Unique and redundant β-catenin regulatory roles of two Dishevelled paralogs during C. elegans asymmetric cell division

Multicellular organisms use a variety of mechanisms and strategies to control gene activation in developmental and homeostatic processes. One widely used scheme is the Wnt/β-catenin signaling pathway, which controls the stability of the transcriptional activator β-catenin and its interaction with the transcription factor TCF. Wnt/β-catenin signaling in mammals is used to control processes as diverse as axial specification, heart development and adult stem cell proliferation (Clevers and Nusse, 2012; Gessert and Kühl, 2010; Hikasa and Sokol, 2013; Houston, 2012; McMahon and Moon, 1989). Although many core and auxiliary proteins involved in Wnt/β-catenin signaling have been identified, the mechanism and significance of the interactions involved in key steps of the Wnt signal cascade are lacking. Additionally, although the identities of the major players of Wnt/β-catenin signaling are widely accepted, how those players interact in vivo to transduce the Wnt signal remains a major unresolved question among and within various tissues and organisms.

Our current understanding of Wnt/β-catenin signaling is that in the absence of the extracellular Wnt signal, cytoplasmic β-catenin is bound, phosphorylated and ubiquitylated by a group of proteins known as the ‘destruction complex’ (Clevers and Nusse, 2012). The canonical destruction complex includes the scaffold proteins APC and Axin and the kinases casein kinase 1α (CKIα, also known as CSNK1A1) and glycogen synthase kinase 3β (GSK3β). TCF proteins bind to regulatory regions of target genes, but in the absence of nuclear β-catenin interact with corepressors and downregulate gene expression (Roose et al., 1998). When a Frizzled (Fz) receptor is bound by the extracellular Wnt ligand, a signaling cascade commences that results in the inactivation of the destruction complex and stabilization of cytoplasmic β-catenin. Increased cytoplasmic levels of β-catenin result in higher nuclear levels of β-catenin, and the interaction between β-catenin and TCF in the nucleus results in the activation of Wnt target genes. Mutations of β-catenin and/or destruction complex members that result in inappropriate β-catenin stabilization lead to constitutive activation of Wnt target genes; in stem cell populations, such as those of the human intestinal lining or breast, constitutive activation of Wnt targets can result in tumor formation and proliferation (Barker et al., 2009; Groden et al., 1991; Khramtsov et al., 2010; Webster et al., 2000). Thus, understanding the mechanisms that control Wnt/β-catenin signal transduction and destruction complex activation have great implications for human health in addition to informing our knowledge of development and stem cell biology.

In Caenorhabditis elegans, most asymmetric cell divisions (ACDs) are controlled by a Wnt pathway that uses β-catenin–TCF interactions to differentially control transcription of target genes in the two daughter cells after ACD. This Wnt/β-catenin asymmetry (WβA) signaling pathway can be most simply described as an adaptation of the canonical Wnt/β-catenin signaling pathway within the context of an asymmetric cell division; a mother cell that has polarized localization of conserved Wnt signaling components divides to produce daughter cells with asymmetric inheritance of Wnt signaling components, which ultimately translates to asymmetric activation of Wnt target genes (Sawa, 2012). The pathway contains two branches: one that more closely resembles the canonical pathway, which controls the stability of SYS-1 (one of the C. elegans β-catenins), and another non-canonical pathway where the nuclear localization of POP-1 (a C. elegans TCF) is regulated by another β-catenin, WRM-1 (Fig. S1) (Phillips and Kimble, 2009). The result is that the anterior daughter nucleus contains low levels of SYS-1 and WRM-1 leading to target gene repression, whereas the posterior daughter nucleus contains high levels of SYS-1 and WRM-1, leading to target gene activation.

Despite some clear mechanistic departures from canonical Wnt/β-catenin signaling, specifically the regulation of nuclear TCF, WβA holds much in common with the canonical pathway. Asymmetric SYS-1 levels are controlled by the conserved destruction complex members APR-1, PRY-1 and KIN-19 (the Caenorhabditis elegans APC, Axin and CKIα, respectively), and the proteasome, indicating that its stability is likely to be regulated similarly to canonical β-catenin, but adapted for ACD (Baldwin and Phillips, 2014; Huang et al., 2007; Phillips et al., 2007; Vora and Phillips, 2015). WRM-1 localization is also controlled by APR-1, PRY-1 and KIN-19, with the difference being that nuclear WRM-1 levels are controlled by protein trafficking. This trafficking is rendered asymmetric by the asymmetric localization of APR-1, which stabilizes microtubules in a conserved fashion (a summary of prior models of WβA signaling can be found in Fig. S1) (Baldwin and Phillips, 2014; Mizumoto and Sawa, 2007; Rocheleau et al., 1999; Sugioka and Sawa, 2012; Takeshita and Sawa, 2005). Finally, both pathways have similar unanswered questions surrounding the transduction of the Wnt signal: in both pathways it is clear that cells are responding to extracellular Wnt cues and transducing those cues into inactivation of the β-catenin destruction complex, but how this transduction occurs remains unresolved in both pathways.

In the canonical Wnt/β-catenin pathway, activation of Wnt signaling results in the Dishevelled (Dvl) scaffold protein binding to a Fz receptor (Wong et al., 2003). Dvl then binds and sequesters Axin, inactivating β-catenin degradation (Cliffe et al., 2003; Fiedler et al., 2011; Itoh et al., 2000). How exactly the destruction complex is inactivated remains controversial, with several models proposing inactivation of β-catenin phosphorylation mediating destruction complex inactivation or dissociation, whereas a recent model proposes that the complex remains intact, although inactivated, during signal transduction (Lee et al., 2003; Li et al., 2012; Mao et al., 2001; Taelman et al., 2010). In the hypodermal seam cells of C. elegans, several Wnt ligands and receptors control the orientation of polarity of WβA activation (Yamamoto et al., 2011). Additionally, the C. elegans Dvl paralogs DSH-2 and MIG-5 have been shown to be involved in WβA signaling, but whether and how they polarize destruction complex inactivation is unclear (Chang et al., 2005; King et al., 2009; Phillips et al., 2007; Sugioka et al., 2011). Thus, given the similar unresolved questions in both canonical and WβA signaling, investigating the role of Dvl in signal transduction in WβA might also provide insights into canonical Wnt/β-catenin signaling activation.

Previously, we have demonstrated that WβA signaling appears to subdivide APR-1 into two functional pools during asymmetric cell division: one pool that regulates SYS-1 localization and one pool that regulates WRM-1 localization (Baldwin and Phillips, 2014). Here, we show that the Dvl proteins DSH-2 and MIG-5 redundantly control cell fate specification during ACD. Additionally, we show that DSH-2 and MIG-5 are both necessary to properly regulate SYS-1 and WRM-1 localization, but that their functions in regulating SYS-1, WRM-1 and APR-1 localization are each specialized. We also demonstrate that neither DSH-2 nor MIG-5 control PRY-1 localization, suggesting that the ultimate target of Dvl regulation in WβA is APR-1. Finally, we show that overexpression of MIG-5 and DSH-2 results in recruitment of SYS-1 into Dvl-containing puncta and SYS-1 stabilization in anterior (unsignaled) daughter cells. In total, we show that DSH-2 and MIG-5 have both redundant and non-redundant roles in the control of β-catenin localization in C. elegans ACDs. This result demonstrates that Dvl paralogs can cooperate to fulfill conserved roles in Wnt/β-catenin signaling while also having individualized functions, providing insight into roles multiple Dvl proteins can have in the Wnt signal cascade.

WβA signaling has been shown to control cell fate in the asymmetric cell divisions of the C. elegans hypodermal seam cells (Banerjee et al., 2010; Gleason and Eisenmann, 2010; Gorrepati et al., 2013; Hughes et al., 2013; Ren and Zhang, 2010; Yamamoto et al., 2011). The C. elegans seam is composed of two lines of hypodermal cells, one each on the left and right sides of the worm. The seam cells are essential for expression of various collagens that assemble the outer cuticle of the worm (Liu et al., 1995; Thein et al., 2003) and also generate the cuticle alae structure during L1 and adult stages (Costa et al., 1997; Singh and Sulston, 1978). Over the course of larval development the seam cells divide several times, generally in an asymmetric pattern that produces an anterior daughter that fuses to the hypodermal syncytium hyp7, and a posterior daughter that retains the seam fate and positioning (Sulston and Horvitz, 1977). This ‘stem-cell-like’ mode of division allows the seam cells to repeatedly contribute additional nuclei to the growing hyp7 syncytium while maintaining a consistent seam cell number and structure (Fig. S2).

In order to determine the roles of DSH-2 and MIG-5 in WβA signaling, we investigated the function of each gene in determining seam cell fate. Although WβA controls many asymmetric divisions in C. elegans, the divisions of the seam cells are easily observable and study of their divisions might provide insights into stem cell biology as well as Wnt/β-catenin signaling. Previous experiments using RNA interference (RNAi) have shown that depletion of dsh-2 or mig-5 does not have any effect on seam cell number (Banerjee et al., 2010). Thus, we utilized the deletion mutants dsh-2(or302) and mig-5(tm2639), hereafter referred to as dsh-2 and mig-5, respectively. Both mutations are sizeable insertions or deletions that also result in frameshifts, likely causing complete loss-of-function (Hawkins et al., 2005). When crossed with wIs51 [expressing GFP under a seam cell promoter (P_scm::GFP)], both dsh-2 and mig-5 resulted in statistically significant alterations in seam cell number (P<0.01, t-test; Fig. 1). Interestingly, both mutants displayed both gains and losses of seam cells when compared to wild type. Gains of seam cells are associated with loss of function in negative regulators of Wnt signaling, such as APR-1 or KIN-19 (Banerjee et al., 2010; Gleason and Eisenmann, 2010; Ren and Zhang, 2010). Conversely, losses of seam cells are associated with loss of function of positive regulators of Wnt target gene activation, such as WRM-1 or LIT-1. Treatments that result in both gains and losses might represent disruption of the machinery that polarizes the WβA function (Yamamoto et al., 2011), resulting in randomized ACD or symmetric division. Thus, as dsh-2 and mig-5 result in both gains and losses of seam cells, they are likely involved in polarizing the seam cell fate decision. However, the dsh-2 mig-5 double mutant displays a significant increase in seam cell number compared to either single mutant and appears to abrogate most losses of seam cells, indicating that in sum the two Dvl paralogs function together as negative regulators of seam cell fate (Fig. 1C, see also Fig. 6 and the Discussion for further interpretation).

A third C. elegans homolog of Dvl, DSH-1, has not been directly implicated in WβA signaling. Previous work has shown that knockdown of DSH-1 results in wild-type numbers of seam cells, although mild suppression of seam cell hyperplasia has been observed in a dsh-1, kin-19 double RNAi experiment (Banerjee et al., 2010). To more definitively resolve a role for DSH-1 in seam cell specification, we quantified seam cell numbers using the dsh-1 deletion allele (ok1445) (Table S1). We did not find any significant changes in seam cell number in dsh-1(ok1445) compared to wild-type worms (Table S1). However, it remained possible that DSH-1 function in the seam cells could be redundant with DSH-2 and MIG-5, so we combined dsh-1(RNAi) with the dsh-2 mig-5 double mutant. The dsh-2 mig-5, dsh-1(RNAi) treatment did not result in any changes in seam cell number compared to dsh-2 mig-5 (Table S1). We thus conclude that DSH-1 is not a major contributor to asymmetric cell division of the seam cells, and we focused the remainder of our experiments on DSH-2 and MIG-5.

Having established that DSH-2 and MIG-5 control seam cell fate, we investigated whether dsh-2 and mig-5 mutants display disruption in the localization of WβA components. In canonical Wnt/β-catenin signaling, Dvl acts downstream of Wnt ligands and Fz receptors and upstream of the β-catenin destruction complex (Clevers and Nusse, 2012). Given that Dvl proteins responds to upstream signals in multiple ways to function in multiple Wnt pathways (Dillman et al., 2013; Gao and Chen, 2010), we focused on the DSH-2- and MIG-5-mediated regulation of downstream Wnt components to determine how they regulate the complex outputs of WβA signaling. Thus, we crossed the uiwIs4(P_sys-1::mCherry::SYS-1) transgene into dsh-2, mig-5, and dsh-2 mig-5 mutants and observed the localization of both transgenes during the terminal seam cell division as we described previously (Baldwin and Phillips, 2014). In this final division during wild-type development, mCh::SYS-1 localizes asymmetrically to seam cell daughter nuclei (Fig. 2A) (Baldwin and Phillips, 2014). In mig-5 mutant worms, we found that average nuclear mCh::SYS-1 fluorescence was significantly reduced in both anterior and posterior daughters compared to wild type. This reduction resulted in daughter cells where anterior and posterior mCh::SYS-1 fluorescence levels were no longer significantly different from each other (P=0.067) (Fig. 2B,E). In dsh-2 mutant worms, average mCh::SYS-1 fluorescence was also significantly reduced in both anterior and posterior daughters compared to wild type, but anterior and posterior mCh::SYS-1 fluorescence levels remained significantly different from each other (P<0.001) (Fig. 2C,E). Given that both single mutants caused reductions of SYS-1 levels, we examined dsh-2 mig-5 double mutants to determine whether the Dvl paralogs had redundant functions. In the dsh-2 mig-5 double mutant, average nuclear mCh::SYS-1 fluorescence in both daughters was significantly reduced compared to wild type (Fig. 2E). Additionally, the dsh-2 mig-5 double mutant posterior daughter displayed significantly reduced SYS-1 compared to posterior daughters of dsh-2 single mutants (Fig. 2E). No significant difference was observed between anterior and posterior mCh::SYS-1 in dsh-2 mig-5 double mutants (P=0.653) (Fig. 2E), indicating that these divisions had been rendered symmetric. Given the reduction of nuclear mCh::SYS-1 in dsh-2 and mig-5 single and double mutants, we conclude that DSH-2 and MIG-5 act as positive regulators of nuclear SYS-1. Additionally, as the dsh-2 mig-5 mutant phenotype significantly differs from dsh-2 but not mig-5 mutants, it is possible that the unique MIG-5 function is genetically downstream of DSH-2 in SYS-1 regulation.

As APR-1 is a known negative regulator of SYS-1 and likely to be downstream of Dvl function in WβA, we crossed both osIs13 (P_apr-1::APR-1::YFP) and uiwIs4 (mCh::SYS-1) into dsh-2 and mig-5 mutants in order to observe whether changes in SYS-1 localization were accompanied by changes in APR-1 localization (Baldwin and Phillips, 2014). As we and others have previously demonstrated, APR-1::YFP localizes to the anterior seam cell cortex, whereas mCh::SYS-1 localizes asymmetrically to daughter cell nuclei (Fig. 2F) (Baldwin and Phillips, 2014; Mizumoto and Sawa, 2007; Wildwater et al., 2011). In mig-5 mutants, levels of cortical APR-1::VNS appear elevated in the posterior daughter when compared to wild type, indicating that MIG-5 is required for cortical polarization of APR-1 (Fig. 2G). Conversely, cortical APR-1 localization is largely unperturbed in dsh-2 mutants, although occasional posterior puncta were observed (Fig. 2H, arrowheads). To better describe changes in APR-1 localization, we scored divisions on whether cortical levels appeared to be higher in the anterior daughter, the posterior daughter or neither (Table 1). mig-5 mutants displayed a significantly altered distribution of APR-1::YFP compared to wild type (P<0.001, Fisher's exact test), whereas APR-1 distribution in dsh-2 mutants was not significantly perturbed (P=0.057, Fisher's exact test). We also note that SYS-1 fluorescence was significantly higher in dsh-2 and mig-5 mutants expressing APR-1::YFP than in those that did not express APR-1::YFP (Fig. 2G,H), although this is likely an artifact of APR-1 overexpression in a Dvl mutant background. A dsh-2 mig-5 double mutant expressing APR-1::YFP was also constructed, but homozygotes died as embryos or arrested as L1 larvae, preventing observation of the final seam cell division (data not shown). Thus, mutation of mig-5 results in clear APR-1 polarity defects, whereas mutation of dsh-2 does not, agreeing with the defects in SYS-1 nuclear asymmetry observed for each mutant.

Given that MIG-5 controls APR-1 localization and APR-1 controls both β-catenin outputs in the WβA pathway, we visualized WRM-1::GFP to functionally analyze the WRM-1-regulating pool of APR-1 in seam cell divisions lacking Dvl (Baldwin and Phillips, 2014; Mizumoto and Sawa, 2007). To do this, we combined both Dvl mutants with the osIs5 (P_scm::WRM-1::YFP) transgene that we and others have previously reported to localize asymmetrically to seam cell daughter nuclei (Baldwin and Phillips, 2014; Hughes et al., 2013; Mizumoto and Sawa, 2007). mig-5 mutants displayed a significant reduction in average nuclear WRM-1 levels in the posterior daughter nucleus, indicating that MIG-5 positively regulates WRM-1 levels in that daughter (Fig. 3B,E). This decrease in nuclear WRM-1 was accompanied by an increase in posterior cortical WRM-1, as would be expected from increased WRM-1 export (Mizumoto and Sawa, 2007; Sugioka et al., 2011). In dsh-2 mutants, average WRM-1 levels in anterior and posterior nuclei are not significantly different from average WRM-1 levels in corresponding wild-type nuclei (Fig. 3E). However, although a statistically significant difference exists between anterior and posterior average nuclear WRM-1 levels in wild type (indicating asymmetry), anterior and posterior average nuclear WRM-1 levels are not significantly different in dsh-2 mutants (Fig. 3E). Thus some alteration to WRM-1 localization has likely occurred in dsh-2 mutants despite the average WRM-1 nuclear levels being unaltered compared to wild type. This difference between the single mutants can be more clearly seen with a ‘polarity index’ calculated by subtracting anterior WRM-1 nuclear fluorescence from posterior in individual daughter pairs. In this analysis, mig-5 mutants show a shift towards zero (i.e. symmetry) compared to wild type (Fig. 3F, white bars versus black bars). Conversely, dsh-2 mutants follow a similar distribution to wild type but one that is slightly ‘flatter’, indicating that mutation of dsh-2 might result in a very mild polarity defect for WRM-1 localization (Fig. 3F). From these data, MIG-5 has the most prominent role in controlling WRM-1 nuclear localization as seen from the significant decrease in posterior nuclear WRM-1 levels, which renders this division symmetric for nuclear WRM-1. Conversely, dsh-2 appears to have little to no role in regulation of nuclear WRM-1 levels.

As the mig-5 and dsh-2 mutant phenotypes for WRM-1 were different, we observed WRM-1::YFP localization in the double mutant. Surprisingly, dsh-2 mig-5 double mutants displayed average WRM-1 levels that were significantly higher in both anterior and posterior daughter nuclei when compared to wild-type, mig-5 and dsh-2 mutant seam cells (Fig. 3D,E). There was no statistically significant difference in WRM-1 fluorescence between the anterior and posterior nuclei in dsh-2 mig-5 double mutants, indicating that loss of both DSH-2 and MIG-5 simultaneously results in WRM-1 polarity defects (Fig. 3E). When a ‘polarity index’ was constructed for dsh-2 mig-5 double mutants, the distribution of asymmetry was broadened, similar to dsh-2 mutants but it also had more reversals of polarity (Fig. 3F). The polarity index also demonstrates that although average WRM-1 levels were the same in anterior and posterior cells in dsh-2 mig-5 mutants, many divisions still displayed a high degree of asymmetry, indicating that polarity of WRM-1 asymmetry is randomized (as can be seen in Fig. 3D). The dsh-2 mig-5 double mutant distribution fails to strongly recapitulate either single mutant, which is consistent with individualized Dvl roles in the regulation of WRM-1 polarity, in addition to overlapping roles in regulation of WRM-1 nuclear levels. We conclude that DSH-2 and MIG-5 redundantly and negatively regulate nuclear levels of WRM-1, whereas MIG-5 is primarily responsible for regulating polarity of WRM-1 asymmetry.

PRY-1 regulates APR-1 cortical localization (Baldwin and Phillips, 2014; Mizumoto and Sawa, 2007). As we have shown in Fig. 2 that mig-5 disrupts cortical APR-1 asymmetry, we investigated whether either mig-5 or dsh-2 mutants could affect the localization of PRY-1. The functional PRY-1 construct osEx229(P_pry-1::PRY-1::GFP) has been previously reported to localize asymmetrically to the cortex of the earlier V5.p seam cell and its daughter cells (Fig. S2) (Mizumoto and Sawa, 2007). Observing the same construct at the later terminal seam cell division, we found that PRY-1::GFP localized symmetrically to the cortex of both daughter cells in 90% of divisions (n=30) (Fig. 4A,B). Rare cases of apparent asymmetry were mild (see middle cell in Fig. 4A) and likely derived from focal plane changes given that the PRY-1 cortical domain is ‘thin’. PRY-1::GFP cortical symmetry is likely biologically relevant because cells with lower levels of expression maintain symmetric PRY-1::GFP expression (Fig. S3A), and because PRY-1::GFP was not observed to induce fate changes in the seam cells, which were all observed to be differentiating asymmetrically (Fig. S3B, n=15). Despite lacking the hallmark asymmetry displayed by most WβA signaling components, we proceeded to determine whether MIG-5 and DSH-2 regulate PRY-1 localization. We found that PRY-1::GFP localized symmetrically to the cortex in 100% of mig-5 mutant divisions (n=27), 93.4% of dsh-2 mutant divisions (n=16), and 100% of dsh-2 mig-5 divisions (n=15) (Fig. 4C–E). As we have previously demonstrated that kin-19(RNAi) disrupts APR-1 cortical asymmetry, we also examined whether KIN-19 controls PRY-1 cortical localization (Baldwin and Phillips, 2014). Again, we found that PRY-1::GFP localized symmetrically to seam daughter cortices in 90% of kin-19(RNAi) divisions (n=10) (Fig. 4F). Finally, we combined the Dvl mutants with kin-19(RNAi) to determine whether PRY-1::GFP localization could be disrupted in combination; however, PRY-1::GFP symmetrically localized to daughter cell cortices in all observed mig-5; kin-19(RNAi) and dsh-2; kin-19(RNAi) divisions (n=7 and 5, respectively) (Fig. 4G,H). From these experiments, we conclude that PRY-1 localizes symmetrically to the cortices of seam cell daughters and that MIG-5, DSH-2 and KIN-19 are unnecessary for this localization.

Having uncovered the roles of DSH-2 and MIG-5 in controlling major downstream WβA signaling components, we turned to the localization of DSH-2 and MIG-5 themselves. Dvl localization is of special interest given the symmetric localization pattern of PRY-1 in the later division – do DSH-2 and MIG-5 also have an unexpected localization pattern in the terminal seam cell division? Previously published fluorescent epitope-tagged DSH-2 and MIG-5 transgenes driven by the seam-cell-specific SCM promoter were used to visualize DSH-2 and MIG-5 localization in the later seam cell divisions (Mizumoto and Sawa, 2007). Both of these constructs had been reported to localize to the posterior cortex of the T seam cell that divides during the first larval phase (Mizumoto and Sawa, 2007).

We crossed osEx233(P_scm::MIG-5::YFP) with uiwIs4 to observe MIG-5 and SYS-1 colocalization. At the very low levels of the osEx233 extrachromosomal array expression, cortical asymmetry as previously recorded was observable (Fig. 5A). When MIG-5::VNS levels were low, SYS-1 asymmetry in the wild-type pattern was observed (Fig. 5A,D). At moderate levels of MIG-5::YFP expression, MIG-5 localized to the cytoplasm but was excluded from the nucleus (Fig. 5B). At this moderate level of MIG-5::YFP expression, SYS-1 levels and asymmetry remained unperturbed (Fig. 5B,D). Interestingly, at high to very high levels of MIG-5::YFP expression, cytoplasmic MIG-5 puncta began to form (Fig. 5C). These puncta were observable in mother cells prior to division and did not appear biased to either pole of the cell (Fig. S4A). With the appearance of MIG-5 puncta in these animals, SYS-1 puncta that colocalized with the MIG-5 puncta also became apparent (Fig. 5C). When SYS-1 nuclear levels were quantified in divisions where MIG-5 puncta were observed, anterior nuclear SYS-1 levels were significantly increased, resulting in SYS-1 nuclear symmetry (Fig. 5C,D). Puncta expressing both MIG-5 and SYS-1 were not observed to localize to the seam cell nuclei, and an increase in SYS-1 levels was not observed in the posterior nuclei, so the increase in anterior nuclear SYS-1 is not an artifact of quantification of punctate SYS-1. Thus, high levels of MIG-5 overexpression result in formation of MIG-5 puncta that then result in SYS-1 stabilization and symmetry.

We also combined osEx225 (P_scm::DSH-2::YFP) with uiwIs4 to observe DSH-2 and SYS-1 colocalization. Similar to with MIG-5::YFP, DSH-2::YFP could be observed to be excluded from the nucleus, and DSH-2::YFP puncta would form in some cells that were observable before and after division (Fig. 5F,G, Fig. S4B). In contrast to MIG-5, clear asymmetric cortical localization of DSH-2::YFP was never observed, and although DSH-2::YFP puncta colocalized with mCherry::SYS-1, their overlap was less frequent (Fig. 5E,G; Fig. S4). DSH-2-overexpressing cell pairs were categorized into three classes: low to non-observable, medium, and high or displaying puncta (Fig. 5E–H). SYS-1 levels in daughter pairs with low to no DSH-2::YFP overexpression showed the wild-type pattern of nuclear SYS-1 asymmetry (Fig. 5E,H). However, in cells with medium levels of DSH-2::YFP overexpression, the average SYS-1 nuclear intensity in the anterior daughters was significantly higher than in cells with low or no DSH-2::YFP overexpression (Fig. 5F,H). Conversely, average SYS-1 nuclear intensity in cells with high levels of DSH-2::YFP overexpression (or displaying puncta) was not statistically different from cells with low or no DSH-2::YFP overexpression; we believe this is due to the rarity of this phenotype, leading to our low N-value and statistical power (Fig. 5G,H). From these data, we conclude that DSH-2 overexpression results in SYS-1 stabilization in anterior seam cell daughters.

Here, we provide insight into how cellular signals are transduced upon Wnt activation by uncovering the roles of the Dvl homologs DSH-2 and MIG-5 in regulating downstream effectors of WβA signaling in the seam cells of C. elegans. Mutation of each Dvl individually results in observable alterations in seam cell number, whereas loss of both Dvls results in increased seam cell fate specification defects compared to either single mutant. Furthermore, DSH-2 and MIG-5 positively regulate SYS-1 nuclear levels, as dsh-2 and mig-5 single mutants as well as the dsh-2 mig-5 double mutant have lower SYS-1 levels compared to wild type. We also show that loss of MIG-5 results in loss of nuclear WRM-1 polarity through a reduction of posterior nuclear levels of WRM-1, whereas loss of DSH-2 appears to have a milder WRM-1 polarity defect. Loss of both MIG-5 and DSH-2 results in an increase in the average amount of nuclear WRM-1 in both anterior and posterior nuclei, indicating that together DSH-2 and MIG-5 negatively regulate nuclear WRM-1 levels. In the final seam cell division, we show that PRY-1 localizes symmetrically to the cortex of seam cell daughters, and that mutation of dsh-2 and/or mig-5 does not alter localization of PRY-1. Finally, overexpression of either MIG-5 or DSH-2 in the seam cells results in SYS-1 stabilization in the anterior (i.e. unsignaled) daughter. Overall, DSH-2 and MIG-5 fit a scheme of Wnt/β-catenin signaling specialized for ACD that is continuing to grow in complexity but offers both interesting similarities and departures from the canonical models of Wnt/β-catenin signaling, as depicted in Fig. 6 and discussed below.

Our initial finding was that DSH-2 and MIG-5 redundantly control seam cell fate, although the defects in seam cell fate specification were mild in either mutant. For example, up to 22 seam cells were observed in mig-5 mutants versus the average of 16 in wild type, a gain of six seam cells (Fig. 1). Such a gain suggests that only six out of the 40 or more asymmetric seam cell divisions that occur from L1 onward and are likely controlled by WβA had defective symmetric divisions in mig-5 mutants. However, we also observed dsh-2 and mig-5 mutant worms with fewer than 16 seam cells, indicating that mutation of either protein can result in losses of seam cells as well as gains. Therefore, it is not accurate to simply label DSH-2 and MIG-5 as positive or negative regulators of seam cell fate; we provide possible hypotheses to explain these unusual phenotypes below.

One explanation for the mild seam cell fate phenotype is that gains and losses of seam cells through symmetric division mask each other in a final count of a large population of mutants. We observe gaps in the seam and clusters of extra seam cells in the same animals (Fig. 1C) suggesting that masking occurs within a single animal, likely causing our average seam cell counts to be an underestimation of the number of disrupted cell fate decisions. Gaps and clusters between seam cells were observed simultaneously in ∼20% of Dvl mutant worms (data not shown), indicating that the phenotype of at least 20% of observed individual Dvl mutant animals would be underestimated by counting seam cells. Additionally, ∼60% of Dvl mutants showed either gaps or clusters, suggesting that the phenotype of the overall mutant population is likely to be underrepresented by analyzing the average seam cell number. Although this explanation accounts for gains and losses of Dvl mutant seam cells as a population, individual seam cell defects can be explained by perturbations in downstream WβA regulators. We propose that the mild cell fate phenotype in dsh-2 and mig-5 mutants is due to the interaction between the SYS-1 and POP-1 branches, making WβA signaling robust to perturbation at the level of Dvl. For instance, the decrease in posterior WRM-1 nuclear levels in mig-5 mutants would be predicted to increase nuclear POP-1 levels, which would then be predicted to repress WβA target genes in the seam cells and reduce seam cell number similar to in a wrm-1 loss-of function (Banerjee et al., 2010; Gleason and Eisenmann, 2010; Ren and Zhang, 2010). The corresponding decrease in SYS-1 nuclear levels observed in mig-5 mutants would be predicted to reduce seam cell numbers even further. However, we also observe that WRM-1 can be increased in the double mutant, which would result in increased seam cell number by moderately lowering nuclear POP-1, as in pop-1(RNAi) (Banerjee et al., 2010; Gleason and Eisenmann, 2010). Hence, the loss of Dvl has uncoupled reciprocal SYS-1 and POP-1 regulation, resulting in a confused seam cell fate that can occasionally cause symmetrical division. This uncoupled regulation of SYS-1 and POP-1 might then result in stochastic target gene expression and in the observed random gains and losses of seam cells in mig-5 and dsh-2 single mutants, as well as the moderate disruptions to seam cell number in the dsh-2 mig-5 double mutant.

Alternatively, the mild Dvl loss-of-function phenotype could be due to upstream redundancy controlling WβA outputs. The dsh-2 and mig-5 mutant cell fate phenotypes were surprising given that loss of downstream WβA components, such as wrm-1, apr-1, kin-19 or pop-1, results in a much more severe seam cell fate defect (Banerjee et al., 2010; Gleason and Eisenmann, 2010; Ren and Zhang, 2010). Thus, the seam cells remain more biased towards asymmetric division in the absence of DSH-2 or MIG-5 than in the absence of more downstream components. Interestingly, simultaneous mutation or knockdown of all known Wnt ligands or receptors in C. elegans still allows for some properly polarized seam cell divisions to occur, indicating that WβA signaling is also robust to disruptions upstream of DSH-2 and MIG-5 (Yamamoto et al., 2011). The reason for this robustness is currently unclear, but it might be the result of multiple signaling pathways converging on cell fate in the seam cells, just as WβA and Src signaling converge on cell fate in the early embryonic EMS cell division (Bei et al., 2002). Cell contacts among seam cells have been previously demonstrated to control asymmetric division of the seam cells, and such a mechanism might be redundantly controlling seam cell fate specification alongside Wnt, Fz and Dvl, but ultimately converging upon downstream WβA regulators such as SYS-1, WRM-1 and POP-1 (Austin and Kenyon, 1994).

Mutation of either dsh-2 or mig-5 individually results in a significant reduction in nuclear SYS-1 levels in both anterior and posterior nuclei, and mutation of both dsh-2 and mig-5 simultaneously results in further reduction of SYS-1 levels in the posterior daughter compared to dsh-2 but not mig-5 mutants (Fig. 2). Still, the dsh-2 mig-5 double mutants exhibit a largely similar SYS-1 decrease phenotype to the dsh-2 and mig-5 single mutants (Fig. 2), so then why and how are both Dvls required for WβA signaling? One answer is that WβA signaling combines not only both the β-catenin stabilization function of canonical Wnt/β-catenin signaling but also polarization of cortical proteins, both of which are conserved functions of Dvl in both canonical and non-canonical Wnt pathways (Gao and Chen, 2010). Hence, a non-overlapping function of the two Dvls is revealed when comparing mig-5 mutants to dsh-2 mutants: loss of mig-5 but not dsh-2 abrogates SYS-1 asymmetry between the anterior and posterior nuclei, and similarly mutation of mig-5 but not dsh-2 results in significant disruption of cortical APR-1 localization. Thus, both DSH-2 and MIG-5 are capable of and necessary for the conserved β-catenin stabilization, but it appears that MIG-5 is responsible for the polarization of β-catenin stabilization (Fig. 6A). An alternative hypothesis that fits our data is that DSH-2 acts upstream of MIG-5 in SYS-1 regulation, posing the interesting problem of how multiple Dvls can act at different steps of an ordered genetic pathway. How these differences in function between DSH-2 and MIG-5 are effected by WβA signaling remains a mystery because DSH-2 and MIG-5 both contain the major conserved Dvl domains (PDZ, DEP and DIX), suggesting that difference in how DSH-2 and MIG-5 regulate and are regulated by WβA are dependent on more subtle differences in sequence and structure (Dillman et al., 2013).

dsh-2 mig-5 dual loss of function has been previously shown to result in loss of global POP-1 asymmetry in the POP-1-dependent asymmetric cell divisions in the early embryo (King et al., 2009). This finding is consistent with our discovery that the dsh-2 mig-5 double mutant results in increased nuclear WRM-1 (Fig. 3), which is expected to result in symmetrically low nuclear POP-1. In contrast with the role of DSH-2 and MIG-5 in positive regulation of SYS-1, DSH-2 and MIG-5 redundantly function in negative regulation of WRM-1 nuclear localization (Fig. 3). What remains unclear is the degree of redundancy between DSH-2 and MIG-5 in polarizing WRM-1 nuclear localization. Our statistics fail to show a significant difference between anterior and posterior WRM-1 nuclear localization in both dsh-2 and mig-5, but despite this dsh-2 does not show a significant posterior reduction in nuclear WRM-1 compared to wild type (Fig. 3E). mig-5 mutants, by contrast, do show a significant WRM-1 reduction. One explanation is that DSH-2 redundantly regulates WRM-1 nuclear levels alongside MIG-5, but that MIG-5 is primarily responsible for polarizing WRM-1 asymmetry (Fig. 6A). This is a notable difference from regulation of SYS-1 levels as described above, where DSH-2 and MIG-5 appear to act non-redundantly.

Previously, we have shown that APR-1 is necessary for negative regulation of both SYS-1 and WRM-1 nuclear levels in the seam cells but PRY-1 is not required for SYS-1 negative regulation. Instead PRY-1 is required for proper polarization of APR, which is then read out as SYS-1 and WRM-1 asymmetry (Baldwin and Phillips, 2014). Furthermore, we have shown that the SYS-1- and WRM-1-regulatory roles of APR-1 are genetically separable, and we hypothesized that Wnt signaling separates APR-1 into distinct functional pools for SYS-1 and WRM-1 regulation (Baldwin and Phillips, 2014). Building on our prior model, we show that MIG-5 controls APR-1 cortical localization and that both DSH-2 and MIG-5 are involved in regulation of SYS-1 and WRM-1 (Figs 2 and 3). Given that dsh-2 and mig-5 mutants do not disrupt PRY-1 localization (Fig. 4), it is likely that DSH-2 and MIG-5 are controlling SYS-1 and WRM-1 localization through APR-1 (Fig. 6B). Furthermore, because the dsh-2 mig-5 double mutant results in increased nuclear WRM-1 but decreased nuclear SYS-1, it is likely that DSH-2 and MIG-5 are involved in the separation of APR-1 into its SYS-1- and WRM-1-regulatory pools. Specifically, it appears that DSH-2 and MIG-5 together activate the WRM-1-regulatory pool of APR-1 and inhibit the SYS-1-regulatory pool of APR-1. Given that the phenotype of kin-19 RNAi closely resembles the phenotype of mig-5 for SYS-1, WRM-1 and APR-1 localization (Baldwin and Phillips, 2014), MIG-5 likely cooperates with KIN-19 to regulate APR-1, potentially phosphorylating APR-1 to change its localization and/or activity.

That PRY-1 appears to localize symmetrically to seam cell cortices during their final division was also a surprising finding (Fig. 4). An earlier report that describes asymmetric PRY-1 cortical localization analyzed the earlier V5.p seam cell division, whereas our experiments examined the later, final seam cell division (Mizumoto and Sawa, 2007). Interestingly, this earlier paper reported that PRY-1 had no function in WRM-1 regulation; our previous work reported that PRY-1 does control the WRM-1 in the later seam cell division, indicating that there might be a profound difference in PRY-1-mediated regulation and function between the V5.p division and the terminal division (Baldwin and Phillips, 2014; Mizumoto and Sawa, 2007). Why these differences in WβA function exist within the seam cells at different times during development is unclear, but it has been hypothesized that differences in seam cell size between earlier and later divisions might account for differences in function of WβA components between earlier and later seam cell divisions (Kanamori et al., 2008). The observation of symmetric cortical localization of PRY-1 calls into question how PRY-1 controls an asymmetric process. We propose that a physical interaction between MIG-5 and PRY-1 on the posterior cortex prevents a PRY-1–APR-1 interaction in the posterior daughter, restricting cortical APR-1 to the anterior daughter. Interestingly, yeast two-hybrid experiments have previously shown that MIG-5 physically interacts with PRY-1; hence, the fact that mutation of either mig-5 or pry-1 results in polarity defects in both APR-1 and WRM-1 localization indicates that the MIG-5–PRY-1 interaction might be important for polarizing cortical APR-1 (Baldwin and Phillips, 2014; Korswagen et al., 2002). Consistent with the hypothesis that MIG-5 is more important than DSH-2 for mediating an APR-1–PRY-1 interaction, the previous yeast two-hybrid experiment failed to show an interaction between PRY-1 and DSH-2 (Korswagen et al., 2002). The biological relevance of a negative yeast two-hybrid result is debatable, but if this reflects how MIG-5 and DSH-2 interact with PRY-1 within a seam cell, it potentially explains how MIG-5 bears the brunt of polarizing APR-1 activity whereas DSH-2 does not.

Our Dvl overexpression experiments provided both expected and unexpected results. That overexpression of Dvl results in stabilization of β-catenin is not so surprising in light of the commonly accepted positive role of Dvl proteins in Wnt signaling and our finding that C. elegans Dvl stabilize SYS-1, but the way by which each Dvl protein accomplished this was intriguing. Significant SYS-1 enrichment was observed in cells overexpressing both MIG-5 and DSH-2, but this increase in SYS-1 was correlated with MIG-5 and not DSH-2 puncta formation (Fig. 5). Given the intense interest in Fz- and Dvl-mediated regulation of the destruction complex, the nature and composition of the MIG-5 puncta is an interesting question for future investigation. Whether they are simply MIG-5 aggregates or perhaps the destruction-complex-containing ‘signalosomes’ described in other reports on Wnt signaling (Bilić et al., 2007; Taelman et al., 2010), the ability of these MIG-5 complexes to stabilize and colocalize with β-catenin is clear. Conversely, DSH-2 overexpression was able to achieve modest β-catenin stabilization without the formation of visible puncta. Whether this represents a legitimate biological difference in the mode of action between MIG-5 and DSH-2 remains to be seen.

The precise role of Dvl in the canonical Wnt/β-catenin signaling cascade remains puzzling mainly because the precise mechanism of β-catenin destruction complex inactivation continues to evade us. In most systems of canonical Wnt/β-catenin signaling, the action of Dvl is presumed to be to sequester and inactivate Axin. Here, we have shown that Dvl loss of function does not alter Axin localization in a live cell that is responding to Wnt signaling. In addition, we show that Dvl proteins are controlling β-catenin localization independently of Axin localization, likely through APC and perhaps dependent on the regulatory state of Axin, rather than the quantity of cortical Axin. Given that vertebrates also possess multiple Dvl proteins that have been shown to have redundant roles in various gross developmental processes, the genetic interactions of MIG-5 and DSH-2 may inform how Dvls cooperate in vertebrate Wnt signaling (Etheridge et al., 2008). We demonstrate how multiple Dvl homologs within a single cell/system could have individualized roles in Wnt signaling despite having a singular function, i.e. control of β-catenin nuclear levels. Dvl has been called ‘the hub of Wnt signaling’, as it is the signaling protein that appears to truly unifies the ‘Wnt’ pathways (Gao and Chen, 2010). Our data presented here describes the complex mechanisms by which Dvls can manage the multiple potential outputs of Wnt signaling in a single cell, better elucidating the function of a class of proteins whose role in Wnt signaling is paradoxically well-known but little understood.

Strains used in this study were: BTP105 [uiwIs4 (P_sys-1::mCherry::SYS-1)]; BTP118 [unc-76V; osIs13 (P_apr-1::APR-1::VNS; unc-76 (+)), uiwIs4]; HS1325 [unc-76(e911) V; osEx229 (P_pry-1::PRY-1::GFP+unc-76(+))]; HS1417 [osIs5 (P_scm::WRM-1::YFP) II]; JR 667 [unc-119(e2498::Tc1) III; wIs51(P_scm::GFP) V]; BTP137 [dsh-2(or302)/mIn1(dpy-10(e128), mIs14) II; osIs13, uiwIs4]; BTP138 [mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; osIs13, uiwIs4]; BTP139 [dsh-2(or302) mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; osIs13, uiwIs4]; BTP143 [mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; ws51V]; BTP144 [dsh-2(or302)/mIn1(dpy-10(e128), mIs14) II; wIs51V]; BTP145 [dsh-2(or302) mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; wIs51V]; BTP148 [dsh-2(or302)/mIn1(dpy-10(e128), mIs14) II; unc-76(e911) V; osEx229]; BTP182 [unc-76(e911) V; uiwIs4; osEx233 (P_scm::MIG-5::YFP+unc-76(+))]; BTP183 [unc-76(e911) V; osEx225 (P_scm::DSH-2::YFP+unc-76(+))]; BTP185 [dsh-2(or302) mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; uiwIs4]; BTP161 [dsh-2(or302),osIs5/mIn1(dpy-10(e128), mIs14) II]; BTP167 [mig-5(tm2639),osIs5/mIn1(dpy-10(e128), mIs14) II]; BTP175 [mig-5(tm2639)/mIn1[dpy-10(e128), mIs14] II;unc-76(e911) V; osEx229]; BTP184 [dsh-2(or302) mig-5(tm2639),osIs5/mIn1(dpy-10(e128), mIs14) II]; BTP194 [mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; uiwIs4]; BTP195 [dsh-2(or302) mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; uiwIs4]; BTP197 [dsh-2(or302) mig-5(tm2639)/mIn1(dpy-10(e128), mIs14) II; unc-76(e911) V; osEx229]; BTP200 [dsh-1(ok1445), wIs51V].

RNAi was performed using standard techniques using the pL4440 feeding vector (Timmons et al., 2001). Synchronized L1 worms were grown at 20°C until late L3 when the terminal seam cell division occurs. L3 worms were then immobilized using 10 µM muscimol on 6% agarose pads for live imaging. Cells and their corresponding nuclei were identified through differential interference contrast (DIC) microscopy prior to fluorescence image capture; n values refer to the number of daughter pairs imaged and compared. Box plots in Fig. 1 were assembled using the Tyers Lab tool (Spitzer et al., 2014).

Images were obtained using Zeiss Axio Imager.D2 compound fluorescent scope and Zeiss Zen software. Raw image data was exported into TIFF format and analyzed using ImageJ software. Nuclei of seam cells were identified in DIC images and then the mean protein fluorescence in fluorescent images was quantified using ImageJ. Background fluorescence was normalized for each experiment by using the same channels and exposures to image seam cells in N2 wild-type worms (n=24 for each exposure setting). All statistical comparisons were performed with either a t-test in Microsoft Excel or the Mann–Whitney test and VassarStats software (VassarStats, Lowry, 2010).

We would like to thank Doug Houston, Jan Fassler, Lori Adams, and the Phillips laboratory for helpful comments during the development of the manuscript. Many strains and transgenes were provided by the Caenorhabditis Genetics Center, which is funded by the NIH Office of Research Infrastructure Programs [grant number P40 OD01440].