Optical coherence tomography-guided Brillouin microscopy highlights regional tissue stiffness differences during anterior neural tube closure in the Mthfd1l murine mutant

Folate (vitamin B9) is an essential nutrient to support healthy cellular development. Folate-dependent one-carbon metabolism is involved in several core biological processes ranging from genome replication and cell division. It is required to produce universal methyl donors, regeneration of redox cofactors, de novo purine and thymidylate synthesis, DNA synthesis and amino acid metabolism (Iskandar and Finnell, 2022; Caiaffa et al., 2023). During the first few weeks of pregnancy, the embryonic stem cells on the blastocyst inner cell mass extensively divide and differentiate into a variety of cell types and tissues to form the axial body plans of an early embryo. At this stage of embryonic development, folate is a crucial requirement as an essential nutrient in the maternal diet (Greene and Copp, 2014; Copp et al., 2015; Finnell et al., 2021). The human body does not produce folate on its own, and to sustain natural cell development and metabolic functions, folate must be obtained from nutrition (National Institutes of Health Fact Sheet, 2022; WHO Executive Board, 2023). Folate deficiency during pregnancy is a major factor leading to the incidence of neural tube defects (NTDs), and adequate intake of folate during the periconceptional period can reduce the prevalence of congenital disabilities (Finnell et al., 2021).

NTDs occur when the neural plate, an ectoderm-derived structure growing adjacent to the notochord and prechordal mesoderm, fails to fold and shape along its anteroposterior axis, resulting in abnormal development of the brain and the spinal cord (Fig. 1). Abnormal neural tube closure can result in conditions such as spina bifida occulta (abnormal closure of the caudal neuropore with no herniation), spina bifida cystica (caudal neuropore fails to close, developing herniation at meninges and neural tissue), anencephaly (abnormal closure of the rostral neuropore leading to brain and skull defects) and several brain abnormalities (Brewer et al., 2002, 2004; Copp et al., 2015; Inman et al., 2018; Tang et al., 2019; Iskandar and Finnell, 2022). Folate is a crucial component of one-carbon metabolism, facilitating the cycling of one-carbon units in the form of methyl groups, formyl groups, formaldehyde or formate, and numerous enzymes and cofactors, including vitamins B6, B9 and B12, regulate their production. One-carbon metabolism is linked to the folate and methionine cycles and is compartmentalized between the cytoplasm, mitochondria and nucleus (Tibbetts and Appling, 2010) (Fig. 1). Generally, carbon units are metabolized from donors such as serine or glycine in the mitochondria and are enzymatically oxidized to formate in a folate-mediated process. The mitochondrial-produced formate is then used in the cytoplasm or nucleus to synthesize purines, pyrimidines and methionine. The methionine cycle occurs primarily in the cytosol to generate and recycle methionine and transfer methyl groups for metabolic reactions and post-translational modifications. In the nucleus, the one-carbon units are used as a source for DNA methylation, which is essential for epigenetic inheritance and regulation of gene expression. Impairment of one-carbon metabolism can lead to a variety of health problems, including NTDs, cardiovascular disease and certain forms of cancer (Ducker and Rabinowitz, 2017).

Methylenetetrahydrofolate dehydrogenase (NADP⁺ dependent) 1-like (Mthfd1l) is a gene encoding a mitochondrial monofunctional enzyme responsible for catalyzing the interconversion of 10-formyl-tetrahydrofolate to formate, which is the last step in the flow of one-carbon units from the mitochondria to the cytoplasm (Fig. 1). Embryos lacking Mthfd1l have a significantly reduced capacity for mitochondrial formate production (Bryant et al., 2018). Mthfd1l is expressed during all stages of mouse embryonic development, presenting higher levels of expression along the neural tube anteroposterior axis, adjacent paraxial mesoderm, craniofacial tissues, brain, limbs and tail buds from embryonic day (E) 9.5 to 13.5 (Pike et al., 2010; Momb et al., 2013). Genetic variants or alterations in the expression levels of Mthfd1l in humans have been associated with cardiovascular disease, neurological conditions, cancer and NTDs (Parle-McDermott et al., 2009; Naj et al., 2010; Franceschini et al., 2011; Minguzzi et al., 2014; Palmer et al., 2014; Hubacek et al., 2015; Xie et al., 2017; Petschner et al., 2018; Wang et al., 2020a; Bischof et al., 2021; Vaughn et al., 2021; Zhao et al., 2021; Chen et al., 2022; Moreira et al., 2022; Sial et al., 2022; Yi et al., 2022; Zhou et al., 2023).

In this context, we focused on studying the effect of mutations of the gene Mthfd1l. Deletion of Mthfd1l impairs normal organogenesis during embryonic development; nullizygous mouse embryos exhibit a characteristic growth delay, aberrant neural tube closure (NTC), abnormal craniofacial development and embryonic lethality by E12.5. The NTD phenotype is variable, including defects such as craniorachischisis, exencephaly and a wavy neural tube (Momb et al., 2013). In addition, Shin and colleagues have shown that E8.5 Mthfd1l mutant embryos present decreased cellular density at the head mesenchyme in an early organogenesis stage where the first point of closure of the hindbrain neuropore is formed (Shin et al., 2019). At this stage, apoptosis and neural crest cell specification were not affected by Mthfd1l ablation (Shin et al., 2019). These findings suggest that NTDs resulting from impairment of one-carbon metabolism may be caused by alterations in local tissue stiffness, disrupting the biomechanical properties that contribute to NTC.

Although our comprehension of the genetic and molecular mechanisms of NTC and NTDs has rapidly grown in recent years, the biomechanical processes leading to proper or aberrant NTC remain largely unknown. Nevertheless, neurulation involves dramatic morphological changes widely accepted to result from a delicate balance between mechanical forces and tissue stiffness (Nikolopoulou et al., 2017), i.e. the resistance to deformation under an applied force. However, it has been extremely challenging to map the biomechanical properties of tissues with high resolution in vivo using a noninvasive method during embryonic development.

Techniques such as atomic force microscopy (AFM) (Park et al., 2012; Chevalier et al., 2016), micro-indentation with optical coherence elastography (Marrese et al., 2019), laser ablation (Galea et al., 2017) and acoustic radiation force elastography (Park et al., 2012) have been previously used to understand embryonic tissue stiffness. However, these techniques are invasive, contact-based, have low resolution, or depth-wise imaging of the entire embryonic neural tube is challenging (e.g. AFM). This study demonstrates a novel, multimodal imaging technique combining optical coherence tomography (OCT) and Brillouin light scattering microscopy to correlate neural tube structural and biomechanical changes in a murine NTD model. Brillouin microscopy is a noninvasive optical imaging technique capable of mapping the Brillouin frequency shift of tissues with high spatial resolution without contact (Scarcelli and Yun, 2007; Palombo and Fioretto, 2019; Prevedel et al., 2019). Imaging structural information is crucial for understanding the physical location of the captured Brillouin frequency shift to correlate embryo morphology with the mechanical information provided by the Brillouin frequency shift. OCT is a well-established technique that can noninvasively provide high-resolution 3D structural details of developing embryos (Raghunathan et al., 2016; Wang et al., 2020b; Scully and Larina, 2022). Earlier studies have demonstrated the feasibility of biomechanical assessment of the developing neural tube in mouse embryos using Brillouin microscopy (Raghunathan et al., 2017; Zhang et al., 2019; Ambekar et al., 2022). However, Brillouin microscopy does not provide structural information, which often results in lengthy alignment and imaging times, which are unsuitable for in vivo studies. Including structural guidance would significantly speed up the imaging time and repeatability of Brillouin microscopy measurements (e.g. imaging the same region in multiple samples). We have therefore developed a co-aligned Brillouin-OCT system with customized instrumentation software (Ambekar et al., 2022). Here, we have adapted, optimized and combined two technologies to build a single Brillouin-OCT instrument for structurally guided tissue stiffness mapping of the neurulation process in Mthfd1l knockout mouse embryos with high resolution.

We demonstrate that tissue stiffness is significantly lower in the neural tube neuroepithelia, at the otic pit, non-neural surface ectoderm and adjacent paraxial mesenchyme of Mthfd1l mutants in comparison with similar-stage wild-type embryos. The neuronal fate of Pax6-positive neural progenitors was disrupted in these embryos, indicating that mitochondrial formate is essential to sustain the contractile properties of the cytoskeleton and general extracellular matrix composition and has a crucial role during mammalian cell differentiation. The importance of mitochondrial formate during cell differentiation was also revealed by Pax3 staining, indicating that somitogenesis is disrupted in Mthfd1l mutants. Our data also indicate decreased Sox10-positive neural crest-derived cells along the embryo anteroposterior axis in the Mthfd1l mutants. Furthermore, we show that maternal supplementation with formate rescues regional tissue stiffness at the neural tube neuroepithelia, otic pit, non-neural surface ectoderm and adjacent paraxial mesenchyme and re-establishes the potential of neural progenitor cell differentiation, somitogenesis and neural crest migration in the supplemented Mthfd1l knockout embryos.

Ablation of the Mthfd1l gene in mice disrupts normal neural tube development (Momb et al., 2013; Shin et al., 2019). Due to a variety of aberrant neural tube phenotypes observed in embryos of this mouse line, we hypothesized that tissue stiffness is an intrinsic component driving failed NTC. Therefore, we took advantage of an innovative multimodal imaging technique, relying on the association of high-resolution OCT and Brillouin microscopy to measure stiffness during anterior NTC. Three-dimensional OCT images (Figs 3A-G and 4A-F) were acquired to locate a region of interest on the embryonic neural tube with 1000 A-lines per B-scans, 1000 B-scans per volume and five frames per position for averaging and improving signal-to-noise ratio (SNR). In addition to structural imaging and morphological phenotyping of the embryos, the OCT images were used to ensure a comparable region was imaged for all embryos. Two-dimensional OCT images consisting of 1000 A-lines per B-scan were also acquired to guide Brillouin imaging at a transverse plane throughout the otic pits in murine embryos at the developmental stages E9.5 (Fig. 2H-N) and E10.5 (Fig. 3G-L). Using the 2D-OCT images as a starting point, a transverse section crossing the otic pits was chosen to capture Brillouin light scattering in terms of Brillouin frequency shift. The dimensions of the Brillouin scans varied based on the size of the cross-section selected using the 2D-OCT image on each embryo. The dimensions of the 2D Brillouin scan for wild-type and heterozygous embryos were ∼1 mm×∼0.6 mm, and for nullizygous embryos, it was ∼0.5 mm×∼0.5 mm, and Brillouin images were acquired in the selected region. The average Brillouin frequency shift was calculated based on the identification of tissues using 2D-OCT images (Fig. S1). The longitudinal modulus derived by the Brillouin shift was calibrated against the Young's modulus obtained by AFM using embryonic tissue. This confirms that the Brillouin shift is a valid metric for assessing tissue biomechanics (Fig. S3).

Early during embryonic development, a relatively soft and flexible extracellular matrix is expected to exist, facilitating neural crest cell migration and a general rearrangement of the embryonic shape. As the embryo develops and the cells further differentiate, the cytoskeleton and the extracellular matrix become more organized, increasing tissue stiffness (Shellard and Mayor, 2021).

The embryos at E9.5 had an average Brillouin frequency shift of 6.18±0.06 GHz (mean±s.d.) for the wild-type embryos (n=4), 6.14±0.06 GHz for heterozygous embryos (n=6) and 6.09±0.05 GHz for nullizygous embryos (n=6) (Fig. 2O-U; Fig. S2; Table S1). The mice embryos at E10.5 had an average Brillouin frequency shift of 6.28±0.09 GHz for the wild-type embryos (n=6). In contrast, heterozygous embryos (n=5) had average shifts of 6.24±0.11 GHz and nullizygous embryos (n=5) presented with shifts of 6.15±0.07 GHz (Fig. 3M-R; Fig. S2; Table S1). Considering the developmental window between E9.5 and E10.5 embryos, the Brillouin frequency shift of the developing neural tube was lower, as expected, in both heterozygous and nullizygous Mthfd1l mutant embryos compared with the wild type. Three-dimensional structural imaging with OCT at E9.5 showed normal NTC in the wild-type embryos (Fig. 2A,B), aberrant neural tube development in heterozygous (Fig. 2C,D) and a characteristic open or wavy neural tube in the nullizygous embryos (Fig. 2E-G). At E10.5, 3D structural imaging with OCT showed the characteristic growth delay expected in Mthfd1l mutants, including aberrant NTC in the mutants (Fig. 3E,F).

Mitochondrial loss of formate production is expected after Mthfd1l gene ablation and has been experimentally reported in Mthfd1l^−/− embryos (Bryant et al., 2018). After determining that aberrant NTC in the Mthfd1l knockout mouse lineage is associated with alterations in tissue stiffness, we sought to determine whether periconceptional maternal formate supplementation would improve tissue stiffness in the mutant embryos. Pregnant dams were given ad libitum access to water containing calcium formate to achieve a calculated dose of 5000 mg calcium formate kg⁻¹ d⁻¹ (Materials and Methods).

At E9.5, the 3D-OCT structural images of formate-supplemented embryos showed normal NTC in all the embryos analyzed (Fig. 4A-F), regardless of genotype. As expected, the 3D-OCT structural images obtained from E10.5 formate-supplemented mutant embryos also showed normal NTC in all the embryos analyzed (Fig. 5A-F). Figs 4G-L and 5G-L show the 2D OCT images of the corresponding planes shown by the red dashed lines in the 3D OCT images in Figs 4A-F and 5A-F. At E9.5, formate-supplemented embryos had an average Brillouin frequency shift of 6.20±0.04 GHz in wild-type embryos (n=3). In contrast, heterozygous embryos (n=3) had a shift of 6.17±0.04 GHz and the nullizygous mutants (n=3) had a shift of 6.17±0.04 GHz. Similarly, at E10.5, the average Brillouin frequency shift in the formate-supplemented wild-type embryos (n=3) was 6.25±0.07 GHz; in the heterozygous embryos (n=3) it was 6.25±0.06 GHz, and in the nullizygous mutant embryos (n=3) it was 6.21±0.06 GHz. Therefore, as quantified by the Brillouin frequency shift, proper tissue stiffness was rescued in the Mthfd1l mutant embryos after formate supplementation (Figs 4M-R and 5M-R).

The complete tissue stiffness analysis is provided in the supplementary Materials and Methods. Imaging by the various methods were performed on numerous litters, but all the embryos analyzed in this work were stage-matched at E9.5 and E10.5 for consistency. For all Mthfd1l supplemented and non-supplemented embryos, the Brillouin frequency shift of the neural tube neuroepithelia, adjacent paraxial mesenchyme, otic pit and non-neural surface ectoderm was statistically tested. The data were tested using Kruskal–Wallis ANOVA and pairwise Dunn's tests to see whether there was a significant difference between genotypes (Table S1). The Brillouin frequency shift of the neural tube neuroepithelia, adjacent paraxial mesenchyme, otic pit and non-neural surface ectoderm was plotted in Fig. 6. At both the E9.5 and E10.5 stages for non-supplemented embryos, the difference in average stiffness in the different regions of the wild-type embryo was more significant in the wild-type and heterozygous embryos than in the nullizygous mutant embryos (Fig. 6). The Brillouin frequency shift of all regions in similarly staged wild-type embryos significantly differed from that of nullizygous mutant embryos. However, for E9.5 stage supplemented Mthfd1l embryos, the neural tube neuroepithelia, adjacent mesenchyme, otic pit area and surface ectoderm were not significantly different between wild-type, heterozygous and nullizygous mutants, showing that formate supplementation overcame the difference in stiffness in these regions. Additionally, formate-supplemented embryos at E10.5 showed no significant difference in stiffness between wild-type, heterozygous and nullizygous mutant embryos, implying full stiffness recovery at this stage (Fig. 6). All plots shown in Fig. 6 depict the inter-sample mean and standard deviation of the Brillouin frequency shift, with data from individual samples plotted alongside. Whole-embryo stiffness quantifications (Table S2) showed the importance of region-wise stiffness characterization. In the case of whole-embryo stiffness, only the difference between the nullizygous embryos showed a difference between the genotype- and stage-matched embryos at E9.5 and E10.5.

The importance of Mthfd1l gene expression during neural plate development and the elevated levels of Mthfd1l expression in the neuroepithelium, together with a variety of aberrant mutant phenotypes associated with decreased tissue stiffness during NTC (Pike et al., 2010; Momb et al., 2013), led us to hypothesize that the population of neural progenitor cells would decrease following the overall growth delay and characteristic asymmetric neuroepithelium bulges observed in the wavy areas of the neural tube in mutant embryos. We anticipated that the neural fate from Pax6-positive progenitor cells would be disrupted along the embryonic anteroposterior axis. We evaluated the neural progenitor fate using immunostaining against Pax6 and Tubb3 (neurofilament-Tuj1) to test this hypothesis. Fig. 7A-C shows that the population of Pax6 positive progenitor cells decreased in Mthfd1l mutant embryos and consequently, the number of differentiated neurons along the body axis significantly decreased. Our data, therefore, demonstrate that formate supplementation, shown in Fig. 7J-L, rescues not only the abnormal levels of Pax6 during neural tube development, but also re-establishes neuronal differentiation from Pax6-positive progenitor cells in E9.5 embryos.

To determine whether the importance of mitochondrial formate was restricted to neuronal cell differentiation or whether it could be correlated with tissue-specific differences in the Mthfd1l ablated mice, we also examined somitogenesis using Pax3 staining and the migration streams of Sox10-positive neural crest cells throughout the embryonic body plan. Pax3 immunostaining indicated that somitogenesis was abnormal in the Mthfd1l mutants. The immunostaining of Sox10-positive neural crest cells indicates decreased neural crest migration along the anteroposterior axis, but the general orientation of dorsoventral migration streams did not change (Fig. 8). Furthermore, we show that maternal supplementation with calcium formate rescues somitogenesis and Sox10-positive neural crest migration along the anteroposterior axis in embryos lacking a functional Mthfd1l gene (Fig. 8).

The extracellular environment composition regulates the synchronized development of the neural tube and adjacent anatomical structures, the activation of gene co-expression network modules, and the biomechanical forces arising from the cytoskeleton, cell-cell adhesions and cell-matrix adhesions. These structures interact mechanically with one another to define an organized shape foundational for embryo development. Understanding and imaging the biomechanical properties underlying embryo morphogenesis at the cellular or tissue level is a challenging problem due to a lack of development of instruments that could be adapted to measure the distribution of mechanical forces in vivo (Shawky and Davidson, 2015; Thompson et al., 2019; Moreira et al., 2022). This study used a multimodal co-aligned OCT-Brillouin system (Ambekar et al., 2022) to measure and map tissue stiffness in Mthfd1l-deficient mouse embryos with and without maternal formate supplementation at two distinct development stages. Structural data obtained using 3D-OCT imaging during hindbrain neurulation revealed genotype-dependent alterations in the distribution of tissue stiffness consistent with the abnormal NTC phenotypes observed in this mouse model (Figs 2E-G and 3D-F).

Following the original description by Momb and colleagues (Momb et al., 2013), our morphological analysis of mutant embryos using 3D-OCT provides a significantly more robust confirmation of the classic Mthfd1l wavy neural tube phenotype (Figs 2E and 3F), in which the neural tube epithelia are compacted in an abnormal wavy line beginning at the cranial region and extending throughout the most caudal areas, disrupting normal NTC. We propose that the existence of these characteristic asymmetric bulges along the embryonic dorsal midline is due to a lack of neuroepithelial stiffness and alterations in stiffness of adjacent tissues, as well as failures located at the biomechanical network interlinking actomyosin contractility, turnover or cytoskeletal assembly, which disrupts the connection of both neuroepithelial walls at their leading edge by the surrounding non-neural surface ectoderm (Zhou et al., 2009; Butler et al., 2019). Interestingly, a similar ‘wavy’ neural tube phenotype (in some cases referred to as a ‘tortuous’ or ‘kinked’ neural tube or spine) has been observed in several other genetic mouse models of NTDs, including mutants of the chromatin regulators Gmnn and Fbxl10 (also known as Kdm2b) (Fukuda et al., 2011; Patterson et al., 2014; Boulard et al., 2016), a methyltransferase necessary for ribosome biogenesis, Emg1 (Armistead et al., 2015), and an apoptosis regulator, Siva1 (Jacobs et al., 2020). Notably, the phenotypes in each model are accompanied by increased apoptosis, decreased cellular proliferation or altered differentiation of neural progenitors, similar to what is observed in Mthfd1l mutants. This information suggests some common mechanisms indicating that cellular density and composition of tissues influence their stiffness during biomechanical processes like NTC.

In this study, we used a new imaging technique using high-resolution 2D-OCT as a starting point to capture Brillouin-shifted scattered light at transversal sections crossing a segment of the embryonic hindbrain. Differences in the Brillouin frequency shift detected in these sections identified the neural tube epithelial walls, surface non-neural ectoderm, otic pits and adjacent mesenchyme. Analysis of the Brillouin frequency shift at E9.5 showed that the surface ectoderm already connected the neural folds at the level of rhombomere 5 in all the embryos analyzed, except for the nullizygous embryos represented in Fig. 2O-U. Tissue stiffness analysis at the neuroepithelial walls, adjacent paraxial mesenchyme, otic pit and non-neural surface ectoderm showed a significant loss of stiffness in the Mthfd1l mutants at E9.5 and E10.5 (Fig. 6). Proper NTC requires an initial contact established between the apposed neural fold tips, which is only possible via cell protrusions and membrane ruffles that will connect the pseudostratified neuroepithelium and the squamous surface ectoderm. The surface ectoderm is also responsible for building a structural backbone of high-tension actomyosin cables to pull the leading edge of both neuroepithelial walls together across the dorsal midline (Rolo et al., 2016; Maniou et al., 2021). Surprisingly, stiffness at the surface ectoderm is lower in Mthfd1l mutants at E9.5 and E10.5, indicating that the opposed neuroepithelial walls cannot present resistance against the biomechanical forces generated by the surface ectoderm actomyosin backbone (Figs 2, 3 and 6).

The Brillouin frequency shift captured from wild-type embryos at E9.5 and E10.5 shows that the otic pit and the neural tube epithelial layers are stiffer than the adjacent mesenchyme and surface ectoderm (Figs 2, 3 and 6). The mesenchyme at the rhombomere 5 transverse section receives a higher contribution of neuroblasts, which are delaminated from the otic pit epithelia, and a much smaller contribution of cells from the neural crest, which may still contribute to inner ear development in small amounts (Trainor et al., 2002; Freyer et al., 2011). At E9.5 and E10.5, the otic pits and the adjacent mesenchyme of mutant embryos were less stiff than those of the wild type. As shown in Figs 4, 5 and 6, formate supplementation also re-establishes stiffness in the otic pits and the adjacent mesenchyme, indicating that these tissues contribute to the balance of stiffness during the development of posterior hindbrain structures.

Shin and colleagues reported in 2019 that deletion of Mthfd1l causes reduced cranial mesenchyme density at E8.5, a stage before hindbrain neuropore formation and subsequent hindbrain neuropore closure (Shin et al., 2019). Tissue stiffness measured at this scale (tens of micrometers) is heavily influenced by cellular density (Weber et al., 2017). Although Brillouin microscopy measurements depend on other factors like hydration, it can still detect tissue stiffness based on other factors (Weber et al., 2017; Ambekar et al., 2022). Earlier results in wild-type embryos have shown a clear difference between the cell-dense neuroepithelial region (higher Brillouin frequency shift) and the less cell-dense mesoderm layer (lower Brillouin frequency shift) (Ambekar et al., 2022; Handler et al., 2023). Our results explicitly show similar observations where the cell-dense neuroepithelial layers had a greater Brillouin shift than the adjacent tissues. In the mutant embryos, there was a reduction in the Brillouin shift in the neuroepithelial region and the mesenchyme layer, indicating a lower stiffness than in the wild-type embryos. Embryos supplemented with formate had greater Brillouin frequency shifts in the mutant than non-supplemented embryos, indicating improvement in the stiffness of the neural tube tissue with supplementation along with a corresponding decrease in the occurrence of NTDs. Therefore, tissues undergoing proper NTC have sufficient stiffness. However, our measurements cannot reveal whether this reduction in stiffness is the cause of the NTDs or an associated phenotype. Further research is needed to elucidate this mechanism and mechanical phenotype, and this work lays a groundwork for such future studies.

The decreased Brillouin shift measured in the adjacent mesenchyme at E9.5 and E10.5 (Figs 4, 5 and 6) supports a scenario where the extracellular matrix and cytoskeletal components derived from mesenchymal cells are softer than the wild type, probably owing to weaknesses in the neuroepithelial walls, which are unable to support the biomechanical forces coming from the actomyosin backbone established by the surface ectoderm. This imbalance of forces established between mutant neuroepithelial walls, the surface ectoderm and the adjacent mesenchyme may explain the appearance of the characteristic neural tube wavy phenotype in mutants of the Mthfd1l lineage (Fig. 2E).

Collectively, considering the developmental window between E9.5 and E10.5 embryos, the Brillouin frequency shift of the developing neural tube was restored after formate supplementation in both heterozygous and nullizygous Mthfd1l mutant embryos, as expected (Fig. 5). Decreased stiffness levels were detected specifically at the neural epithelial walls, which can be correlated with the importance of Mthfd1l for proper neural plate development. Previous work in this mouse lineage indicated that the Mthfd1l gene is highly expressed in the mouse embryonic neuroepithelium (Pike et al., 2010; Momb et al., 2013). Considering the aberrant neural tube phenotypes associated with a significant decrease in the neuroepithelial stiffness during NTC in Mthfd1l mutants, we decided to validate the connection between stiffness and cell fate aberrations in Mthfd1l knockout neural progenitor cells. We chose immunostaining against Pax6 and Tubb3 (Tuj1 monoclonal serum against neurofilaments). Pax6 and Tubb3 provide information about neuronal cells and their differentiation and development of neurofilaments. Fig. 7A-C indicates that neuronal differentiation was disrupted at the cranial ganglia derived from rhombomeres 2, 4, 6, 7 and 8, and along the embryonic posterior axis without mitochondrial formate. Lower neural differentiation was also detected along the body axis using the Tubb3 antibody, as shown in Fig. 7D-F. The Pax6 and Tubb3 results corroborate with the lower stiffness measurements in the Mthfd1l mutants. Neuronal differentiation from Pax6-positive progenitor cells was re-established along the entire body axis after formate supplementation of Mthfd1l mutants (Fig. 7J-O), validating the recovered stiffness observed in Brillouin results.

It is well known that cranial neural crest cell migration happens at the even rhombomeres 2 and 4 anteriorly located in the mouse hindbrain, avoiding the constricted areas in the odd rhombomeres 1, 3 and 5 (Serbedzija et al., 1992; Sechrist et al., 1993). The otic pit is an embryonic structure located adjacent to rhombomere 5, a hindbrain area where the mesenchyme does not receive a significative contribution of the neural crest population of cells (Trainor et al., 2002) but instead is infiltrated by neuroblast cells delaminated from the otic pit (Freyer et al., 2011). Recent publications have demonstrated the role of tissue stiffness in regulating the migration of neural crest cells (Coutiño and Mayor, 2022). Tissue stiffening is a mechanical cue for neural crest migration initiation (Barriga et al., 2018; Shellard and Mayor, 2021), and the ultimate specified fate of post-migration neural crest appears to be determined by the stiffness of the surrounding environment (Li et al., 2011, 2020; Zhu et al., 2019). Considering the decreased tissue stiffness detected in the mesenchyme of Mthfd1l embryonic mutants, we prepared immunostaining experiments to check for alterations in the tissues derived from Sox10-positive neural crest cell migration throughout rhombomeres 2 and 4 and along the embryonic anteroposterior body plan. Sox10-positive neural crest migration is disrupted in the absence of Mthfd1l, but the general orientation of dorsoventral migration streams did not change. Shin and colleagues reported that deletion of Mthfd1l does not affect neural crest cell specification, using immunostaining against Sox9 at E8.5; however, our data suggest that subsequent migration of neural crest cells at later stages is altered. Sox9 is a broader neural progenitor marker, and its expression pattern is not restricted to the neural crest population. Sox9 is expressed all along the developing neural tube axis and activates genes that will induce the migration of neural crest cells (Cheung and Briscoe, 2003; Shin et al., 2019). Sox10-positive neural crest cell migration was also rescued by formate supplementation (Fig. 7).

Considering the importance of Mthfd1l for proper neural differentiation, we also examined somitogenesis to explore a hypothesis where the importance of mitochondrial formate could be correlated with tissue-specific differences in the Mthfd1l-ablated mice. Pax3 immunostaining indicates that somite segmentation is altered at E9.5, following the overall developmental delay observed in Mthfd1l mutants (Momb et al., 2013). Typical somite segmentation could be observed along the embryonic body plan after maternal formate supplementation (Fig. 8). Although it has been shown that the Mthfd1l-ablated mice show a developmental delay of ∼0.75 days (Momb et al., 2013), the Brillouin shift difference observed due to the ablation of Mthfd1l is greater than that observed from a full day of development based on the results in this work and our previous work (Ambekar et al., 2022).

We have also tried indirect fluorescent Phalloidin derivatives to stain for F-actin. However, Phalloidin is not cell-permeable and should not be used on living cells, and the embryos had to be fixed in paraformaldehyde before imaging. At the resolution of our spinning disk confocal images, we did not detect any significant changes in fluorescence. Unfortunately, the mouse embryo is less reliable for fluorescence time-lapse studies than other model organisms like zebrafish, Xenopus or chick embryos. However, our future work is focused on using other methods for potentially imaging actomyosin in embryos, such as chemical imaging, e.g. Raman microscopy (Herrero, 2008), or non-linear imaging methods, e.g. second harmonic generation microscopy (Mohler et al., 2003). Raman microscopy has already been demonstrated as compatible with Brillouin microscopy (Traverso et al., 2015).

During these experiments, we were able to characterize the biomechanical properties of four individual tissue compartments during hindbrain neuropore closure in mouse embryos lacking Mthfd1l. These data suggest that mitochondrial one-carbon metabolism is functionally crucial to modulating embryonic tissue stiffness required for NTC. There are several reasons why this may be the case. Mitochondrial formate is essential for the synthesis of thymidine and the replication of DNA during cellular division. It has also been demonstrated in mouse embryonic stem cells that up to 75% of carbon units entering the methylation cycle are derived from mitochondrial formate (Pike et al., 2010) and, as mentioned earlier, transgenic mouse lines with disruptions in certain demethylases or methyltransferases present with similar neural tube phenotypes as Mthfd1l knockouts. Additionally, mitochondrial one-carbon metabolism has been previously shown in other contexts to mediate redox homeostasis and promote cell survival. Finally, we demonstrate here that certain cell types do not correctly differentiate in the absence of Mthfd1l. Thus, impaired mitochondrial one-carbon metabolism may alter embryonic tissue composition and density through various mechanisms, ultimately disrupting their biomechanical properties. Our results show, for the first time, the distribution of tissue stiffness during this process and that the absence of formate significantly hinders neurulation, resulting in significant defects. Moreover, formate supplementation dramatically reduced the observed stiffness and phenotypic abnormalities, further enforcing the importance of mitochondrial formate in the mechanical processes driving neural tube closure. These correlative measurements show that Brillouin microscopy can image localized tissue stiffness. However, future work is still needed to determine whether the changes in stiffness are a consequence or a cause of NTDs.

All animals were maintained on the C57BL/6 background and housed in a 16 h light/8 h dark cycle. All experiments followed Institutional Animal Care and Use Committees approved protocols at the Baylor College of Medicine and the University of Houston. Mthfd1l, encoding an essential enzyme for tetrahydrofolate synthesis inside the mitochondria, was ablated in this lineage (Pike et al., 2010; Momb et al., 2013). Male and female heterozygous mice were mated overnight, and the existence of the vaginal plug was inspected every morning. The morning when the vaginal plug was detected was determined as E0.5. The pregnant mice were euthanized at E9.5 and E10.5, embryos were dissected, and the embryos were kept in 100% rat serum at Baylor College of Medicine. The fresh embryos were transferred to the University of Houston within 30 min. The yolk sac around the embryos was removed, and the embryos were mounted on a 1.5% agarose plate filled with the culture media. The embryos were aligned with their neural tube side up to acquire Brillouin-OCT measurements. Embryos were imaged in a random order, and pregnant dams were randomly selected for formate supplementation. Brillouin analysis was blinded until needed, i.e. data plotting and statistical analysis.

A previous article described the multimodal Brillouin-OCT system (Ambekar et al., 2022). The home-built system consists of a swept-source OCT subsystem, a Brillouin microscopy subsystem with a dual virtually imaged phase array (VIPA) spectrometer and a combined scanning arm.

Brillouin microscopy is based on spontaneous Brillouin light scattering, an optical phenomenon arising from photons and inherent acoustic phonons (thermodynamic fluctuations) in a sample. The resulting scattered light experiences a frequency shift (Brillouin frequency shift ω_B), which is related to the high-frequency longitudinal modulus M′ (in the GPa range) of the material mechanical properties (Randall and Vaughan, 1982). The Brillouin subsystem used a 660 nm single-mode laser source. The incident power on the sample was 35 mW. The collected backscattered light from the sample was transferred to the dual VIPA spectrometer, and an electron-multiplying charge-coupled device camera was adapted to detect the Brillouin frequency shift of the sample. The camera acquisition time was 0.2 s. The system was calibrated with standard water, acetone and methanol. The sample was imaged with an achromatic doublet with 0.25 NA, resulting in an axial resolution of ∼36 μm and lateral resolution of ∼3.8 μm. The performance of Brillouin microscopy has been previously tested (Ambekar et al., 2020). The swept source OCT subsystem had a central wavelength of ∼1310 nm, a scan rate of 50 kHz, a scan range of ∼105 nm and ∼8 mW incident power on the sample. The lateral and axial resolutions were ∼17.5 μm and ∼10 μm in air, respectively. Light from both systems was combined using a dichroic mirror, and galvanometer-mounted mirrors scanned the beam across the sample. For Brillouin imaging, the sample was stepped by a motorized vertical stage. Custom instrumentation software was developed to use the OCT structural image to guide Brillouin imaging (Ambekar et al., 2022).

The embryo yolk sac tissues were used to perform genotyping to differentiate the embryos as wild-type (Mthfd1l^+/+), heterozygous (Mthfd1l^+/−) and nullizygous mutant (Mthfd1l^−/−) embryos. After imaging on the Brillouin-OCT system, the embryos were fixed in a 2-h treatment in 4% paraformaldehyde, washed in PBS, dehydrated in ethanol and stored for further imaging.

Calcium formate at 200 mM concentration was added to the drinking water of the female Mthfd1l^+/− mice for 10 days before the first attempt at mating. Considering an average intake of 5 ml water per day for an average 25 g mouse, the calculated dose is ∼5000 mg calcium formate kg⁻¹ d⁻¹. After 10 days of supplementation, these mice were set up for timed mating, and at E9.5 and E10.5 the embryos were imaged using the Brillouin-OCT system.

Pax6 (1:500, 42-6600, Thermo Fisher Scientific; RRID: AB_2533534) and Tubb3 (1:500, 2H3-Tuj1, Developmental Studies Hybridoma Bank; RRID: AB_531793), Sox10 (1:500, 89356, Cell Signaling Technology; RRID: AB_2792980) and Pax3 (1:500, Developmental Studies Hybridoma Bank; RRID: AB_528426) whole-mount immunostaining was performed on E9.5 embryos, after fixation in 4% paraformaldehyde and dehydration to 100% methanol. Endogenous peroxidase activity was inhibited by 1 h incubation in Dent's bleach at room temperature. Embryos were rehydrated in TBS containing 0.1% Tween-20 (TBST) and incubated in a 1:500 dilution of anti-Pax6 and anti-Tuj1 overnight at room temperature. Embryos were washed five times in PBST (phosphate-buffered saline with 1% Tween-20) and then incubated with 1:1000 dilution of goat anti-rabbit IgG (H+L) Highly Cross-Adsorbed secondary antibody, Alexa Fluor Plus 488 (ThermoFisher, A32731) or donkey anti-mouse IgG (H+L) Highly Cross-Adsorbed secondary antibody, Alexa Fluor Plus 647 (ThermoFisher, A32787). For general cell visualization, the embryos were incubated overnight in DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride; 2 µg/ml) to label all the nuclei. Embryos were cleared in 25% glycerol and imaged using a Nikon CSU-W1 Yokogawa spinning disc confocal microscope. The obtained z-stacks were projected at maximum intensity and exported as TIFF files (Inman et al., 2018).

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202475.reviewer-comments.pdf