MEIS-WNT5A axis regulates development of fourth ventricle choroid plexus

The choroid plexus (ChP) is a sheet of predominantly epithelial cells that produces cerebrospinal fluid (CSF), secretes factors important for brain development and forms a crucial brain barrier (Chau et al., 2015; Fame and Lehtinen, 2020; Ghersi-Egea et al., 2018; Lehtinen et al., 2011; Silva-Vargas et al., 2016). ChP tissue is specified during the early stages of brain development (Hunter and Dymecki, 2007), and forms at, or near, the dorsal midline in each brain ventricle (lateral ventricle, LV; third ventricle, 3V; fourth ventricle, 4V; Currle et al., 2005). As progenitor cells proliferate and mature into epithelial cells, the epithelial sheet extends in a conveyor belt-like manner into the ventricles (Dani et al., 2021; Liddelow et al., 2010). Interactions between the maturing epithelial cells and the surrounding cellular network of mesenchymal (Wilting and Christ, 1989) and vascular (Nielsen and Dymecki, 2010) cells transform the growing epithelium to adopt its mature form.

The ChP tissues are regionally patterned, and they harbor distinct transcriptomes resulting in ventricle-specific secretomes (Dani et al., 2021; Lun et al., 2015). These findings raise the possibility that local CSF environments are tailored to instruct development of adjacent brain areas. Indeed, the 4V ChP expresses high levels of WNT5A, which is released into the CSF, associates with lipoprotein particles and influences hindbrain morphogenesis (Kaiser et al., 2019). In mice, the ChP also appears morphologically distinct in different ventricles; the LV ChP consists of two nearly flat, leaf-like epithelial sheets, whereas in the 3V and 4V the ChP adopts a more complex, frond-like structure (Dani et al., 2021). Despite a general understanding of the key steps required for ChP formation, surprisingly little is known about the molecular mechanisms underlying differences in ChP morphogenesis in each of these ventricles.

WNT5A signaling represents one compelling candidate signaling pathway for regulating regional ChP development. Wnt5a transcription and production are enriched in distally located mature epithelial cells in 4V ChP (Kaiser et al., 2019), whereas mesenchymal Wnt5a expression is predominantly localized to the proximal region extending to the core of all ChP (Dani et al., 2021; Langford et al., 2020). Systemic Wnt5a deficiency was recently reported to impair ChP development in the LV, 3V and 4V ChP (Langford et al., 2020), but the specific role of WNT5A in ChP epithelial cells has not been studied. As such, the molecular mechanisms that regulate the regionalized expression of Wnt5a in the ChP and its time- and cell type-restricted functions remain to be elucidated.

Overall, WNT signaling encompasses highly conserved signaling pathways involved in the regulation of numerous physiological processes during embryonic development and in adulthood (Saito-Diaz et al., 2013). WNT proteins (WNTs) represent a large family of lipid-modified glycoproteins acting as extracellular ligands that activate either a canonical cascade, mediated by active β-catenin, or the non-canonical branches of the WNT pathway, which depend on ligand binding to a large repertoire of cognate receptors (Niehrs, 2012). Among the WNTs, WNT5A represents a prototypical non-canonical WNT ligand, predominantly linked to the activation of the planar cell polarity (WNT/PCP) pathway (Kumawat and Gosens, 2015), where it influences various aspects of tissue patterning and establishment of cell polarity (Humphries and Mlodzik, 2018). Notably, WNT5A mediates conserved roles in the outgrowth of body structures, including budding tentacles in Hydra or limbs and distal digits in mouse. In these tissues, WNT5A exhibits similar spatial patterns of expression with upregulation in the most distal regions of the outgrowing structures (Philipp et al., 2009; Yamaguchi et al., 1999). As such, WNT5A, with its expression mainly restricted to the epithelium of 4V ChP, is an ideal candidate factor for regulating the formation of the complex frond-like shape of the 4V ChP.

Here, we elucidate dual roles of WNT5A in 4V ChP development. First, at early stages of ChP development, Wnt5a temporal expression and dosage are essential for establishing the blueprint for ChP morphogenesis and size. Second, at later stages of ChP development, WNT5A expression and secretion are enriched at the distal tips of the 4V ChP epithelium where WNT5A acts either locally to activate components of the WNT/PCP pathway in an autocrine manner or is released into the CSF for long-range signaling. At this later developmental stage, MEIS1 binds to the Wnt5a proximal promoter to regulate Wnt5a expression in the 4V ChP. Collectively, our findings reveal two distinct developmental stages at which WNT5A plays important roles in establishing 4V ChP form and function.

First, we confirmed our previous observation showing that, during development [embryonic day (E) 14.5], Wnt5a expression is enriched in the mouse 4V ChP epithelium in contrast to the LV ChP (Kaiser et al., 2019; Lun et al., 2015) (Fig. 1A,B). Wnt5a expression was most prominent at the distal tips of epithelial villi, with protein signal also present in mesenchymal cells of the stromal space (Fig. 1C, filled and unfilled arrowheads; white dashed line demarcates the distal part of each ChP villus, red line demarcates base of the villi). In 4V ChP, we observed segregation of WNT5A signal to the apical cytoplasm, with aquaporin 1 (AQP1) staining highlighting epithelial membrane (Fig. 1D, top, arrowheads), and to the basolateral cell membrane (stained with collagen IV) and mesenchymal cells (Fig. S1A). Using a ChP epithelium-based transwell system, we confirmed bi-directional secretion of WNT5A apically and basally by embryonic 4V ChP epithelia (Fig. S1B). Analysis of human gestational week 9 (GW9) brain specimens confirmed WNT5A expression on the apical and basolateral sides of the 4V ChP epithelium (Fig. 1D, bottom, filled and unfilled arrowheads, respectively) suggesting that Wnt5a may have evolutionarily conserved roles in the ChP.

We analyzed the functional consequences of Wnt5a deficiency employing several different mouse models. Phenotypic analysis of Wnt5a null mutants (Wnt5a^KO) (Yamaguchi et al., 1999) revealed severe impairment of 4V ChP development as demonstrated by decreased size and reduced branching morphology compared with wild-type control embryos (Wnt5a^WT) (Fig. 1E-H; Fig. S1C), in agreement with others (Langford et al., 2020). In the most severe Wnt5a^KO cases, the 4V ChP lacked its characteristic convoluted structure and resembled instead the simpler sheet-like shape of the LV ChP (Fig. S1D). We also confirmed the previously reported disruption of LV ChP morphogenesis in Wnt5a^KO embryos (Langford et al., 2020), which was characterized by the collapsed growth and impaired protrusion of the tissue into the lumen of the ventricle (Fig. S1E) accompanied by general shortening of the tissue (Fig. S1F,G). We investigated whether the observed morphological differences in 4V ChP were due to altered proliferative or apoptotic activity. No differences were observed in the 5-ethynyl-2′-deoxyuridine (EdU) incorporation rate (delivered at E11.5 and analyzed 24 h later) of mature epithelial cells (AQP1⁺) in the 4V ChP epithelium of Wnt5a^KO ChP (Fig. S1H,I). Similarly, no changes in apoptotic activity analyzed by cleaved caspase 3 (CASP3) staining were detected in the embryonic 4V ChP between Wnt5a^KO and control littermates (Fig. S1J). This suggests that any possible contribution of altered proliferation and apoptosis to the observed phenotypes occurred earlier in development.

Next, we conditionally deleted Wnt5a (Wnt5a^cKO) in ChP epithelial cells using an inducible Foxj1 promoter-driven CRE^ERT2 system (Kaiser et al., 2019) (Fig. S2A,B). In line with the reported properties of the Cre-lox recombination system (Nakamura et al., 2006), we observed the loss of WNT5A protein approximately 24 h following tamoxifen injection (Fig. S2C). When tamoxifen was delivered at E12.5, tissue-specific WNT5A ablation in 4V ChP epithelium of Wnt5a^cKO versus control embryos persisted at least until E16.5 (Fig. S2D). WNT5A was readily detected in the lower rhombic lip (LRL) domain adjacent to the 4V ChP where Foxj1 is not expressed in both E16.5 Wnt5a^WT and Wnt5a^cKO embryos (Fig. S2E). We also observed loss of WNT5A immunoreactivity in the stromal compartment at the tips of distal villi (Fig. S2D) as well as in the mesenchyme of more proximal regions (Fig. S2E) in the Wnt5a^cKO 4V ChP. As such, WNT5A signal detected in the mesenchymal cells located at the distal tip of ChP villi likely represents predominantly epithelial-derived WNT5A. We failed to observe disrupted AQP1 localization in 4V ChP epithelial cells, as suggested in other studies (Langford et al., 2020).

The specific timing of Wnt5a ablation had strikingly different effects on the developing 4V ChP. Induction of Cre recombination by tamoxifen injection at E11.5 profoundly disrupted the overall size and branching morphology of the 4V ChP in Wnt5a^cKO (Fig. 1I-L), when analyzed at E16.5. In contrast, induction of Cre recombination only 24 h later, at E12.5, failed to induce a detectable phenotype in the E16.5 4V ChP (Fig. 1I-L). The morphological defects observed following E11.5 Cre-recombination were regionalized within the 4V ChP, being most pronounced in the ventral part of the 4V ChP, adjacent to the LRL region compared with the more lateral region of the ChP adjacent to the upper rhombic lip (URL) region (Fig. S2F, filled arrowheads). Consistent with the systemic Wnt5a^KO embryos and previous studies (Langford et al., 2020), no differences in either proliferation (EdU⁺ cells) or apoptotic cell death (cleaved CASP3) markers were observed at E12.5, despite the disrupted 4V ChP morphogenesis (Fig. S2G-I). In contrast to findings in systemic Wnt5a^KO embryos (Fig. S1E-G), we failed to detect morphological changes in the developing LV ChP in Wnt5a^cKO embryos (Fig. S2J-L). Taken together, these data demonstrate that Wnt5a expression between E11.5 and E12.5 is essential for establishing the blueprint for 4V ChP morphogenesis and size, and Wnt5a deficiency at this early stage has profound, long-lasting consequences on ChP morphology.

Wnt5a dosage, demonstrated by both ablation and ectopic overexpression studies, disrupts the morphogenesis of the small intestine (Cervantes et al., 2009; van Amerongen et al., 2012), a simple columnar epithelium with similarities to the ChP (Grosse et al., 2011). To determine the consequences of supplemental WNT5A on 4V ChP, we employed a previously established model for Wnt5a overexpression (Wnt5a^OE) (van Amerongen et al., 2012). In these mice, despite being driven by the Rosa26 promoter, higher levels of FLAG-WNT5A were detected in embryonic 4V ChP compared with the surrounding regions (Fig. 2A, arrowheads). We also observed FLAG-WNT5A expression in other brain regions that typically express Wnt5a, including in close proximity to the cortical hem (Fig. 2B, arrowheads; Fig. S3A, arrowheads). In the 4V ChP epithelium, FLAG-WNT5A localized to the cytoplasm of epithelial cells and revealed limited local spreading within the surrounding 4V ChP extracellular space (Fig. 2C, filled and unfilled arrowheads). The reasons for spatially restricted overexpression of WNT5A despite using the Rosa26 promoter remain unknown but this has been observed in other tissues (personal communication, R.v.A.). These observations may result from post-transcriptional as well as post-translational regulation of WNT5A in cells that either express or do not express Wnt5a endogenously.

Wnt5a overexpression had detrimental consequences on the formation of the 4V ChP that were characterized by considerable reduction of its overall size and decreased complexity of its branched, villous structure (Fig. 2D-G). These effects likely stemmed from the overactivation of WNT5A signaling early in differentiating AQP1⁺ epithelial cells (Fig. S3B). As with knockout experiments, we failed to observe any changes in Ki67 or CASP3 immunoreactivity in the E14.5 4V ChP of Wnt5a^OE embryos compared with Wnt5a^WT littermate controls (Fig. S3C,D, arrowheads). We also noted increased FLAG-WNT5A expression in LV ChP epithelial cells of Wnt5a^OE embryos compared with controls (Fig. 2H, filled and unfilled arrowheads). The effects of Wnt5a overexpression on LV ChP size and length were similar to 4V ChP but slightly weaker (Fig. 2I-K). Taken together, our data demonstrate that 4V ChP morphogenesis relies on tightly regulated WNT5A dosage, likely stipulated at the earliest stages of ChP development.

At E16.5, WNT5A localizes to the distal tips of the 4V ChP epithelium (Fig. 1C) and can be released apically into the CSF for long-range signaling to instruct hindbrain development (Kaiser et al., 2019). In addition, WNT5A localization suggests that it can be released basally (Fig. 1B-D; Fig. S1A,B) for potential local signaling within the ChP. To address the extent of WNT5A signaling activation, we performed biochemical analyses of LV and 4V ChP tissue lysates. WNT5A activation leads to phosphorylation of several downstream signaling pathway components; namely, the WNT5A receptors ROR1 and ROR2 (Grumolato et al., 2010; Ho et al., 2012; Yamamoto et al., 2007) and a key signaling mediator, dishevelled 2 (DVL2) (Bryja et al., 2007). This activation can be detected by western blotting as a phosphorylation-dependent mobility shift. Using this approach, we discovered considerably stronger activation of WNT signaling in the embryonic 4V ChP than in LV ChP during late embryogenesis (Fig. 3A; Fig. S4A). Wnt5a deficiency in HEK293 cells mimicked the differences between 4V and LV ChP, suggesting that activation of these readouts depended on WNT5A (Fig. S4B). Indeed, Wnt5a ablation in both Wnt5a^KO and Wnt5a^cKO ChPs resulted in the loss of induction of non-canonical WNT pathway components, thus demonstrating that their activation was mediated exclusively by 4V ChP epithelium-derived WNT5A (Fig. 3B-D; Fig. S4C). Conversely, we observed that LV ChP, which is normally devoid of endogenous WNT5A signaling (Kaiser et al., 2019), exhibited relatively higher activation of various markers of canonical WNT signaling, including upregulation of the downstream target genes Axin2 (Jho et al., 2002) and Lef1 (Hovanes et al., 2001) or phosphorylation of LRP6 (Tamai et al., 2000) compared with 4V ChP (Fig. 3E-G; Fig. S4D-F). We attribute these effects to the capacity of WNT5A to suppress the canonical WNT pathway during development (Topol et al., 2003). In support of this model, ablation of Wnt5a using both Wnt5a^KO and Wnt5a^cKO mouse models resulted in a partial shift in the balance of WNT signaling from non-canonical to canonical signaling (Fig. 3H-J; Fig. S4G). The progressive decrease of the canonical WNT signaling target genes Axin2 and Lef1 expression along the proximal-distal (P-D) axis was correlated with gradually increasing expression levels of Wnt5a within 4V ChP epithelium (Fig. S4H, filled and unfilled arrowheads). The regionally restricted activation of non-canonical signaling supports a role of WNT5A-mediated signaling that is specific for embryonic development of 4V ChP.

We analyzed in greater detail the spatial distribution of WNT pathway signaling components in the ChP. Expression of genes encoding non-canonical WNT pathway components, including Dvl2 and both Ror1 and Ror2, was mostly restricted to the epithelial cell layer with only low expression in the stromal compartment of the LV and 4V ChP (Fig. 4A,B; Fig. S5A-B′). Given the central role of WNT5A in the WNT/PCP pathway, we also inspected the expression pattern of Vangl2, a downstream mediator of WNT5A-dependent signaling in the WNT/PCP pathway (Gao et al., 2011). Vangl2 expression exhibited a distinct spatial pattern with higher levels detected in the embryonic LV ChP than the 4V ChP (Fig. 4C). In addition, Vangl2 expression was enriched in ChP epithelial cells compared with the stroma, in agreement with our observations regarding expression of genes encoding other non-canonical WNT components (Fig. 4D; Fig. S5B″). VANGL2 antibody was validated in Vangl2^−/− knockout animals (Vangl2^KO) (Fig. S5C) and a Vangl2 knockout cell line (Mentink et al., 2018). With this antibody, we observed that VANGL2 expression was enriched in epithelial cells in both mouse and human embryonic LV ChP (Fig. 4E-G; Fig. S5D). In contrast to a recent report (Langford et al., 2020), we found that VANGL2 exhibited a strong basolateral distribution typical for its role in mediating cell-to-cell signaling within the developing epithelium (Sittaramane et al., 2013). We also showed overlap of VANGL2 with the dedicated VANGL2 binding partner Scrib (SCRIBBLE; Kallay et al., 2006) in the ChP epithelium during mouse and human embryonic development (Fig. S5E,F; arrowheads). Taken together, our data support a model in which VANGL2 is involved in the control of the WNT/PCP signaling in embryonic ChP.

We next sought to identify the mechanisms that control the expression of Wnt5a in the 4V ChP epithelium. The MEIS family of transcriptional co-activators has been previously linked to the patterning of vertebrate limb or hindbrain, which are also developmentally regulated by WNT5A-mediated signaling (Dibner et al., 2001; Kaiser et al., 2019; Mercader et al., 2009; Yamaguchi et al., 1999). Moreover, several studies have established a direct interaction between MEIS factors and the WNT signaling pathway in the nearby regions of developing 4V ChP, such as hindbrain (Elkouby et al., 2012; Stephens et al., 2010). Importantly, MEIS1 is implicated in the regulation of Wnt5a expression in branchial arches (Amin et al., 2015) and is upregulated in the 4V ChP compared with LV ChP during development (Lun et al., 2015). We thus decided to investigate the possible role of MEIS1 in the regulation of Wnt5a expression in the 4V ChP.

We confirmed Meis1 expression in 4V ChP during late embryogenesis (Fig. 5A) with Meis1 transcripts being restricted to ChP epithelial cells marked by Ttr, a signature gene of ChP epithelial cells (Fig. 5B) (Kato et al., 1986). We confirmed the presence of MEIS1 protein in 4V ChP from E14.5 to E17.5 (Fig. 5C,D) using validated antibodies against MEIS1 (Fig. S6A,B). We identified the potential MEIS1-binding site(s) upstream of Wnt5a by analyzing chromatin immunoprecipitation followed by sequencing (ChIP-seq) on E18.5 4V ChP. MEIS1 binding was enriched at the genomic region containing the proximal Wnt5a promoter (Fig. 5E-G). We confirmed MEIS1 binding to the Wnt5a promoter by ChIP-qPCR (Fig. 5H). Consistent with these data, the spatial distribution of Meis1 expression partially overlapped with Wnt5a expression in 4V ChP epithelial cells (Fig. 5I,J). Further supporting these findings, Meis1 was co-expressed with Wnt5a from very early stages of embryonic 4V ChP development (Fig. S6C-E). At the protein level, MEIS1-positive cells also displayed strong immunoreactivity for WNT5A in both embryonic mouse and human 4V ChP epithelium (Fig. 5K,L), particularly at the distal tips of ChP villi (Fig. 5M,N, filled and unfilled arrowheads). At later embryonic and early postnatal stages, the expression of both MEIS1 and WNT5A progressively decreased in the 4V ChP (Fig. S6F-H).

We adopted an adeno-associated viral (AAV) approach using a serotype with tropism for ChP epithelial cells (Cui et al., 2020; Haddad et al., 2013; Xu et al., 2021) to test whether supplemental Meis1 expression would be sufficient to influence 4V ChP morphology (Fig. S7A,B). We validated efficient intracerebroventricular delivery and protein induction by robust induction of GFP signal in the epithelium of both LV and 4V ChP (Fig. S7C,D). We then successfully applied this approach to overexpress MEIS1 at E10.5 in developing 4V ChP (Fig. S7E,F) to test the effect of MEIS1 enrichment on WNT5A signaling during the crucial developmental time window. qPCR analysis revealed marked upregulation of Wnt5a expression in E16.5 4V ChP following Meis1 overexpression induced at E10.5, which we further confirmed by in situ hybridization (Fig. 6A,B; filled and unfilled arrowheads). Meis1 overexpression also resulted in higher levels of WNT5A protein in the 4V ChP epithelium at E16.5 (Fig. 6C,D) and in 4V ChP tissue lysate (Fig. 6E). Nevertheless, we did not observe any change in the overall morphology of the tissue (Fig. 6F). These results suggest that MEIS1-driven WNT5A upregulation was not sufficient to induce changes in gross morphology, possibly because it was unable to induce activation of the non-canonical WNT pathway monitored by phosphorylation of ROR1 or DVL2 (Fig. S7G). MEIS1 overexpression in embryonic LV ChP epithelium did result in a small increase of Wnt5a expression (Fig. S7H-J) that, however, did not translate into a detectable increase in WNT5A protein level or activation of the non-canonical WNT pathway (Fig. S7K,L).

To extend the functional analysis of the link between Meis1 and Wnt5a expression, we used the Wnt1-Cre2 mouse line (Lewis et al., 2013) to conditionally delete Meis1 (Meis1^cKO) in the 4V ChP epithelium (Fig. S8A,B). We confirmed efficient reduction of Meis1 expression in 4V ChP epithelial cells (Fig. S6B) but not in adjacent regions where Wnt1 was not expressed (Fig. S8A,C, filled and unfilled arrowheads). Importantly, immunofluorescence analysis of Meis1^cKO embryos revealed a prominent reduction of WNT5A production in 4V ChP epithelial cells and reduced size of the 4V ChP compared with Meis1^WT littermates (Fig. 6G-I). Meis2, another member of the MEIS transcription factor family, is also expressed in the 4V ChP epithelium (Lun et al., 2015). To test for possible redundancy between Meis1 and Meis2, we first confirmed that Meis2 expression was restricted to the developing 4V ChP epithelium similar to Meis1 (Fig. 6J). Next, we conditionally deleted both Meis1 and Meis2 (Meis1/2^dKO) in the 4V ChP. We were unable to obtain more than three Meis1/2^dKO embryos, but even this limited sample set indicated that Meis2 deficiency further decreased WNT5A expression in Meis1/2^dKO embryos compared with Meis1 knockout alone (Fig. 6K,L). Notably, analysis of Meis1/2^dKO embryos was also suggestive of further impairment in the morphogenesis and decreased overall size of the 4V ChP compared with wild-type and Meis1^KO littermates (Fig. 6M). Collectively, these findings demonstrate a direct regulation of Wnt5a expression by MEIS1 in the developing 4V ChP and suggest redundancy of MEIS1 and MEIS2 in this process.

Despite large overlap in their cellular architecture and secretory capacity, ChPs adopt dramatically different morphologies during development. Here, we report a novel role for epithelium-derived WNT5A in the regulation of tissue branching morphology within the 4V ChP. WNT5A-mediated signaling, established as an important regulator of morphogenesis in various tissues (Yamaguchi et al., 1999), has been recently recognized to play an important role during ChP embryogenesis (Kaiser et al., 2019; Langford et al., 2020). Using temporally controlled conditional deletion of Wnt5a directed to the ChP epithelium, we identified two roles of WNT5A, separated in time and space. Our data agree and expand upon recently published findings showing a crucial role of WNT5A in the development of embryonic ChPs (Langford et al., 2020).

By integrating information about WNT5A expression domains and functional data from various Wnt5a mouse mutants, we propose a model in which WNT5A has two distinct roles in mouse ChP formation (Fig. 7). Early on, Wnt5a is expressed in the precursor domains that are directly adjacent to and feed the growth of 4V, 3V and LV ChP epithelium (Langford et al., 2020; Dani et al., 2021). As development proceeds, Wnt5a is expressed in the maturing ChP epithelium in a spatially restricted fashion enriched in the 4V ChP (Kaiser et al., 2019). Systemic Wnt5a deletion impairs development of all ChPs (Langford et al., 2020). In contrast, conditional deletion of Wnt5a in time and space within Foxj1-expressing epithelial cells (Wnt5a^cKO) restricts the growth phenotype to the 4V and not LV ChP. However, we are unable to comment on 3V ChP, which was not examined in this present study. The effect is age-dependent: strong when recombination is induced at E11.5 and non-detectable when induced at later stages. Our data show that the tamoxifen injection at E11.5 results in the loss of WNT5A protein at E12.5. Thus, WNT5A appears to be essential during this developmental period when branching of the 4V ChP begins. A similar temporal requirement for Otx2 expression has also been described (Johansson et al., 2013).

We identify MEIS1 as one transcription factor that participates in the control of Wnt5a expression in the 4V ChP epithelium. Our data further suggest that MEIS1 and MEIS2 together regulate Wnt5a expression and 4V ChP development. Regionalized and conserved expression of Wnt5a, Meis1 and Meis2 has previously been reported to occur during both human and mouse ChP embryogenesis (Kaiser et al., 2019; Lun et al., 2015). Nevertheless, despite our identification of this MEIS-WNT5A axis, the full transcriptional factor network that defines 4V ChP epithelium cell fate remains elusive. Potential co-regulation with HOXA2, which is also enriched in the 4V ChP epithelium (Awatramani et al., 2003; Lun et al., 2015), may be possible. HOXA2 binds the Wnt5a promoter (Donaldson et al., 2012), and in the branchial arches, MEIS1 and HOXA2 synergize to positively regulate Wnt5a expression (Amin et al., 2015). Future studies may reveal whether a similar evolutionarily conserved mechanism contributes to the development of the ChP.

WNT5A controls cell polarity (Humphries and Mlodzik, 2018) as well as morphogenesis and tissue outgrowth in many developing tissues, including mammary gland (Kessenbrock et al., 2017), kidney (Pietilä et al., 2016), lung (Li et al., 2005) and prostate gland (Huang et al., 2009), which all exhibit complex branched morphology. The role of Wnt5a in regulating the morphogenesis of distal regions of branched structures appears to also be evolutionarily conserved as it can be observed in various developmental contexts, including lung and prostate (Huang et al., 2009; Li et al., 2005), embryonic digits (Yamaguchi et al., 1999) and budding tentacles in Hydra (Philipp et al., 2009). Basolateral secretion of WNT5A by epithelial cells has also been shown to participate in the process of lumen formation in kidney epithelium (Yamamoto et al., 2015). These findings point to a more universal role of epithelial WNT5A in the regulation of branching and outgrowth of epithelial sheets. Importantly, WNT5A appears to be necessary for the initial stages of ChP development, but it is not essential after E12.5 despite continuing 4V ChP tissue growth and folding that progress with development.

WNT5A is known to contribute to the balance between WNT/β-catenin (canonical) and non-canonical WNT signaling, which is essential for proper tissue morphogenesis and homeostasis (Alexander et al., 2012). Crosstalk between WNT5A and other WNTs may differ for the two expression domains of Wnt5a in the developing ChP (see Fig. 7). As Wnt5a expression domains are located adjacent to all embryonic ChPs (Langford et al., 2020) and overlap with the expression domains of WNT ligands that activate β-catenin pathway, it is possible that WNT5A acts in concert with other WNT genes, for example the typical canonical ligand WNT2b, which is known to be an important patterning factor orchestrating proper development of ChPs in the cortical hem (Grove et al., 1998). Importantly, the Wnt5a domain also overlaps with an R-spondin (R-SPO)-positive progenitor domain (Dani et al., 2021). R-SPO acts as an amplifier of WNT/β-catenin signaling, which suggests that in the progenitor domain WNT5A can synergize or coordinate with WNT/β-catenin signals. In contrast, in the 4V ChP epithelium WNT5A seems to be a signaling factor suppressing canonical WNT signaling, as reported in mammary gland development (Roarty et al., 2009). Higher WNT/β-catenin signaling (as shown by upregulation of target genes Axin2 or Lef1 or phosphorylation of LRP6) was detected in the embryonic LV ChP epithelium with low/undetectable Wnt5a expression or upon deletion of Wnt5a in the 4V ChP. Notably, in comparison with Wnt5a, both Axin2 and Lef1 show a reverse pattern of expression along the P-D axis in the embryonic 4V ChP. This observation suggests that the spatial control of WNT/β-catenin signaling is achieved by increasing levels of WNT5A within the ChP epithelium (Pourreyron et al., 2012; Sato et al., 2010). Canonical and non-canonical WNT pathways have been implicated in regulating different aspects of epithelial morphogenesis, such as apical constriction, which is crucial for inducing branching in epithelial monolayers (Choi and Sokol, 2009; Fumoto et al., 2017). The striking difference observed between mouse LV and 4V ChP morphologies, together with distinct activation of WNT pathway signaling, raises the possibility that tight regulation of canonical versus non-canonical WNT signaling contributes to induction or modulation of epithelial folding, as has been recently proposed for developing Drosophila wing disc epithelium or in mouse embryonic submandibular gland (Gou et al., 2021; Sui and Dahmann, 2020). Although Wnt5a contributions to 4V ChP morphogenesis may be conserved, it is important to note that the process of ChP morphogenesis will likely prove to involve multiple pathways, as human LV ChP, in contrast to rodent LV ChP, exhibits a branched structure.

WNT5A has also been implicated in the regulation of endothelial cell migration and microvasculature assembly (Carvalho et al., 2019; Ramakrishnan et al., 2016). Thus, WNT5A production at the distal tips of 4V ChP epithelium may contribute to the interplay between the ChP epithelium and the stromal compartment, where it may facilitate vascular outgrowth within the villi as has been shown for SHH, another morphogen specifically produced in the developing 4V ChP (Nielsen and Dymecki, 2010). Given the lack of detected changes in proliferation or cell death upon Wnt5a manipulation, these newly revealed biological functions of WNT5A represent avenues for future research to elucidate further the mechanisms underlying disrupted 4V ChP morphogenesis upon loss of WNT5A.

In summary, our data reveal the complex functions of WNT5A in regulating ChP development. From a clinical perspective, rare neoplasms of the ChP have higher incidence in children and display regional specificity (Sun et al., 2014). Thus, elucidating the regionalized signaling pathways underlying ChP development may expose molecular mechanisms that lend one ventricle more or less susceptible to cancer. Because the ChP represents a tissue amenable to genetic manipulation (Chen et al., 2020), and it adopts distinct, ventricle-specific morphologies owing to inherent differences in activation of WNT5A-driven PCP signaling, the ChP provides a tractable model for future research investigating the mechanistic basis of WNT5A signaling in tissue morphogenesis.

Conditional Meis1^fl/fl mice were generated from the embryonic stem cell clone HEPD0632_4_H07 purchased from EUCOMM. The Frt-flanked lacZ/neo cassette was removed by ACTFLPe (strain #005703). LoxP sites flank exon ENSMUSE00000655363 encoding the homeobox region of the Meis1 gene. Genotyping primers CTGCGCTTCCTACATCACTG and CACTTCAGCGTCACTTGGAA produce a 227-bp fragment for the wild-type allele and a 262-bp band for the floxed allele.

Conditional Meis2^fl/fl strain with loxP sites around exons 2-6 was described previously (Machon et al., 2015). Genotyping primers GCAAGGGTGCTGAGGTTAAA and TCAGACCCAGGAATTTGAGG produce a 235-bp fragment in the wild-type allele and a 324-bp fragment in the floxed allele.

The Wnt1-Cre2 mouse strain was purchased from The Jackson Laboratory (strain #022137) and it was used for specific deletion of the Meis1^fl/fl gene in 4V ChP (referred to as Meis1^cKO in this article). Genotyping primers GCATTTCTGGGGATTGCTTA and CCCGGCAAAACAGGTAGTTA amplify a 241-bp Cre fragment.

The reporter line mT/mG was purchased from The Jackson Laboratory (strain #007676).

The Wnt5a^−/− (Wnt5a^KO) mouse strain used in this article corresponds to Wnt5a^tm1Amc (Yamaguchi et al., 1999). Mouse strain Wnt5a^tm1.1Krvl/J (referred to as Wnt5a^flox/flox in this article) (Ryu et al., 2013) was purchased from Jackson Laboratories; Foxj1^{tm1.1(cre/ERT2/GFP)Htg} (referred to as Foxj1-creERT2 in this article) (Muthusamy et al., 2014) and Gt(ROSA)26Sor^{tm14(CAG−tdTomato)Hze} (referred to as tdTomato in this article) (Madisen et al., 2010) were shared with the Karolinska Institutet, Sweden through a collaboration agreement. All mouse strains were housed, bred and treated in Czech Centre for Phenogenomics (Institute of Molecular Genetics, CAS) in accordance with protocols approved by the animal work committee of the Institute of Molecular Genetics, CAS and Central Commission for Animal Welfare of Ministry of Agriculture Czech Republic (PP-90-2015, PP-64-2018). Induction of conditional knockout or the tdTomato reporter was induced by single dose of tamoxifen (Sigma-Aldrich) intraperitoneal injection of pregnant female mice at a concentration of 4.5 mg of tamoxifen dissolved in sterile sunflower oil per 20 g weight of mouse.

Overexpression of Wnt5a (Wnt5a^OE) was induced in embryos carrying both an inducible Wnt5a transgene and a Rosa26rtTA driver as described earlier [van Amerongen et al., 2012; strains are available from Jackson Laboratories: FVB/N-Tg(tetO-Wnt5a) 17Rva/J, stock number 022938; B6.Cg-Gt(ROSA)26Sortm1(rtTA*M2)Jae/J, stock number 006965]. Wnt5a^OE was induced by administering doxycycline to pregnant mice from E10.5 onwards by dissolving doxycycline in the drinking water (1-2 mg/ml) ad libitum. Ror2; Vangl2 mice were obtained from Dr Yingzi Yang (Gao et al., 2011). Mice were crossed as double heterozygotes (Ror2^+/−; Vangl2^+/−) due to homozygote lethality. All mice were used according to the rules and regulations of the local ethical committee (Stockholm Norra Djurförsöksetisks Nämd: N273/11, N326/12, N158/15; and the Animal Welfare Committee of the University of Amsterdam).

Reagents used are listed in Table S1.

All AAV injection in utero experiments were performed under protocols approved by the IACUC of Boston Children's Hospital (BCH). Timed-pregnant CD-1 dams were obtained from Charles River Laboratories, were anesthetized with isoflurane inhalation, and laparotomy was performed at E10.5 or E13.5. AAV-GFP or AAV-Meis1 (1 µl containing 1∼5×10¹² gc/ml) was injected into a lateral ventricle using a pulled glass pipette (Drummond Scientific Company). Analgesic (meloxicam 5 mg/kg) was provided subcutaneously following surgery. For E10.5 injections, uterine horns were bathed in sterile PBS for improved visualization of ventricles by ultrasound. E16.5 embryos were harvested and drop-fixed in 4% paraformaldehyde (PFA; Millipore) before tissue processing or snap-freezing for qPCR analysis.

The murine Meis1 gene was obtained from the MEIS1A-MIY (Addgene, plasmid #21013) by EcoRI digestion and ligated into an AAV-CMV-IRES-hrGFP (Agilent) vector digested with EcoRI. AAV5-GFP and AAV5-Meis1 were produced by the viral core at Boston Children's Hospital (BCH).

Reagents used are listed in Table S1. Constructs used are listed in Table S2.

ChIP-seq protocol was adapted from Laurent et al. (2015). Briefly, 4V ChP from E18.5 embryos were dissected into cold HBSS. Tissues were crosslinked in 1% formaldehyde at room temperature for 10 min while rotating. Glycine (Sigma-Aldrich) was added to a final concentration of 125 mM and incubated at room temperature for 5 min. Tissues were centrifuged at 4°C at 2000 g and washed twice with 1×PBS with protease inhibitors (Thermo Fisher Scientific). Tissues were snap-frozen in liquid nitrogen and stored at −80°C.

Frozen tissues were resuspended in lysis buffer with protease inhibitors (50 mM Tris pH 8.1, 10 mM EDTA pH 8.0, 1% SDS). Tissues were homogenized using an insulin syringe and then fragmented with an ultrasonic sonicator (Qsonica). After sonication, dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl pH 8.0 and 167 mM NaCl) with protease inhibitors were added to give a final SDS concentration of 0.1%.

Chromatin samples were pre-cleared with Protein A beads (Thermo Fisher Scientific) under rotation at 4°C for 1 h prior to incubation overnight with anti-Meis1 or anti-IgG antibody (15 µg of antibody per sample). Protein A beads were added to precipitate antibody complexes and rotated for 1 h at 4°C and then washed with low salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8.0, 150 mM NaCl), high salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8.0, 500 mM NaCl), LiCl buffer (0.25 M LiCl, 1% IGEPAL-CA630, 1% deoxycholic acid, 1 mM EDTA, 10 mM Tris pH 8.0) and TE buffer (10 mM Tris-HCl, 1 mM EDTA). Antibody complexes were eluted with 1% SDS in 100 mM NaHCO_3, and crosslinks were reversed with NaCl and proteinase K (New England Biolabs). DNA was recovered by phenol-chloroform extraction and ethanol precipitation and quantified with Qubit (Invitrogen).

ChIP-seq libraries were prepared using NEBNext DNA library preparation reagents (E6000) as described in the Illumina Multiplex ChIP-seq DNA sample Prep Kit. Libraries were indexed using a single indexed PCR primer. Libraries were quantified by Qubit (Invitrogen) and sequenced using a HiSeq 2500 (Illumina) to generate 50 bp single-end reads. Reagents used are listed in Table S1.

ChIP-seq samples were sequenced in single-end on the Illumina HiSeq2500 platform. Raw reads were mapped to mouse reference genomes (mm9) downloaded from the UCSC website (www.genome.ucsc.edu). Bowtie (v1.0.0) (Langmead et al., 2009) was used for alignment with ‘–m 1’ and other default parameters. The ‘–m 1’ allows reads uniquely aligned to the genome. MACS (v2.0.10) (Zhang et al., 2008) was used to identify the enriched regions with default parameters. These enriched regions were annotated with an R package, ChIPpeakAnno (v3.6.5) (Zhu et al., 2010), in Bioconductor. The promoters were defined as 2 kb from up- to downstream of transcriptional start sites. Reagents used are listed in Table S1.

ChIP-qPCR was performed on immunoprecipitated DNA after amplification with NEB Next DNA library preparation (New England Biolabs). qPCR was performed with SYBR green (Roche) for detection on a LightCycler 480 system (Roche) according to the manufacturer's instructions. The results were calculated as relative fold enrichment over the input. Reagents used are listed in Table S1. Primers are listed in Table S5.

ChP tissue was collected from E14.5 embryos isolated from sacrificed pregnant CD1 mice and choroid plexus epithelial cells (CPECs) were isolated from LV ChP and 4V ChP. During isolation, extracted tissue was kept at room temperature (RT) in HBSS solution (Sigma-Aldrich). After isolation, extracted tissue was briefly centrifuged (200 g, 10 s at RT). Following aspiration of supernatant, 500 µl of 2 mg/ml solution of Pronase (Sigma-Aldrich) was added to the extracted tissue and incubated for 5 min at 37°C. The solution was then transferred to DMEM/F-12 medium containing 10% fetal bovine serum (FBS; Thermo Fisher Scientific) and centrifuged (300 g, 2 min at RT). Tissue was transferred to complete culture medium consisting of DMEM/F-12 supplemented with 10% FBS, 10 ng/ml EGF (Invitrogen), 20 µM cytosine arabinoside (Sigma-Aldrich) 50 U/ml penicillin and 50 U/ml streptomycin. Cells were mechanically dissociated through a 21-gauge needle for six to eight forced passages, followed by gentle repeated re-suspension with a 200 µl pipette. Finally, cells were seeded onto Transwell-0.4 µm thick clear filter inserts (Sigma-Aldrich). Inserts were pre-coated on their upper side with laminin (Sigma-Aldrich) as described by the manufacturer. To achieve higher purity of epithelial cells, an adhering-off method was applied to reduce fibroblast contamination. After the initial seeding, supernatant containing unadhered cells was transferred to a new laminin-coated well thus removing from culture fibroblasts characterized by a higher adherence affinity.

In order to produce conditioned medium (CM), CPEC primary cultures were maintained in complete culture medium. CM was collected every 48 h up to 10 days after seeding. Supernatant was subjected to sequential centrifugation steps of 200 g for 5 min (to remove viable cells), 1500 g for 10 min (to remove dead cells) and 6000 g for 15 min (to remove cell debris). Reagents used are listed in Table S1.

Ethical approval allowing human fetal tissue acquisition and analysis was provided by the National Research Ethics Service Committee East of England–Cambridge Central, UK (ethics number 96/085).

HEK293T cells were seeded in complete DMEM medium containing 10% FBS, 2 mM L-glutamine, 50 U/ml penicillin and 50 U/ml streptomycin (Thermo Fisher Scientific) on 10 cm dishes 24 h prior to transfection. The cells were transfected with a total of 5 µg of DNA at ∼40% confluency in DMEM medium only. The transfection reaction mixture was prepared using OptiMEM (Thermo Fisher Scientific) and Lipofectamine 2000 (Thermo Fisher Scientific), at a ratio of 1 µg DNA:2 µl Lipofectamine 2000, followed by incubation with cells for 4-6 h. Afterwards, the transfection medium was exchanged for the complete medium. Reagents used are listed in Table S1.

Plasmid-encoding guide RNA targeting the human WNT5A gene was used. The gRNA sequence was cloned into plasmids pSpCas9(BB)-2A-GFP (PX458) (Addgene plasmid #41815) or pU6-(BbsI) CBh-Cas9-T2A-mCherry (Addgene plasmid #64324). T-REx-293 cells (Invitrogen) were cultured according to the manufacturer's instructions and transfected by Lipofectamine 2000 DNA Transfection Reagent (Thermo Fisher Scientific) with plasmids encoding guide RNAs targeting human WNT5A gene. Next day, transfected cells were single cell sorted and grown as single colonies. Selection of WNT5A knockout (^KO) clones was carried out by PCR. Genomic DNA was isolated by DirectPCR Lysis Reagent (Viagen Biotech) and then the fragment of genomic DNA was amplified by PCR using DreamTaq DNA Polymerase (Thermo Fisher Scientific). Then, the PCR product was cut by TaqI (Thermo Fisher Scientific). WNT5A ablation efficiency was assessed by western blot analysis using WNT5A antibody (MAB645, R&D Systems).

Reagents used are listed in Table S1. Constructs used are listed in Table S2. Primers are listed in Table S5.

Samples were subjected to SDS-PAGE, electrotransferred onto a Hybond-P membrane (GE Healthcare), immunodetected using appropriate primary and secondary antibodies and visualized by ECL (Millipore) or Supersignal Femto solution (Thermo Fisher). Signal intensities were calculated using ImageJ. Briefly, the area of the peak intensity for the protein of interest was divided by the corresponding values of peak intensity obtained for the control protein.

Reagents used are listed in Table S1. Antibodies used are listed in Table S3.

Embryos were fixed for 1 week in 4% PFA and stained with Lugol's Iodine solution for 2 weeks or longer. Stock solution (10 g KI and 5 g I₂ in 100 ml H₂O) was diluted to a 25% working solution in H₂O. Stained specimens were removed from contrast agent, rinsed with PBS and embedded in 2.5% low gelling temperature agarose dissolved in water. Scanning was performed using a SkyScan 1272 high-resolution microCT (Bruker, Belgium), with voxel size set up for 3 µm.

For mouse embryo immunofluorescence analysis and in situ hybridization mice were dissected and isolated embryos were transferred into ice-cold PBS, fixed in 4% PFA in PBS for several hours followed by several washes in ice-cold PBS and finally cryoprotected by sequential incubation in 15% and then 30% sucrose solutions. Embryos were next frozen in optimum cutting temperature (OCT) compound (Sakura FineTek) on dry ice. Serial 14-µm-thick coronal sections were used for immunofluorescence analysis. Human fetal tissue for cryosectioning was immersion-fixed overnight in 4% PFA at 4°C, then cryoprotected in sucrose before embedding in OCT compound and then 14-µm-thick sections were cut using a Leica CM1850 cryostat. Human fetal tissue samples were then processed using an identical immunofluorescence protocol as for the mouse samples.

For immunofluorescence analysis, all the sections underwent antigen retrieval by direct boiling for 1 min at 550 W in the microwave followed by 10 min incubation at 85°C in a water bath, using antigen retrieval solution (Dako). Sections were washed in PBT (PBS with 0.5% Tween-20) and blocked in PBTA (PBS, 5% donkey serum, 0.3% Triton X-100, 1% bovine serum albumin). Samples were incubated overnight at 4°C with primary antibodies diluted in PBTA. Following washes in PBT, samples were incubated with corresponding Alexa Fluor secondary antibodies (Invitrogen, Abcam) for 1 h at RT, followed by 5 min incubation at RT with DAPI (Thermo Fisher). Finally, samples were mounted in DAKO mounting solution (Dako).

EdU (Life Technologies) was injected 24 or 72 h before the embryos were harvested at a concentration of 65 mg/g. Cells with incorporated EdU were visualized using a Click-iT EdU Alexa Fluor 555 Imaging Kit (Life Technologies).

For immunofluorescence analysis of ChP primary culture, cells grown on laminin-coated cover slips were first washed several times in ice-cold PBS, followed by fixation for 15 min in ice-cold 4% PFA. Later, cells were washed several times in PBT, blocked with PBTA for 30 min and incubated overnight at 4°C with primary antibodies. Following repeated washing in PBT, cells were incubated for 1 h at RT with appropriate secondary antibodies, DAPI for 5 min and mounted in DAKO mounting medium (Dako).

Reagents used are listed in Table S1. Antibodies used are listed in Table S3.

Quantitative analysis of immunofluorescence signal was carried out in a randomized single-blinded manner. Only the AQP1⁺ area of embryonic 4V ChP was used for the signal intensity analysis. Embryos from the same litters were used for the comparative analysis and all the samples were processed using the same immunofluorescence protocol (described above; see Table S3) and same settings (for laser wavelength 561: gain 800 V, laser intensity 1.2%; for laser wavelength 640: gain 800 V, laser intensity 0.5%) of the confocal microscope (LSM880, Zeiss). Images were not further post-processed or enhanced. Signal was analyzed from three consecutive sections obtained from each embryonic 4V ChP. Normalization was carried out by averaging WNT5A signal intensity from each individual control AAV-GFP embryonic 4V ChP as indicated in the corresponding graphs (Fig. 6D,H).

For morphological analysis of embryonic LV and 4V ChP, only the AQP1⁺ ChP area was included in the analysis.

In situ analysis of the gene expression was performed on 14-µm cryosections of embryos at various stages of embryonic development isolated from CD1 mice. After isolation, embryos were immediately transferred and kept in fresh 4% PFA for 2 h, washed briefly in ice cold PBS, incubated for 6 h in 30% sucrose solution at 4°C, and frozen at −80°C. Transcripts were detected using an adapted protocol for the RNAscope 2.0 assay for fixed frozen tissue (Advanced Cell Diagnostics). Staining was performed using the RNAscope Fluorescent Multiplex Reagent Kit (Advanced Cell Diagnostics).

Indicated in situ images were adopted from Allen Institute for Brain Science: Allen Developing Mouse Brain Atlas (Lein et al., 2006) (http://developingmouse.brain-map.org) or from the Eurexpress database (Diez-Roux et al., 2011) (http://www.eurexpress.org/ee/).

Reagents used are listed in Table S1. Probes used are listed in Table S4.

RNA was isolated from WT CD1 embryos collected at different developmental stages using RNAeasy kit (Qiagen). Samples were treated with DNase (Qiagen) to prevent contamination with genomic DNA. The specificity of primers was determined by BLAST run of the primer sequences. Annealing temperature was 57°C for all primers.

qPCR reactions were performed once for every gene/sample in triplicate. PCR was performed according to the manufacturer's protocol using Lightcycler 480 SYBR Green 1 Master Mix (Roche). The following thermo cycling program parameters were used for qPCR analysis: incubation step at 95°C for 5 min, then 45 cycles 95°C for 10 s, 57°C for 10 s and 72°C for 10 s. qPCR analysis was carried out using a LightCycler 480 Instrument II (Roche).

ΔCp values were calculated in every sample for each gene of interest with Actb as the reporter gene. Relative change of expression level for the analyzed gene (ΔCp) was performed by subtraction of the gene expression levels in the LV ChP or 4V ChP from the gene expression level of the housekeeping gene (Actb). Next, the ratio of the gene expression level of Actb (Figs 3F, 4F, 5A and Fig. S4F), or of Rn18s for qPCRs with Taqman probes (Fig. 6A, Fig. S7E,I,J), and the gene of interest in either 4V ChP or LV ChP was calculated using the following formula: 2^^−ΔCp.

Reagents used are listed in Table S1. Primers are listed in Table S5.

Replicates represent independent experiments. Data in Figs 3F, 4C, 5A, 6A and Fig. S4E and S7E,I,J are represented in columns showing the mean with standard deviation (s.d.). Significance was measured using paired or unpaired two-tailed Student's t-test with unequal sample variance. Biological replicates per condition are indicated in the corresponding graphs. Data in Fig. 5H are represented as columns showing the standard error of the mean (s.e.m.).

Data in Figs 1F-H,J-L, 2E-G,J,K, 6D,F,H,I,K,L and Figs S1F,G,I and S2H,K,L are expressed as columns showing the mean with s.d. Significance was measured using unpaired or paired two-tailed Student's t-test with unequal sample variance. Biological replicates used per condition are indicated in the corresponding graphs.

Images used for quantitative analysis reported in Figs S4A,C,D,G and S5D are expressed as columns showing the mean with s.d. Significance was measured using paired two-tailed Student's t-test with unequal sample variance. Biological replicates per condition are indicated in the corresponding graph.

Except for Fig. 1D (human fetal sample), Fig. 4G,K and Figs S2C and S6A, all the displayed immunostaining images and western blots are representative of at least three independent experiments.

We thank members of Bryja and Lehtinen labs for their help and suggestions; Dr Marcela Buchtová and Mgr Marie Šulcová (Masaryk University) for their valuable assistance with RNAscope hybridization; Dr Sean Li (Boston Children's Hospital) for sharing Wnt5a knockout mice with the Lehtinen lab; and the BCH Viral Core. We are also grateful to Nikodém Zezula (Masaryk University) for his assistance with the figure graphic design and Mgr Monika Novákova (Masaryk University) for help with the animal work. We thank MEYS CR for support to the following core facilities: CELLIM of CEITEC supported by the Czech-BioImaging large RI project (LM2018129 funded by MEYS CR), Czech Centre for Phenogenomics (LM2018126), Higher quality and capacity of transgenic model breeding (by MEYS and ERDF, OP RDI CZ.1.05/2.1.00/19.0395), Czech Centre for Phenogenomics: developing towards translation research (by MEYS and ESIF, OP RDE CZ.02.1.01/0.0/0.0/16_013/0001789), BCH viral core for AAV production.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.192054