Celsr1 suppresses Wnt5a-mediated chemoattraction to prevent incorrect rostral migration of facial branchiomotor neurons

During vertebrate brain development, differentiating neurons frequently move considerable distances in order to establish complex circuits required for cognitive and motor function. Within the hindbrain, facial branchiomotor (FBM) neurons, which innervate the facial and jaw muscles, migrate caudally from rhombomere 4 (r4) to form a motor nucleus in r6 (Chandrasekhar, 2004; Guthrie, 2007; Song, 2007). Several genes, including many encoding core components of the Wnt/planar cell polarity (PCP) pathway (Gray et al., 2011; Butler and Wallingford, 2017), are necessary for caudal migration of FBM neurons (Vivancos et al., 2009; Qu et al., 2010; Glasco et al., 2012; Yang et al., 2014). In contrast, only one gene thus far has been shown to regulate directionality. In mice deficient for the atypical cadherin and PCP component Celsr1, some FBM neurons inappropriately migrate rostrally into r3 and r2 instead of caudally (Qu et al., 2010). Tissue-specific knockouts indicate that Celsr1 function is not required in FBM neurons (Qu et al., 2010), and that it functions non-cell autonomously within the ventricular zone rostral to r4 to prevent inappropriate rostral migration (Glasco et al., 2016). Dye-labeling experiments suggest that rostral migration of FBM neurons in Celsr1 mutants is not the consequence of a random loss of cell polarity, but instead results from the loss of a local guidance cue (Glasco et al., 2016). Intriguingly, Wnt5a, which encodes a putative chemoattractant of FBM neurons (Vivancos et al., 2009), exhibits an overlapping expression pattern with Celsr1 in the rostral hindbrain. Here, we test an unconventional model for how Celsr1 regulates Wnt5a to ensure caudal migration of FBM neurons.

Celsr1 is expressed in a restricted fashion in the hindbrain at E10.5 and later ages (Fig. S1A) (Qu et al., 2010). At E11.5, Celsr1 is expressed at all rostro-caudal levels of the hindbrain (Fig. 1A), but only within a small domain of the ventricular zone immediately dorsal and adjacent to the floor plate (Fig. S1C) (Qu et al., 2010). Importantly, Celsr1 is not expressed in FBM neurons (Fig. 1A), consistent with a non-autonomous role for Celsr1 in regulating FBM neuron migration (Qu et al., 2010). Wnt5a is expressed at a low level and in a rostrally restricted domain at E10.5 (Fig. S1B). This expression pattern is more defined by E11.5, with a rostro-medial domain extending up to the r3-r4 boundary, and a caudal domain evident from the r4-r5 boundary (Fig. 1B; Fig. S1D) (Vivancos et al., 2009; Glasco et al., 2016). These data indicate that Celsr1 and Wnt5a are expressed in overlapping patterns in midline tissues in r3 and r2 before the onset (E10.5) and during the early stages (E11.5) of FBM neuron migration (Fig. 1C; Fig. S1). Even though the rostral Wnt5a domain extends up to and abuts the rostral most FBM neurons in r4 (Fig. 1B), these neurons never migrate into r3 (toward the putative chemoattractant) in wild-type embryos but do so extensively in Celsr1 knockouts (Qu et al., 2010). We propose that putative chemoattraction of FBM neurons located near the r3-r4 boundary toward the rostral source of Wnt5a in r3 is normally suppressed by Celsr1 (Glasco et al., 2016) (Fig. 1C). In Celsr1 knockouts, this suppression is relieved, resulting in FBM neurons migrating into r3 and r2 (Fig. 1D; Fig. 2A,C). This model predicts that the rostral migration seen in Celsr1 single mutants would not occur in Celsr1; Wnt5a double mutants because the putative Wnt5a chemoattractive source is lost (Fig. 2D). We have tested and confirmed this prediction (Fig. 2).

E12.5 embryos were collected from Celsr1^+/KO; Wnt5a^+/KO double heterozygote crosses, genotyped and processed for Tbx20 in situ hybridization. The four phenotypic classes were recovered in roughly Mendelian ratios (Fig. S2), with a slight over-representation of double mutants, which often exhibited open neural tubes characteristic of Wnt/PCP mutants such as Looptail (Vangl2^−/−) (Curtin et al., 2003; Glasco et al., 2012). FBM neurons migrated caudally in Wnt5a^KO/KO embryos obtained from single (Fig. S3F,G) or double heterozygote crosses (Fig. 2F,I) in a similar fashion to wild-type embryos (Fig. 2E,I), although there was a slight broadening of the lateral pathway caudal to r5, suggestive of precocious dorsolateral migration (Vivancos et al., 2009). FBM neurons migrated rostrally into r3 and r2 in Celsr1^KO/KO; Wnt5a^+/+ embryos obtained from these crosses (Fig. 2G,I), as in Celsr1^KO/KO embryos obtained from single heterozygote matings (Fig. S3C,G) (Qu et al., 2010). Reducing Wnt5a dose by removing one wild-type allele slightly decreased the incidence of rostral migration in Celsr1^KO/KO; Wnt5a^+/KO embryos (Fig. 2I). Importantly, eliminating Wnt5a function almost completely suppressed inappropriate rostral migration in Celsr1^KO/KO; Wnt5^KO/KO embryos (Fig. 2H,I) as predicted by our model (Fig. 2D). Suppression of rostral migration in double mutants may reflect the inability of FBM neurons to migrate into r3 due to potential defects in the anterior hindbrain of the mutants. However, the efferent neuron marker Gata3 (Karis et al., 2001) and rhombomere marker Egr2 (Schneider-Maunoury et al., 1993) were both expressed normally in r3 in double mutants (Fig. S4), ruling out non-specific causes for the loss of rostral migration. We conclude that FBM neurons migrate rostrally in Celsr1^KO/KO embryos due to an unmasking of the Wnt5a chemoattractant in r3.

To test further whether Wnt5a can act as a chemoattractant in the anterior hindbrain, we placed beads coated with recombinant Wnt5a in the rostral hindbrains of E11.5 SE1::GFP wild-type explants cultured in vitro. Similar to previous observations (Vivancos et al., 2009), FBM neurons migrated rostrally in significant numbers from r4 towards the Wnt5a beads in r3 within 24 h (strong attraction) in the majority of explants (Fig. S5D,G), while PBS-coated beads elicited no effect (Fig. S5B,G). Importantly, the number of explants exhibiting rostral migration (strong attraction) was substantially reduced when cultured from SE1::GFP Dvl2^KO/KO embryos (Fig. S5F,G), bordering on significance (P∼0.07), suggesting that the observed rostral migration is dependent on Disheveled (Dvl) function downstream of Wnt5a-Frizzled signaling. The incomplete effect on rostral migration in Dvl2 mutant explants may reflect functional redundancy with Dvl1 and Dvl3, which are also expressed in E12.5 hindbrains (Wang et al., 2006). We addressed this issue further by comparing rostral migration towards Wnt5a beads in a sensitized background by treating wild-type and Dvl2 mutant explants with the ROCK inhibitor Y27632, which attenuates Wnt signaling downstream of Dvl and reduces FBM neuron migration towards Wnt5a beads in a similar explant assay (Vivancos et al., 2009). As expected, rostral migration toward Wnt5a beads was strongly reduced in Y27632-treated Dvl2 mutant explants compared with Y27632-treated wild-type explants and bordered on significance (Fig. S5G).

The rostral migration of FBM neurons in the Wnt5a bead explant experiments is essentially a Wnt5a gain-of-function phenotype and is consistent with our model (Fig. 1D), as excess Wnt5a released by the beads presumably overcomes the suppressive effects of Celsr1 expressed rostral to r4. Therefore, we predicted that Wnt5a gain of function in the rostral hindbrain in vivo would mimic the bead experiment phenotype, but more importantly enhance the rostral migration of FBM neurons seen in Celsr1^KO/KO embryos (Fig. 3B,C). To generate Wnt5a gain of function in vivo (Wnt5a^GOF), we employed a Wnt5a loxP-STOP allele inserted into the ROSA26 locus (Cha et al., 2014) crossed to Egr2-Cre (Voiculescu et al., 2000), resulting in ectopic and robust Wnt5a expression throughout r3 and r5 (Fig. S6C). In Celsr1^+/+; Wnt5a^GOF embryos, a significant number of Tbx20-expressing FBM neurons were located in r3 (Fig. 3G compared with control Fig. 3D). In Wnt5a^GOF embryos, the r2-r3 boundary was demarcated by the caudal edge of the trigeminal motor nucleus, and the rostro-caudal length of r3 was defined as the average length of r3 measured in control embryos (Fig. 3D,E; Fig. S6A,B). In Celsr1^+/KO; Wnt5a^GOF embryos, rostral migration of FBM neurons was even more evident with neurons spilling over into r2 (Fig. 3H compared with Fig. 3E). In Celsr1^KO/KO; Wnt5a^GOF embryos, FBM neurons migrate extensively into r3 (Fig. 3I compared with Fig. 3F). Interestingly, although FBM neurons often migrate rostrally into r2 and beyond in Celsr1^KO/KO; Wnt5a^+/+ embryos (Fig. 3F), rostrally migrating neurons, although more numerous, appear to remain largely within r3 in Celsr1^KO/KO; Wnt5a^GOF embryos (Fig. 3I), likely because r3 is a potent source of Wnt5a chemoattractant due to ectopic expression (Fig. S6C). We quantified the extent of rostral migration in various genotypes (Fig. S7) and noticed a strong additive effect (Fig. 3J). Although ectopic Wnt5a expression in r3 did not lead to substantially increased rostral migration in Celsr1^+/KO compared with Celsr1^+/+ embryos, there was a dramatic increase in rostral migration in Celsr1^KO/KO compared with Celsr1^+/KO embryos (Fig. 3J), demonstrating that Celsr1 normally suppresses the chemotactic activity of Wnt5a to prevent inappropriate rostral migration.

Our genetic studies establish definitively a role for Wnt5a as a chemoattractant for FBM neurons. Although previous studies using hindbrain explants and coated beads suggested that Wnt5a may function as a chemoattractant to guide FBM neurons caudally into r5 and r6, the Wnt5a knockout phenotype was subtle and suggestive (Vivancos et al., 2009). Our Wnt5a loss- and gain-of-function genetic experiments demonstrate conclusively that FBM neurons can be chemoattracted in vivo to a Wnt5a source in the rostral hindbrain. However, our data do not provide further insight into whether Wnt5a also functions as a chemoattractant to induce FBM neurons to migrate caudally, as we did not observe pronounced effects on the caudal migratory stream of FBM neurons in the Wnt5a loss- and gain-of-function conditions. There was a marked reduction in the size of the caudal migratory stream in Celsr1^KO/KO; Wnt5a^GOF embryos, likely due to a substantial increase in the number of FBM neurons migrating rostrally into r3. However, caudally migrating neurons did not accumulate preferentially in r5 in Wnt5a^GOF embryos, as might be expected due to overexpression of Wnt5a in r5 (Fig. S6C). Importantly, enhancement of Wnt5a-mediated chemoattraction in Celsr1^KO/KO; Wnt5a^GOF embryos further validated the role of Celsr1 in suppressing Wnt5a function in the rostral hindbrain.

Wnt5a has demonstrated roles in the nervous system as a chemoattractant for hindbrain commissural growth cones acting via Fzd receptors (Lyuksyutova et al., 2003; Onishi et al., 2013) and as a chemorepellent for corticospinal and callosal growth cones via Ryk receptors (Hutchins et al., 2011). We provide strong genetic evidence for a chemoattractive function for Wnt5a in the migration of neuronal cell bodies. Our Wnt5a bead studies suggest that this function depends on Dvl signaling, which is downstream of the Fzd receptor. Although a chemoattractive role for Wnt5a in regulating neuronal migration has not been previously reported, Wnt5a-mediated signaling has been implicated in the migration of several non-neuronal cell types such as primordial germ cells (Laird et al., 2011), cardiomyocytes (Moyes et al., 2013), T-cells (Ghosh et al., 2009), and breast cancer and leukemia cells (Hasan et al., 2017; Kim et al., 2020).

Although Wnt5a is a chemoattractant for FBM neurons, they never migrate rostrally out of r4 towards the source of Wnt5a in r3 in wild-type embryos. Here, we have tested and confirmed that Celsr1, the expression of which overlaps that of Wnt5a in r3 and r2, suppresses chemoattraction toward Wnt5a, ensuring that FBM neurons exclusively migrate caudally out of r4. The mechanism of suppression of Wnt5a activity or function by Celsr1 remains to be determined, and could be direct or indirect, extracellular or intracellular, and cell-autonomous or non-autonomous. Elucidating this mechanism will be a topic for future study.

Mouse colony maintenance and embryo collection was carried out using protocols approved by the Animal Care and Use Committee (ACUC) at the University of Missouri, Columbia, MO, USA. In cages used for timed matings, dams were checked every morning; noon on the day that a copulation plug was detected was defined as embryonic day (E) 0.5. Embryos were staged using standard morphological criteria (Nagy, 2003) before fixation.

The Celsr1 knockout (KO) line (Ravni et al., 2009), Tg(Isl1-EGFP)2Slp (SE1::GFP) transgenic line (Shirasaki et al., 2006) and the Egr2^tm2(cre)Pch (Egr2-Cre, previously Krox20^Cre) Cre line (Voiculescu et al., 2000) were cryo-recovered from frozen sperm by in vitro fertilization using C57B6NCrl (Celsr1^KO and SE1::GFP lines) and C57B6NJ (Egr2-Cre line) donor oocytes. Strain B6;01W55-Wnt5aGOF carrying the conditional Wnt5a^GOF allele, generated by inserting a floxed Wnt5a expression construct at the ROSA26 locus (Cha et al., 2014), was a kind gift from Dr Terry Yamaguchi (National Cancer Institute-Frederick, MD, USA). B6;129S7-Wnt5a^tm1Amc (Wnt5a^KO) (Yamaguchi et al., 1999) and 129S-Dvl2^tm1Awb (Dvl2^KO) (Hamblet et al., 2002) mice were purchased from Jackson Laboratory. Genotyping of various lines was carried out as previously described (Yamaguchi et al., 1999; Hamblet et al., 2002; Shirasaki et al., 2006; Cha et al., 2014; Glasco et al., 2016).

Synthesis of digoxygenin and fluorescein labeled probes and whole-mount in situ hybridization was carried out as described previously (Song et al., 2006; Qu et al., 2010; Glasco et al., 2012; Thoby-Brisson et al., 2012). Demarcation of rhombomere boundaries was performed as described previously (Glasco et al., 2012).

Explant culture protocol was adapted from a previously described method (Vivancos et al., 2009). Embryos collected at E11.5 were screened for the presence of GFP-expressing FBM neurons, and hindbrains were dissected in cold L-15 media (Gibco, 11415064). As Dvl2^KO/KO embryos display no overt morphological defects (Hamblet et al., 2002), handplates were saved for genotyping. The dissected hindbrains were placed on laminin-coated filters in six-well plates (laminin from Sigma, L2020; 8μm polycarbonate membrane inserts from Corning costar, 3428), neurobasal medium [Gibco, 21103049, supplemented with 0.001% GDNF (R&D Systems, 212-GD), 1% antibiotic antimycotic (Sigma, A5955), 1% GlutaMAX (Gibco, 35050061) and 2% B-27 (Gibco, 17504044)] was added to the well and incubated at 37°C. After a 30-min incubation, Affi-gel blue agarose beads (Bio-Rad, 153-7301), treated overnight with PBS or Wnt5a (R&D Systems, 645-WN), were placed rostral to the r3/r4 boundary by visualizing the FBM neurons in r4. After a 30-min incubation at 37°C, samples were imaged under GFP fluorescence (designated as 0 h). Explants were imaged again at 24 h and 48 h to visualize the distribution of GFP-expressing FBM neurons and compared with the 0 h timepoint. As migration was evaluated before genotyping, rostral migration phenotypes were essentially scored in a blinded fashion. Explants in which FBM neurons migrated rostrally by more than one bead diameter were scored as exhibiting ‘strong attraction’ (to the beads), while explants with FBM neurons less than one bead diameter of the r3/r4 boundary were scored as exhibiting ‘weak attraction’. The ROCK inhibitor experiments were carried out by adding Y27632 (Tocris, 1254 in DMSO, final concentration 15 µM) to the neurobasal explant medium.

To quantify the proportion of rostral migration observed among embryos with increased Wnt5a expression, the r3/4 boundary was first estimated based on control samples (non-Wnt5a gain of function; wild type and heterozygote for Celsr1). All measurements were carried out using the ‘Multi-measure’ function in Fiji (ImageJ) open-source image analysis software. After establishing the r3/4 boundary for each experimental sample, four ROIs were defined by outlining the Tbx20 signal (specific to FBM neurons) rostral and caudal to the putative r3/4 boundary for the two sides. After exporting measurements of both the area and intensity, the percentage of signal found rostral to the r3/4 boundary was calculated (Fig. S6).

For all comparisons, a one-tailed, unpaired Student's t-test with equal variance was used to determine whether differences were statistically significant (P<0.01).

We thank Dr Terry Yamaguchi (National Cancer Institute-Frederick, USA) for providing the Wnt5a^GOF mouse, and Drs Fadel Tissir and André Goffinet (KU Leuven, Belgium) for discussion. We thank Dr Whitney Pike (Agilent Technologies) for establishing the hindbrain explant culture protocol in the lab. Some of the text and figures in this paper formed part of D.H.’s PhD thesis presented to the Faculty of the Graduate School at the University of Missouri in May 2022.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200553.reviewer-comments.pdf.