The Rho guanine nucleotide exchange factor Trio is required for neural crest cell migration and interacts with Dishevelled

Directed cell migration controls a plethora of crucial events in embryonic development and adult homeostasis, including morphogenetic cell movements, wound healing and immune function, as well as tumor cell migration. Neural crest (NC) cells are embryonic multipotent cells, which have been likened to cancer metastasis and provide an excellent example for analyzing the molecular mechanisms that control cell migration (Theveneau and Mayor, 2011). NC cells migrate over long distances throughout the embryo where they contribute to the formation of various tissue types, including neurons, glia, cartilage, smooth muscle and pigment cells. Thus, NC cell migration needs to be precisely controlled, as defects can lead to severe congenital malformations, including craniofacial and heart defects, hearing loss and mental retardation (Trainor, 2014).

Collective chemotaxis, repellent guidance cues along with dynamic cell-matrix and cell-cell interactions synergize to regulate the directional migration of NC cells (Shellard and Mayor, 2016; Szabó and Mayor, 2018). Migrating cells extend protrusions in the form of lamellipodia and filopodia in the direction of migration, while they retract them at the trailing edge. Cell migration is directed by attractive signals such as Sdf1 or VEGFA (McLennan et al., 2010; Theveneau et al., 2013) as well as repulsive cues like semaphorins and ephrins (Gammill et al., 2007, 2006; Smith et al., 1997). These, sometimes contradictory, signals have to be integrated to allow for directional migration (Bajanca et al., 2019). Although attractive and repulsive cues originate from the surrounding tissue, including the placodes, NC cells also exchange information via transient cell-cell contacts. For example, they show contact inhibition of locomotion, a phenomenon whereby cells change their direction upon cell-cell contact formation (Abercrombie and Heaysman, 1953; Roycroft and Mayor, 2018). In the past decade, contact inhibition of locomotion has been shown to be crucial for directional migration of cranial NC cells (Carmona-Fontaine et al., 2008; Matthews et al., 2008; Theveneau et al., 2010, 2013). A number of molecular regulators of cranial NC migration have been identified, including: adhesion molecules, such as N-cadherin and cadherin-11; effectors of Wnt planar cell polarity (PCP) signaling, such as Frizzled 7 (Fz7) and Dishevelled (DVL); and the cell polarity protein Par3 (Becker et al., 2013; Carmona-Fontaine et al., 2008; Moore et al., 2013; Theveneau et al., 2010). These molecules synergize to control the activity of small GTPases of the Rho family, which then leads to retraction of protrusions at the cell-cell contact site and formation of new protrusions in the direction of migration.

The complex signaling network affecting directional migration of NC cells converges at the level of small GTPases of the Rho family. RhoA, Rac1 and Cdc42 are crucial regulators of cellular morphology and locomotion, and are activated by binding of GTP. The exchange of GDP to GTP is controlled by guanine nucleotide exchange factors (GEFs), which determine the temporal and spatial activity of Rho GTPases (Nguyen et al., 2018). The Rho GEF Trio is especially well suited to dynamically control protrusive activity of cranial NC cells. The protein was named Trio as it features two GEF domains of distinct specificity as well as a serine/threonine kinase domain (Debant et al., 1996). The N-terminal GEF1 domain activates Rac1 and RhoG, while the C-terminal GEF2 domain specifically acts on RhoA (Bellanger et al., 1998; Blangy et al., 2000; Debant et al., 1996). Furthermore, Trio exhibits an N-terminal putative lipid-transfer SEC14 domain, a variable number of spectrin-like domains, two Src-homology 3 domains, one immunoglobulin-like domain, and a C-terminal serine/threonine kinase domain. Thus, Trio is well suited to function as a signal integrator of various molecular stimuli during development and disease. Indeed, Trio has been shown to have an evolutionarily conserved role in the development of the nervous system and its expression is upregulated in different tumors (reviewed by Schmidt and Debant, 2014). Furthermore, de novo mutations of Trio have been identified in individuals with a combination of neurodevelopmental, skeletal and craniofacial abnormalities (Ba et al., 2016; Pengelly et al., 2016), the last being suggestive of defects in the development of the neural crest. In fact, a function of Trio in the control of NC migration is supported by its expression in migrating Xenopus cranial NC cells (Kratzer et al., 2019). Indeed, a number of molecules that interact with Trio and control cell migration have been identified (reviewed by Schmidt and Debant, 2014; van Rijssel and van Buul, 2012), including focal adhesion kinase, cell-adhesion molecules of the cadherin family and proteins involved in cytoskeletal dynamics (Backer et al., 2007; Bellanger et al., 2000; Charrasse et al., 2007; Kashef et al., 2009; Medley et al., 2003; Timmerman et al., 2015; Vanderzalm et al., 2009; Yano et al., 2011). Moreover, Trio interacts with Cadherin-11 (Cad11) and can rescue protrusion formation and migration of cranial NC cells in Cad11 morphant embryos (Kashef et al., 2009; Li et al., 2011). Recently, Trio has been shown to interact with the polarity protein Par3 (Landin Malt et al., 2019; Moore et al., 2013), which is required for contact inhibition of locomotion during NC migration (Moore et al., 2013). At NC cell-cell contacts, Par3 promotes microtubule catastrophe, likely by inhibiting Trio-dependent activation of Rac1. As Trio interacts with Cad11, which is localized in NC cell protrusions as well as at NC cell-cell contacts (Kashef et al., 2009), Trio may likely exert additional functions in the control of NC migration. Thus, in this study we further dissected the role of this unique molecule in NC migration.

First, we revisited Trio function in Xenopus NC cell migration by repeating previously published loss-of-function experiments (Moore et al., 2013). Trio morpholino antisense oligonucleotides (Trio MO) were targeted to cranial NC cells and NC development was analyzed by in situ hybridization for AP-2α, a known marker of cranial NC cells. As expected Trio loss of function severely impaired NC migration (Fig. 1A), whereas neither injection of the Trio MO nor a control MO (Co MO) affected the induction of AP-2α-positive NC cells (Fig. S1). Compared with the uninjected side of the embryo, the Trio MO-injected side showed severe defects in NC cell migration, while NC cell migration was not affected by injection of a control MO. This effect is specific to Trio loss of function as the cranial NC migration defects could be rescued by co-injection of a human Trio DNA, which lacks the sequence targeted by the Trio MO (Fig. 1A,B; Moore et al., 2013). Overexpression of Trio in control MO-injected embryos did not cause NC migration defects (data not shown). These results were further verified using a second Trio MO: Trio MO2 (Fig. 1A,B). In addition, we confirmed that both Trio MO inhibited embryonic Trio expression (Fig. 1C,D). Furthermore, to test whether Trio function is required in NC cells, we performed transplantation experiments. Transplants injected with control MO exhibited directional migration towards the pharyngeal pouches (Fig. 1E,F). In contrast, the migration of transplants from Trio morphants was severely inhibited (Fig. 1E,F). These defects were rescued by co-injection of Trio, which restored the directional migration of cranial NC cells.

One of the major derivatives of cranial NC cells is the craniofacial cartilage. Therefore, we asked whether Trio knockdown affects craniofacial morphogenesis as a consequence of disturbed NC cell migration. Indeed, Trio morphant embryos displayed severe defects in head cartilage morphology, as shown by Alcian Blue staining (Fig. 1G). Whereas control MO injected embryos exhibited bilateral symmetric cartilage structures, Trio MO injection led to loss of Meckel's cartilage and reduced posterior cartilage structures. Co-injection of human Trio rescued the craniofacial phenotype (Fig. 1G). Taken together, these results demonstrate that Trio is required for cranial NC cell migration in a cell-autonomous manner and is important for the formation of cartilage head structures in Xenopus embryos.

In order to identify the cellular function of Trio, we performed in vitro time-lapse analysis of cranial NC cells explanted on fibronectin. Wild-type NC cells injected with GAP43-GFP to mark the membrane (green) exhibited intense protrusive activity by forming filopodia and lamellipodia (Fig. 2A; Movie 1). In contrast, NC cells from Trio morphants completely lost their ability to form cell protrusions and showed membrane blebbing instead (Fig. 2A; Movie 2). Co-expression of human Trio DNA in morphant NC cells restored cell protrusion formation (Fig. 2A; Movie 3). This is also reflected by a significant increase in cell circularity of morphant NC cells compared with controls, which is rescued by co-expression of human Trio (Fig. 2B).

As Trio is an exchange factor for Rac1 and RhoA (Bellanger et al., 1998; Blangy et al., 2000; Debant et al., 1996), which modulate F-actin dynamics that are important for filopodia and lamellipodia formation (Hall, 1998), we co-injected NC cells with lifeact-mCherry to visualize F-actin dynamics. In wild-type NC cells, F-actin was enriched in stress fibers and in cell protrusions at the leading edge (Fig. 2C). In contrast, in Trio morphant NC cells, F-actin staining appeared more diffuse, but was also detected outlining the cell cortex of the blebs, consistent with a function in stabilization of the extending bleb (reviewed by Fackler and Grosse, 2008) (Fig. 2C, yellow arrowheads). In contrast, NC morphant cells co-injected with human Trio displayed the same cellular morphology as wild-type NC cells (Fig. 2C). Thus, Trio seems to be crucial for filopodia and lamellipodia formation in NC cells. This is further supported by Trio localization in explanted NC cells. NC cells expressing Trio-GFP together with GAP43-mCherry, showed Trio localization at the plasma membrane at cell-cell contacts, as well as cell protrusions (Fig. 2D, arrows). Taken together, the Trio loss-of-function phenotype as well as its subcellular localization supports a requirement of Trio for cell protrusion formation that is essential for cranial NC migration.

As cell protrusion formation depends on the activity of small GTPases of the Rho family, like Rac1, RhoA and Cdc42 (Jaffe and Hall, 2005), and as Trio is a known GEF, these small GTPases likely act downstream of Trio. Therefore, we first analyzed whether Trio activates Rac1, a known activator of cell protrusions (Hall, 1998). Rac1 pull-down assays were performed using Xenopus embryos that were injected with full-length Trio, the Trio GEF1 or GEF2 domain, or Trio MO (Fig. 3A,B). Overexpression of Trio and GEF1 significantly increased Rac1-GTP; however, the GEF2 domain also led to a minor increase in Rac1 activity. In contrast, knockdown of Trio by MO injection significantly decreased Rac1 activity (Fig. 3A,B).

Next, to elucidate whether the small GTPases of the Rho family act downstream of Trio, we tested whether constitutively active forms of RhoA, Rac1 or Cdc42 are, per se, able to rescue the Trio morphant phenotype. Embryos were injected with Trio MO in combination with either constitutive active RhoA, Rac1 or Cdc42, and NC cell migration was analyzed in whole embryos using AP-2α in situ hybridization (Fig. 3C,D). While RhoA and Rac1 alone, or in combination, were able to rescue the Trio morphant phenotype, Cdc42 did not. Overexpression of constitutive active RhoA, Rac1 or Cdc42 alone did not cause craniofacial defects (data not shown). Similar results were also obtained by analyzing protrusion formation in explants co-injected with GAP43-GFP and H2B-mcherry RNA to visualize the plasma membrane or nucleus, respectively. Co-injection of RhoA and Rac1 restored protrusion formation in Trio morphant NC cells, while Cdc42-co-injected cells showed a blebbing phenotype (Fig. 3E). Taken together, these data indicate that RhoA and Rac1, but not Cdc42, act downstream of Trio in protrusion formation and migration of cranial NC cells.

The Rho GEF Trio has two GEF domains: the N-terminal GEF1 domain mediates the exchange of GDP to GTP of Rac1 and RhoG, whereas the C-terminal GEF2 domain activates RhoA (Bellanger et al., 1998; Blangy et al., 2000; Debant et al., 1996). Given that RhoA as well as Rac1 can rescue the Trio morphant phenotype, one would expect the GEF1 as well as the GEF2 domain to be essential for migration and cell protrusion formation. To determine this, we co-injected the Trio GEF1 or GEF2 domain together with Trio MO, respectively. In situ hybridization for AP-2α revealed that co-injection of GEF2, but not GEF1, significantly restored in vivo NC migration (Fig. 4A,B). Moreover, co-injection of GEF2 restored lamellipodia and filopodia formation of Trio morphant NC cells, whereas co-expression of the GEF1 domain resulted in blebbing cells, reminiscent of the morphant NC cells (Fig. 4C). Taken together, these results indicate that the GEF2, but not the GEF1, domain of Trio is essential for filopodia and lamellipodia formation, and for cranial NC cell migration.

In search of molecules that may relay Trio signaling in NC migration, we explored the role of DVL. DVL is a downstream effector of non-canonical Wnt PCP signaling, which has previously been shown to control NC cell migration by affecting RhoA and Rac1 activity (Carmona-Fontaine et al., 2008; De Calisto et al., 2005). Therefore, we analyzed whether DVL is able to rescue the Trio morphant phenotype (Fig. 5). DVL harbors three major conserved domains, the DIX, PDZ and DEP domain, which have been implicated in different downstream signaling events (Boutros and Mlodzik, 1999; Wallingford and Habas, 2005). The DIX domain is used for canonical Wnt signaling, whereas the PDZ domain interacts with canonical and non-canonical Wnt effectors (Boutros et al., 1998; Habas et al., 2003; Itoh et al., 2000; Rothbächer et al., 2000). The DEP domain has been shown to activate PCP signaling, but recent data also indicate functions in canonical Wnt signaling (Gammons et al., 2016; Paclikova et al., 2017). To analyze whether different DVL activities are able to substitute for Trio function, embryos were injected with Trio MO in combination with either wild-type xDvl2 or the respective deletion constructs. Indeed, wild-type xDvl2, as well as a xDvl2 construct lacking the DIX domain (ΔDIX), were able to rescue Trio loss-of-function NC cell migration defects (Fig. 5A,B). In contrast, xDvl2 constructs lacking the PDZ (ΔPDZ) or DEP (ΔDEP) domain failed to rescue the Trio morphant phenotype. Overexpression of the different Dvl2 constructs alone did not affect NC migration (Fig. S2). Similar results were obtained analyzing explanted NC cells. Protrusion formation of Trio morphant NC cells was restored by co-expression of xDvl2 or the ΔDIX mutant, while cells co-expressing ΔPDZ or ΔDEP mutants retained a blebbing phenotype (Fig. 5C). Furthermore, as we had shown that Trio activates Rac1 (Fig. 3A), we analyzed whether DVL rescues the Trio morphant phenotype by activating Rac1. Indeed, DVL expression significantly rescues Rac1 activity in Trio morphants (Fig. 5D,E). Thus, these data suggest that DVL acts downstream of Trio – likely by activating Rac1 signaling.

Previously, we have shown that Trio interacts with Cad11, a calcium-dependent cell-adhesion molecule required for migration and protrusive activity of cranial NC cells (Kashef et al., 2009). Furthermore, Trio expression was able to rescue the Cad11 morphant phenotype (Kashef et al., 2009). Thus, we analyzed whether DVL is also able to rescue the Cad11 loss-of-function phenotype. As previously published (Kashef et al., 2009), Cad11 loss of function inhibited NC cell migration in whole embryos and caused blebbing of explanted NC cells (Fig. 6A-C). Co-expression of xDvl2 significantly rescued the migration defects of Cad11 morphant cells (Fig. 6A,B). Injection of the ΔDIX construct improved the migration defects of Cad11 morphants, but this effect was not significant. In contrast, co-expression of the ΔPDZ or ΔDEP mutants did not rescue the Cad11 morphant phenotype (Fig. 6A-C). Similar effects were observed if cranial cartilage formation was analyzed by Alcian Blue staining of tadpole embryos. While Cad11 MO-injected embryos showed severe defects in cartilage formation on the injected side, these defects were rescued by co-injection of xDvl2 or its ΔDIX mutant (Fig. 6D). In contrast, co-injection of the ΔPDZ or ΔDEP mutants did not restore cranial cartilage formation in Cad11 morphants. Thus, these data indicate that signaling via the DEP and PDZ domains of DVL is able to rescue Cad11 loss of function.

Activation of Wnt PCP signaling is accompanied by membrane recruitment of DVL, and Xenopus ectodermal explants have previously been used to demonstrate Fz-mediated membrane recruitment of DVL (Axelrod et al., 1998; Medina et al., 2000; Medina and Steinbeisser, 2000; Rothbächer et al., 2000; Sheldahl et al., 2003). Therefore, we used this system to analyze whether Cad11 interacts with DVL and recruits this protein to the plasma membrane. Ectodermal explants were injected with myc-tagged xDvl2 and GAP43-mCherry RNA to label the plasma membrane (Fig. 7A). xDvl2 localization was detected in the cytoplasm using confocal microscopy. Co-expression of Fz7-mCherry efficiently recruited xDvl2 to the plasma membrane and intracellular xDvl2 was significantly reduced compared with ectodermal explants expressing only xDvl2 (Fig. 7A,B). In contrast, Cad11-mCherry failed to translocate xDvl2 to the plasma membrane (Fig. 7A,B).

As DVL recruitment to the membrane correlates with its hyperphosphorylation (Rothbächer et al., 2000), we used this readout to further verify our findings. Lysates of ectodermal explants expressing myc-tagged xDvl2 alone or in combination with Cad11 were analyzed by western blotting. Explants expressing xDvl2 alone or together with Cad11 showed a xDvl2 band representing non- or poorly phosphorylated xDvl2 (Fig. 7C). However, a prominent high molecular weight band, representing hyperphosphorylated xDvl2, was detected if Fz7 was co-expressed (Fig. 7C). Hyperphosphorylated xDvl2 was also detected in lysates injected with Fz7 RNA in combination with Cad11 MO or control MO. Thus, Cad11 gain or loss of function does not seem to affect DVL hyperphosphorylation. These data suggest that, although DVL rescues the Cad11 phenotype, these proteins likely do not directly interact. To analyze whether the Cad11 protein physically interacts with DVL, co-immunoprecipitation experiments were performed. Myc-tagged Dvl2 was co-expressed with GFP-tagged Cad11 or β-catenin, and Cad11 was precipitated using anti-GFP antibodies. Although β-catenin co-precipitated with Cad11, Dvl2 did not, further confirming the theory that Cad11 does not directly interact with DVL (Fig. 7D).

As DVL does not seem to intersect with the Cad11/Trio signaling cascade at the level of Cad11, we analyzed whether DVL interacts with Trio. Indeed, co-immunoprecipitation experiments showed that HA-tagged Trio was able to co-precipitate Flag-tagged DVL2 in HEK293 lysates (Fig. 8A). In order to map the interaction domain in Trio, we performed further immunoprecipitation experiments expressing its GEF1 or GEF2 constructs (Fig. 8B). Intriguingly, both (i.e. GEF1 or GEF2) domains of Trio revealed some degree of interaction; however, the GEF2 domain appeared to be more effective in precipitating Dvl2 in comparison with GEF1 (Fig. 8B). To further characterize the interaction of DVL with Trio and its truncated variants, we performed immunofluorescence analysis in human HEK293 cells. Full-length DVL is well known to localize into cellular puncta and has a capacity to recruit interacting proteins into these puncta (Schwarz-Romond et al., 2005; Smalley et al., 2005). In absence of DVL, Trio as well as its GEF1 and GEF2 domains show an even cytoplasmic localization in HEK293 cells (Fig. 8C,D). However, co-expression of DVL2 completely recruited full-length Trio as well as the Trio GEF2, but not the Trio GEF1 domain, to DVL2-positive dots (Fig. 8E). This suggests that DVL interacts predominantly with the Trio GEF2 domain. In line with the co-immunoprecipitation experiment in Fig. 8B, some degree of colocalization was also detected between GEF1 and DVL2 (white arrows for even localization and yellow arrows for the slight punctae phenotype in Fig. 8E). Taken together, the GEF2 domain of Trio seems to be sufficient for the interaction with DVL, but the GEF1 domain presumably also plays a non-negligible role.

In order to map the interaction domain in DVL, we employed a set of DVL3 truncation mutants (Fig. 8F) (Angers et al., 2006). Trio can efficiently pull down all N-terminally truncated constructs (Fig. 8G), suggesting that Trio interacts with the DVL DEP domain and C-terminus. To further confirm this point for DVL2, we co-transfected DVL2 DEP/C-terminus construct (amino acids 433-736 in human DVL2) with full-length Trio, the Trio GEF1 or the Trio GEF2 domain (Fig. 8H). Indeed, we found that the DEP/C-terminus is sufficient to bind Trio or Trio GEF2 but, unsurprisingly, not GEF1. In summary, these results propose that DVL can interact with Trio and that this binding is mediated mainly by the Trio GEF2 and DVL DEP/C-terminus. Nevertheless, the supportive role of the GEF1 domain of Trio and other parts of DVL (especially the region between PDZ and DEP) cannot be omitted. Taken together, our data suggest that Trio interacts and signals via DVL to control protrusion formation and migration of cranial NC cells.

Migrating NC cells receive multiple signaling inputs, which together need to be relayed to the cytoskeleton to enable directional migration. The Rho GEF Trio is ideally suited to fine-tune signaling, as it contains three different signaling hubs – hence the name Trio – that mediate activation of a number of downstream targets, including different members of the Rho GTPase family. In this study, we have analyzed the function of the Rho GEF Trio in NC development and demonstrate that it is required for protrusion formation and migration of cranial NC cells, as well as development of the craniofacial cartilage. We propose a model (Fig. 9) whereby Trio enables protrusion formation by interacting with DVL to activate Rac1.

(1) We find that Trio interacts with DVL via the DEP/C-terminal domain of DVL, as well as the GEF2 domain of Trio.

(2) DVL rescued protrusion formation as well as in vivo migration of Trio morphant NC cell. This is likely mediated by DVL-mediated activation of Rac1, because DVL restores Rac1 activity in Trio morphant embryos.

(3) Rescue experiments demonstrate that the Trio GEF2 domain, which binds DVL, is sufficient to rescue Trio loss of function.

The Trio/DVL/Rac1 signaling module likely functions downstream of Cad11, because all these components are able to rescue Cad11 loss of function and a biochemical interaction of Cad11 and Trio has previously been demonstrated (Kashef et al., 2009). Thus, these data reveal a novel signaling function for Trio via DVL in protrusion formation of NC cells.

Both Trio and DVL have previously been implicated in the control of NC migration; however, functional interaction of both proteins has not been demonstrated yet. DVL is a key molecule of the Wnt signaling pathway and a known player of Wnt PCP signaling (Gao and Chen, 2010). In Xenopus, xDvl1 and xDvl2 are expressed in migratory cranial NC cells (Gray et al., 2009) and different DVL deletion mutants have been used to demonstrate a role for non-canonical Wnt signaling in the regulation of NC migration (De Calisto et al., 2005). Furthermore, membrane recruitment of DVL – indicative of activation of non-canonical Wnt PCP signaling – has been demonstrated in NC cells undergoing contact inhibition of locomotion (Carmona-Fontaine et al., 2008). Like DVL, Trio has also been implicated in the control of NC cell migration. Trio is expressed in Xenopus migratory cranial NC cells (Kratzer et al., 2019) and has been shown to interact with Cad11, a known regulator of NC cell migration (Becker et al., 2013; Borchers et al., 2001; Kashef et al., 2009; Langhe et al., 2016). Moreover, Trio interacts with Par3, a polarity protein required for contact inhibition of locomotion (Moore et al., 2013). In the presented study, we provide evidence for a functional and physical interaction of Trio with DVL. First, we note that DVL rescues Trio loss of function, an activity requiring the PDZ and DEP domains of DVL. Second, we show that both proteins interact in human cells via the DEP/C terminus of DVL and the GEF2 domain of Trio. This interaction likely takes place in the cytoplasm, as full-length Trio and its GEF2 domain localized to intracellular DVL-positive punctae – possibly representing signalosomes (Gammons et al., 2016) – in HEK293 cells. Furthermore, we did not find evidence for the formation of a trimeric complex of Cad11/Trio and DVL, as Cad11 did not interact with DVL or recruit it to the plasma membrane. In addition, we also failed to observe hyperphosphorylation of DVL – either in presence of Cad11 or Trio (Fig. S3). Thus, the Trio/DVL interaction likely does not require Cad11; however, Cad11 likely activates the Trio/Dvl module by interacting with Trio.

The multifunctional Rho GEF Trio may play different roles, depending on subcellular localization. In cell protrusions Trio can be part of a complex with Cad11, which is prominently localized in cell protrusions (Kashef et al., 2009). As we find that Trio activates Rac1 at stages of cranial NC cell migration, this complex likely supports protrusion formation by activating Rac1. Furthermore, DVL interacts with Trio and rescues protrusion formation and migration of Cad11 and Trio morphants, indicating that Trio signaling via DVL and Rac1 is active in protrusion formation. Moreover, Trio loss-of-function decreases embryonic Rac1 activity and this can be rescued by DVL expression. Indeed, DVL has been shown to interact and activate Rac1 (Cajanek et al., 2013; Habas et al., 2003), e.g. by interacting with the Rac1 GEF Tiam (Cajanek et al., 2013). Thus, DVL may activate Rac1 by acting as a scaffold to bring Rho GEFs, such as Trio, in close proximity to Rac1. This may also explain why Trio alone cannot rescue the decrease in Rac1 activity of DVL morphants (Fig. S4). Surprisingly, although the GEF1 domain activates Rac1, it is not sufficient to rescue the Trio morphant phenotype. In contrast, the GEF2 domain, which binds DVL, is able to rescue Trio loss of function. This suggests that – in addition to activation of Rac1 signaling – interaction with DVL may have additional qualities that are required to stimulate NC migration. For example, NC migration is probably not exclusively mediated by Rac1 activation, but may also involve activation of RhoA, because constitutive active Rac1 as well as RhoA rescue protrusion formation and migration of Trio morphant NC cells. Whether Rac1 and RhoA activation are two independent events or are caused by the ability of small GTPases of the Rho family to bidirectionally affect their activity (Burridge and Wennerberg, 2004) is currently unclear. However, we observed that embryonic RhoA activity is relatively low at stages of cranial NC migration. Thus, loss of function of Trio does not lead to a prominent decrease, although ectopic expression of Trio or its GEF2 domain activates RhoA signaling (Fig. S5). Taken together, these data suggest that Trio promotes NC protrusion formation by interacting with DVL. This leads to activation of Rac1, but may also affect other yet unknown signaling mechanisms.

Although our data suggest Trio activity is required in nascent NC cell protrusions, the situation may be different at NC cell-cell contact zones. Moore et al. showed that Par3 is required for contact inhibition of locomotion and promotes microtubule catastrophe by inhibiting Trio Rac1-GEF activity at cell-cell contacts (Moore et al., 2013). Thus, at the cell-cell-contact zone, Par3 interacts with Trio, thereby preventing it from activating Rac1. As Rac1 promotes microtubule stability, inhibition of Trio results in microtubule catastrophe at cell-cell contact sites, a reversal of cell polarity and a change in the direction of migration. Although Moore et al. suggest that Par3 inhibits Trio GEF activity, a recent report proposes that Par3 also exerts positive effects on Trio (Landin Malt et al., 2019). In the inner ear, Par3 is essential for planar cell polarity of hair cells and interacts with Trio to activate a Rac1-Pak signaling pathway. In this scenario, Par3 regulates the activity of microtubule-associated proteins and contributes to the stability of microtubules. Whether Par3/Trio signaling intersects with Cad11/Trio signaling and whether DVL plays a role in such a signaling event is currently unknown. Taken together, these data suggest that Trio has different functions in NC cell migration depending on the subcellular localization. Future research will have to dissect the underlying distinct signaling activities, as well as the molecular mechanisms that accomplish their control. As Trio, Cad11 and DVL are expressed in a variety of different cell types, our results may uncover a general function of these molecules in mediating cell migration, and may contribute to a better understanding of cell migration and cell invasion in a variety of diseases.

Trio and Cad11 morpholino antisense oligonucleotides were used as published previously (Kashef et al., 2009). An additional Trio MO, Trio MO2 (5′-ATCCTTAGAGTTCCCCAACCCTCCA-3′), targeted to the 5′-UTR region was generated by Gene Tools. As a control, a standard control MO from Gene Tools was used. The following plasmids were used for RNA or DNA Xenopus injection: Cad11 (Borchers et al., 2001), Cad11-myc (Langhe et al., 2016), Cad11-GFP (Kashef et al., 2009), Fz7 (Medina et al., 2000), Fz7-myc (Winklbauer et al., 2001), GAP43-mCherry and GAP43-GFP (Moriyoshi et al., 1996), H2B-mCherry (Kashef et al., 2009), Lifeact-mCherry (Riedl et al., 2008), constitutively active RhoA, Rac1 and Cdc42 (Schambony and Wedlich, 2007), Trio-HA (Kashef et al., 2009), Trio-GEF1-HA and Trio-GEF2-HA (Debant et al., 1996), Trio-GFP (Moore et al., 2013), xDvl2-GFP (Yang-Snyder et al., 1996), xDvl2-myc (Sokol, 1996), and xDvl2ΔDIX, xDvl2ΔPDZ and xDvl2ΔDEP (Miller et al., 1999) cloned in pCS2+. For transfection of human cells, DVL3 wt-Flag and truncated variants (Angers et al., 2006) and DVL2-Flag (Narimatsu et al., 2009) were used.

For cloning of the Xenopus GEF constructs, the coding sequence, containing the DH, PH and SH3 domain of Trio-GEF1 (3784-5197 bp) or GEF2 (5791-8016 bp), was amplified by PCR from cDNA of Xenopus laevis embryos. C-terminal eGFP-tagged Xenopus GEF constructs were generated using the following primers with ClaI/XhoI restriction sites: GEF1-forward, 5′-ATCGATATGGGTTCCGAAGTGAAGCTTCG-3′; GEF1-reverse, 5′-CTCGAGGACAGAGAGGGAATCTTTGTGGT-3′; GEF2-forward, 5′-ATCGATATGGGTGACAGTAGTAGCCCATCG-3′; GEF2-reverse, 5′-CTCGAGAAACTCTGGGGGAGCATCATA-3′. The PCR product was cut with ClaI and XhoI, and cloned into the respective sites of pCS2+/eGFP.

Xenopus laevis embryos were obtained by in vitro fertilization and staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967). All procedures were performed according to the German animal use and care law (Tierschutzgesetz) and approved by the German state administration Hesse (Regierungspräsidium Giessen). RNA was synthesized in vitro using the mMessage mMachine Kit (Ambion). If not indicated otherwise, injections were performed into one of the animal dorsal blastomeres of 8-cell stage embryos to target the NC. Fluorescein-dextran (4 ng 10,000 MW, ThermoFisher Scientific), GAP43-GFP RNA (500 pg), GAP43-mcherry RNA (500 pg) or H2B-mcherry RNA (400 pg) were used as lineage tracer, membrane or nuclear marker, respectively. For whole-mount in situ hybridization (Harland, 1991), digoxigenin-labeled antisense RNA probes were generated according to the manufacturer's instructions (Roche) using AP-2α (Luo et al., 2003). NC cell migration was rated as defective when the hyoid and branchial branches did not migrate at least half of the distance compared with the control side. Analysis were performed double-blind, whereby the person scoring the neural crest defects was not aware of the respective treatments of the embryos. Significances were tested with one-way ANOVA followed by Tukey's or Dunnett's post-hoc multiple comparison test using Prism8 (GraphPad). NC transplantations, NC explants and cartilage staining were performed as previously described (Kashef et al., 2009).

Ectodermal explants (animal caps) were dissected at stage 8/9 and fixed at stage 11 in 4% paraformaldehyde. Immunostaining was carried out as previously described (Yang-Snyder et al., 1996) with anti-Myc antibodies (9E10, M5546 Sigma, 1:2000) and anti-mouse Alexa Fluor 594 (A-11005 ThermoFisher Scientific, 1:400). Dvl fluorescence intensity was quantified using ImageJ. To this end, a straight line was drawn from the cell membrane (indicated by the mCherry or RFP signal in the red channel) to the cytoplasm, the cell nucleus was avoided. The line was transferred to the green channel of the Dvl-myc signal and the fluorescence intensity was measured. A line was drawn of three random cells per explant. Statistical analysis was performed with Prism using a two-way ANOVA. For western blot analysis, 20 animal caps were homogenized at stage 11 in 200 µl lysis buffer [50 mM Tris (pH 7.5), 150 mM NaCl, 0.5% Nonidet P40] containing protease and phosphatase inhibitor cocktail tablets (Roche). Myc-tagged DVL protein was detected by western blot using primary anti-Myc (9E10, M5546 Sigma, 1:2000) and anti-actin (MAB1501, Sigma, 1:1000) and secondary anti-mouse-IRDye 680CW (LI-COR, 926-32212, 1:7500). The infrared signals were detected by LI-COR Odyssey Fc System and analyzed by Image Studio Software (LI-COR).

Trio expression in whole embryos was analyzed by western blotting using lysates of stage 12 embryos. Trio protein was detected using anti-Trio antibodies (HPA064664, Sigma, 1:50). As a loading control, actin was detected using an anti-actin (MAB1501, Sigma, 1:1000).

Embryos were imaged using a Leica MZ16F stereo-microscope. NC and ectodermal explants were analyzed using spinning disc confocal microscopy [Axio Observer Z1 with 63× or 40× plan apochromat NA 1.4 oil objective using AxioVision 4.8.2 or ZEN software (Zeiss)] or fluorescence microscope [Axio Observer Z1 with Apotome and 63× plan apochromat NA 1.4 oil objective using AxioVision 4.8 (Zeiss)]. Cell circularity of NC cells was measured by ImageJ using the formula 4π(area/perimeter²). Statistical analysis was performed using Prism8 (GraphPad). Normality of datasets was tested using D'Agostine and Pearson test, Shapiro-Wilk test and Kolmogorov–Smirnov test. Datasets that did not show normal distribution were compared using non-parametric ANOVA with Kruskal–Wallis multiple comparison test.

HEK293 cells were cultivated on 25 cm² cell culture flasks and transfected with DNA plasmids using JetPEI (Polyplus Transfection) according to the manufacturer's instructions. 48 h after transfection, the cells were washed twice with ice-cold 1× PBS, scraped and lysed in NOP-lysis buffer [150 mM NaCl, 10 mM HEPES (pH 7.4), 2 mM EDTA, 1% Nonidet P40 and 0.1% SDS] containing protease and phosphatase inhibitor cocktail tablets (Roche). Co-immunoprecipitation with Protein A-Sepharose CL-4B (GE Healthcare) (Cad11-Dvl2 Co-IP) was carried out as previously reported (Berger et al., 2017) using anti-GFP (Abcam, ab290, 1 µg) for antigen-antibody reaction. Co-immunoprecipitations with Pierce Co-Immunoprecipitation Kit (GEF-DVL Co-IPs) were performed according to the manufacturer's instructions using anti-GFP (Abcam, ab290, 1:200) antibodies. Antibody-resin coupling was carried out for 1 h and antigen-antibody reaction for 1 h at 4°C. Proteins were detected by western blot using different antibodies: anti-GFP (Abcam, ab290, 1:2000), anti-Myc (9E10, M5546 Sigma, 1:2000), anti-Flag (Sigma, F7425, 1:1000) and anti-HA.11 (MMS-101R, Covance, 1:1000) as primary antibodies and anti-rabbit-IRDye 800CW (926-32213, LI-COR, 1:7500), anti-mouse-IRDye 680RD (926-68072, LI-COR, 1:7500) and anti-rabbit IgG (whole molecule) peroxidase antibody (Sigma, A0545 1:5000) as secondary antibodies. Western blot samples were developed digitally using the chemiluminescence documentation system FusionSL (Vilber-Lourmat) or Infrared signals were detected by LI-COR Odyssey Fc System and analyzed by Image Studio Software (LI-COR).

Embryos were injected at the one-cell stage with 500 pg Cad11-GFP, 75 pg xDvl2-myc and 100 pg β-catenin-myc RNA (Fagotto et al., 1996), and cultured until stage 11. For co-immunoprecipitation, 50 embryos of each condition were lysed in 500 µl NOP-lysis buffer containing protease inhibitor (Roche). Co-immunoprecipitation was performed with Dynabeads Protein G Immunoprecipitation Kit according to the manufacturer's protocol using anti-GFP antibodies (Abcam, ab290, 1:250). Proteins were detected by western blot using the following antibodies: anti-GFP (Roche, 11814460001, 1:1000) and anti-Myc (Abcam, ab19234, 1:1000), anti-mouse-IRDye 680CW (LI-COR, 926-32212, 1:7500) and anti-goat-IRDye 800CW (LI-COR, 926-32214, 1:7500). Infrared signals were detected by LI-COR Odyssey Fc System and analyzed by Image Studio Software (LI-COR).

HEK293 cells (HEK293T ATCC-CRL-11268, tested regularly for mycoplasma contamination) were seeded on gelatin-coated coverslips in 24-well plates and were transfected the next day with 100 ng of each plasmid. 24 h later, cells were fixed in fresh 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, blocked with PBS/BSA/Triton/Azide buffer (PBTA) [3% (w/v) BSA, 0.25% Triton, 0.01% NaN₃] for 1 h, and incubated overnight with primary antibodies in PBTA at 4°C. The next day, the coverslips were washed in 3× PBS and incubated for 2 h with secondary antibodies conjugated to Alexa Fluor 488 (Invitrogen, A11001, 1:1000) and Alexa Fluor 594 (Invitrogen, A11058, 1:1000). Subsequently, the samples were washed 4× with PBS and 1× DAPI (1:5000 in PBS), and the coverslips were mounted on microscopic slides. The samples were then visualized on an Olympus Fluoview 500 confocal laser scanning microscope IX71 using a 100× oil objective. The primary antibodies used were as follows: anti-FLAG M2 (Sigma, F1804, 1:1000) and anti-HA (Abcam, ab9110, 1:1000).

Xenopus embryos were co-injected with RNA or MO together with 4 ng fluorescein-dextran to distinguish injected from uninjected embryos. Control embryos were injected only with fluorescein-dextran. Twenty-five stage 20-22 embryos were collected for each condition and lysed in 250 µl Rac-RIPA buffer [50 mM Tris (pH 7.2), 150 mM NaCl, 10 mM MgCl₂, 1% Triton X-100, 0.1% SDS, complete protease inhibitor mix (Roche)]. As a positive control, lysates of 25 wild-type embryos of stage 20/21 were incubated with GTPγS (Sigma-Aldrich) according to the manufacturer's instructions. Rho or Rac pull-down assays were performed as described previously (Benard et al., 1999; Ren et al., 1999). Samples were loaded on 12% SDS PAGE gels and detection of proteins was carried out using anti-Rac1 (1:1000, mouse, BD Biosciences 610651), anti-RhoA (1:500, mouse, Santa Cruz sc-418) and anti-mouse-IRDye 800 (1:7500, Li-COR 926-68072) antibodies. Significances were tested with one-way ANOVA.

We thank Marisa Heipel for excellent assistance with the GTPase assays, and Roberto Mayor and Dietmar Gradl for providing plasmids.