Novel Mode of Defective Neural Tube Closure in the Non-Obese Diabetic (NOD) Mouse Strain

Failure to close the neural tube results in birth defects, with severity ranging from spina bifida to lethal anencephaly. Few genetic risk factors for neural tube defects are known in humans, highlighting the critical role of environmental risk factors, such as maternal diabetes. Yet, it is not well understood how altered maternal metabolism interferes with embryonic development, and with neurulation in particular. We present evidence from two independent mouse models of diabetic pregnancy that identifies impaired migration of nascent mesodermal cells in the primitive streak as the morphogenetic basis underlying the pathogenesis of neural tube defects. We conclude that perturbed gastrulation not only explains the neurulation defects, but also provides a unifying etiology for the broad spectrum of congenital malformations in diabetic pregnancies.

genetic background of the NOD strain, but arise from the severe maternal hyperglycemia (Supplemental Table 1) common to both experimental models.
Imaging by 2-photon confocal microscopy, three-dimensional reconstruction of whole embryos from optical sections (Fig. 1i), and subsequent generation of single-plane views, allowed closer examination of the juncture between a protrusion and the embryo (Fig. 1j,k). We detected contiguity between the protrusion and the neural plate, with the outer layer resembling neuroepithelium. The core of a protrusion displayed lower cell density, reminiscent of mesenchymal character. This suggested that protrusions are not exclusively composed of neuroepithelial cells.
To determine the origin of protrusions, we performed gene expression profiling on microdissected protrusion tissue from the FVB model, using 3'-expression tag sequencing. For comparison, we a small one at the mid/forebrain region, and a larger one in the trunk area. Insert: virtual sections through unaffected (top) and affected (bottom) neural plate areas. Compared to the unaffected region, the area affected by a protrusion displays an overall normal organization with the exception of a bulge at the midline. This bulge has an outer layer similar to and contiguous with the neuroepithelium, and a core where cell nuclei are sparse. (g,h), Embryos from diabetic pregnancies of the FVB strain display protrusions similar to those observed in the NOD strain. The embryo in panel g has two such malformations: a small one at the caudal end of the prospective hindbrain, and a larger one in the caudal trunk region. (i-k), Images derived from 3-dimensional reconstruction of confocal imaging data of whole embryos stained with DAPI. (i), Same embryo as in panel e, shown with a surface (purple) as calculated from volume data. The protrusion at the hindbrain level is clearly discernible. (j), Two aspects of a virtual parasagittal plane of section in 'openbook' presentation. The protrusion is contiguous with the neuroepithelium, with the core exhibiting a lower density of DAPI-stained cell nuclei, reminiscent of mesenchymal character. (k), 'Open-book' presentation of the FVB embryo shown in panel h. The morphological appearance of protrusions is similar in diabetesexposed NOD and FVB embryos.
laser-microdissected open neural tube immediately anterior of closure site 1 (Supplemental Fig. 2a-c). Analysis of Noto gene expression indicated that neural tube samples were free of potentially contaminating notochord (Supplemental Fig. 2d-h). We identified 799 genes with statistically significant differential expression in protrusions compared to open neural tube (Fig. 2a, Supplemental Figure 2j), and confirmed the sequencing-based observations by quantitative RT-PCR for selected genes (Fig. 2b). Hierarchical clustering demonstrated a clear distinction of expression profiles between protrusions and open neural tube (Supplemental Fig. 2i). Annotations for the 570 genes with predominant expression in protrusions were significantly enriched for the GO terms "mesoderm formation" and "mesoderm development".
For 85 of these genes, expression patterns at Theiler stages 11 to 13 (which correspond to gestational days E7.5 and E8.5) have been reported previously (MGI, http://www.informatics.jax.org/): 22 genes are known to be expressed in neuroectoderm and mesoderm, 33 are limited to mesoderm, 10 are expressed in the node, and 28 genes in the primitive streak (Supplementary Table 4). These results demonstrate that protrusions contain mesoderm, and they link protrusions to molecular networks that are active during gastrulation: Protrusions featured expression of genes known to be involved in early gastrulation, such as Nodal and Furin, which is required for Nodal activation 20 , and of components involved in maintenance of the primitive streak 21 , such as Fgf8, Wnt3a, and T/brachyury. We also detected targets of T, e.g. Cdx2, Axin2, and Lef1, together with 57 genes in the regulatory networks driven by T 22,23 , as well as 153 downstream targets of Cdx2 22 (Supplemental Table 5); these included known axial, paraxial, and lateral mesoderm markers. Presence of these regulatory networks indicates that the mesenchymal cells in the protrusions have undergone the entire currently known mesoderm specification program.
Analyses of NOD embryos by in situ hybridization at E8.5 ( Fig. 3) revealed parallels between protrusions arising from spontaneously initiating and chemically induced maternal hyperglycemia, respectively. Sox2, a marker for the epiblast/neuronal lineage 24 , was present throughout the outer layer of the protrusion, whereas T, a marker for primitive streak and nascent mesoderm 25 , was extended into the proximal core of the protrusion. Tbx6, a marker for committed mesoderm 24,26 , was found in the primitive streak were done using DESeq. Data are shown as a volcano plot, with statistical significance expressed as -log10(padj) plotted against the expression ratio (protrusion vs. neural tube). Gray color represents genes that did not reach statistical significance (Benjamini-Hochberg correction); red labels genes with expression prevalence in the protrusions; among these genes, blue color indicates genes with a known role in mesoderm development. Green indicates genes with prevalence in the open neural tube, with orange color indicating two genes of interest. Genes where expression was completely absent on one side of the paradigm were plotted at a log2 of either 10 or − 10. (b) Validation of the expression difference of selected genes by quantitative real-time PCR. Eight genes with prevalence in protrusions (Acvr1b, Dll3, Mixl1, Nodal, Noto, T, Tbx6, and Wnt3a) were chosen for validation, and two genes (Fgfr2, Zic1) with expression predominant in the open neural tube. The same samples were analyzed from which the sequencing data were derived; thus, we measured gene expression levels in four technical replicates each of three protrusions, and in four technical replicates each of 4 neuroepithelium samples (from 4 individual embryos). Bars represent expression ratio between protrusions and neural tube expressed as log2. Statistical significance in a t-test is indicated by a star symbol. Expression ratios obtained from the sequencing experiment were confirmed for all candidate genes, except for Nodal, which exhibited the expected direction for the expression difference, but did not reach statistical significance. and migrating cells of the mesodermal wings. Tbx6 expression in the core of a protrusion was reminiscent of proper mesoderm development 27 : expression was strongest where T expression had already been extinguished. Overall, these data indicate that migration of mesodermal cells is not completely blocked, but impaired locally around the protrusion.
The ectopic mesoderm in the protrusions could result from altered proliferation of newly generated mesodermal cells, or from disoriented migration of nascent mesoderm. Histological analyses support the second possibility: we did not find evidence for increased cell proliferation, as staining for the mitosis marker Phospho-Histone 3 did not reveal enrichment in protrusions compared to the rest of the embryo. Instead, we detected deposition of Laminin between mesodermal cells within and at the base of a protrusion (Fig. 4). This is paralleled by protrusion-prevalent expression of Laminin α 5, Nidogen 2 (components of the basal lamina 28,29 ), and Integrin α 6, a receptor for Lama5 30 , and is consistent with a previous report of elevated expression of extracellular matrix components in rat embryos exposed to hyperglycemia conditions 31 . These data implicate altered cell adhesion or impaired migration of mesodermal cells in the formation of protrusions.
Evidence for impaired migration came from explant cultures (Fig. 5) under conditions that support migration and differentiation of mesoderm, confirmed by virtue of staining for Vimentin 32 (Fig. 5b,c). In these cultures, outgrowth from posterior tissue explants of diabetes-exposed NOD embryos at E7.5, or at E8.5, was significantly reduced compared to migration from explants of embryos from normoglycemic NOD pregnancies (Fig. 6a). Furthermore, for diabetes-exposed embryos, we compared cultures of dissected protrusions to explants of the adjacent posterior tissue at E8.5. Cells from protrusions either failed to migrate away from the explant, or migrated significantly less compared to the outgrowth observed from the corresponding posterior tissue explant (compare within green frames: Fig. 5g,h,i to j,k,l, and Fig. 5m,n,o to p,q,r, respectively; and Fig. 6b). The reduced migration of cells away from protrusions cannot be attributed to developmental immaturity, as cells from earlier embryos exhibit comparable migratory capacity in these assays (Supplemental Fig. 3a); the extent of outgrowth was also uncorrelated to size of the starting explant (Supplemental Fig. 3b). We therefore conclude that the exposure to maternal diabetes is responsible for the impaired migratory capacity of mesoderm in protrusions. Finally, outgrowth of explants from diabetes-exposed NOD E8.5 embryos was reduced in comparison to explants  ,g,k); red); merged color images are shown in (d,h,l) respectively. Abbreviations: ac, amniotic cavity; da, dorsal aorta; en, endoderm; ep, epiblast layer; hf, headfold; hg, hindgut; mw, mesodermal wing; np, neural plate; p, protrusion; so, somatopleura; spl, splanchnopleura. All protrusions, from the early stage to the mature stage, feature internal accumulations of laminin in the mesenchyme (yellow triangles). Furthermore, the outer cell layer is typically separated from the internal mesenchyme by a laminin-positive layer of extracellular matrix. of diabetes-exposed E8.5 embryos of the FVB strain (Fig. 6c), which could explain the higher incidence of protrusions in the NOD model compared to FVB.
These results also demonstrate that the culture conditions, including supportive extracellular matrix and growth factors present in fetal calf serum, are not sufficient to rescue the cell migration deficiencies during the 26 hours of culture of the explants. Culture at lower glucose concentrations had no effect on outcome (p = 0.23 for E7.5, p = 0.22 for E8.5). Thus, the explant cultures confirm our conclusion that the exposure in utero to maternal diabetes causes impaired cell migration and reduces egress from the primitive streak, which in the most severely affected individuals creates protrusions from the neural plate.
Intriguingly, protrusions formed at discrete anterior-posterior locations rather than along the entire primitive streak. Consistent with prior data 55 , we found that two thirds of NTDs in NOD embryos (~26% of all embryos) involve the trunk region of the embryo; this rate is comparable to a protrusion incidence of 25%, of which almost all appear in the territory covered by the primitive streak. In the FVB model, half the NTDs (~11% of all embryos) involve the trunk (unpublished observations), a rate that again parallels the protrusion incidence (12.9%). Thus, in both diabetes models, defective mesoderm migration can account for the vast majority of trunk and caudal neural tube defects in mouse diabetic pregnancies. Protrusions in more anterior locations were detected only occasionally (e.g. Fig. 1F and insert). Even within the primitive streak territory, protrusions were limited to unique locations, possibly indicative of a limited time window for perturbations that contribute to the formation of protrusions.
There are three possibilities how protrusions can cause neural tube defects: (i) by preventing formation of the medial hinge point that is required for initial bending of the neural plate 56,57 , (ii) by compromising elevation and bending of future neuroepithelium 57 due to diminished cell migration into the underlying mesoderm, and (iii) by physically interfering with the closure of the neural tube at the dorsal midline. Given that these alternatives are not mutually exclusive, they require further investigation. Close examinations and histological analyses of embryos with protrusions (Fig. 7) revealed properly closed neural tubes rostral to the protrusion, with neurulation failure caudal to the protrusion, indicating that in these cases the protrusions physically interfered with closure of the neural tube.
In this work, we have identified impaired mesoderm migration as the morphogenetic failure underlying the pathogenesis of NTDs. Since these NTDs are caused by an environmental risk factor, maternal Figure 6. Exposure to maternal diabetes decreases cell migration in explant cultures. Quantification of outgrowth from the explants as the net outward distance migrated by cells at the margin of each explant between hours 6 and 26 after initiation of culture. Asterisks mark statistical significance in a t-test (p < 0.05). (a) Posterior tissue explants from diabetes-exposed (exp., orange) NOD embryos display significantly reduced outgrowth compared to explants from normal (n, white) pregnancies at E7.5 (normal n = 10, exposed n = 7; p = 1.7 × 10 −3 ; 96.7% power at alpha = 0.05) and E8.5 (normal n = 8, exposed n = 39; p = 6.7 × 10 −4 ; 99.7% power at alpha = 0.05). (b) Explants of dissected protrusions (red) produced significantly less outgrowth than the corresponding posterior tissue (orange) from the same embryo (n = 12 pairs; p = 6.2 × 10 −5 ; 99.9% power at alpha = 0.05) at E8.5. Note the difference in scale of the Y-axis. (c) Posterior tissue explants from diabetes-exposed E8.5 NOD embryos (orange; same data as in (a); n = 39) exhibit reduced outgrowth compared to explants from diabetes-exposed FVB embryos at E8.5 (purple; n = 6; p = 4.4 × 10 −7 ; 100% power at alpha = 0.05).
metabolic disease, our findings imply that mesoderm migration is sensitive to metabolic state. Mesoderm migration also appears to be responsive to composition of the maternal diet, as we previously demonstrated that diet modulates the rate of NTDs in the FVB model 19 . In the NOD strain, NTD incidence is reduced by supplementation of folinic acid, as shown above, similar to the beneficial effects of folic acid in STZ-induced diabetic mouse pregnancies 58 . These findings support the conclusion that metabolic factors can affect mesoderm formation and migration, and -together with the results from our molecular analyses-identify novel cellular and molecular targets for the prevention of neural tube and other birth defects.
The most characteristic congenital malformations in human diabetic pregnancies are heart defects, neural tube defects, and caudal growth defects 7-10 , and have been postulated to arise before the 7th week of pregnancy 7 . Our results support the proposition that perturbed mesoderm migration during gastrulation is the common etiology for these seemingly heterogeneous birth defects 59,60 : (i) neural tube defects arise as a consequence of impaired mesoderm migration, as demonstrated here; (ii) early heart progenitors originate and migrate from the primitive streak 61,62 ; and (iii) caudal growth defects are also consistent with altered mesoderm formation and migration 63,64 in the posterior primitive streak. Similarly, mesodermal deficiencies are believed to underlie the vertebral, cardiac, renal and limb malformations of VACTERL 65 and axial mesodermal dysplasia 66 phenotypes, which have been linked to maternal diabetes [67][68][69][70] . Thus, our discovery of aberrant mesoderm migration during gastrulation in two different mouse models of Type I diabetes provides a unifying cellular mechanism that can explain both the developmental timing and the morphogenetic origin of the most common structural anomalies in diabetic embryopathy.

Methods Summary
All animal experiments were performed with prior approval of the Pennington Biomedical Research Center IACUC and in accordance with the "Guide for the care and use of laboratory animals" of the United States National Institutes of Health. Embryos were prepared at 8.5 days of gestation for histological or molecular analysis from hyperglycemic NOD or FVB dams, as well as from strain-matched normoglycemic control dams. Embryos with malformations were fixed, stained with DAPI, and imaged by 2-photon confocal microscopy. Optical sections were used to generate three-dimensional reconstructions of individual embryos using Imaris software. To determine etiology and identity of protrusion tissue, we performed gene expression profiling by expression tag sequencing on an AB SOLiD 5500XL sequencer; expression profiles were compared between protrusion tissue and open neural plate prepared by laser microdissection immediately anterior of neural tube closure site 1. Sequence reads were mapped (RefSeq RNA, mm9) using SOLiDSAGE to generate count data for each gene. Differential gene expression was determined using DESeq, with validation of select genes by qPCR. In situ hybridizations and immunohistochemical analyses were performed on cryosections following established protocols. Migratory capacity of cells in protrusions and posterior embryonic tissue was assessed in explant culture, using time-lapse video and phase contrast microscopy.