To determine the role of anaerobic glycolysis in NTC, we initially confirmed the mRNA expression of the hypoxia-inducible glycolytic genes Aldoa, Ldha, and Pfkfb3 using whole-mount in situ hybridization. The expression of Aldoa, Ldha, and Pfkfb3 was abundant in the anterior and posterior NP of E8.0 (three to five somite stage) mouse embryos (Figure 1A, dorsal view). Next, we analyzed the mRNA expression profiles using transverse sectioning. We found that Aldoa mRNA localized to the apical surface of neuroepithelial and non-neural ectodermal cells, Ldha mRNA was expressed in neuroepithelial and mesenchymal cells adjacent to the non-neural ectoderm, and Pfkfb3 mRNA was detected in neuroepithelial cells (Figure 1A, section). Thus, our finding revealed that NP expression was common among these three genes. We examined whether the expression of glycolytic genes is dependent on hypoxia. We developed E8.0 (3-5 somite stage) mouse embryos under normoxia (20% O₂) or hypoxia (5% O₂) and then isolated RNA at E8.5 (9–11 somite stage). The results of RT-qPCR revealed that glycolytic gene expression decreased by ∼ 50% under normoxia compared with hypoxia (Figure 1B). The expression of NAD(P)H dehydrogenase (quinone) 1 (Nqo1) that encodes the electron respiratory chain enzyme is independent of, and was not altered by hypoxia. These results indicated that glycolytic genes are expressed in NP and that this part depends on hypoxia. Furthermore, we determined whether hypoxia-dependent glycolytic gene expression is tissue-specific. To address this, we performed in situ hybridization (Figure 1C). Aldoa mRNA expression was diminished in the apical surface of neuroepithelial and non-neural ectodermal cells by normoxia. Unexpectedly, Aldoa mRNA turned out to be expressed in cranial mesenchyme of embryos cultured under normoxia. Ldha mRNA expression was significantly reduced in neuroepithelial cells, but slightly increased in cranial mesenchyme by normoxia. Furthermore, Ldha mRNA was highly expressed in cranial mesenchymal cells localized near the basal side of NP. Pfkfb3 mRNA expression was observed in embryonic cranial mesenchyme, but not in neuroepithelial cells under normoxia (Figure 1C). Collectively, these findings demonstrated that glycolytic gene expression is dependent on hypoxia in neuroepithelial cells of early stage embryos.

Inhibiting glycolysis using the glucose analog 2-deoxy-D-glucose (2-DG) prevents NTC (Hunter and Tugman, 1995). Therefore, we reevaluated the role of glycolytic enzyme activity in NTC. First, we confirmed that inhibiting glycolysis prevented NTC in our ex utero whole-embryo culture system. E8.0 (3-5 somite stage) embryos were cultured with 2-DG for 24 h, then the gross morphology of the embryos was analyzed. We found that NTC was completely impaired by 2-DG, whereas untreated embryos formed normal NTs (data not shown). Next, we incubated E8.0 (3-5 somite stage) embryos with 2-DG or the lactate dehydrogenase inhibitor sodium oxamate (oxamate) for 12 h, then removed the inhibitors by changing the culture medium. The embryos were cultured for a further 24 h (Figure 2A). Figure 2A also shows the developmental events that occurred during exo utero whole-embryo culture. Neural tubes formed normally in control cultures (81.2%, 27/33 embryos). In contrast, NTC was completely prevented by the glycolytic inhibitors (2-DG; 100%, 21/21 embryos. Oxamate; 100%, 15/15 embryos), whereas developmental defects were not found in other tissues (Figure 2B, yellow arrowheads). Incubation with glycolytic inhibitors for 12 h (E8.0-8.5, 3–11 somite stage) was sufficient to prevent NTC, suggesting that glycolytic activity is required for the early phase of NTC process. Moreover, NTC is prevented by exo utero whole-embryo culture under normoxia (Figure 2B, 20% O₂) (75.0%, 6/8 embryos). This result shows that a hypoxic condition is required for glycolytic gene expression during embryogenesis. The TCA cycle and ETC, are the major metabolic pathway for ATP production after E8.5. Hence, incubating E8.5 embryos with FCCP causes NTC defects (Miyazawa et al., 2018), indicating that the ETC, is involved in the late phase of the NTC. We examined if incubation with an ETC, inhibitor for 12 h (E8.0–8.5) affects NTC progression. To examine this, E8.0 embryos (3-5 somite stage) were incubated with the ATP synthase inhibitor oligomycin, or the electron transport chain complex II inhibitor 3-nitropropionic acid (3-NP). Oligomycin and 3-NP obviously retarded embryonic growth and caused abnormal gross morphology, whereas NTC was not impaired in oligomycin-treated (74.0%, 17/23 embryos) and 3-NP-treated embryos (80.0%, 16/20 embryos) (Figure 3). These results indicated that the ETC, is not required during the early phase of NTC progression, at least during E8.0-8.5.

During NTC in mouse embryos, NP undergoes a dynamic morphological transformation comprising the following four spatially and temporally overlapping stages: (1) NP shaping (transverse view of NP is M-shaped), (2) NP folding (V-shaped), (3) NP elevation (U-shaped), and (4) neural fold fusion (O-shaped) (Figure 4A) (Colas and Schoenwolf, 2001). Next, we examined the stage at which the NTC is compromised by glycolytic inhibition using exo utero whole-embryo culture. We initially confirmed that the NTC process was precisely simulated through the appropriate steps in control cultures (Figure 4B, control) (6 h: 100%, 8/8 embryos; 12 h: 100%, 9/9 embryos; 24 h: 87.5%, 7/8 embryos). Notably, the NP retained the M shape in embryos incubated with 2-DG for 24 h (Figure 4B, 2-DG) (6 h: 100%, 10/10 embryos; 12 h: 100%, 10/10 embryos; 24 h: 100%, 10/10 embryos), suggesting that glycolytic activity is required for the onset of NP folding. The NP transformed from the M to the V shape during the first 6h in control cultures. Hence, somite segmentation is controlled by a segmentation clock with a 2-h cycles (Pourquie, 2022), when E8.0 embryos at the 3-5 somite stage develop into the 6-8 somite stage during 6 h of culture. This indicated that NP folding is sensitive to glycolytic inhibition during the 3-5 somite stage. To examine the time window of sensitivity for 2-DG exposure on NP folding, we incubated embryos with 2-DG at the 4, 5, 6, and 7 somite stages for 12, 10, 8, and 6 h, respectively. After the exposure to inhibitors, removed the inhibitors by changing the culture medium. The embryos were cultured for a further 24 h. As a result of glycolytic inhibition, 93.8% (15/16 embryos), 100% (26/26 embryos), and 71.4% (10/14 embryos) of the 4, 5, and 6 somite stage embryos, respectively, had defective NTC. Importantly, 2-DG did not affect the NTC of 7 somite stage embryos (0/13 embryos) (Figure 4C). Taken together, these results suggest that anaerobic glycolysis functions within a short time frame during neural tube development.

We investigated whether defective NTC in 2-DG-treated embryos was associated with cell proliferation, viability, and differentiation. Proliferating and apoptotic cells were detected by immunofluorescent staining using anti-phospho-histone H3 and anti-cleaved caspase3 antibodies, respectively. The number of phospho-histone H3+ mitotic cells moderately decreased in neuroepithelial cells incubated with 2-DG (Figure 5). The numbers of apoptotic cells did not significantly differ between neuroepithelial cells incubated with or without 2-DG-treated (Supplementary Figure S1A). Cell differentiation, including dorso-ventral patterning of the NP and cranial mesenchyme formation, plays crucial roles in NTC (Yamaguchi and Miura, 2013). The expression of Wnt1 mRNA, a marker of the dorsal edge of the NP, and Twist1 mRNA, a marker of cranial mesenchymal cells, was not altered by 2-DG (Supplementary Figure S1B). These findings suggested that glycolysis regulates NTC through the proliferation of neuroepithelial cells.

Neuroepithelial cells undergo dynamic morphological changes during the early phase of NP folding. The thickness of the NP is initially increased by the elongation of neuroepithelial cells along the apico-basal polarity, accompanied by microtubule extension. Thereafter, the apical F-actin ring constricts dependently on the activation of myosin via the phosphorylation of myosin light chain 2 (Suzuki et al., 2012). We investigated whether morphological changes in neuroepithelial cells are impaired by inhibiting glycolysis. Microtubule extension was comparable between NPs incubated with and without 2-DG (Figure 6A, tubulin). The F-actin ring also formed normally in NPs incubated with 2-DG (Figure 6A, F-actin). In contrast, the amount of phosphorylated myosin light chain 2 (pMLC2) was significantly reduced at the apical surface of neuroepithelial cells in 2-DG-treated embryos, suggesting impaired apical F-actin constriction (Figure 6A, pMLC2, and Figure 6B). The amount of pMLC2 was also reduced at the apical surface of neuroepithelial cells in embryos cultured under normoxia (Figure 6C). This defect is probably due to the downregulation of glycolytic gene expression in NP by normoxia (Figure 1C). We evaluated the size of the F-actin ring on the apical side of dissected phalloidin-labeled NPs using confocal microscopy. The size of the apical F-actin ring was substantially reduced in NPs incubated with 2-DG compared to NPs not incubated with 2-DG (Figure 6D). The apical area of individual cells was quantified, and its distribution was quantified. The results showed that the proportion of cells with large apical areas was significantly increased in embryos incubated with, than without 2-DG (Figure 6E). These findings indicated that glycolysis regulates myosin light chain 2 phosphorylation and subsequent apical constriction.

Local ATP synthesis contributes to the dynamic rearrangement of F-actin at the leading edge of migrating cells (De Bock et al., 2013; Cruys et al., 2016; Cunniff et al., 2016). Furthermore, some glycolytic enzymes, such as Aldoa, and Pfkfb3 protein, directly bind to F-actin, and this interaction is required for the local activation of glycolysis (Roberts and Somero, 1987; 1989; De Bock et al., 2013; Hu et al., 2016). Therefore, we examined the cellular localization of Aldoa, Ldha, and Pfkfb3 protein in neuroepithelial cells at E8.25 (six to eight somite stage). These glycolytic enzymes were localized at the apical surface of the neuroepithelial cells, where pMLC2 was also localized. Both Ldha and Pfkfb3 proteins were detected in the cytosol as well (Figure 7). Notably, Aldoa, Ldha, and Pfkfb3 proteins formed punctate structures and resided in close proximity to pMLC2; however, most of these were not colocalized (Figure 7). These results implied that ATP is locally produced to regulate apical constriction.

All animal experiments were performed following the Guidelines for the Care and Use of Laboratory Animals of Kanazawa Medical University. A minimum sample size of five individuals was used in each assay unless otherwise stated. ICR mice were obtained from Sankyo Lab Service. For embryonic staging, the morning on which the vaginal plug was observed was designated as E0.5.

Exo utero whole-embryo culture was performed as previously described (Takahashi et al., 2008; Sakai et al., 2016; Ogoh et al., 2017). Figure 2A shows a schema of pharmacological inhibition. Briefly, E8.0 (three to five somite stage) embryos were dissected and cultured under a 5% O₂; 5% CO₂; 90% N2, 37°C atm in DR50 (50% rat serum; 50% DMEM/F-12, 2% glucose) with or without 0.1 mM 2-deoxy-D-glucose (2-DG), 28 mM oxamate, 0.1 and 0.5 mM oligomycin, or 0.1 and 0.5 mM 3-nitropropionic acid (3-NP). After 12 h of culture (corresponding to E8.5), the culture medium was changed to remove inhibitors and the cells were cultured for 24 h (corresponding to E9.5).

We analyzed gene expression using RT-qPCR as described (Sakai et al., 2022). Briefly, total RNA extracted from E8.5 (9–11 somite stage) embryos was reverse-transcribed into cDNA, which was then amplified by qPCR using SYBR Green on a LightCycler Nano System (Roche). Gene expression was normalized to that of Gapdh. All samples were analyzed at least in triplicate. Relative fold change was calculated using the 2^−ΔΔCT method. The expression of Aldoa, Ldha, Pfkfb3, Nqo1, and Gapdh was detected using the following primers: Aldoa FW: TGG​GAA​GAA​GGA​GAA​CCT​GA and Aldoa RV: GAC​AAG​CGA​GGC​TGT​TGG; Ldha FW: GGCACTGACGCAGACAAG and Ldha RV: TGA​TCA​CCT​CGT​AGG​CAC​TG; Pfkfb3 FW: AAC​AGC​TTT​GAG​GAG​CGT​GT and Pfkfb3 RV: CGG​GAG​CTC​TTC​ATG​TTT​TG; Nqo1 FW: AGC​GTT​CGG​TAT​TAC​GAT​CC and Nqo1 RV: AGT​ACA​ATC​AGG​GCT​CTT​CTC​G; Gapdh FW: CAT​GTT​CCA​GTA​TGA​CTC​CAC​TC and Gapdh RV: GGC​CTC​ACC​CCA​TTT​GAT​GT.

Some mouse Aldoa, Ldha, and Pfkfb3 sequences were amplified by PCR using the following primers: FW-Aldoa: TCT​GAC​ATC​GCT​CAC​CGC​ATT and RV- Aldoa: AAG​AGA​GAT​TCA​CTG​GCT​GCG, FW-Ldha: TGA​AGA​ACC​TTA​GGC​GGG​TG and RV- Ldha: TGT​GTC​TCA​GAG​ACA​GTG​GG, FW-Pfkfb3: TCA​CCA​GGC​TGT​TCT​ACG​CT and RV- Pfkfb3: GTT​GTC​TTT​GCC​ACC​CCA​AC. The PCR products were cloned into the pGEM-T Easy vector (Promega) to synthesize the cRNA probes. Plasmids for the synthesis of cRNA probes against Wnt1 and Twist1 were gifts from Dr. Paul Trainor (Stowers Institute for Medical Research, USA).

Whole-mount in situ hybridization proceeded as described (Sakai et al., 2012).

Mouse embryos were fixed with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 3 h at 4°C and cryopreserved in 30% sucrose in PBS. Brains were embedded in optimal cutting temperature compound (OCT) and stored at −80°C until further use. The cryostat sections (10 μm thick) were adhered to glass slides and washed with PBS. Antigens were retrieved by incubation with 10 mM citric acid (pH 6.0) for 30 min at 80°C. After a brief wash with PBS, the sections were incubated with 0.5% Triton X-100 in PBS for 15 min at room temperature. Non-specific antigen binding was blocked by incubation with a blocking buffer (3% BSA in TBST) for 30 min at room temperature. The sections were then incubated at 4°C overnight with primary antibodies against phospho-histone H3 (ser28) (06-570, Upstate; 1/500), cleaved caspase 3 (9661, Cell Signaling; 1/200), Sox2 (AF 2018, R&D systems; 1/200), acetylated α-tubulin (6-11B-1)(T6793, Sigma; 1/2000), phospho-MLC2 (3671, Cell Signaling; 1/100), Aldoa (sc-12059, Santa Cruz; 1/200), Ldha (sc-27230, Santa Cruz; 1/200), and Pfkfb3 (60,241-1, Protein Tech; 1/500).

The sections were washed three times with TBST for 10 min and then incubated with the appropriate secondary antibodies conjugated with Alexa 488 or 546 (A11001, A21208, Invitrogen; 1/300) for 1 h at room temperature. Nuclei were stained with DAPI. Sections were assessed using a BX51 fluorescence microscope equipped with a DP30BW CCD camera (Olympus) and 10× and 20× objective lenses. Images were acquired using the DP controller software (Olympus).

We outlined apical membranes by staining the F-actin ring with Alexa Fluor-488 phalloidin (A12379, Thermo Fisher; 1/1000). Sections were assessed using a LSM PASCAL confocal fluorescence microscope (Carl Zeiss) and a 40× objective lens. Confocal optical slices were collected, and maximum-intensity projections of 0.3 mm stacks were generated using Zeiss LSM5 software. The apical surface area was calculated using ImageJ software.

Data were statistically analyzed using two-tailed Student's t-tests and Chi-square test. Values with p < 0.05 indicated statistically significant differences.