Cellular mechanisms underlying Pax3-related neural tube defects and their prevention by folic acid

Neural tube defects (NTDs) such as anencephaly and spina bifida are among the most common birth defects worldwide, affecting more than 250,000 pregnancies every year (Copp et al., 2013; Zaganjor et al., 2016). Maternal use of supplemental folic acid (FA) prevents a proportion of NTDs (Berry et al., 1999; Czeizel and Dudás, 1992; MRC Vitamin Study Research Group, 1991), and FA food fortification programmes, which now exist in many countries, have generally been associated with a reduction in NTD prevalence, compared with historical pre-fortification frequencies (Berry et al., 2010; Castillo-Lancellotti et al., 2013). A key aim for public health efforts is now to ensure that women who might become pregnant achieve recommended daily intakes of FA, in order to lower NTD rates (Martinez et al., 2018). However, there is still a gap in our knowledge concerning the mechanism by which FA prevents NTDs in the developing embryo. FA is the synthetic form of folate, a term that refers to a group of molecules, based on a tetrahydrofolate (THF) backbone, that carry one-carbon (1C) units in folate 1C metabolism. FA is converted via dihydrofolate to THF, with subsequent addition of 1C groups derived principally from serine, glycine and formate (Tibbetts and Appling, 2010; Leung et al., 2017).

NTDs result from incomplete closure of the neural tube during embryonic development (Greene and Copp, 2014; Nikolopoulou et al., 2017). Among mouse genetic models, exencephaly and/or spina bifida arise in splotch mice, carrying mutations of the paired-box-domain-containing transcription factor Pax3 (Epstein et al., 1991; Greene et al., 2009). Notably, Pax3 mutants (Pax3^Sp2H, Pax3^Sp) are among the very few models in which NTDs have been found to be both preventable by supplemental FA and exacerbated by maternal folate deficiency, imposed by diet and administration of antibiotic to remove folate-synthesizing gut bacteria (Burren et al., 2008; Fleming and Copp, 1998; Wlodarczyk et al., 2006). Hence, NTDs in this model are both FA responsive and folate sensitive. In addition to mutation of the Pax3 gene itself, suppression of Pax3 expression in mouse embryos is also proposed to contribute to NTDs induced by environmental factors, such as maternal diabetes (Fine et al., 1999; Machado et al., 2001) and polycyclic aromatic hydrocarbons (Lin et al., 2019).

Mutations of the human PAX3 coding sequence have been identified in some individuals with NTDs (Hart and Miriyala, 2017) and may contribute to a minority of NTDs. Altered methylation of PAX3 has also been identified in NTD cases, suggesting that altered expression could potentially play a contributory role (Lin et al., 2019).

Understanding the mechanisms by which Pax3 loss of function prevents neural tube closure will not only give insight into possible causes of NTDs but could also provide an opportunity to better understand the means by which FA prevents NTDs. It has been proposed that Pax3-related NTDs (specifically the Pax3^Sp allele) result from excess apoptosis: NTDs were prevented by genetic or pharmacological suppression of p53 function (Pani et al., 2002), leading to the hypothesis that Pax3 functions to suppress p53-dependent apoptosis in the neuroepithelium. A p53-dependent excess of apoptosis has also recently been proposed to underlie NTDs associated with zinc deficiency (Li et al., 2018). Both excess and insufficient apoptosis have been associated with exencephaly in other mouse mutants, although – in most cases – a causal relationship has not been definitively proven (Greene and Copp, 2014; Nikolopoulou et al., 2017).

Other studies of apoptosis in splotch (Pax3^Sp) embryos have produced differing findings. For example, increased TdT-mediated dUTP-biotin nick-end labelling (TUNEL) staining was reported in the neural tube at embryonic day (E) 10.5 (Pani et al., 2002), whereas no change in the number of pyknotic nuclei in the spinal neuroepithelium was detected at E9.5 (Kapron-Brás and Trasler, 1988). Moreover, although aberrant levels of apoptosis have been demonstrated in Pax3^Sp/Sp and Pax3^Sp2H/Sp2H mutants in the dermomyotome of the developing somites, increased apoptosis was not observed in the neural tube at E9.5 or later stages (Borycki et al., 1999; Mansouri et al., 2001). In the current study, we sought to address the question of the possible contributory role of apoptosis to NTDs in the Pax3^Sp2H/Sp2H model and to investigate other potential causative cellular abnormalities. Having identified a tissue-specific defect in cellular proliferation, we went on to ask whether this abnormality was corrected by FA supplementation in association with prevention of NTDs.

NTDs in splotch embryos result from a cell-autonomous defect in the neuroepithelium (Goulding et al., 1991; Li et al., 1999). Therefore, if excess apoptosis is the cause of cranial NTDs in Pax3 mutants, this should be detectable prior to and/or during closure of the cranial neuroepithelium, which has not previously been examined.

The initiation of neural tube closure, at the hindbrain-cervical boundary (Closure 1; five to six somites; E8.5) and in the posterior forebrain (Closure 2; nine to ten somites; E9.0), occurs similarly in Pax3^Sp2H/Sp2H embryos and wild-type littermates. However, progression of ‘zippering’ forwards from Closure 1 and backwards from Closure 2 fails in those mutants that develop midbrain/hindbrain exencephaly (Fleming and Copp, 2000). In the current study, exencephaly, characterised by persistently open cranial neural folds, arose in 65% of Pax3^Sp2H/Sp2H mutants (n=46), whereas all Pax3^+/+ embryos (n=54) completed closure by the 16-somite stage at E9.5 (P<0.001; chi-square test).

In Pax3^+/+ and Pax3^Sp2H/Sp2H embryos, TUNEL-positive cells were detected in the rostral forebrain, in the midline of the closed forebrain neural tube and in the hindbrain neural folds (Fig. 1A-F), corresponding to known sites of apoptosis in wild-type embryos (Massa et al., 2009; Mirkes et al., 2001). However, we did not observe an increase in the number or location of TUNEL-positive cells in the neural folds of Pax3^Sp2H/Sp2H embryos at any stage of closure in either the cranial or spinal region (Fig. 1; Fig. S1). Consistent with the results of TUNEL staining, the number of cleaved caspase-3-positive, apoptotic cells in the cranial neural folds did not differ between genotypes (Fig. 1G).

Although not supported by our TUNEL and cleaved caspase-3 analysis, we reasoned that if apoptosis is indeed causally related to splotch NTDs then direct inhibition of apoptosis would potentially rescue these defects. Therefore, as a complementary approach, embryos were cultured through the period encompassing cranial neurulation (40 h from E8.5) in the presence of the pan-caspase inhibitor Z-VAD-FMK, which we previously showed to effectively suppress apoptosis in neurulation-stage embryos (Massa et al., 2009) (Fig. 2). Contrary to predictions, inhibiting apoptosis caused a dose-dependent increase in exencephaly incidence among Pax3^Sp2H/Sp2H embryos. These defects were also observed among some heterozygotes, which are unaffected in the absence of Z-VAD-FMK (Fig. 2D). In addition, the posterior neuropore (PNP) length in Sp^2H/+ embryos treated with 200 µM Z-VAD-FMK was greater than that among vehicle-treated controls, suggesting suppression of spinal neural tube closure (Table S1). Wild-type embryos were unaffected by the inhibitor. Moreover, there was no apparent effect of Z-VAD-FMK on viability, growth or developmental progression in Sp^2H/+ embryos treated with 200 µM Z-VAD-FMK compared with vehicle-treated controls, showing that increased exencephaly frequency does not result from generalised non-specific teratogenicity (Table S1). Hence, inhibition of apoptosis does not prevent NTDs and may even interact with genetic mutation of Pax3 to increase the susceptibility to NTDs in embryos carrying the Sp^2H allele. Overall, our data do not support the hypothesis that NTDs result from increased apoptosis in Pax3 mutant embryos.

We next asked whether the frequency of cranial NTDs in Pax3^Sp2H/Sp2H embryos can be lowered by maternal treatment with pifithrin-α, as described for embryos homozygous for the Sp allele of Pax3 (Pani et al., 2002). We found a lower frequency of cranial NTDs (Fig. 3A,B) among Pax3^Sp2H/Sp2H offspring of pifithrin-α treated mice compared with untreated control litters (Fig. 3A; Table S2).

Pifithrin-α was originally identified in a screen for inhibitors of p53 (Murphy et al., 2004), suggesting this as a mode of action in Pax3 mutants. Surprisingly, however, among litters generated by intercross of Pax3^Sp2H/+; Trp53^+/− mice, we found no effect of p53 genotype on the susceptibility to NTDs in Pax3^Sp2H/Sp2H embryos (Table 1). Cranial NTDs occurred among some Trp53^−/− embryos that were wild type, as predicted from previous reports (Sah et al., 1995), demonstrating effectiveness of the Trp53 mutant allele. Moreover, Trp53 mutation caused an increased rate of cranial NTDs in Pax3^Sp2H/+ heterozygotes (Table 1). Lack of replication of the reported NTD prevention in Pax3^Sp/Sp; Trp53^−/− embryos (Pani et al., 2002) could potentially reflect an allele-specific difference between Sp, which primarily causes spinal NTDs, and Sp^2H, which causes both cranial NTDs (∼60% of homozygotes) and almost fully penetrant spina bifida (Burren et al., 2008; Greene et al., 2009).

Possible cellular responses to p53 activity include not only apoptosis but also cell-cycle arrest, senescence and differentiation (Vousden and Lane, 2007). Therefore, in view of the inconsistent findings with apoptosis inhibition, pifithrin-α usage and genetic knockout of p53, we asked whether cellular differentiation and/or cell-cycle progression may be dysregulated in the neuroepithelium of Pax3 mutant embryos.

In wild-type embryos, the onset of neuronal differentiation correlated closely with the timing of cranial neural tube closure at E9.5 (Fig. 3). Few cells were positive for the early neuronal marker, β-tubulin type III (TuJ1; also known as Tubb3), in the midbrain neuroepithelium at the 13- to 15-somite stages, prior to closure (Fig. 3G,K). Neuronal differentiation then increased concomitantly with completion of cranial neural tube closure, which is achieved in wild-type embryos by the 16-somite stage (Fig. 3K). In contrast, Pax3^Sp2H/Sp2H embryos showed a pronounced shift in the timing of neuronal differentiation in the midbrain, compared with wild-type embryos, with 83% and 210% increases in the number of Tuj1-positive cells at the 14- and 15-somite stages, respectively (Fig. 3H,I,K). These supernumerary neuronal cells arose in the dorsal region of neuroepithelium corresponding to the Pax3 expression domain (Fig. 3C,D; Fig. S2), suggesting that differentiation occurred as a result of loss of Pax3 function. Hence, excess neuronal cells were present in the midbrain neuroepithelium prior to failure of closure and at a stage when they are scarce in wild-type controls. Forebrain closure occurs successfully in Pax3 mutant embryos, the forebrain being a region in which Pax3 expression is weaker than at other axial levels of the neural tube in wild types (Fig. 3B,C). This region exhibited low numbers of Tuj1-positive cells at E9.5 with no difference between genotypes (Fig. 3E,F,L).

Among pifithrin-α treated embryos, concomitant with prevention of exencephaly in Pax3^Sp2H/Sp2H embryos, there was a significant reduction in the number of TuJ1-positive cells in the midbrain neuroepithelium (Fig. 3J,K). Thus, our findings suggest that normalisation of neural tube closure in Pax3 mutant embryos following pifithrin-α treatment may be mediated, not via prevention of p53-dependent apoptosis, but through suppression of premature neuronal differentiation.

We investigated the possibility that premature neuronal differentiation could be associated with altered proliferation in the neuroepithelium of Pax3^Sp2H/Sp2H embryos. At neurulation stages, Pax3^Sp2H/Sp2H embryos had a comparable number of somites and crown-rump length to those of wild-type littermates, showing that there is no overall growth or developmental retardation (Table S3). However, we observed a significant deficit in the proportion of phospho-histone H3 (PHH3)-labelled cells at late-G2/M phase in the dorsal neural plate of Pax3^Sp2H/Sp2H embryos (Fig. 3M). Hence, a dorsal-high, ventral-low proliferation differential was observed in the midbrain neuroepithelium of wild-type embryos at E9.5, consistent with findings in the spinal neural tube (Kim et al., 2007; McShane et al., 2015; Megason and McMahon, 2002), whereas this proliferation differential was absent in Pax3^Sp2H/Sp2H mutants (Fig. 3M).

The observations of premature neuronal differentiation and diminished proliferation in the dorsal cranial neuroepithelium at the stage of closure suggest potential causative mechanisms for cranial NTDs. We therefore asked whether comparable phenotypes were present in the spinal neural folds. Spina bifida occurs in almost all Pax3^Sp2H/Sp2H embryos, and results from failure of PNP closure at E10.5. Measurements of PNP length showed that spinal neural tube closure was delayed from soon after initiation of closure, with an enlarged PNP evident at E9.0 (12- to 13-somite stage) and becoming particularly prominent from E9.5 (14- to 16-somite stage) onwards (Fig. 4A; Fig. S2E). The PNP did not dramatically increase in size as development progressed, unlike in mutants such as Grhl3^−/− (de Castro et al., 2018), indicating that closure progresses in Pax3 mutants. However, this is at a diminished speed such that closure ultimately fails to be completed in the low spine.

Arguing against a primary effect in which the presence of neuronal cells causes spina bifida, abnormal neuronal differentiation was not apparent in the spinal neuroepithelium of Pax3 mutants at E9.5. Neuronal differentiation proceeds in a rostro-caudal temporal wave and, in embryos of all genotypes, Tuj1 staining (whole mount and on sections) was apparent only in the closed neural tube rostral to the site of PNP zippering. Neurons were never observed in the open PNP region of Pax3^+/+ (n=6), Pax3^Sp2H/Sp2H (n=7) or Pax3^Sp2H/+ (n=7) embryos (Fig. 4B,C; Fig. S3). Moreover, the caudal axial level at which Tuj1-positive cells was detected in the closed neural tube did not differ between genotypes (Fig. S3).

As observed in the cranial region (Fig. 3M), the dorsal neuroepithelium of Pax3^Sp2H/Sp2H embryos exhibited an apparent proliferation defect. Hence, at the 14- to 15-somite stage (E9.5), soon after delay of PNP closure became apparent, immunostaining for PHH3 at the rostral level of the PNP revealed a significant decrease in the proportion of cells at late-G2/M phase in the dorsal neural plate (Fig. 4D-E), corresponding to the region of Pax3 loss of function (Fig. S2B). This proliferation deficit abolished the dorsal-ventral proliferation difference that is present in wild-type embryos at this stage (Fig. 4E; Fig. S4A,B) and led to a significant reduction in the number of cells in the neuroepithelium (Fig. 4F), including in both dorsal and ventral subregions (Fig. S4C-E). In parallel with this finding, the cross-sectional area of the neuroepithelium was diminished in Pax3 mutants compared with wild-type embryos (Fig. 4G), as were the dorsoventral height and mediolateral width of the neural tube, but not their relative ratio (Fig. S4G-J). The significant deficit of cells and size of the Pax3 mutant neural plate, detected in 14- to 15-somite-stage embryos, was still evident at the 19- to 20-somite stage (Fig. 4F,G; Fig. S4). These findings show that suppression of PNP closure is associated with a proliferation defect in the dorsal neuroepithelium.

Overall, Pax3 loss of function in the spinal region causes diminished proliferation without induction of neuronal differentiation. It seems likely that Pax3 has comparable functions at cranial and spinal levels. Hence, the effect of Pax3 on the balance between proliferation and differentiation in the cranial region, where the neuroepithelium is primed to differentiate, may be principally mediated through regulation of cell-cycle progression rather than direct inhibition of differentiation.

splotch (Pax3^Sp2H) was the first mouse model in which NTDs were found to be preventable by supplemental FA (Fleming and Copp, 1998), but – as in human NTDs – the protective mechanism has not been defined. FA is reduced via dihydrofolate to THF, which acts to carry 1C groups in folate 1C metabolism (FOCM), a key function of which is to provide 1C units for nucleotide biosynthesis. Our previous analysis showed no effect of Pax3^Sp2H genotype on embryonic folate content at neurulation stages (Burren et al., 2008). Here, mass-spectrometry-based analysis of the individual folate species that are involved in FOCM did not reveal significant perturbation in their relative abundance in Pax3^Sp2H/Sp2H embryos compared with wild type (Fig. 5A). This is consistent with our finding that the relative abundance of s-adenosylhomocysteine and s-adenosylmethionine did not differ with Pax3^Sp2H genotype (Burren et al., 2008). Moreover, unlike loss-of-function mutants of the mitochondrial FOCM components Mthfd1L and Gldc, or the mitochondrial folate transporter SLC25A32 (Momb et al., 2013; Pai et al., 2015; Leung et al., 2017; Kim et al., 2018), NTDs were not prevented by maternal supplementation with formate, a 1C donor, in Pax3 mutants (Table S4).

Cranial neurulation appears highly sensitive to disturbance of cellular proliferation, with exencephaly often being induced by teratogens that cause generalised retardation of embryonic growth or developmental progression (Copp, 2005). Together with our finding of a proliferation defect in the Pax3-deficient cranial and spinal neuroepithelium, this prompted the question of whether FA status impacts cell proliferation in developing Pax3^Sp2H/Sp2H embryos. Indeed, FA is required for proliferation of cultured neural progenitor cells (Ichi et al., 2010, 2012). We therefore carried out further analysis of cell-cycle progression in the Pax3 mutant neural plate in order to determine whether maternal FA treatment could influence cell-cycle progression. Dams were treated with FA using the protocol that we found to reduce the frequency of cranial NTDs among Pax3^Sp2H/Sp2H offspring (Burren et al., 2008; Fleming and Copp, 1998). Embryos were analysed during the period of cranial neurulation at E9.5 (14- to 15-somite stage), with a pulse of bromodeoxyuridine (BrdU) preceding embryo collection to allow analysis of cell-cycle progression in the midbrain neuroepithelium (Fig. 5B).

The proportion of cells in G2 and M phase of the cell cycle were analysed on the basis of the characteristic appearance of PHH3 immunostaining (de Castro et al., 2012). A lower proportion of non-mitotic PHH3-positive (early-G2 to late-G2) cells was present in the dorsal neuroepithelium of Pax3^Sp2H/Sp2H embryos compared with +/+ littermates (Fig. 5C), corresponding with the observed deficit of cells at late-G2/M phase (Fig. 3M). In contrast, the mitotic (M-phase) population did not differ between genotypes (Fig. 5D). Interestingly, there was a small but significantly higher proportion of G2-phase cells in the ventral neuroepithelium of Pax3 mutant compared with +/+ embryos, which suggests a possible non-cell-autonomous effect of Pax3 loss of function.

The BrdU-labelling index (reporting cells in S phase) was also significantly diminished in the dorsal neuroepithelium of Pax3^Sp2H/Sp2H embryos (Fig. 5E), whereas the ventral region did not differ from wild type (Fig. 5E). Hence, both BrdU labelling and PHH3 immunostaining showed the presence of a dorsal-high to ventral-low proliferation differential in wild-type embryos that is absent, or even reversed, in Pax3 mutants (Fig. 5D,E).

FA treatment had a striking effect on cell-cycle parameters, particularly in the dorsal midbrain neuroepithelium of Pax3^Sp2H/Sp2H embryos, where an FA-induced increase in the proportion of PHH3-positive, G2-phase cells was observed (Fig. 5C). Similarly, although FA did not significantly alter the proportion of S-phase, BrdU-labelled cells, there was a trend towards increased labelling in all regions (Fig. 5E). Most notably, in FA-treated Pax3^Sp2H/Sp2H embryos, we observed a significant increase in the proportion of double-positive BrdU/PHH3 cells; i.e. cells that were in S phase during the pulse of BrdU labelling and then progressed into G2 by the time of embryo collection (Fig. 5F). Hence, the rate of transition from S phase to G2 was increased by supplemental FA.

In the dorsal and ventral neuroepithelium of Pax3^+/+ embryos, FA treatment led to non-significant trends towards increased labelling for both PHH3 (Fig. 5C) and BrdU (Fig. 5E). As a result, the dorsal-ventral proliferation differential was absent in FA-treated +/+ embryos. In contrast, the proportion of cells in M phase did not differ with FA treatment in either Pax3 genotype (Fig. 5D).

The Pax3 mutant (Sp^2H) mouse provides a model in which to investigate the cause of NTDs that are sensitive to folate deficiency and responsive to prevention by FA. Analysis of apoptosis markers, together with lack of NTD prevention by inhibition of apoptosis, are incompatible with the idea that increased cell death contributes to failure of neural tube closure. Instead, our findings suggest that Pax3 is required for cell-cycle progression and suppression of neuronal differentiation in the dorsal neuroepithelium until completion of cranial neural tube closure. The close correlation between onset of neuronal differentiation and completion of cranial neural tube closure that is observed in wild-type embryos is lost in Pax3 mutants. Consistent with these observations, anti-sense downregulation of Pax3 in cultured neuronal ND7 cells resulted in morphological differentiation, without apparent cell death (Reeves et al., 1999). Similarly, during melanogenesis, Pax3 induces the melanocyte fate but acts to repress terminal differentiation until additional differentiation signals are present (Lang et al., 2005). Moreover, during myogenesis, Pax3 is required in muscle progenitors but must be downregulated for myoblast differentiation (Goljanek-Whysall et al., 2011). Thus, in different cellular contexts, including the neuroepithelium, Pax3 may function to maintain committed progenitor cells in an undifferentiated, proliferative state until an appropriate developmental stage.

The relationship between Pax3 mutation, cell-cycle regulation and folate status is intriguing. Function of FOCM depends on adequate abundance of the THF ‘backbone’ (from maternal diet and/or microbiota) and supply of 1C units. Supplemental FA can enhance supply of THF but does not carry a 1C unit. During neurulation, the primary requirement for 1C units is met by catabolism of serine and glycine in mitochondrial FOCM (Leung et al., 2017; Tibbetts and Appling, 2010). This generates formate, which is transferred to cytoplasmic FOCM, in which the intermediates 5,10-methylene THF and 10-formyl THF act as the 1C donors in thymidylate and purine biosynthesis, respectively. Conditions of folate deficiency further impose a requirement for serine-dependent generation of 5,10-methylene THF by the action of SHMT1 (Beaudin et al., 2011), acting in the cytoplasm and/or nucleus (Herbig et al., 2002; MacFarlane et al., 2011).

Pax3 genotype does not affect embryonic folate content at neurulation stages (Burren et al., 2008) or the relative proportion of different folate intermediates (the current study). This contrasts with mouse embryos carrying genetic defects in FOCM-associated enzymes such as Mthfr (Leung et al., 2017) and Gldc (Pai et al., 2015), or with methotrexate exposure (a FOCM inhibitor) (Leung et al., 2013). Unlike in Pax3 mutants, NTDs resulting from disruption of the 1C supply from mitochondrial FOCM (e.g. by mutation of Gldc or Amt) are neither FA preventable nor responsive to folate deficiency (Narisawa et al., 2012; Pai et al., 2015). Furthermore, although NTDs in Gldc- and Mthfd1L-null embryos are preventable by maternal supplementation with formate as a 1C donor (Leung et al., 2017; Momb et al., 2013; Pai et al., 2015), we found no apparent protective effect of this treatment in Pax3 mutants. These findings highlight the potential for non-equivalent mechanisms underlying FA-mediated prevention of NTDs compared with prevention by 1C donors (e.g. formate) of NTDs resulting from disruption of FOCM.

Overall, it appears unlikely that NTDs in Pax3 (splotch) embryos result from a deficit in folate uptake, 1C supply or interconversion of folate intermediates in FOCM. Significantly, however, Pax3 mutant embryos and embryonic fibroblasts preferentially use the nucleotide salvage pathway over de novo thymidylate biosynthesis, suggesting a possible impairment of FOCM that could underlie sensitivity to maternal folate status (Beaudin et al., 2011; Burren et al., 2008; Fleming and Copp, 1998). Moreover, the mutual NTD-exacerbating pairwise interactions of folate deficiency, Shmt1 loss of function and Pax3 mutation further implicate thymidylate synthase as an FA-sensitive process for neural tube closure (Beaudin et al., 2011; Burren et al., 2008; Martiniova et al., 2015). For example, folate deficiency or Pax3 mutation can cause NTDs in Shmt1-null embryos, which do not exhibit NTDs on a folate-replete diet (Beaudin et al., 2011). The apparent deficit of de novo thymidylate synthesis in Pax3^Sp2H mutants, under normal dietary conditions, is not due to a limiting pool of 5,10-methylene THF, but may instead relate to post-transcriptional downregulation of thymidylate synthase, as found in Pax3^Sp mutant embryos (Beaudin et al., 2011). A lowering of THF availability in folate-deficient conditions could then further suppress thymidylate synthesis and exacerbate NTDs. This would also be consistent with the increased frequency of Pax3-related NTDs caused by loss of function of SHMT1, the translocation of which to the nucleus favours use of 5,10-methylene THF in thymidylate synthesis over reduction to 5-methyl THF for methylation of homocysteine (Herbig et al., 2002).

We previously found that maternal dietary folate deficiency limits embryonic growth and developmental progression of wild-type and Pax3 mutant embryos without dissociating these parameters (Burren et al., 2008). Here, we find that supplemental FA stimulates cell-cycle progression, and that this effect is mediated via an increased rate of transition through S phase and into G2 phase. An effect of FA in the DNA-synthesis phase of the cell cycle correlates with a proposed mechanism of FA action in stimulating thymidylate biosynthesis.

Notably, the effect of FA in driving cell-cycle progression was greatest in Pax3 mutant regions, such that the overall effect was to normalise the dorsal-ventral proliferation gradient in Pax3^Sp2H/Sp2H embryos. It appears that dorsal proliferation, as opposed to the dorsal-ventral gradient specifically, is associated with NTD prevention, as FA-treated wild-type embryos no longer exhibit a proliferation gradient, but all complete closure. The particular requirement for cellular folate in the dorsal neuroepithelium is highlighted by the notable dorsal (overlapping the Pax3 expression domain) versus ventral enrichment of mRNA for the folate receptor, Folr1 (Saitsu et al., 2003), the knockout of which also causes NTDs (Piedrahita et al., 1999). Overall, we hypothesise that supplemental FA contributes to neural tube closure by promoting S-phase cell-cycle progression, particularly in the neuroepithelial component of the dorsal neural folds.

Pax^Sp2H (Sp^2H; splotch) mice were maintained as a closed, random-bred (heterozygous with wild type) colony. The Sp^2H mutation is carried on a mixed background, which includes CBA/Ca, 101 and C3H/He. Experimental litters were generated by intercross of heterozygous mice. Embryos were genotyped by PCR of genomic DNA (Conway et al., 1997). To generate mice lacking p53, Trp53^flox/+ mice (Jonkers et al., 2001) were used to generate Trp53^+/− heterozygotes for intercross to Pax3^Sp2H. The Pax3^cre allele was used as a null allele in crosses with the p53-null allele (Engleka et al., 2005). Animal studies were carried out under regulations of the Animals (Scientific Procedures) Act 1986 of the UK Government, and in accordance with the guidance issued by the Medical Research Council, UK in Responsibility in the Use of Animals for Medical Research (July 1993).

Pifithrin-α (Calbiochem) treatment at a dose of 2.2 mg/kg was achieved by intraperitoneal injection with 0.22 mg/ml solution [diluted with phosphate buffered saline (PBS) from a 5 mg/ml stock in sterile dimethyl sulfoxide (DMSO)], at E8.5 and E9.5 (Pani et al., 2002). FA (20 mg/kg) was administered by intraperitoneal injection of dams at E7.5, E8.5 and E9.5, using a dose that we have found to lower the frequency of NTDs (Burren et al., 2008; Fleming and Copp, 1998). Embryos were collected 3 h after the final injection. Formate treatment was administered by addition of sodium formate (30 mg/ml) in the drinking water of the dam from E0.5 until the collection of litters (Pai et al., 2015). Litters were dissected from the uterus in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal calf serum. Embryos were rinsed in PBS and fixed in 4% paraformaldehyde (PFA). Embryos were dehydrated in a methanol series and stored at −20°C prior to in situ hybridisation or TUNEL analysis.

For immunohistochemistry, BrdU (Invitrogen) was administered by maternal intraperitoneal injection at 50 mg/kg from a stock solution of 10 mg/ml. Litters were collected after 15 min, embryos explanted in ice-cold DMEM, immediately fixed in 4% PFA, dehydrated through an ethanol series and processed for wax embedding.

Embryos were explanted at E8.5, leaving the yolk sac and ectoplacental cone intact and cultured for 40 h in rat serum at 38°C (Dunlevy et al., 2006; Pryor et al., 2012). Stock solutions of 50 mM and 100 mM Z-VAD-FMK (Sigma-Aldrich) in DMSO were added to cultures as 0.2% (v/v) additions. An equivalent volume of vehicle was added to control cultures, and embryos were randomly allocated to treatment groups.

TUNEL analysis of whole embryos (Dunlevy et al., 2006) was performed on a minimum of eight embryos of each genotype at E8.5, E9.0, E9.5 and E10.5.

Whole-mount in situ hybridisation (de Castro et al., 2018) was performed using sense and anti-sense digoxygenin-labelled riboprobes for Pax3 (Conway et al., 1997), generated using a digoxygenin RNA-labelling kit (Roche) and purified on Chroma spin columns (Clontech).

Immunostaining using antibodies for PHH3 (06-570 Merck Millipore) and activated caspase-3 (Cell Signaling Technology) was performed on paraffin-embedded coronal (cranial) or transverse (spinal) sections. Numbers of positive cells were counted in defined areas, blind to genotype (Dunlevy et al., 2006). The number of positive cells per section was normalised to the area of neuroepithelium (cells/1000 µm²) or the total number of neuroepithelial cells on the section as detected by 4′,6-diamidino-2-phenylindole (DAPI) staining (% labelling).

Double staining for PHH3 and BrdU was performed following the deparaffinisation of slides by HistoClear. Sections were rehydrated through an ethanol series and antigen retrieval was performed in 0.01 M citric acid (pH 6.0). After washing in PBS, 0.1% Triton X-100, slides were blocked in 5% sheep serum, 0.15% glycine, 2% bovine serum albumin (BSA), followed by addition of anti-PHH3 (1:300). Secondary antibody was Alexa Fluor 568 (A11077, Life Technologies), diluted 1:500 in blocking solution. Nuclear stain was DAPI, diluted in PBS (1:10,000). After post-fixation in 4% PFA, slides were incubated in 1 M HCl, followed by 2 M HCl, to expose the BrdU antigen, and neutralised by washing in 0.1 M sodium borate (pH 8.5). Slides were blocked in 10% sheep serum and incubated with anti-BrdU (347583, BD Bioscience; 1:100). After incubation with secondary antibody (Alexa Fluor 488, A11070, Life Technologies; 1:500), slides were mounted [12.5% Mowiol 4-88 (Sigma-Aldrich), 0.2 M Tris base, 0.2 M Tris-HCl, 30% glycerol, pH 6.8]. Images were acquired on an inverted LSM710 confocal microscope (Zeiss).

For TuJ1 (β-tubulin type III) staining, 15 µm coronal cryosections were prepared (Burns and Douarin, 1998). Sections were blocked (PBS, 10% heat-inactivated sheep serum, 0.1% Triton X-100) and treated with primary antibodies against TuJ1 (BAbCO; 1:1000 in blocking solution) or Pax3 (American Type Culture Collection; 1:50), followed by secondary antibody (goat anti-mouse Alexa Fluor 488, Southern Biotech; 1:500). Sections were mounted in Vectashield with DAPI (Vector Laboratories). The number of TuJ1-positive cells in three adjacent matched sections of forebrain and midbrain was counted for each embryo.

For whole-mount immunostaining, embryos were fixed in methanol/DMSO (4:1) and incubated in methanol/DMSO/30% hydrogen peroxide (4:1:1). Embryos were blocked in PBS containing 2% non-fat dried milk, 0.5% Triton X-100 and 10% sheep serum, and then incubated in the same solution with antibody against TuJ1 (1:1000) followed by secondary antibody (donkey anti-mouse Alexa Fluor 488, Invitrogen; 1:500).

Analysis of multiple folates was performed by ultra-pressure liquid chromatography–tandem mass spectrometry as described previously (Leung et al., 2017; Pai et al., 2015). Folates were measured by multiple-reaction monitoring with optimised cone voltage and collision energy for precursor and product ions as described (Cabreiro et al., 2013; Leung et al., 2013).

Statistical analysis was performed using Sigmastat version 3.5 (Systat Software).