Mosaic neuroepithelial Vangl2 deletion counteracts posterior neuropore closure
Induced deletion of Vangl2 in a minority of neuroepithelial cells is sufficient to stall PNP closure, as we found previously when Vangl2 deletion in a small proportion of cells using the constitutive Grhl3Cre produced distal spina bifida13. Grhl3Cre recombines conditional alleles ubiquitously in the surface ectoderm and mosaically in the neuroepithelium (Supplementary Figure 1A), precluding attribution of Vangl2 requirement to either tissue. To our knowledge, no currently available constitutive Cre drivers selectively yet persistently lineage trace the ventral PNP neuroepithelium. Tamoxifen-inducible CreERT2 drivers are limited because early tamoxifen administration by intra-peritoneal injection prevents embryo implantation and is teratogenic45. We recently validated an oral tamoxifen administration protocol which robustly activates embryonic CreERT2 without impeding NT closure46. Using this method to induce Sox2CreERT2 led to extensive lineage tracing of neuroepithelial cells in the closed NT, but only to a very limited extent in the open PNP (Supplementary Figure 1B)47. In contrast, the neuromesodermal progenitor marker Nkx1.2CreERT2 lineage traces cells in both ventral PNP and closed NT48 (Figure 1A-C). Moreover, it produces mosaic neuroepithelial recombination similarly to Grhl3Cre (Supplementary Figure 1A, C).
In Nkx1.2CreERT2/+; Vangl2Fl/-; Rosa26mTmG/+ embryos (henceforth called Cre;Fl/-), CreERT2- induced EGFP reporter expression identifies cells in which Vangl2 has been deleted, 24 h after tamoxifen treatment (Figure 1D,E). Relative Vangl2 reduction is significantly greater when one allele is pre-deleted in Cre;Fl/- embryos compared with Cre;Fl/Fl (Figure 1D). This produces a patchy, mosaic pattern of Vangl2 deletion in the neuroepithelium (Figure 1F).
The PNP fails to close, producing pre-spina bifida lesions, in 54% (7/13) of Cre;Fl/- embryos at E10.5 compared with 0% (0/23, Fisher’s exact test p < 0.001) of Cre;Fl/Fl embryos. Cre;Fl/Fl embryos occasionally develop dorsally flexed tails suggestive of delayed PNP closure (Supplementary Figure 2A). In this study, Cre-negative littermate embryos were used as controls for the Cre;Fl/Fl and Cre;Fl/- genotypes. Vangl2-deleted cells, lineage traced with EGFP, are more abundant in the dorsal closed NT of Cre;Fl/- embryos with pre-spina bifida lesions than those which achieve PNP closure (Figure 1G, Supplementary Figure 2C). Even so, there is substantial overlap between these groups and embryos which fail to close their PNPs only lose Vangl2 in around 16% of neuroepithelial cells (Supplementary Figure 2C).
Pre-spina bifida lesions reflect a pathological state which is difficult to compare to control embryos with physiological, closed NT. Morphometric analyses were therefore performed at earlier developmental stages to identify the first quantifiable phenotype before failure of PNP closure. Dorsolateral and medial hinge points do not differ overtly between the PNPs of Cre;Fl/- and control embryos (Figure 1H), whereas PNP length is significantly longer in Cre;Fl/- embryos than in controls at late stages of closure (Figure 1I). Preceding this, the neural folds are less elevated in Cre;Fl/- embryos than in controls (Figure 1J,K). Thus, the first morphometrically quantifiable tissue-level consequence of mosaic Vangl2 deletion is failure of neural fold elevation. Neither PNP length (Supplementary Figure 2B) nor neural fold elevation (Figure 1K) were significantly different between Cre;Fl/Fl embryos and Cre-negative controls, and there was no significant difference in neuroepithelial thickness between genotypes (Figure 1L).
At the cellular level, mediolateral orientation of cell apical surfaces is a readily quantifiable planar-polarised phenotype in the mouse PNP neuroepithelium13. Cells in control embryos, as well as both EGFP and tdTom cells in Cre;Fl/Fl embryos had preferentially mediolaterally- oriented apical surfaces (median orientation 52-53o, Figure 2A-C). In contrast, neither Vangl2- deleted (EGFP) nor Vangl2-replete (tdTom) cells showed preferential mediolateral orientation in Cre;Fl/- embryos (median orientation 42o each, Figure 2B). The proportions of deleted and replete cells in each orientation bracket were significantly different from control embryos (Figure 2B). Apical orientations of Vangl2-deleted and -replete cells did not differ significantly from each other, suggesting non-autonomous disruption of apical planar polarity. However, median apical cellular areas were smaller in Vangl2-deleted cells than in Vangl2-replete cells in the same Cre;Fl/- embryos, and smaller than cells in control embryos (Figure 2D). Vangl2 deletion did not alter neuroepithelial proliferation (Supplementary Figure 2D-E).
Mosaic Vangl2 deletion diminishes neuroepithelial apical constriction
In both mouse and chick embryos, neural fold elevation requires apical tension generated by actomyosin-dependent apical constriction21,25 and variable apical areas in Cre;Fl/- embryos suggest differential constriction within the mosaic neuroepithelium. Cellular mechanical tension is commonly inferred from recoil of cell borders immediately following laser ablation. We previously demonstrated that annular laser ablations in the apical neuroepithelium produce actomyosin-dependant, rapid initial shrinkage of the cluster of cells within the annulus as they are untethered from the surrounding tissue21. Here, the reduction in apical area of cells or cell clusters following ablation will be referred to as “retraction” to differentiate it from spontaneous apical “constriction” during live imaging. Apical retraction is smaller in some neuroepithelial cells in Cre;Fl/- embryos (Figure 3A-C), but not in Cre;Fl/Fl (Supplementary Figure 3A), compared with controls.
In Cre;Fl/- embryos, the reduction in apical retraction is limited to cell clusters which include Vangl2-deleted cells. Selective annular ablations which do not include Vangl2-deleted cells in Cre;Fl/- embryos produce equivalent retractions to control embryos (Figure 3C). This shows apical tension is diminished locally in association with the deleted cells.
To test whether the local diminution of neuroepithelial apical retraction around Vangl2-deleted cells occurs cell autonomously, apical areas of Vangl2-deleted EGFP+ cells were analysed separately from their Vangl2-replete tdTom+ neighbours. Whereas Vangl2-deleted cells were found to retract similarly to cells in control embryos, their Vangl2-replete neighbours retracted less (Figure 3E-G). Both EGFP+ cells and tdTom+ neighbours retracted similarly to controls in Cre;Fl/Fl embryos (Supplementary Figure 3B). Thus, the diminution of apical retraction following mosaic Vangl2 deletion is non-cell autonomous: Vangl2-replete neighbours of Vangl2-deleted cells fail to undergo apical retraction, whereas the Vangl2-deleted cells themselves continue to retract.
Potential explanations for differential retraction following laser ablation include changes in cell adhesion, material properties, or actomyosin-dependent constriction. Vangl2-deleted cells and their neighbours continue to assemble adherens junctions labelled with N-cadherin and active β- catenin, as well as tight junctions labelled with ZO-1 and an apical F-actin cortex (Supplementary Figure 4A-B). These findings suggest that neither neuroepithelial cell-cell adhesion nor cortical actin differences explain differential apical retraction differences. We therefore sought to directly assess apical constriction through visualisation in live-imaged embryos. The length of live-imaged sequences was limited by substantial changes in tissue morphology as the PNP continued to narrow (Supplementary Figure 5A) and time constraints of imaging at least one control and one comparable Cre;Fl/- embryo from each litter, while avoiding prolonged culture. Live-imaged sequences were therefore limited to 20 minutes. Over this time, neuroepithelial cells were found to vary their apical surfaces in an asynchronous, frequently pulsatile manner characteristic of apical constriction in other cell types (Supplementary Figure 5B-D). Individual cells could be in constriction or dilation phases at the start of imaging. In order to rationalise the data, the apical sizes of individual cells were temporally aligned by their largest observed apical area. This produced averaged traces of dilation followed by constriction (Supplementary Figure 5B-D). A pilot study analysing a wild-type embryo showed that 24 cells need to be analysed to detect a 20% difference in apical area reduction (p = 0.05, power = 0.8).
Vangl2-deleted cells in Cre;Fl/- embryos dilate faster than cells in control embryos, but subsequently constrict at a similar rate (Figure 4A-C). This pattern could be explained by either a change in constriction frequency or magnitude, but we cannot discriminate between these with the available data (Supplementary Figure 5E). In contrast, Vangl2-replete neighbours of Vangl2- deleted cells dilate similarly to controls, but then fail to constrict (Figure 4D). Vangl2-replete cells which do not contact Vangl2-deleted cells in Cre;Fl/- embryos (“Distant” cells in Figure 4D) dilate and constrict similarly to cells in control embryos. These findings corroborate the laser ablation analyses by implicating non-autonomous failure of apical constriction in Vangl2- replete “neighbour” cells, which leads to diminution of overall neuroepithelial apical tension.
Three subgroups of cells were defined for further analysis: 1) Vangl2-/EGFP+ cells which constrict, 2) Vangl2+/EGFP- “neighbouring” cells which do not constrict and, 3) Vangl2+/EGFP- “distant” cells which do constrict (Figure 4d).
Mosaic Vangl2 deletion alters the actomyosin and microtubule cytoskeletons
Cytoskeletal regulation by Vangl2/PCP signalling is well established in other contexts, and both actomyosin and microtubule changes may underlie differential apical constriction in the neuroepithelium following mosaic Vangl2 deletion22-24,36,49,50. Neuroepithelial cells assemble apical phosphorylated, active non-muscle myosin light chain-II around their cell cortex (Figure 5A). In addition, phospho-myosin light chain (pMLC)-II decorates the apical cap of individual neuroepithelial cells (Figure 5A). This pattern is accentuated when total myosin heavy chain (MHC)-IIb is visualised, producing marked differences in localisation to the cell cortex, with or without staining on apical caps (Figure 5B). The same pattern is observed with a second commercial anti-MHC-IIb antibody (Supplementary Figure 6A).
The effect of mosaic Vangl2 deletion on both cortical and apical cap MHC-IIb was assessed sequentially. Cortical MHC-IIb staining between neuroepithelial cells with varying contractility behaviour was defined as above. MHC-IIb staining visualised on cell borders is the average between adjacent cells: 1) “EGFP/neighbour” borders are between a contractile and a non-contractile cell, 2) “neighbour/neighbour” borders are between two non-contractile cells, and 3) “distant/distant” borders are between two contractile cells. Average MHC-IIb staining intensity was found to be highest along the most contractile distant/distant borders of Cre;Fl/- embryos, whereas EGFP/neighbour and neighbour/neighbour borders had significantly lower intensity that did not differ between them (Figure 5C).
MHC-IIb is enriched on the apical cap of approximately half of PNP neuroepithelial cells in wild-type embryos (Figure 5D-E). Apical cap MHC-IIb forms a sarcomere-like pattern of punctate staining or linear arrangements resembling stress fibres (Figure 5D “cap”). However, Vangl2-deleted cells show primarily cortical, rather than apical cap MHC-IIb staining (Figure 5E). This effect is cell-autonomous as neither neighbouring nor distant EGFP-/Vangl2+ cells in Cre;Fl/- embryos are different from controls (Figure 5E).
Similar apical sarcomere-like arrangements have been described in insect cells, and are enhanced in mammalian cells with diminished microtubule turnover51. Vangl2 deletion alters microtubule organisation in other mammalian cells36,37. Non-mitotic PNP neuroepithelial cells were found to assemble microtubules in two predominant patterns: apically-enriched radiating fibres versus apicobasally-elongated “tails” (Figure 5F). Both apical pools and the elongated tails stain positively for the stable microtubule marker acetylated tubulin (Supplementary Figure 6B). Both of these microtubule arrangements influence apical constriction in other contexts: apical networks counteract, whereas apicobasal tails promote constriction33-35,52.
Apical radiating microtubule fibres were mainly seen in neuroepithelial cells with apical cap myosin, whereas elongated tails were found to be associated with a primarily cortical myosin distribution (Figure 5F). The small proportion of neuroepithelial cells in mitosis at any one time have a tubulin network which is apical to and distinct from the centrosomal network (Supplementary Figure 6C). Both these microtubule patterns are preserved in neuroepithelia with mosaic Vangl2 deletion, but the length of microtubule tails was found to be shorter in Cre;Fl/- embryos than in wild-type littermates (Figure 5G). This effect appears non-autonomous as Vangl2-deleted cells have longer tails than their neighbours (Figure 5H-I). Conversely, Vangl2- deleted cells have less abundant apical microtubule networks than their neighbours (Figure 5J,K).
Thus, Vangl2-deleted cells preferentially localise myosin to the contractile cell cortex in a cell- autonomous manner, while both autonomously and non-autonomously altering microtubule organisation in a pattern expected to favour their constriction (Figure 5L).
Non-autonomy amplifies mosaic Vangl2 deletion
Having demonstrated non-autonomous suppression of apical constriction by Vangl2-deleted cells, we sought to quantify the potential for effect amplification as a result of each deleted cell inhibiting multiple neighbours. Lewis’ law predicts that the apical area of epithelial cells should correlate linearly with the number of cells they neighbour, and that the average cell should have six apical neighbours53,54. In agreement with this law, Vangl2-deleted cells have 5.7 total neighbours on average, of which 5.3 are Vangl2-replete (Figure 6A-B). The proportion of Vangl2-replete cells neighbouring a Vangl2-deleted cell is greater in embryos with more recombined cells within the range of recombination achievable with Nkx1.2CreERT2 (Figure 6C). However, cells share neighbours, so the number of unique neighbours per Vangl2-deleted cell decreases as the proportion of deleted cells increases (neighbours/deleted cell shown in Figure 6D). In other words, as more cells lose Vangl2 they each have fewer unique neighbours. Thus, neighbour-sharing limits amplification of effect size through non-autonomy and individual Vangl2-delete cells have the greatest impact in embryos with low levels of recombination.
To test this, we investigated correlations between the proportion of Vangl2-deleted cells versus neural fold eversion, the earliest phenotype detected in Cre;Fl/- embryos. The proportion of Vangl2-deleted neuroepithelial cells does not correlate with magnitude of neural fold eversion (Figure 6E). However, the proportion of Vangl2+EGFP- neighbour cells is significantly correlated with neural fold eversion (Figure 6F). This supports a model in which non- autonomous inhibition of neighbour cells’ apical constriction underlies failure of PNP closure following mosaic Vangl2 deletion.