at 25°C for 24 h to induce transgene expression prior to filming (unless otherwise noted). Crosses were performed at 15°C. Filming was at 20°C. For GFP induction using the NMD system, see the supplementary materials and methods.

T7 and T3 sites were added to full-length wve-1 cDNA derived from wild-type worms by PCR with primers forward TAATACGATCACTATAGGTCCATCAACAATCCATCGTG and reverse CCCTTTAGTGAGGGTTAATTCCCATTCATCATCATCAGC. From the PCR product, MegaScript kits (Ambion) were used to synthesize double-stranded wve-1 RNA, which was injected (Zipperlen et al., 2001) into the psuedocoelom at 0.5 μg/μl. As described previously for feeding RNAi, injection RNAi elicited partial knockdown of wve-1 (Patel et al., 2008). For ced-10(Q61L/CA) suppression, animals were incubated at a semi-permissive temperature of 18°C for 6 h prior to mounting. For mig-2(gm103/CA) suppression, animals were incubated at 20°C for 16 h. For vav-1 RNA, see the supplementary materials and methods.

A rescuing unc-73 GEF1 construct, pNW248 (Punc-73::unc-73B::GFP) (Steven et al., 2005), was microinjected into the gonads of wild-type animals (Mello and Fire, 1995) at 10 ng/μl to create the extrachromosomal array jcEx224[P_unc-73::UNC-73::GFP), rol-6(su1006)].

lbp-1 is expressed predominantly in the epidermis at embryonic stages (Plenefisch et al., 2000). crml-1 cDNA was cloned into an existing P_lbp-1::GFP vector, pSL500 (Fridolfsson and Starr, 2010), using Gibson assembly (Gibson, 2011). The following primers were used (forward and reverse; uppercase letters indicate regions complementary to template and lowercase letters indicate overhangs complementary to vector): pSL500, ATGAGTAAAGGAGAAGAACTTTTCAC and CATGGTTGGCTAGGTCGATGAATG; crml-1 cDNA, tcgacctagccaaccatgTCCATGTCGAGGAGCACC and agttcttctcctttactcatTTTTTGGAAAATTCTAGCCATATCAGC. The resulting construct (pEWS38) was injected into crml-1(gm326) mutants at 10 ng/μl. crml-1(gm326) can be detected using PJV98/PJV99 (Vanderzalm et al., 2009) and AccI digestion. Three independently derived lines were analyzed. Intensity measurements of CRML-1::GFP were performed using the Plot Profile tool with congruent rectangular selections outlining each dorsal cell in ImageJ. Cells came from at least five different embryos.

Some strains were provided by the CGC, which is funded by the NIH Office of Research Infrastructure Programs [P40 OD010440]. Bethany Lucas created the wve-1 cDNA clone used for RNAi. We thank the laboratories of E. Lundquist and G. Garriga for graciously providing many worm strains used in this study and R. Steven for providing pNW248 DNA. The SAGE data were produced at the Michael Smith Genome Sciences Centre with funding from Genome Canada; the embryonic RNA sequencing data were produced by the Waterston laboratory at the University of Washington. We are grateful to members of the J.H. laboratory for helpful input.