Introduction

Neural tube defects (NTDs), such as spina bifida and anencephaly, affect approximately 1:1,000 births globally¹. These severe neurodevelopmental malformations are caused by abnormal morphogenesis of the neural tube during embryonic development. The most severe NTDs result from failure of the physical process of neural tube closure, which in humans is completed by approximately 30 days gestation^2,3. Its completion during largely inaccessible stages of pregnancy has limited our ability to study neural tube closure in humans. This challenge is compounded by the multigenic causation of NTDs, as well as interactions of both maternal and embryonic genetics with environmental factors^4–6. Consequently, few patients receive an interpretable genetic diagnosis.

A few patient-identified mutations have been functionally validated in transgenic animal models^5,7–9. Mice and other vertebrates have been invaluable models for understanding the process of neural tube closure. Both mice and humans close their open neural tube regions – called neuropores – through progressive elevation of neural folds which meet at the dorsal midline and zipper along the body axis^2,10. Progression of closure requires the action of different cell types, particularly the surface ectoderm (future epidermis) and neuroepithelium (future brain and spinal cord)³. These cells coordinate multiple behaviours within and between each other. Neuroepithelial cells thicken along their apicobasal axis as they differentiate into a pseudostratified epithelium which undergoes interkinetic nuclear migration synchronised with cell cycle progression, superimposed on active apical constriction^11–17.

Neuroepithelial apical constriction is an evolutionarily conserved force-generating mechanism which contributes to neural tube closure in chordates including amphibians, birds and mammals³. It requires targeted localisation and phosphorylation of non-muscle myosin II in the cell’s apical domain. Myosin binds filamentous (F-)actin, cross-linking filaments and generating contractile mechanical force that pulls actin filaments together through power strokes¹⁸. Neuroepithelial apical enrichment of F-actin and phosphorylation of myosin both require Rho-associated kinase (ROCK)¹⁹, which is in turn apically enriched at mediolateral cell borders at least in part by signalling through the non-canonical wingless homology (Wnt)/planar cell polarity (PCP) pathway^20–22. PCP components promote mediolaterally oriented neuroepithelial apical constriction in Xenopus frogs^23,24, chickens²² and mice^15,20, but this pathway has been difficult to study in human epithelial models due to lack of quantifiable readouts relevant to its functions in vivo.

PCP was originally described in Drosophila as a regulator of polarised cell features such as hair cell bristle orientation^25–27. Pathway components were later shown to be conserved in vertebrates and contribute to neural tube closure^28–31. The core PCP pathway involves signalling through asymmetric receptor complexes on adjacent cells, pairing van Gogh-like (VANGL) and Cadherin EGF LAG seven-pass G-type receptor (CELSR) proteins on one cell, with frizzled (FZD)/CELSR complexes on the adjacent cell³². Both VANGL and FZD branches of this pathway are involved in neural tube closure^33,34. VANGL2 can be activated through phosphorylation following binding of PCP Wnt ligands^35,36.

PCP signalling drives two functionally distinct processes in neuroepithelial cells essential for mammalian NT closure: early convergent extension and later apical constriction. Global genetic deletion of Vangl2 precludes medial convergence and rostro-caudal elongation necessary to initiate neural tube closure in mouse embryos³⁷, although this effect may be partly mediated by the adjacent mesoderm³⁸. At later developmental stages, however, VANGL2 promotes cytoskeletal reorganisation involving F-actin planar polarisation and phosphorylation of non-muscle myosin IIB, likely by ROCK^20,39–41. However, we previously reported that VANGL2 is not cell-autonomously required for apical constriction of mouse neuroepithelial cells²⁰.

Mutations in PCP genes including VANGL2^42–47 have been identified in patients who have NTDs and damaging variants in actomyosin-related cytoskeletal regulation are enriched in individuals with spina bifida⁴⁸. However, it has been challenging to study the regulation of apical constriction in human neuroepithelial cell models. A previous study described elegant modelling of neural tube closure by differentiation of induced pluripotent stem cells (iPSCs) on micropatterned surfaces⁴⁹. That model’s formation of neuroepithelium enveloped within a larger cyst limits access to its apical surface for functional measurements, such as inference of mechanical tension through laser ablation assays. Similarly, previous studies have demonstrated that disruption of actomyosin regulation dilates apical surfaces on the inside of human iPSC-derived neuroepithelial spheres^50,51. Several models have also been described which enable studies into later stages of neuroepithelial function as they differentiate into neural progenitors and then neurons^52–55. While these neurogenesis models provide insights into later differentiation, they do not capture cell behaviours crucial for neural tube closure.

We recently characterised a minimal model of neuroepithelial morphogenesis in which a monolayer of cuboidal iPSCs apicobasally elongate into a complex, pseudostratified neuroepithelium with a planar morphology¹⁷. Importantly, this model presents readily accessible neuroepithelial apical surfaces which are directly comparable in size with those of mouse embryos¹⁷. Moreover, the same protocol can be extended to differentiate neuroepithelial cells into cortical neurons⁵⁶. This enables sequential assessment of neuroepithelial functions relevant to generating mechanical deformations which close the neural tube and their subsequent differentiation which normally occurs after completion of closure. Here, we show that neuroepithelial development in this model is reproducible, even when comparing non-congenic iPSC lines. We validate use of the model to study PCP components including ROCK and VANGL2, finding that a VANGL2 point mutation previously identified in a patient⁴⁶ diminishes neuroepithelial apical constriction. Extension of the model to two iPSC lines derived from individuals who have spina bifida identifies line-specific deficiencies, including apicobasal elongation and neural cell differentiation, potentially relevant to their malformation. Strikingly, all three NTD-relevant lines – VANGL2 point mutation and two patient-derived lines – present failures of distinct neuroepithelial functions, recapitulating the heterogeneity of these conditions.

Results

Lineage-specific marker expression precedes morphological changes during iPSC to neuroepithelium differentiation

Intra-line variability often limits the use of iPSCs to model multigenic diseases, such as NTDs, in which congenic controls are often impractical. In contrast, neuroepithelial differentiation through dual SMAD inhibition (Figure 1a)^17,56 is highly reproducible between different iPSC lines. We illustrate this by using three previously reported iPSC lines from healthy individuals, selected to represent different origins and reprogramming methods (Figure 1b). For each of the three iPSC lines, we analysed three independently differentiated plates, each representing a separate biological replicate. Line H0193b was reprogrammed from male (46XY) amniocytes through microfluidic delivery of modified mRNA^57,58. SFC086-03-01 was reprogrammed from adult female (46XX) fibroblasts using non-integrating Sendai virus^17,59. 802-30F was from endothelial cells of a Hispanic female (46XX) reprogrammed through StemRNA⁶⁰. Despite their varied origins, all three undergo equivalent, temporally predictable changes in epithelial morphology during neuroepithelial differentiation (Figure 1a-h).

Figure 1:

These cells begin their differentiation as an apicobasally polarised cuboidal monolayer approximately 25 µm thick with large apical areas approximately 60 µm² (‘day 2’, Figure 1c-g). During differentiation, they double their apicobasal length into a pseudostratified epithelium¹⁷ approximately 55 µm thick, with small apical surfaces around 10 µm² (‘day 8’, Figure 1c-g, Supplementary Figure 1a-c). Apicobasal elongation increases total epithelial volume, accommodating ongoing cell division¹⁷, yet the volume of individual cells decreases from 1,550 µm³ at the start of differentiation to 500 µm³ at the end (Figure 1h). This reduction in total cell volume is partly due to a halving of nuclear volume (Supplementary Figure 1d-e).

Neuroepithelial apical size and apicobasal thickness – properties whose dynamic control enables neural tube closure in vivo^20,22,61,62 – are indistinguishable between non-congenic iPSC lines at each stage of differentiation analysed (Figure 1c-g). Equivalent neuroepithelial thickness and apical areas are also obtained in cultures differentiated by different investigators (Supplementary Figure 2). All lines show a non-linear morphological evolution during differentiation: apical area and apicobasal thickness remain similar between days 2 and 4 of differentiation, but then change rapidly from day 5 (Supplementary Figure 1a-c) until the end of differentiation on day 8 (Figure 1e-g). In contrast, their expression of pluripotency markers such as NANOG and OCT4 are rapidly lost between days 2 and 4 of differentiation (Figure 1i,j)¹⁷. The cadherin CDH1 is similarly down-regulated early during differentiation, switching to expression of the neuroepithelial CDH2 from day 4 of differentiation (Figure 1i), before morphological changes. The neuroepithelial marker PAX6 is detectably expressed from day 4 and its expression increases between day 4 and 8 (Figure 1i,k).

iPSC-derived neuroepithelial cells actively generate mechanical tension through apical constriction

While reproducible between cultures, apical areas are variable within individual cultures (Figure 1e): a hallmark of asynchronous apical constriction in vivo⁶³. Through live imaging, we observe that iPSC-derived neuroepithelial cells spontaneously and dynamically alter their apical area (Figure 2a). Pharmacological inhibition of the myosin-activating kinase ROCK increases neuroepithelial apical area (Figure 2b-c), as previously observed in mouse embryos with equivalent ROCK inhibition in vivo¹¹.

Figure 2:

Acutely expanding apical cell areas produced corrugation of the apical domain in ROCK-inhibited cultures (Supplementary Figure 3a-b), suggestive of buckling due to expansion within fixed boundaries imposed by the culture dish.

The functional outcome of apical constriction is generation of mechanical tension which is transmitted between adjacent cells and pulls on the surrounding tissue^20,64.

Contractile neuroepithelial apical tension can be inferred through annular laser ablation assays^11,20, in which release of a small cluster of cells from their neighbours causes rapid retraction of the cell borders within the ablated ring (Figure 2d). The ablated surface remained within the focal plane after ablation, indicating minimal movement along the apical-basal axis. Their tension is significantly diminished following a short period of ROCK inhibition (Figure 2e), confirming that it is actively established through apical constriction.

In vivo, the Rho/ROCK/myosin pathway can be activated by PCP/VANGL2 signalling^20,22, and is well-established to promote apical constriction^20,23,65. We observe that VANGL2 and its binding partner scribbled (SCRIB) are expressed at low levels from early stages of neuroepithelial differentiation and are significantly up-regulated by the end of the differentiation period (Figure 2f,g). Both are co-localised on the cell’s apical cortex (Supplementary Figure 4). Serine-79/82/84 tri-phosphorylation of VANGL2 also increases significantly during differentiation, with a particularly large up-regulation between days 4 and 6 (Figure 2h), coinciding with the reduction in apical area (Figure 1e). Thus, iPSC-derived neuroepithelial cells express the cellular machinery required for physiological control of apical constriction and actively establish apical tension in vitro.

An NTD-associated VANGL2 mutation impairs neuroepithelial apical constriction

To our knowledge, no genetic mutations identified in human patients have previously been functionally associated with failure of apical constriction. The VANGL2-R353C mutation (Figure 3a) was initially reported in a fetus which had both anencephaly and spina bifida⁴⁶, and impairs other functions of PCP signalling in fruit flies⁶⁶. This point mutation causes a basic to polar amnio acid switch in a C-terminal region (Supplementary Figure 5) which is highly conserved across species and mediates interactions with pathway adaptor proteins^46,66. We therefore considered it a high priority target for testing in a human system.

Figure 3:

We find that VANGL2-R353C mutation does not alter overall VANGL2 protein levels, phosphorylation (Figure 3b,c), or apical localisation (Supplementary Figure 6) in two iPSC clonal knock-in lines homozygous for this mutation compared with a congenic control line. Knock-in and congenic control lines retain their ability to differentiate into different germ layers (Supplementary Figure 7), achieve equivalent changes in neuroepithelial apicobasal thickness (Supplementary Figure 8a-c) and nuclear pseudostratification (Supplementary Figure 8d,e) during differentiation. The VANGL2-R353C mutation does not significantly alter cell volume, nor does it diminish extension of sub-apical protrusions which are characteristic of neuroepithelial cells⁶⁷ (Supplementary Figure 9).

VANGL2-R353C knock-in lines have significantly larger apical areas (Figure 3d,e) and less apical localisation of phosphorylated myosin regulatory light chain (Figure 3f,g) than their congenic controls. Laser ablation assays confirm that apical tension is significantly lower in VANGL2-R353C knock-in lines than their congenic controls (Figure 3h,i). Thus, this VANGL2 mutation associated with human NTDs selectively impairs neuroepithelial apical constriction. This is consistent with previous reports in Xenopus that Vangl2 knockdown diminishes apical constriction, but we had previously observed that mouse Vangl2 knockout cells remain able to apically constrict when surrounded by cells which express this gene. We therefore tested whether global loss of Vangl2 impairs neuroepithelial apical constriction in a mammalian model. As expect, mouse embryos globally lacking Vangl2 fail to close their entire spinal neural tube, developing craniorachischisis (Supplementary Figure 10a). Vangl2-null neuroepithelial cells show an equivalent shift towards larger apical areas in embryos (Supplementary Figure 10b-d) as seen in vitro (Figure 3e). Neuroepithelial apical areas are 39±8% (mean ± 95% CI) larger in Vangl2 knockout embryos versus wild-type littermates (Supplementary Figure 10d). VANGL2-R353C knock-in iPSC-derived neuroepithelial apical areas were 98±32% (mean ± 95% CI across both knock-in clones with three independent experiments each) larger than their wild-type counterpart. This suggests the VANGL2-R353C mutation can explain all the impairment of neuroepithelial apical constriction caused by loss of Vangl2 action.

Establishment of patient-derived models of neuroepithelial dysfunction

As well as mechanistic and reverse genetic studies, an additional application of this highly reproducible model is patient-specific testing of key neuroepithelial behaviours (Figure 4a). Importantly, short-loop reprogramming of amniotic fluid stem cells with microfluidic delivery of modified mRNA⁵⁷ enables rapid generation of patient-specific models in the ∼4 months between spina bifida diagnosis and birth (Figure 4a). As proof of principle, we used short-loop reprogramming to compare two iPSC lines derived from amniotic fluid cells of individuals who do not have spina bifida (here called GOC1-2 for simplicity) with two lines of individuals treated at Great Ormond Street Hospital who underwent foetal surgery for spina bifida (GOSB1-2). All four lines express comparably high levels of the iPSC marker NANOG before differentiation, and differentiate to form endoderm, ectoderm and mesoderm (Supplementary Figure 7).

Figure 4:

Both control lines and the first spina bifida line - GOSB1 - achieve indistinguishable apical areas and thicknesses (Figure 4b-d), further testament to the excellent reproducibility of this modelling system when comparing non-congenic lines. In contrast, the spina bifida line GOSB2 fails to apicobasally elongate as much as other lines and retains larger apical areas (Figure 4b-d). We consider this is unlikely to represent an apical constriction defect: laser ablation assays show comparable retraction in the GOSB2 line compared with the other lines tested (Supplementary Figure 11). The epithelia GOSB2 cells form include mitotic cells which are not apically-located, in contrast to all other cells studied which are pseudostratified and localise their mitoses apically (Supplementary Figure 12). Unlike all other lines tested, differentiated GOSB2 cultures also show abnormally heterogenous expression of CDH2 while partially retaining lamin A/C (Figure 4e) at the end of the standard differentiation protocol.

Attempts to adapt the protocol to ‘rescue’ this phenotype are beyond the scope of this study given the excellent reproducibility of differentiation in the seven other lines presented here.

We performed whole exome sequencing of all four amniocyte-derived lines to identify coding variants of potential clinical relevance in GOSB1 and GOSB2. Variants were compared with a list of genes previously associated with NTDs in mice and/or humans (Figure 4f, Supplementary Table 1).

GOSB1 harbours 851 variant loci within 747 genes (Supplementary Table 2; Figure 4g). Of all genes with variants, 27 are NTD-associated, such as ZEB2 (variant 2-144389425-T-C) and ZIC5 (13-99970996-G-T). None of the variants in NTD-associated genes are predicted to have a high transcript impact and all are heterozygous.

GOSB2, which has a more severe neuroepithelial phenotype, has 1122 variant loci in 939 unique genes of which 43 are in NTD-associated genes (Supplementary Table 3; Figure 4g). In this line, six NTD-associated genes harbour variants predicted to have a high impact on transcript sequence: CSNK2A1, FAM20C, LFNG, MED24 and PCSK5.

Of these, CSNK2A1 (20- 503643-G-GCATATTT), MED24 (17-40019203-C-CACAT) and PCSK5 (9-76175222-AAATGGAATGGAATGAAATGG-A) have gnomAD variant allele frequencies <0.1. MED24 (formerly called TRAP100) is of particular interest because its deletion in mice results in a thin neuroepithelium which is not pseudostratified⁶⁸, equivalent to our observations during GOSB2 differentiation. This gene is part of the mediator complex involved in RNA polymerase II transcriptional co-activation. GOSB2 cells have compound heterozygous frameshift mutations in the same repetitive region of the final exon of this gene. We additionally confirmed that the same MED24 variants are present in independent exome sequencing of amniocytes from which GOSB2 had been derived (not shown). Both frameshifts are in the 3’ untranslated region but are predicted to alter RNA structure, causing loss of a multi-branch loop (Figure 4h), and may therefore diminish binding of translation-regulating proteins. To test this, we compared MED24 protein levels between GOSB2 iPSCs and three control lines, finding reduced protein abundance compared to all three controls (Figure 4i,j). Although genetic rescue experiments are beyond the scope of this study, particularly given the likelihood of gene-gene interactions, equivalence of the cellular phenotype we observed with a published mouse model and confirmation of reduced protein abundance suggest the MED24 variants in this line are pathogenic.

As expected, given poor neuroepithelial morphogenesis of cells from the GOSB2 patient line, further differentiation results in only sparse differentiation of PAX6-positive neural progenitor cells with few neurites positive for the neuronal marker TUJ1 (Supplementary Figure 13). In contrast, GOSB1 formed prominent neuronal rosettes equivalently to the control lines (Figure 5a). However, by day 35 of neuronal differentiation, GOSB1 cultures show diminished TUJ1 positivity, without a change in the proportion of cells positive for the progenitor marker SOX2, compared with the two control lines differentiated in parallel (Figure 5b-d). Poor neuronal maturation in cells from the GOSB1 line is clearly evident by day 45 of differentiation, when control lines form dense aggregates interconnected by axons while GOSB1 cells form sparse and poorly-connected aggregates (Figure 5e,f). Thus, both the spina bifida patient-derived lines tested, in addition to VANGL2-R353C knock-in model, demonstrate quantifiable yet distinct abnormalities in neuroepithelial functions (Figure 5g).

Figure 5:

Discussion

We present a multifaceted yet tractable and reproducible model of neuroepithelial morphogenesis which is sufficiently sensitive to identify NTD-associated dysfunctions in both forward and reverse genetic studies. All three of the NTD-associated models we compare – knock-in of a patient-identified VANGL2 mutation and two patient-derived iPSC lines – identify abnormalities which are specific to that line, and which are not present in a panel of control lines tested. The abnormalities observed relate to apical constriction, apicobasal elongation, and neuronal maturation.

Neuroepithelial apical constriction is required for closure of both the cranial and spinal neural tube^11,14,15,21. Mouse and Xenopus models establish that expression of PCP signalling components enhances apical constriction^20,23,65 but, to our knowledge, this had never been shown in a human system and no human genetic mutations have been linked to apical constriction as a likely pathogenic mechanism. The mechanisms by which VANGL2 promotes apical localisation and activation of myosin²⁰ remain incompletely understood. They may include both PCP-specific functions of VANGL2 equivalent to those originally identified for its orthologue in Drosophila, and potentially vertebrate-specific roles^70,71. The neuroepithelial model presented here does not produce planar features commonly used as characteristic readouts of PCP signalling in vivo^26,40. Nonetheless, our observation of increasing VANGL2 poly-phosphorylation suggests ligand-induced pathway activation during differentiation³⁵. This contrasts with decreasing VANGL2 phosphorylation previously reported during in vitro differentiation of lactogenic cells⁷⁰.

The diminished apical constriction phenotype we document following VANGL2-R535C mutation occurs despite achieving wildtype-equivalent VANGL2 expression. Nonetheless, comparison with mouse embryos globally lacking Vangl2 suggests VANGL2-R535C is sufficient to explain all of the dilation in apical area caused by deletion of this protein. Other mutations in the protein’s C-terminus domain which disrupt VANGL2 function sufficiently to stop neural tube closure have previously been reported in mice^20,39,40. VANGL2’s C-terminus contains the prickle-binding domain (directly affected by the R353C mutation) and PDZ domain, which interacts with other components of the PCP pathway including SCRIB⁷². SCRIB also enhances apical constriction in mice^15,73.

We observe that apical constriction correlates with apicobasal epithelial thickening during iPSC-to-neuroepithelial differentiation. This inter-relation is well-established in Drosophila ventral furrow formation^74,75, and can be replicated following optogenetic induction of apical constriction in cultured mouse cells⁷⁶. In mice, neuroepithelial apicobasal thickness is spatially patterned with shorter cells at the midline, under the influence of SHH signalling^14,77,78. Apicobasal thickness of the cranial neural folds increases from ∼25 µm at E7.75 to ∼50 µm at E8.5⁷⁹: closely paralleling the elongation between days 2 and 8 of differentiation in our protocol. The rate of thickening is non-uniform, with the greatest increase occurring during elevation of the neural folds⁸⁰, also paralleled in our model by the rapid increase in thickness between days 4-6 as apical areas decrease. Elevation requires neuroepithelial apical constriction and these cells’ apical area decreases between E7.75 and E8.5 in mice⁷⁹, but we and others have recently shown that this reduction is both region and sex-specific^14,81. Specifically, apical constriction occurs in the lateral (future dorsal) neuroepithelium: this corresponds with the identity of the cells generated by the dual SMAD inhibition model we use⁵⁶. More recently, Brooks et al⁸² showed that the rapid reduction in apical area from E8-E8.5 is associated with cadherin switching from CDH1 (E-cadherin) to CDH2 (N-cadherin). This is also directly paralleled in our human system, which shows low-level co-expression of CDH1 and CDH2 at day 4 of differentiation, immediately before apical area shrinks and apicobasal thickness increases.

Conservation of cells’ volume has been suggested to link apical constriction with apicobasal elongation⁷⁵. However, we provide two lines of evidence that this inter-relation during neuroepithelial differentiation is not simply due to volume conservation. Firstly, cell volume decreases during neuroepithelial differentiation as thickness increases. Secondly, neither epithelial thickness nor cell volume are significantly altered in VANGL2-R535C neuroepithelial cells with diminished apical tension. Mechanistic separability of apical constriction and apicobasal elongation is consistent with biomechanical modelling of Xenopus neural tube closure showing that both are independently required for tissue bending⁶¹. Nonetheless, neuroepithelial apical constriction and apicobasal elongation are co-regulated in mouse models: for example, deletion of Nuak1/2⁸³, Cfl1⁸⁴, and Pten⁷⁹ all produce shorter neuroepithelium with larger apical areas. Neuroepithelial cells of the GOSB2 line described here, which has partial loss of MED24, similarly produces a thinner neuroepithelium with larger apical areas. Although apical areas were not analysed in mouse models of Med24 deletion, these embryos also have shorter and non-pseudostratified neuroepithelium⁶⁸.

Our GOSB2 line – which retains readily detectable MED24 protein – is clearly less severe than the mouse global knockout of this gene, and the clinical features of the patient from which this line was derived are milder than the phenotype of knockout embryos⁶⁸. Mouse embryos lacking one of Med24’s interaction partners in the mediator complex, Med1, also have thinner neuroepithelium and diminished neuronal differentiation but successfully close their neural tube⁸⁵. As general regulators of polymerase activity, MED proteins have the potential to alter the timing or level of expression of many other genes, including those already known to influence pseudostratification or apicobasal elongation. MED depletion also causes redistribution of cohesion complexes⁸⁶ which may impact chromatin compaction, reducing nuclear volume during differentiation. MED family proteins are increasingly implicated in human neural tube defects: MED12 and MED13L likely pathogenic variants have been reported in patients^9,87. Their functions during neural tube closure, including through known interactions with Wnt and PCP pathways⁸⁸, remain incompletely understood and future work will be needed to determine whether MED24 mutation is the only genetic change relevant to neuroepithelial morphogenesis in GOSB2. MED protein contributions to neurogenesis and neuronal maturation are more well-established^85,89–91.

Another limitation of our model, shared by many in vitro systems, is the poor ability to replicate maternal environmental influences. For example, the validated cell culture media used for this model contain 3.4 mg/L folic acid instead of folate. This may mask NTD predispositions caused by maternal folate deficiency or genetically impaired placental transport, while accentuating cellular deficiencies which diminish folic acid conversion into a one-carbon donor⁹². We cannot exclude in vitro accentuation of the cellular phenotype of the GOSB2 line, as it is difficult to imagine that such severe failure of neuroepithelial morphogenesis would produce a sufficiently normal spinal cord to benefit from fetal surgery performed on the patient these cells were derived from. Alternatively, compensation may be possible in vivo, potentially through interactions between tissue types which promote robustness of morphogenesis.

Some of these limitations, potentially including inclusion of environmental risk factors, can be addressed by using alternative iPSC-derived models^93,94. For example, if patients have suspected causative mutations in genes specific to the surface (non-neural) ectoderm, such as GRHL2/3, 3D models described by Karzbrun et al⁴⁹ or Huang et al⁹⁵ may be informative. Characterisation of surface ectoderm behaviours in those models is currently lacking. These models are particularly useful for high-throughput screens of induced mutations⁹⁵, but their reproducibility between cell lines, necessary to compare patient samples to non-congenic controls, remains to be validated. Spinal cell identities can be generated in human spinal cord organoids, although these have highly variable morphologies^96,97. As such, each iPSC model presents limitations and opportunities, to which this study contributes a reductionist and highly reproducible system in which to quantitatively compare multiple neuroepithelial functions.

A key benefit of our system is the natural progression from cell functions relevant to neurulation, to neurogenesis. Others previously reported that iPSC lines derived from individuals who have spina bifida are less able to differentiate into neurons than those from healthy controls⁹⁸. This is consistent with our findings in the GOSB1 patient-derived line. Although gene rescue experiments are beyond the scope of this study, potentially deleterious variants identified in the GOSB1 patient line, including a ZIC5 variant, may impair neuronal development. The reduction in TUJ1 positive cells observed in this line compared with non-congenic controls is at a point in differentiation which is highly comparable between individuals and produces deep-layer cortical neurons⁹⁹. Individuals with spina bifida commonly have thinning of the neocortex and other supratentorial brain abnormalities¹⁰⁰. What is less clear is whether diminished neurogenesis, which normally occurs after closure of the neural tube, may relate to causation of NTDs. The converse – that open spina bifida impairs neural development –is directly supported by experimental evidence¹⁰¹. Nonetheless, impaired neuronal differentiation of patient-derived iPSC lines raises the potential for shared genetic/epigenetic causation of the spinal lesion and cortical brain abnormalities, potentially compounded by abnormalities in cerebrospinal fluid drainage and amniotic fluid exposure.

In summary, our findings in a human neuroepithelial model corroborate animal studies identifying diminished apical constriction as a cause of NTDs, implicate VANGL2 acting upstream of myosin phosphorylation in this behaviour, identify diminished neuroepithelial thickness in a patient-specific disease model and corroborate the importance of MED family proteins known to enhance this process as an emerging genetic contributor to human NTDs. The multi-genic nature of neural tube defect susceptibility, compounded by uncontrolled environmental risk factors (including maternal age and parity¹⁰²), mean that patient-derived iPSC models are unlikely to provide mechanistic insight. They do provide personalised disease models which we anticipate will enable functional validation of genetic diagnoses for patients and their parents’ recurrence risk in future pregnancies, and may eventually stratify patients’ postnatal care. We also envision this model will enable quality control of patient-derived cells intended for future autologous cell replacement therapies, as is being developed in post-natal spinal cord injury¹⁰³. Thus, the highly reproducible modelling platform we evaluate – which is robust to differences in iPSC reprogramming method, sex and ethnicity – represents a valuable tool for future mechanistic insights and personalised disease modelling applications.

Materials and methods

Human iPSC culture maintenance

All iPSCs lines were grown in E8 medium (ThermoFisher, A1517001), using 0.5% Matrigel Growth Factor Reduced (Corning, 354277) coated plates, according to a previously published protocol¹⁰⁴. Cells were routinely grown in 6 well plates (SARSTEDT, 833920300), unless otherwise stated. E8 medium was fully replaced daily. At ∼80% confluency cells were washed with PBS without CaCl₂/MgCl₂ (ThermoFisher,14190094) and dissociated with 0.5 mM EDTA (ThermoFisher, 15575020) for 5 min. Cells were resuspended in E8 medium and re-plated at a 1:10 split ratio. All cell lines were tested monthly for mycoplasma contamination, using the MycoAlert^® kit (Lonza, LT07-318). A summary with the characteristics and reprogramming technologies of the hiPSCs lines used on this study can be found in Table 1.

Table 1:

Unless otherwise stated, for each iPSC line three independently differentiated plates were generated and analysed, with each plate representing a separate differentiation experiment performed on different days. For patient-specific iPSC lines, one independent iPSC line was obtained per patient following microfluidic mmRNA reprogramming⁵⁷.

VANGL2 knock-in cell line generation

The selection of the mutation was made according to a previously identified gene alteration in a patient described as having anencephaly with occipital and cervical spina bifida⁴⁶. The specific point mutation was found in heterozygosity and was located on chromosome 1, at nucleotide 160,421,174 in the genomic sequence (Supplementary Figure 5). Although other Vangl2 mutations have been found to exert dominant negative effects¹⁰⁵, we generated homozygous mutant lines to facilitate interpretation of results obtain. In collaboration with SYNTHEGO, an established hiPSC line 802-30F was genetically engineered to introduce a homozygous point mutation by replacing a cytosine (C) with a thymine (T) (CGC>TGC), at a specific genetic locus. After high knock-in efficiency was confirmed, specific sgRNA with Cas9 were electroporated to the cells. VANGL2 knock-in lines were generated using CRSIPR-Cas9 homology directed repair editing by Synthego (SO-9291367-1). The guide sequence was AUGAGCGAAGGGUGCGCAAG and the donor sequence was CAATGAGTACTACTATGAGGAGGCTGAGCATGAGCGA AGGGTGTGCAAGAGGAGGGCCAGGTGGGTCCCTGGGGGAGAAGAGGAGAG.

Sequence modification was confirmed by Sanger sequencing before delivery of the modified clones, and Sanger sequencing was repeated after expansion of the lines (Supplementary Figure 5) as well as SNP arrays (Illumina iScan, not shown) confirming genomic stability. For the generation of the isogenic control line (WT), the cells were electroporated with SpCas9 only and confirmed to be unedited at target locus.

Approximately two weeks later the cells colonies have expanded and genotyping of the pool of clones was undertaken. Individual clones were then isolated and clonally expanded to confirm the presence of the knock-in genetic alteration. Those clones positive for the genetic alteration were further Sanger sequenced to confirm stable VANGL2 editing.

Adherent neural induction protocol

Direct differentiation of hiPSCs to neuroepithelial sheets was as we previously reported¹⁷. Briefly, hPSCs were lifted with EDTA and plated at 100% confluency, ∼300,000 cells/cm², in E8 medium. It is critical to ensure that 100 % confluency is reached at seeding (considered ‘day 1’). After 24 hours, cells were washed twice with PBS without CaCl₂/MgCl₂ and the medium was replaced with 2 ml of neural maintenance media (NMM, see Table 2), supplemented with 1 μM Dorsomorphin (Tocris, 3093) and 10 μM SB431542 (Tocris, 1614). NMM with these supplements is referred as neural induction medium (NIM) and was refreshed daily until day 8. At that timepoint the cultures were terminated or passaged for neuronal differentiation.

Table 2:

Neuronal differentiation was performed as previously described⁵⁶. The neuroepithelial sheets were passaged at day 8, using Dispase (ThermoFisher, 17105041) for 15-20 minutes at 37°C. After washing with PBS, neuroepithelial clumps were plated on 1% Matrigel-coated 6-wells in NIM, ensuring sufficiently sparse seeding to give clumps space to form rosettes. After 24 hours of passaging, the medium was changed to NMM supplemented daily with 20 ng/ml of FGF2 (PeproTech, AF-100-18B) for the next four days. After day 12, neural rosettes were passaged to 1% Matrigel coated plates, using Dispase. NMM was then changed every 48 hours. This process was repeated at least three-times, when confluency reaches 80 %, until day 25 when the neural progenitor cells mature. At that stage, once the confluency reached 80 %, cells were passaged with Accutase (Sigma, A6964) to dissociate the rosettes into single cells. At that stage (∼day 30) cells can be cryopreserved or plated for the last time at 25,000-75,000 cells/cm² on Laminin-521 (Gibco, A29249) coated plates.

Generation of Vangl2 knockout mouse embryos

Studies were performed under the regulation of the UK Animals (Scientific Procedures) Act 1986 and the National Centre for the 3Rs’ Responsibility in the Use of Animals for Medical Research (2019). Mice were mated overnight, and the next morning was considered to be E0.5 if a plug was found. Vangl2 knockout mouse embryos were as previously described¹⁰⁶.

Immunofluorescent staining and confocal microscopy

Adherent cultures were stained for a maximum of 20 minutes in cold 4% paraformaldehyde (PFA) and immunofluorescent staining as previously described¹⁷. Primary antibodies used were rabbit anti-ROCK (Abcam antibody ab45171), rabbit anti-Ser19 pMLC (Cell Signalling Technology antibody #3671), rabbit anti-SOX2 (Merck antibody ab5603), mouse anti-CDH2 (Cell Signaling Technology antibody #14215), mouse anti-Lamin A/C (Santa Cruz antibody sc7292), goat anti-Lamin B1 (Santa Cruz antibody sc6216), rabbit anti-ZO1 (Thermo Fisher Scientific antibody #40–2200), goat anti-scribble (Santa Cruz antibody sc-11049), rat anti-VANGL2 (Merck antibody, MABN750), rabbit anti-Pax6 (Biolegend antibody 901301), and mouse anti-TUJ1 (BioLegend MMS-435P), and detected with Alexa Fluor™ conjugated secondary antibodies (Thermo Fisher Scientific). All primary and secondary antibodies were used in 1:200 and 1:500 dilution, respectively. Nuclei were labelled with DAPI (Severn Biotech Ltd 17057) and F-actin was labelled with Alexa Fluor™-647 conjugated phalloidin (Thermo Fisher Scientific). Confocal salt and pepper noise was removed using a ‘remove outliers’ filter and linear adjustments to brightness and contrast were applied to all parts of each figure panel equally.

Confocal images were captured on a Zeiss Examiner LSM880 confocal, using a 20x/NA1.0 Plan Apochromat dipping objective with AiryScan Fast. Widefield images of the neural rosettes and neuronal cultures were captured on a Zeiss Observer inverted using 10x/NA0.45 objective or a Nikon Ti2 Eclipse widefield microscope using 10x/NA0.30 or 20x/NA0.45 Plan Apochromat λD objectives.

Laser ablations and live imaging

Annual laser ablations were performed as previously reported^11,20 on day 8 neuroepithelial sheets. Briefly, cells were homogenously stained with CellMask membrane dye (Invitrogen, C10046) and a single apical z-slice was selected for the target region. The ablation circle was randomly placed in the stained neuroepithelial sheet and three ablations in different fields of view were performed per dish then averaged (no predictable change in tension was observed between repeated ablations). The percentage reduction in cell area within the ablated annulus immediately after ablation was quantified. The imaging order for each cell line was alternated between each independent replicate to ensure no bias in the analysis and recoils were analysed blinded to genotype/treatment.

For live-imaging sessions, day 8 neuroepithelial sheets were stained for 5 minutes with CellMask Deep Red Actin Tracking stain (ThermoFisher, A57245) in NIM. Imaging was performed with the Zeiss LSM880 confocal microscope using 20x/NA1.0 Plan Apochromat dipping objective. The microscope chamber was pre-warmed at 37 °C and CO₂ was set at 5 %. The water-immersive objective was cleaned with 70 % ethanol solution and let to air dry to remove any potential contaminants. A plexiglass chamber was assembled around the culture plate to maintain stable culture condition. Damp cotton wool was put in the plexiglass set up to create a humidified chamber and avoid media evaporation. ZEN Black v2.3 was used to operate the microscope and capture images. Apical region images were acquired every 5 minutes.

Particle intensity velocimetry (PIV) analysis

Particle Image Velocimetry (PIV) examined the flow of apical areas of day 8 neuroepithelial sheets after laser ablations. Briefly, we identify changes in pixel flow before and after the ablation. For the PIV analysis used in this report a pre-installed plug in FIJI was used¹⁰⁷. The analysis considers the displacement of pixels within a specific region, termed ‘interrogation window’. This area in the present study was specified between two frames, before and immediately after the circular laser ablation. For the analysis a ‘Pure Denoise’ filter (Cycle-spins: 4, Multiframe: 2) was used to ensure that non-specific random noise is not interpreted as movement.

Quantitative cell morphology analysis

Apical area analyses were performed in F-actin stained neuroepithelial sheets, using the polygon selection tool in FIJI and outlining the apical cap of individual cells and saving them as ROIs. At least 100 contiguous cells were analysed from each image.

Epithelial apicobasal thickness, confocal images were optically resliced and their cross-section outline manually annotated. Thickness was calculated using the ‘bounding rectangle’ (width) tool in ImageJ/Fiji. This was performed in three images across each dish and the average height was calculated for each culture.

For fluorescent signal polarisation analysis, the plot profile command in ImageJ/FIJI was applied to an optically-resliced section of equal thickness within each dataset.

TUJ1 and SOX2 neuronal quantification

For the quantification of TUJ1+ axons of neural progenitor, the total area of axons was quantified relative to the area imaged. The TUJ1 signal was binarized using the ‘Threshold’ command, and the total area of the axons was summed. Finally, the TUJ1+ area was normalised to the total number of cells per field of view.

For the quantification of SOX2 nuclear marker in neuronal progenitor cultures, a multi-step pipeline was used. Initially, all nuclei were identified in the DAPI-stained channel and using the command ‘Find Maxima’ in FIJI. A binarized image was generated, with individual points within each nucleus. Each point was enlarged and used to generate a ROI within each nucleus The individual points were used as a reference to quantify the SOX2 intensity within each cell. A threshold was set manually after identifying the nucleus with the lowest positive signal. Finally, the percentage for positive nuclei was calculated after subtracting the total cell numbers with the positive proportion and dividing by the overall cell number.

Western blotting and quantification

Western blotting was performed as previously described¹⁰⁸. Neuroepithelial cells were lysed on ice for 30 min in 50 mM Tris Base pH 7.6, 150 mM NaCL, 1% Triton X-100, 0.02% Sodium Azide, 1x protease inhibitor cocktail (Merck Life Science UK Ltd), 1 mM sodium orthovanadate, 25 mM sodium fluoride. The protein concentration was determined using Pierce BCA Protein Assay kits (Thermo Fisher). Proteins were resolved by SDS-PAGE using 10-12% polyacrylamide gels and transferred to PVDF membranes. Membranes were blocked in TBS (15.4 mM Trizma-HCL, 4.62 mM Tris-base, 150 mM NaCl, pH 7.6) containing 10% w/v milk powder (Merck Life Science UK Ltd). Membranes were incubated with primary antibody overnight and then with horseradish peroxidase (HRP)-conjugated secondary antibodies (Agilent Technologies, Stockport, UK) for 1 hour. Primary antibodies were used rabbit anti-pVangl2 (Merck antibody SAB5701946), mouse anti-NANOG (Abcam antibody ab173368), rabbit anti-OCT4 (Abcam antibody ab19857), mouse anti-CDH1 (R&D FAB18381A), mouse anti-GAPDH (Merck antibody MAB374), rabbit anti-MED24 (BETHYL antibody A301-472A). HRP was detected using ECL Prime (Cytiva, Amersham, UK). To strip and re-probe, membranes were incubated in 0.2 M NaOH for 20 min at 37 °C and then 20 min at room temperature before re-blocking and re-use.

For the quantification of western blot films, the band area method was applied using ImageJ/FIJI. A box was drawn around the bands and after ensuring all bands fit within it, and the intensity profile was plotted. The total area between the line graph and x axis, after removing background, was quantified for each sample. The protein level was normalised to the housekeeping protein, LMNB or GAPDH. The normalised protein levels were compared between samples, and a control value was normalised to 1. Protein quantifications were performed in paired replicates or in three independent cell lines.

Whole exome sequencing

Approximately 500 ng of DNA from each cell line, namely GOC1, GOC2, GOSB1, and GOSB2, were purified. The subsequent steps of fragmentation, library preparation, Illumina sequencing were performed using established pipelines by Azenta, UK. Upon receipt of the samples, a quality control (QC) screening was executed, and all samples successfully pass the QC threshold including concentration and maintaining high DNA integrity number (DIN, >7.8). FASTQ data QC, genomic alignment, variant colling and annotation was also performed by Azenta, UK. For all samples, >97.5% of target regions achieved a read depth ≥20X. Variants were annotated using the Sequence Ontology consensus terms. The following were considered to have a high impact: transcript_ablation (SO:0001893), splice_acceptor_variant (SO:0001574), splice_donor_variant (SO:0001575), stop_gained (SO:0001587), frameshift_variant (SO:0001589), stop_lost (SO:0001578), start_lost (SO:0002012), and transcript_amplification (SO:0001889).

NTD associated gene sorting workflow

Lists of genes with annotated variants in GOSB1 and GOSB2 were compared with GOC1 and GOC2. Only variants not shared with GOC1/2 were considered in order to eliminate batch effects and very common variants. Each of resulting gene lists was then cross-referenced with the most current compilation of mouse models associated with NTDs, as documented in Lee and Gleeson⁵. Two additional gene lists were generated based on searches in ClinVar database for entries linked to "Neural Tube Defect" and "Spina Bifida" (Supplementary Table 1). Collectively, these identified a subset of variant genes associated with NTDs. These were subsequently filtered by sequency ontology severity score, and the allele frequencies of high impact variants was determined using gnomAD v4.1.0¹⁰⁹.

Statistical analysis and reproducibility

Comparisons between two groups were by Student’s t-test accounting for homogeneity of variance in SPSS (IBM Statistics 25). Comparison of multiple groups was first tested for normality and homogeneity of variance in Origin 2020 (Origin Labs). According to the result if samples were normally distributed an ANOVA test was performed, while for non-parametric datasets a Kruskal-Wallis ANOVA was used in Origin 2020 (Origin Labs). Post-hoc pairwise analysis was performed to determine the statistical significance. P values <0.05 were considered statistically significant. All graphs were generated in Origin 2020. Replicates were defined as technical replicates (same iPSC line cultured at the same time) or biological replicates (different iPSC lines or same line cultured independently). In the case of VANGL2 knock-in lines, two clones were compared with their congenic control line. To avoid bias, blinding and randomisation was performed were possible by asking a lab member to rename copied files, without disclosing cell line information prior to analysis. However, changes of neuroepithelial thickness across time points days was too dramatic to not know which day was analysed and morphological differences of GOSB3 also precluded blinding.

Data availability

"Ampartzidis_SourceData.xlsx" contains the numerical data used to generate graphs in the figures.

Acknowledgements

The authors would like to thank Prof F Livesey and Dr F Paonessa for their assistance in establishing the dual-SMAD inhibition protocol in our lab and provision of SFC086-03-01 as a reference line. We thank Jessica Reeves for technical assistance with western blots, Louis Hamilton for help with apical area segmentation, Dr Teresa Pereira da Silva for scientific discussions and cell culture assistance, Dr Timo N. Kohler and Clara Munger for their critical reading of the manuscript, and Dr Dale Moulding for assistance with imaging. IA was funded by the NIHR GOSH Biomedical Research Centre. Views expressed in this manuscript are solely those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. GLG acknowledges funding from the Royal Society (RG\R2\232082), Rosetrees (2023\100266), NC3Rs (NC/Y50063X/1) and Wellcome Trust (211112/Z/18/Z).

Additional information

Author contributions

IA, EMT, YG, AK and GLG performed and analysed experiments. GLG, PDC and NE conceived the study and designed experiments with IA. GLG, PDC and IA secured funding. GLG, PDC, NE and EM supervised staff. GLG and IA wrote the manuscript with input from all co-authors.

Funding

Royal Society (RG\R2\232082)

• Gabriel L Galea

Rosetrees Trust (2023\100266)

• Gabriel L Galea

National Centre for the Replacement Refinement and Reduction of Animals in Research (NC/Y50063X/1)

• Gabriel L Galea

Wellcome Trust

https://doi.org/10.35802/211112

• Gabriel L Galea

Great Ormond Street Hospital for Children

• Gabriel L Galea

Additional files

Supplementary Figures

Supplementary Tables 1-3