The polarity protein VANG-1 antagonizes Wnt signaling by facilitating Frizzled endocytosis

Wiring of the nervous system during development is established first by neuronal migration, followed by navigation of axon growth cones towards their innervation targets and subsequent axon branching and synapse formation. The migration of neurons and axon growth cones is regulated by highly dynamic signals in the environment that function as attractive or repulsive cues (Tessier-Lavigne and Goodman, 1996). Over the past three decades, several classes of conserved neuronal guidance cues have been uncovered, including Netrin, Slit, semaphorins, ephrins, Sonic Hedgehog and Wnt glycoproteins (Tessier-Lavigne and Goodman, 1996; Blockus and Chédotal, 2016; Dudanova and Klein, 2013; Jongbloets and Pasterkamp, 2014; Sun et al., 2011; Yam and Charron, 2013). Although substantial efforts have been dedicated to deciphering the signaling pathways through which these factors instruct neuronal and growth cone migration, little is known about the regulatory mechanisms that downregulate them to prevent prolonged, aberrant signaling that could lead to guidance errors.

The availability of receptors on the plasma membrane is a well-established regulatory mechanism that modulates the magnitude and duration of signaling. In Drosophila embryos, trafficking of the Robo receptor to the growth cone membrane of pre-crossing commissural axons is prevented by the Golgi protein Commissureless, which traps Robo in the Golgi (Keleman et al., 2002; Tear et al., 1996; Kidd et al., 1998). In the commissureless mutant, constitutive Robo availability on the axonal membrane inadvertently transmits repulsive signaling by binding the Slit ligand, preventing the commissural axons from crossing the midline and hence causing the ‘commissureless’ axonal phenotype (Keleman et al., 2002; Kidd et al., 1998; Tear et al., 1996). Recent studies suggest that internalization of Frizzled proteins, seven-pass transmembrane proteins that function as Wnt receptors, modulates Wnt signaling and regulates axon guidance in C. elegans and rodents. Depending on the cellular contexts, Frizzled internalization may activate or terminate Wnt signaling. For example, internalization of Frizzled proteins triggered by PLR-1 E3 ligase-mediated ubiquitylation downregulates Wnt signaling to ensure proper axon development in C. elegans (Moffat et al., 2014), consistent with the vertebrate PLR-1 homologs ZNRF3 and RNF43 downregulating Frizzled proteins and inhibiting Wnt signaling in the mouse intestinal epithelium (Koo et al., 2012). Frizzled internalization could also activate Wnt signaling. In the mouse spinal cord, Vangl2, a four-pass transmembrane protein of the planar cell polarity (PCP) pathway, facilitates Frizzled endocytosis by antagonizing hyperphosphorylation of Frizzled proteins mediated by Dvl1/Dishevelled. In this context, Frizzled internalization activates Wnt signaling and promotes proper axon guidance (Shafer et al., 2011). Consistent with this, we recently showed that C. elegans VANG-1 promotes Wnt signaling by facilitating Frizzled internalization to instruct neurite branching patterns in the PLM mechanosensory neuron (Chen et al., 2017).

The aforementioned findings reveal an important, cell-autonomous function of Vangl2 in promoting Wnt signaling, in contrast to its complex non-autonomous requirement in Drosophila embryonic epithelia that involves interaction with Frizzled proteins in the neighboring cells (Bailly et al., 2017; Devenport, 2014). To explore how Vangl2 tunes Wnt signaling to regulate neuronal development, and whether Vangl2-dependent Frizzled internalization occurs in broader, more diverse cellular contexts, we systemically analyze the functions of C. elegans vang-1 gene in multiple classes of migrating C. elegans neurons. In contrast to the regulation of Frizzled proteins in the non-migrating PLM neuron, we find that VANG-1 antagonizes Wnt signaling by facilitating Frizzled endocytosis in all these migrating neuronal classes. These observations expand the phenotypic repertoire through which VANG-1/Vangl2 modulates Wnt signaling to pattern neuronal migration precisely.

The C. elegans Q neuroblasts are born in early first larval (L1) stage at equivalent positons on the right (hereafter referred to as QR) and the left side (hereafter referred to as QL) of the worm body. The posterior migration of QL descendants (QL.d) during early L1 requires the Abdominal-A family Hox gene mab-5, expression of which in the QL lineage is controlled by a β-catenin-dependent Wnt signaling pathway (Fig. 1A,B) (Clark et al., 1993; Wang et al., 1993; Harris et al., 1996; Whangbo and Kenyon, 1999; Korswagen, 2002). Genes in this pathway that are essential for mab-5 activation include egl-20/Wnt, mig-1/Frizzled, lin-17/Frizzled, mig-5/Dishevelled, bar-1/β-catenin, and the TCF transcription factor pop-1 (Harris et al., 1996; Eisenmann and Kim, 2000; Korswagen et al., 2000). Loss-of-function mutations of these genes diminish mab-5 expression and result in reversed, anterior migration of the QL.d, including the PVM (QL.paa) (Fig. 1C,D).

Interestingly, we found that the tm1422 deletion mutation of vang-1, the C. elegans homolog of the PCP gene Van Gogh/Strabismus/Vangl2, suppressed PVM migration defects of the mig-1 mutant, but it did not affect PVM migration in the otherwise wild-type background (Fig. 1C,D). Similar suppression of PVM defects was seen with another vang-1 deletion allele, ok1142, and also with vang-1 feeding RNAi (Fig. S1A). In contrast to its significant suppression of the mig-1 phenotypes, the vang-1(tm1422) mutation failed to suppress PVM migration defects in the egl-20, mig-5 and bar-1 mutants (Fig. 1D). egl-20, mig-5 and bar-1 each represent the single homolog for Wnt, Dishevelled and β-catenin, respectively, and are required for the posterior migration of QL.d. We hypothesize that vang-1 antagonizes Wnt signaling specifically at the level of Frizzled receptors. Therefore, PVM defects caused by the absence of Wnt ligand (egl-20) or signals downstream of Frizzled receptors (mig-5 and bar-1) are refractory to the effects of vang-1 mutations. By contrast, LIN-17, another Frizzled protein, is also required to promote QL.d migration (Harris et al., 1996; Zinovyeva et al., 2008), and we speculate that restoration of PVM migration in the mig-1; vang-1 double mutant could be mediated by Wnt signaling through the remaining LIN-17 receptor. This hypothesis predicts that the vang-1 mutation is not able to suppress QL.d defects in the mig-1 lin-17 double mutant, which lacks all relevant Frizzled receptors for QL.d migration. Consistent with this hypothesis, the vang-1 mutation improved PVM defects of the lin-17 single mutant but failed to do so in the mig-1 lin-17 double Frizzled mutant (Fig. 1D; Fig. S2).

Expressing vang-1 in the Q lineages of the mig-1; vang-1 mutant, using the egl-17 promoter, completely reversed the suppression of PVM mismigration conferred by the vang-1 mutation (Fig. 1E). Furthermore, overexpressing vang-1 in the QL.d resulted in PVM migration defects in the wild type and enhanced the PVM defects of the mig-1 mutant (Fig. 1E). These results suggest that vang-1 antagonizes Wnt signaling cell-autonomously in the QL.d neuroblasts. We confirmed this result by using somatic clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 gene editing (Fig. S1B) (Shen et al., 2014). In brief, we expressed Cas9 under the egl-17 promoter together with ubiquitously expressed vang-1 sgRNAs. Mutations of vang-1 in the Q lineages significantly suppressed PVM defects of the mig-1 mutant, confirming that vang-1 acts in the Q lineage (Fig. S1B).

On the basis of the hypothesis that vang-1 antagonizes mig-1 functions, we predicted that the expression of mab-5, the transcriptional target of Wnt signaling in QL.d migration, would be restored in the mig-1; vang-1 double mutant. This is indeed the case: mab-5 expression in the QL.d was significantly increased in mig-1; vang-1 compared with that in the mig-1 mutant, which was barely detectable (Fig. 1F,G). By contrast, the vang-1 mutation failed to restore mab-5 expression in the egl-20 mutant (Fig. 1F,G). Of note, the vang-1 mutant showed significantly higher level of mab-5 expression compared with the wild type, consistent with vang-1 being a negative regulator of Wnt signaling (Fig. 1F,G). Taken together, we conclude that VANG-1 antagonizes Wnt signaling cell-autonomously at the level of Frizzled receptors during QL.d migration.

We hypothesize that the antagonistic function of vang-1 in Wnt signaling is broadly conserved in early C. elegans neuronal development. Below we test this idea in three additional neuronal contexts: QR descendant (QR.d) migration, hermaphrodite-specific neuron (HSN) migration, and ALM touch neuron polarization.

In contrast to QL.d, QR.d migrate to the anterior, and these anterior migration events depend on Wnt signaling genes that are largely distinct from those involved in QL.d migration (Harris et al., 1996; Forrester et al., 2004; Zinovyeva et al., 2008). Specifically, in addition to egl-20, cwn-1/Wnt, mom-5/Frizzled and dsh-2/Dishevelled promote QR.d anterior migration, whereas mig-1, mig-5 and bar-1 are not required (Maloof et al., 1999; Zinovyeva et al., 2008). Additional Wnt receptors, such as LIN-17 and CFZ-2, are hypothesized to function with MOM-5 in QR.d migration, as QR.d migration defects in the mom-5 mutant are less severe than those of the cwn-1; egl-20 double mutant (Forrester et al., 2004; Zinovyeva et al., 2008). Consistent with previous studies, we found that AVM, one of the QR.d, terminated their anterior migration prematurely in the cwn-1, egl-20 and mom-5 mutants (Fig. 2A,B). The vang-1 mutation ameliorated AVM migration defects in the egl-20, cwn-1 and mom-5 mutants, and caused the AVM to be more anterior in an otherwise wild-type background (Fig. 2A,B). Improvement of QR.d migration by the vang-1 mutation likely resulted from Wnt signaling mediated by remaining Wnt ligands or Frizzled receptors in respective single mutants, similar to the suppression of QL.d defects in the mig-1 mutant (see above). By contrast, the vang-1 mutation failed to suppress AVM defects caused by the mutation of mig-14, the C. elegans homolog of the Wnt secretory factor Wntless (Fig. 2A,B) (Bänziger et al., 2006; Bartscherer et al., 2006; Goodman et al., 2006). In the mig-14 mutant, Wnt secretion in general is profoundly depleted and very little extracellular Wnt protein is available for signaling. These results are consistent with the model that phenotypic improvement conferred by the vang-1 mutation depends on the Wnt ligands. Similar to its effects on QL.d, overexpression of VANG-1 in the QR.d caused modest but significant AVM migration defects (Fig. 2A,B). This suggests that vang-1 antagonizes Wnt signaling in QR.d migration cell-autonomously.

It was previously shown that mom-5 mutations compromise anterior migration of the QR.p neuroblast, whereas vang-1 mutations move QR.pap (future AVM) further anterior (Mentink et al., 2014). To test whether vang-1 also plays a role during QR.p migration, we examined the positions of QR.p at a defined developmental time points during its anterior migration in L1 larvae. The V5R seam cell divides after the QR.a and QR.P start their anterior migration. We scored QR.p positions in L1 animals in which the V5R seam cell has divided but the two V5R daughter cells remain in contact (Fig. 2C). This experiment confirmed that the anterior migration of QR.p was compromised in the mom-5 mutant, and the vang-1 mutation significantly suppressed the QR.p defects (Fig. 2C,D). These data indicate that vang-1 antagonizes Wnt signaling during QR.p neuroblast migration.

Because QL and QR neuroblasts share similar developmental history and functional property, we investigated whether vang-1 antagonizes Wnt signaling in a completely unrelated neuronal type. The HSNs are bilaterally symmetric serotonergic motor neurons that control egg-laying muscles (Desai et al., 1988). HSNs migrate from the posterior towards the future vulval primordium during embryogenesis under the control of multiple Wnt and Frizzled genes (Fig. 3A) (Garriga et al., 1993; Pan et al., 2006). Various Wnt pathway mutants, including egl-20, mig-14, mig-1, cwn-1; egl-20 or mig-1 mom-5 double mutants, showed migration defects of HSNs, which fall short of their wild-type positions near the vulval primordium (Fig. 3B-D). Similar to its effects on AVM migration, the vang-1 mutation suppressed HSN migration defects in the egl-20, mig-1 or cwn-1; egl-20 double mutants (Fig. 3B-D). The suppression effect of the vang-1 mutation was not seen in the mig-1 mom-5 double mutant or when secreted Wnts were largely eliminated in the mig-14 mutant (Fig. 3C,D). Restoration of vang-1 in the HSNs, using the unc-86 promoter, completely reversed the suppression of HSN migration defects in the egl-20; vang-1 double mutant, making the HSN defects similar to those of the egl-20 mutant (Fig. 3D). This rescue effect was abolished by vang-1 RNAi, suggesting that the reversal of suppression is the action of functional vang-1 transcripts rather than impairment of HSN migration by use of the unc-86 promoter (Fig. 3E). These results confirm that vang-1 antagonizes Wnt signaling cell-autonomously in HSN migration.

Could vang-1 antagonize Wnt signaling in developmental contexts other than neuronal migration? To explore this possibility, we examined the ALM, a unipolar touch neuron morphology of which is regulated by Wnts (Hilliard and Bargmann, 2006; Prasad and Clark, 2006; Pan et al., 2008; Zheng et al., 2015). The bilaterally symmetric ALM neurons extend a long neurite to the anterior. In the mig-14 and cwn-1; egl-20 double mutants, many ALM neurons grew a single posterior neurite that could reach as far as the lumbar ganglia, and some ALMs showed bipolar morphology with both anterior and posterior neurites (Fig. S3A,B). The vang-1 mutation partially suppressed ALM defects in the cwn-1; egl-20, but not in the mig-14 mutants (Fig. S3B). Moreover, vang-1 overexpression in the touch neurons caused ALM defects similar to those seen in the cwn-1; egl-20 mutants in a dose-dependent manner (Fig. S3C). Considering all data from the QL.d, QR.d, HSN and ALM experiments, we conclude that vang-1 broadly antagonizes Wnt signaling in multiple events of early neuronal development in C. elegans.

Our genetic analyses of neuronal migration raise the possibility that vang-1 negatively regulates Wnt signaling at the level of Frizzled receptors. A previous study of growth cone steering suggests that Vangl2 facilitates Frizzled endocytosis by antagonizing Dvl1/Dishevelled-mediated Frizzled hyperphosphorylation (Shafer et al., 2011). We previously found that the localization of MIG-1 to RAB-5(+) early endosomes was decreased in the non-migratory C. elegans PLM neuron in the vang-1 mutant, and that VANG-1 overexpression increased MIG-1 distribution to the early endosomes (Chen et al., 2017). On the basis of these observations, we hypothesize that VANG-1 antagonizes Wnt signaling by promoting Frizzled endocytosis, thereby terminating signal transduction induced by binding of Wnt ligands to the Frizzled receptors. This model makes three predictions. First, blocking endocytosis should suppress neuronal migration defects caused by a partial deficiency in Frizzled receptors. Second, suppression of neuronal defects conferred by the vang-1 mutation is not further enhanced by blocking Frizzled endocytosis. Third, loss of VANG-1 impairs Frizzled endocytosis in the migrating neurons and results in higher Frizzled levels at the plasma membrane.

We tested the first prediction by examining animals that carried a deletion in the arr-1 gene, which encodes the adaptor protein β-arrestin2 essential for Frizzled endocytosis in mammalian cells (Chen et al., 2003). We recently confirmed that arr-1 also mediated MIG-1 endocytosis in the non-migratory C. elegans PLM neuron (Chen et al., 2017). Although the arr-1 mutation did not affect PVM migration in the wild type, similar to the vang-1 mutation, it significantly suppressed PVM defects in the mig-1 mutant (Fig. 4A). Likewise, defective AVM migration in the mom-5 mutant and HSN undermigration of the egl-20 mutant were also suppressed by the arr-1 mutation (Fig. 4B,C). These results suggest that in the migrating C. elegans neurons, Frizzled endocytosis negatively regulates Wnt signaling. To test whether vang-1 acts in the same genetic pathway as arr-1, we constructed multiple mutants that harbored both the vang-1 and arr-1 mutations, together with the mig-1 or the egl-20 mutation. PVM migration defects in the mig-1; arr-1 vang-1 triple mutant were comparable to those of the mig-1; vang-1 or the mig-1; arr-1 mutants (Fig. 4A). This observation suggests that arr-1 and vang-1 act in a shared genetic pathway to temper Wnt-Frizzled signaling in QL.d migration. Consistent with this, HSN undermigration was not further worsened in the egl-20; arr-1 vang-1 triple mutant, compared with that in the egl-20; arr-1 or egl-20; vang-1 double mutant (Fig. 4C).

We further investigated the order of vang-1 and arr-1 function by genetic epistasis, taking advantage of the neuronal defects caused by vang-1 overexpression. PVM and AVM migration defects caused by vang-1 overexpression were ameliorated in the arr-1 mutant (Fig. 4A,B). Moreover, the arr-1 mutation also significantly reduced the ALM polarity defects induced by excess vang-1 activity (Fig. S3C). These results imply that arr-1 acts downstream of vang-1 to antagonize Wnt signaling.

Our previous work indicates that VANG-1 promotes the endocytosis of MIG-1 in the PLM touch neuron (Chen et al., 2017). Taken with results from the genetic experiments that VANG-1 attenuates Frizzled-dependent Wnt signaling and acts upstream of ARR-1, we hypothesized that VANG-1 facilitates endocytosis of the Frizzled receptors in the Q.d and other neuronal types. To test this, we investigated whether VANG-1 regulates Frizzled localization in the Q.d. For unknown reasons, fluorescence signals from our MIG-1::GFP transgene that fully rescued the PVM defects of the mig-1 mutant, could not be detected in the migrating QL.d (data not shown). Therefore, we used the previously characterized mom-5 transgene, zuIs145(Pnmy-2::MOM-5::GFP) (Park et al., 2004), and observed MOM-5::GFP distribution with or without vang-1 in QR.d. Because zuIs145 seemed to be integrated to the sex chromosome and failed to recombine with the sex chromosome-located vang-1(tm1422) allele, we used the CRISPR-Cas9 method to generate the vang-1(twn3) null mutation in the zuIs145 background (Fig. 5A; Fig. S1C). We first confirmed that the vang-1(twn3) mutation suppressed PVM migration defects of the mig-1 mutant to levels comparable to other vang-1 alleles (Fig. S1C). In the QR.d of the wild type, MOM-5::GFP signals were distributed at the plasma membrane and intracellularly (Fig. 5B). By contrast, in the vang-1(twn3) mutant, intracellular MOM-5::GFP signals in QR.d were significantly reduced with reciprocal accumulation of MOM-5::GFP at the cell membrane (Fig. 5B,C). We further confirmed this result by performing vang-1 RNAi, which resulted in the enrichment of MOM-5::GFP signals at the plasma membrane of QR.d (Fig. S4A,B). Moreover, overexpression of vang-1 diminished membrane MOM-5::GFP, with reciprocal increase of MOM-5::GFP signals in the cytosol (Fig. 5B,C). An arr-1 mutation engineered by CRISPR-Cas9 resulted in MOM-5::GFP accumulation at the cell membrane, confirming that these changes in MOM-5::GFP distribution indicate endocytosis blockade (Fig. 5B,C). We verified that the GFP-fused MOM-5 proteins are functional, as zuIs145 significantly rescued AVM undermigration of the mom-5 mutants (Fig. S4C). Together, these results indicate that VANG-1 mediates MOM-5 endocytosis cell-autonomously.

We previously showed that VANG-1 complexed with MIG-1 when expressed in mammalian cells (Chen et al., 2017), and this was again confirmed in the current study (Fig. 6A,B; Fig. S5A). To investigate whether VANG-1 also forms complexes with MOM-5, we expressed HA-tagged VANG-1 and FLAG-tagged MOM-5 in HEK293 cells. Pull-down of MOM-5 by FLAG antibodies co-immunoprecipitated VANG-1, suggesting that VANG-1 and MOM-5 form protein complexes (Fig. 6A,C; Fig. S5B). To gain further insight into the interaction between VANG-1 and Frizzled proteins, we generated truncated MIG-1 and MOM-5 proteins that lacked the Wnt-binding, cysteine-rich domain (ΔCRD). Removal of the CRD profoundly weakened the ability of either MIG-1 or MOM-5 to co-immunoprecipitate VANG-1, whereas deletion of the cytoplasmic domain (ΔC) of MIG-1 had no effect (Fig. 6B,C; Fig. S5). We were not able to express MOM-5 lacking the cytoplasmic domain in mammalian cells for unknown reason. These results are consistent with the observation from a prior Drosophila study reporting that the Frizzled CRD binds Van Gogh (Wu and Mlodzik, 2008), and indicate that the CRD of Frizzled proteins is important for VANG-1-Frizzled interaction.

In addition to its well-established function in the development of Drosophila eye and wing discs, PCP signaling also plays important roles in various aspects of vertebrate neuronal development, such as neural tube closure, polarization of inner hair cells, neuronal migration and axon guidance (Goodrich, 2008; Tissir and Goffinet, 2013; Zou, 2012). In neural tube closure, PCP signaling generates molecular asymmetries within and between neuronal precursors that polarize these cells uniformly along the tissue plane to enable convergent extension, a crucial process that elongates the vertebrate embryo. Compared with these early morphogenetic events, our understanding of mechanisms by which PCP signaling controls long-range neuronal migration and axon guidance is incomplete. The current study establishes that VANG-1 acts cell-autonomously in migrating C. elegans neurons and tempers Wnt signaling by facilitating Frizzled endocytosis. These findings extend current understanding of how PCP signaling controls neuronal development and shapes the connectivity of the nervous system.

vang-1, the C. elegans Vangl2 homolog, is involved in several morphogenetic events that bear some similarity to Drosophila or vertebrate PCP, such as the assembly of ventral nerve cord during which immature neurons display transient rosette formation similar to that in vertebrate convergent extension movement (Shah et al., 2017), or the repositioning of endodermal descendants during intestinal morphogenesis (Asan et al., 2016). vang-1 also controls neuronal polarization (Sanchez-Alvarez et al., 2011), neurite outgrowth (Zheng et al., 2015) and neurite branching (Chen et al., 2017). Mentink et al. reported that vang-1 controls the final positions of two QR.d, QR.paa (future AVM) and QR.pap (Mentink et al., 2014). In the vang-1 mutant, QR.paa and QR.pap are positioned more anteriorly compared with those in the wild type, a phenotype that is opposite to that caused by loss of egl-20, cwn-1 or mom-5 (Mentink et al., 2014). This observation is consistent with our model that vang-1 antagonizes a Wnt signaling cascade composed of egl-20, cwn-1 and mom-5 in QR.d. These results contrast with those in our prior study, which showed that vang-1 promotes Wnt signaling to pattern neurite branching in the non-migratory PLM neuron (Chen et al., 2017). Because in both studies evidence suggests that vang-1 facilitates Frizzled endocytosis, we speculate that cellular contexts determine whether VANG-1-mediated Frizzled internalization upregulates or diminishes Wnt signaling. In immature neurons that undergo long-range migration or cellular polarization, Frizzled internalization terminates Wnt signaling. By contrast, when neurons mature and begin to grow collateral branches in their neurites, as in the case of PLM, Frizzled endocytosis transduces Wnt signaling (Chen et al., 2017). These observations expand the spectrum of Wnt-PCP pathways, implying that VANG-1 directly tunes a Wnt signaling cascade previously thought to be distinct from the PCP pathway, thus blurring the boundaries between the β-catenin and PCP branches of Wnt signaling.

Anterior displacement of QL.d in the egl-20 mutants has led to the hypothesis that these mutant QL.d have acquired properties of QR.d, which migrate anteriorly. This hypothesis gains some support from the observation that a cwn-1 mutation shifts the positions of QL.d in the egl-20 mutants towards the posterior, which is explained by the fact that cwn-1 drives QR.d anterior migration (Zinovyeva et al., 2008). It is therefore intriguing why the migration defects of QR.d, but not those of QL.d, are suppressed by vang-1 mutations in the egl-20 mutants. However, QL.d in the cwn-1; egl-20 sometimes occupies a position that is even more posterior than QL.d in the wild type. The distribution of QL.d in the cwn-1; egl-20 double mutant is also more posterior compared with that of QR.d in the same mutant (Zinovyeva et al., 2008). These observations suggest that QL.d in the egl-20 mutants bears some phenotypic similarity to the wild-type QR.d, but their molecular properties could still be different. One of these differences is likely to be the distinct subsets of Frizzled receptors expressed or functioning in QL.d or QR.d. mig-1 or lin-17 are required in QL.d, but mutations of these Frizzled genes cause no undermigration or even slight overmigration in QR.d (Zinovyeva et al., 2008; Mentink et al., 2014). On the other hand, mom-5 is important for QR.d migration yet dispensable for QL.d migration. We speculate that QL.d and QR.d may express different Frizzled receptors, and these Frizzled receptors engage different signaling cascades. All these potentially contribute to the observed difference in QL.d and QR.d regarding their sensitivity to the vang-1 mutation in the egl-20 mutant background.

We noted that the distribution of PVM becomes somewhat bimodal when vang-1 mutations or RNAi are added to the mig-1 mutant (Fig. 1D; Fig. S1A,B). It has been shown that EGL-20 signaling regulates QL lineage migration independent of its distribution (Whangbo and Kenyon, 1999). We speculate that mab-5 is activated and normal posterior QL.d migration is initiated when Wnt signaling exceeds a threshold in QL.d. Variability in mab-5 expression increases in the mig-1 and lin-17 mutants, and it has been proposed that Wnt signaling acts to reduce variability of mab-5 expression (Ji et al., 2013). We speculate that the vang-1 mutations restore Wnt signaling intensity in general but do not improve all the regulatory steps within the Wnt signaling network, which in part explains low or no mab-5 expression in some mig-1; vang-1 mutant animals and a bimodal distribution of PVM positions.

In vertebrate cells, endocytosis of the Frizzled receptors could be mediated by β-arrestin2 or by ZNRF3-dependent ubiquitylation (Chen et al., 2003; Koo et al., 2012). Although Frizzled endocytosis by ubiquitylation clearly downregulates Wnt signaling in both mammalian intestinal epithelium and C. elegans neurons (Koo et al., 2012; Moffat et al., 2014), Frizzled internalization by β-arrestin2 could promote Wnt signaling both in vertebrate models of tumorigenesis and C. elegans neuronal development (Bonnans et al., 2012; Fereshteh et al., 2012; Seitz et al., 2014; Chen et al., 2017). In the present study, we show that VANG-1 and ARR-1 act in a common genetic pathway to internalize Frizzled receptors, and this downregulates Wnt signaling. It remains to be determined whether ZNRF3 and Vangl2-β-arrestin2 cooperate to internalize Frizzled proteins, or if they represent distinct pathways to recycle Frizzled proteins into separate intracellular compartments. Physical interaction between VANG-1 and Frizzled ligands requires the Frizzled CRD but not the cytoplasmic domain. One possibility is that binding of VANG-1 to the Frizzled-CRD triggers Frizzled internalization mediated by the β-arrestin2 machinery. A prior study showed that Van Gogh, the Drosophila Vangl2, binds Frizzled-CRD, and when addressed in the context of epithelial cell polarization, this represents an interaction in trans between Frizzled proteins and Van Gogh from neighboring cells (Wu and Mlodzik, 2008). By contrast, our genetic and biochemical data suggest that in C. elegans migrating neurons, the interaction between VANG-1 and Frizzled-CRD can occur in cis, i.e. in the same cell. Although the crystal structure of the Frizzled-8-CRD complexed with the Xenopus Wnt-8 ligand has been determined recently (Janda et al., 2012), the structure of Vangl2 or its homologous proteins remains unknown. In the future, it will be important to decipher the structure of Vangl2 by itself and that of Vangl2 complexed with the Frizzled CRD. This will offer important insight into how binding of diffusible (such as Wnts) or membrane-tethered ligands (such as Vangl2) to the Frizzled-CRD initiates Frizzled endocytosis and tunes Wnt signaling.

Wnt signaling has been shown to promote the migration of neurons and growth cones by acting as either attractive or repulsive cues (Yoshikawa et al., 2003; Lyuksyutova et al., 2003; Pan et al., 2006). A recent report in C. elegans suggests that signaling through the Frizzled receptor MIG-1 and LIN-17 terminates the anterior migration of QR.paa and QR.pap neurons through the β-catenin BAR-1 (Mentink et al., 2014). Adding to the complexity of Wnt signaling in controlling neuronal migration, Mentink et al. further showed that vang-1 also acts as a brake to prevent QR.pax from overmigrating. As the migration of QR.p was not affected in the vang-1 mutant, they concluded that vang-1 functions independently of Wnt-β-catenin signaling, although no genetic experiments were performed to test the interaction between vang-1 and the Wnt pathway genes (Mentink et al., 2014). By contrast, we show here that vang-1 antagonizes Wnt signaling by removing Frizzled receptors from the cell membrane. As Frizzled proteins transduce Wnt signaling through both β-catenin-dependent and β-catenin-independent pathways, our results suggest that vang-1 has a broad, negative impact on Wnt signaling in early neural development. This conclusion is supported by observations made in four different neuronal classes (QL.d, QR.d, HSN and ALM). Unlike QL.d, for which the target of Wnt signaling for neuronal migration, the Hox gene mab-5, has been identified, genes that Wnt-Frizzled signaling targets to promote migration in QR.d and HSN or polarization in ALM remain unknown. Identification of such Wnt target genes in neuronal development helps to clarify the role of vang-1 in fine-tuning Wnt signaling. Our current study, together with a prior report that PLR-1/ZNRF3 E3 ligase downregulates Wnt signaling by internalizing Frizzled proteins (Moffat et al., 2014), suggests that timely internalization of Frizzled proteins is a crucial step to prevent aberrant neuronal wiring caused by prolonged Wnt signaling. Understanding the molecular and structural detail of VANG-1-mediated endocytosis of Frizzled proteins will provide important insights into the modulation of Wnt signaling in early neuronal development.

Key experimental materials, including C. elegans and bacterial strains, cell lines and antibodies, are listed in Table S1. C. elegans strains were cultured and maintained as described (Brenner, 1974). The HEK293 cell line used in this study is from the American Type Culture Collection (ATCC) and was authenticated and tested for contamination. The antibodies used in this study were previously validated (Duval et al., 2014; Sanna et al., 2018; Schafer and Braun, 1995).

We performed germline CRISPR as previously described with some modifications (Dickinson et al., 2013). In brief, we co-injected into worms a mixture of the following three plasmids: pDD162(Peft-3::Cas9), PU6::vang-1 sgRNA(GACACGAGGAGTTGCGTT) or PU6::arr-1 sgRNA(GGTGACTATGTCGACTTGA), and unc-22 sgRNA (twitcher phenotypes, as a co-CRISPR marker) (Kim et al., 2014). Twitcher F1 transgenics were subjected to T7 endonuclease digestion to identify nucleotide mismatches caused by Cas9-mediated editing. Homozygotes of the vang-1(twn3) thus generated were confirmed by DNA sequencing.

Somatic CRISPR was performed as previously described (Shen et al., 2014). In brief, we co-injected the mixture of the following plasmids: Pegl-17::Cas9 (Q/V5 lineage-specific), PU6::vang-1 sgRNA(GACACGAGGAGTTGCGTT) and the co-injection markers Pgcy-8::mCherry (for isolation of transformants) and Pmec-7::mCherry. To optimize the expression of transgenes, animals were cultivated at 25°C, and PVM positons scored with Pmec-7::mCherry, which labels the six touch mechanosensory neurons.

We used standard molecular biology techniques for cloning and plasmid construction. Constructs used for generating transgenes of the twnEx series used the pPD95.77 Fire vector as their backbone. The constructs for expression in the HEK293 cells were in the pcDNA, RK5F vector backbone. Primer sequences for cloning vang-1, mig-1 and mom-5 are listed in Table S2. Germline transformation by microinjection was performed as described (Mello et al., 1991).

Feeding RNAi was performed as described (Kamath et al., 2001). In brief, vang-1 RNAi bacteria were induced by 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 1.5 h before seeding on NGM plates supplemented with 1 mM IPTG. Gravid hermaphrodites were bleached on RNAi plates and newly hatched L1 larvae were allowed to develop to adulthood. This procedure was repeated one more time before gravid hermaphrodites were collected and bleached to obtain synchronized early L1 larvae arrested in M9 buffer. These arrested L1 larvae were then released on fresh vang-1 RNAi plates and scored at appropriate stages for individual phenotypes (2-3 h post-hatching L1 for MOM-5::GFP quantification, and late L4 for PVM positions).

The transgene zdIs5(Pmec-4::gfp) labels PVM and AVM, and integrated array mgIs42(Ptph-1::gfp) or extrachromosomal array twnEx199(Punc-86::gfp) mark HSN. Late L4 animals were anesthetized with 1% sodium azide. Neurons were imaged under the 10× objective of the AxioImager M2 imaging system (Carl Zeiss). For quantification of AVM and PVM positions, we set the nose tip as zero and the tail tip as one. AVM and PVM positions were defined as fractions of the nose-to-tail length. For quantification of HSN positions, we set the vulva opening as zero and the tail tip as one, as normally the HSNs do not migrate anterior to the vulva. HSN positions are defined as fractions of the vulva-to-tail distance. Pixel-wise measurement of length was performed with ImageJ.

The transgene zdIs5(Pmec-4::gfp) also labels the ALM and PLM touch receptor neurons. L4 animals were anesthetized with 1% sodium azide. Neurons were imaged under the 10× objective of the AxioImager M2 imaging system (Carl Zeiss). In the Wnt pathway mutant, a bipolar polarity defect was defined as ALM neurons with the normal anterior process and a posterior process longer than ten times the cell diameter; this definition is the same as previously published methods (Pan et al., 2008). For VANG-1 overexpression, a bipolar polarity defect was defined as ALM neurons with the normal anterior process and a posterior process longer than five times cell diameter. A reversed polarity defect was defined as ALM neurons having a long posterior process without the normal anterior process.

The transgene muIs16(Pmab-5::gfp) was used to observe mab-5 expression, with QL.a and QL.p marked by rdvIs1(Pegl-17::Myri::mCherry). We synchronized early L1 larvae by hatching them in M9 buffer, and scored them after resuming feeding for 2-3 h on regular OP50 Escherichia coli plates. L1 larvae were observed using the 100× objective of the AxioImager M2 imaging system (Carl Zeiss) and AxioCam MRm CCD camera (Carl Zeiss) under identical imaging parameters. GFP pixel density OF muIs16 in QL.d, the boundary of which is defined by rdvIs1, was quantified using ImageJ.

We used rdvIs1(Pegl-17::Myri::mCherry), which highlights the plasma membrane, to label the Q cell lineages, and zuIs145(Pnmy-2::mom-5::gfp) to observe MOM-5::GFP signal (Ou et al., 2010; Park et al., 2004). Note that chromosomes are also labeled by rdvIs1. Early L1 larvae (post-hatching 1∼2 h) were examined under the 100× objective of the LSM 700 Confocal System (Carl Zeiss), and series of z-stack projection images were acquired using Zeiss Zen software under identical parameters. For this analysis, we focused on QR.a and QR.p. Pixel-wise quantification for GFP fluorescence intensity was quantified using ImageJ for individual optical slices, and summation of all optical sections from the same z-stack series was represented. To determine the subcellular distribution of MOM-5::GFP, cell membrane MOM-5::GFP signals were first quantified using rdvIs1 as a reference for QR.d membrane. Cytosolic MOM-5::GFP signals were derived by subtracting membrane MOM-5::GFP signals from total MOM-5::GFP signals.

Gene constructs were transfected into HEK293 cells with Lipofectamine (Invitrogen) and grown for 2 days before lysis with buffer (50 mM Tris, 150 mM NaCl, 2 mM EDTA pH 8.0, 0.5% sodium deoxylcholate, 10 mM phenylmethylsulfonyl fluoride, 1 M dithiothreitol, 1% NP-40), protease inhibitor cocktail (Roche) and 20 mM sodium fluoride. For co-immunoprecipitation, we used anti-HA (Invitrogen) or anti-FLAG (Sigma) beads to pull down proteins from cell lysates. After proteins (1 µg) had bound to the beads, we added 4× sample buffer with 10% β-mercaptoethanol and incubated samples at 37°C for 20 min to prevent aggregation of transmembrane proteins. Samples for western blot analysis were electrophoresed in 10% SDS-polyacryalamide gel and transferred onto the PVDF membrane. We used anti-HA, ant-FLAG or anti-β-actin antibodies as described in the Table S1 to detect protein signals.

For all samples, n represents the number of cells scored. Cells were excluded only when they could not be unambiguously identified or visualized. One-way ANOVA followed by Bonferroni's multiple comparison test was used for multiple comparisons between conditions within a dataset. For data in Fig. 2D, Fisher exact test was used. Statistics was performed in Prism.

We thank Gian Garriga and the C. elegans Genetics Center (CGC) for worm strains. CGC is funded by NIH Office of Research Infrastructure Programs (P40 OD010440).