Dietary folic acid prevents peripheral neuropathy in mouse models of neural tube defects and type 2 diabetes

Significance

This study revealed a shared genetic etiology between folic acid–responsive neural tube defects and the development of peripheral neuropathy (PN) in mice with reduced Shmt1 expression. Furthermore, we demonstrate that high-dose dietary folic acid prevents PN in both Shmt1-deficient and Lepr^db/db diabetic mice, indicating that the diabetic state increases folic acid dietary needs to prevent PN. These results indicate a special nutritional requirement for folic acid in diabetes and implicate impaired de novo thymidylate synthesis in PN.

Abstract

Folate-mediated one-carbon metabolism is implicated in several pathologies including neural tube defects (NTDs), cancer, and neurodegenerative disorders, whereas diabetes is associated with NTDs and peripheral neuropathy (PN). The development of peripheral neuropathy was assessed in Shmt1^+/− and Shmt1^−/− mice, which are models of human folic acid–responsive NTDs, and diabetic (Lepr^db) mice to determine whether NTDs and PN have a shared etiology. From 6 wk of age, male and female mice with reduced Shmt1 expression exhibited PN, with greater severity in females compared to males. The neuropathic progression was distinct from diabetic peripheral neuropathy (DPN) observed in Lepr^db mice. Excess dietary folic acid prevented PN in both Shmt1^−/− and Lepr^db/db mice, whereas dietary uridine caused demyelinating PN in mice independent of genotype and folate status. The transcriptome from L3-L5 dorsal root ganglia (DRG) exhibited distinct sex-specific differences in glial cell gene expression when comparing Shmt1^+/+ and Shmt1^−/− mice. DRG sensory neurons exhibited changes in the expression of solute carriers and ion channels involved in nociception, neurotransmission, and structural support. We conclude that reduced thymidylate synthesis causes folic acid–responsive NTDs and PN in mice and that diabetes sensitizes mice to folic acid–responsive PN. Diabetes induces a special nutritional requirement for high intake of folic acid to prevent PN.

Peripheral neuropathy (PN) refers to conditions that decrease the function of the peripheral nervous system. PN is classified as either axonal or demyelinating and often presents as a comorbidity of a primary disease with a wide range of etiologies and heterogeneous clinical symptoms (1, 2). In the United States, PN prevalence is 2 to 8% of the population, with the actual prevalence likely much higher as the disease is often asymptomatic and undiagnosed. PN prevalence exceeds 14% of the population over age 40, and nearly 30% of cases are associated with diabetes (3, 4). Major risk factors for PN include type 2 diabetes mellitus (T2DM), genetics, chemotherapy, autoimmune disease, chronic malabsorption-induced inflammatory and metabolic stress, and B-vitamin deficiencies.

B-vitamin deficiencies, including B1, B6, folate, and B12, can cause neuropathy and dietary supplementation that addresses the deficiency can restore nerve function (5–7). Folates are essential cofactors for proper neuronal and glial cell function by supporting nucleotide synthesis and homocysteine remethylation to methionine, which are outputs of the folate-mediated one-carbon metabolic (FOCM) network (8). Impaired FOCM has been linked to numerous pathological conditions including neural tube defects (NTDs), cancer, and cardiovascular and neurodegenerative disorders (9). The demand for one-carbon units is highest during fetal development, and folate deficiency can lead to NTDs and congenital heart anomalies in sensitized individuals (10, 11).

Serine hydroxymethyltransferase1 (SHMT1) is a folate-dependent enzyme that participates in and regulates de novo thymidylate biosynthesis and homocysteine remethylation (12, 13). Cytoplasmic SHMT1 is SUMOylated at the G1/S boundary of the cell cycle and transported into the nucleus where it functions as a scaffold essential for the assembly of the de novo thymidylate synthesis complex at the replication fork during DNA replication. The complex also includes the enzymes TYMS, MTHFD1, and DHFR (14, 15). Shmt1^+/− and Shmt1^−/− mice are viable and thrive, but are sensitized to low-penetrant folic acid–responsive NTDs, colon cancer, and impaired neurogenesis in the CNS (16–18). Shmt1 deficient mice are the only folic acid–responsive NTD model resulting from mild genetic impairment of a folate-dependent enzyme (17). NTDs in this model are also prevented by dietary deoxyuridine. In contrast, dietary uridine causes NTDs in mice independent of genotype or folate status (19). The Shmt1 mouse model phenocopies human NTD pathogenesis by exhibiting low penetrance, subtle alteration in metabolic biomarkers, and folic acid responsiveness and hence serves as a humanized mouse model to study other neurological disorders.

Maternal type 2 diabetes mellitus (T2DM) is a risk factor for diabetic peripheral neuropathy (DPN), NTDs, and other fetal congenital anomalies (20, 21). DPN manifests with sensory axonal damage and progresses to effect the entire neuron driven by hyperglycemia-induced cell death. Several pathways have been implicated in the etiology of DPN including PARP-mediated cell death activated by neuromodulins (GAP43), β-tubulins, and HSPs, uncontrolled extracellular hyperglycemia-induced oxidative damage, mitochondrial dysfunction, ER stress, and abnormal posttranslational modifications (22). DPN treatment mainly involves pain management, lifestyle/dietary modifications, and strict glycemic control. Glycemic control is an effective strategy for DPN treatment associated mainly with type 1 diabetes (23). DPN etiology and progression have been well characterized using the Lepr^db mouse model. In the current study, we investigated whether NTDs and PN share common etiologies in the Shmt1 mouse models and whether dietary folic acid or nucleosides could prevent PN.

Results

Mice Lacking Shmt1 Exhibit Sex-Specific Peripheral Neuropathy.

At 6 mo of age, Shmt1^−/− mice exhibited significantly reduced motor nerve conduction velocity (NCV) in the sciatic nerve compared to Shmt1^+/+mice, both fed the control diet. The reduction was more pronounced in females (34%, p_adj =0.013) compared to males (20.6%, p_adj = 0.02) (Fig. 1 A and B). Dietary folic acid (FA) supplementation (8 mg/kg) partially rescued (75.1%; p_adj = 0.032) the NCV deficit in female mice, but not in male mice. Deoxyuridine supplementation did not rescue in NCV among Shmt1^−/− mice of either sex (Fig. 1B).

Fig. 1.

Electrophysiology and in vitro characterization of peripheral neuropathy observed in 6-mo-old Shmt1^+/+ and Shmt1^−/− mice. (A) Work flow showing measurement of nerve conduction velocity (NCV) in mice that were randomly weaned to various diet groups. The control group consisted of Shmt1^+/+ mice weaned to AIN93G (2 mg/kg folic acid) only. Shmt1^−/− mice were randomly weaned to AIN93G diet or AIN93G + FA (8 mg/kg folic acid) or AIN93G + 0.1% dU (2 mg/kg folic acid, 0.1% deoxyuridine) for 6 mo followed by electrophysical and biochemical measurements. (B) Sciatic motor nerve conduction velocity comparison between female (Left) vs. male (Right) mice on different diets. Relative NCV (m/s) was plotted on the Y-axis. Each data point represents mean NCV per mouse, calculated from both left and right limbs using the formula: $\mathrm{NCV}\left(\frac{\mathrm{m}}{\mathrm{s}}\right)=\frac{\mathrm{Distance} \mathrm{between} \mathrm{proximal} \mathrm{and} \mathrm{distal} \mathrm{stimulation}}{\mathrm{Latency} \mathrm{of} \mathrm{response} \mathrm{for} \mathrm{proximal} \mathrm{stimulus}-\mathrm{Latency} \mathrm{of} \mathrm{response} \mathrm{for} \mathrm{distal} \mathrm{stimulus}}$ Latency of response times were recorded for three independent proximal and distal stimulations (24). On the X-axis, the diet/genotype group corresponding to NCV data are plotted. Data are shown as mean NCV ± SEM (n = 14 to 16 mice/group). Adjusted p-values were represented only where the comparisons were statistically significant (P < 0.05). (C) Additional electrophysiology parameters were extracted for each functional response curve resulting from distal and proximal stimulations. Compound motor action potential (CMAP) is represented by the amplitude (mV) of functional response curve, while the duration of the response curve was also measured. Panel C (Top and Bottom) represent data from female and male mice, respectively. No change in CMAP or duration of response were observed for either male or female mice across the diet and genotype groups. (D) Sciatic nerve lysates extracted from mice on respective diet and genotype groups were probed with neuron-specific antibodies: neurofascin 155 (NF155, 155kD), myelin associated glycoprotein (MAG, 100kD), myelin basic protein (MBP, 18-21 kD) and internal loading control β3 tubulin (50 kD). Panel D (Left) shows semiquantitative immunoblots from N = 3 biological replicates of each diet/genotype group. Panel D (Right) shows densitometric analysis of the immunoblots on ImageJ. Relative quantitative expression, normalized to β3 tubulin was plotted on the Y-axis whereas, the diet/genotype groups were plotted on the X-axis. Bar graphs on the Right indicate relative expression ± SEM (n = 3 mice/group). Statistical significance, where applicable, was shown with asterix for *p_adj = 0.016 and ***p_adj = 0.0001. Changes were seen in MAG but none were observed for MBP or NF155 protein expression across diet/genotype groups for either sex. (E) Sciatic nerve mRNAs were extracted from Shmt1^+/+ and Shmt1^−/− female and male mice on AIN93G diet for 6 mo. RT-qPCR was performed to test transcriptional changes in the expression of Schwann cell transcription factors (Sox10, Sox2, Krox20, Oct6, Mbp, and Mpz) normalized against endogenous control 18S RNA. Data represent mean normalized mRNA expression ± SEM (n = 3 mice/group for female and n = 4 mice per group for male). Only expression of the Krox20 gene in Shmt1^−/− female was reduced (*P = 0.036).

Despite the reduced NCV, Shmt1^−/− mice fed the control diet showed no significant changes in the compound motor action potential (CMAP) or duration of response (Fig. 1 C, Top and Bottom). In addition, histological analysis of sciatic nerves stained with myelin–specific Luxol fast blue revealed no structural abnormalities (SI Appendix, Fig. S1). These results indicated that the reduced NCV was not due to major structural damage of motor fibers in the sciatic nerve.

Relative quantitative immunoblots of sciatic nerve lysates revealed expression of myelin-associated glycoprotein (MAG) was significantly reduced in the Shmt1^−/− female (53.6%, p_adj = 0.016) and male (70.6%, p_adj = 0.0001) mice fed the control diet (Fig. 1 D, Top) compared to Shmt1^+/+mice fed the control diet. Folic acid supplementation partially elevated MAG expression in Shmt1^−/− mice, but the changes were not statistically significant for either sex (female, 19.2% p_adj = 0.28; male, 32.3% p_adj = 0.14) (Fig. 1 D, Bottom).

Messenger RNA levels of transcription factors (TF) critical for Schwann cell differentiation (Sox10, Sox2, Krox20, Pou3f1, Mbp, and Mpz) (25) were quantified by RT-qPCR from sciatic nerve mRNA in wild-type and Shmt1^−/− mice fed the control diet for 6 mo. Only Krox20 (or Egr2) was significantly downregulated in Shmt1^−/− female mice compared to Shmt1^+/+mice (37.2% p_one-tail = 0.025) (Fig. 1E). Krox20 is known to regulate MAG expression and Schwann cell maturation (26).

Peripheral Neuropathy in Shmt1^−/− Mice Presents Early in Life.

In order to determine whether PN in Shmt1^−/− mice was congenital or develops progressively, NCV was assessed in the sciatic nerve of 6-wk-old mice, which is prior to full myelin maturation. Shmt1^+/+ and Shmt1^−/− mice were weaned to control diets for 3 wk (Fig. 2A) followed by NCV measurements. Significant reductions in NCV were observed in both female (46.8% reduction, P < 0.001) and male (29.1% reduction, P < 0.001) Shmt1^−/− mice compared to wild-type mice fed the control diet (Fig. 2B). The magnitude of these reductions mirrored those observed at 6 mo, indicating the neuropathy is likely congenital and not the result of progressive demyelination. As with older mice, CMAP amplitude and response duration remained unchanged across genotypes and sexes at 6 wk (Fig. 2C), further indicating that the peripheral nerve conduction is functionally impaired without overt axonal loss or degeneration.

Fig. 2.

Characterization of peripheral neuropathy observed in 6-wk-old Shmt1^+/+ and Shmt1^−/− mice. (A) Workflow of study showing Shmt1^+/+ and Shmt1^−/− mice were weaned to AIN93G diet (2 mg/kg folic acid) for 3 wk before measurement of motor nerve conduction velocity on the sciatic nerve. (B) Motor NCV comparisons between Shmt1^+/+ and Shmt1^−/−male (Left) and female (Right) mice. The Y-axis shows relative motor NCV (m/s) calculated for proximal and distal stimulations. Each data point represents mean NCV for one mouse, calculated from both left and right limbs using the same method and replicates as discussed previously. Data are shown as NCV ± SEM (n = 15 to 16 mice/group). (C) Additional electrophysical parameters are represented for functional response curves. The Y-axis shows maximal amplitude of the response (compound motor action potential, CMAP, mV) and the mean duration of response (ms), corresponding to either a distal (near) or proximal (far) stimulation. Each data point represents mean CMAP and mean duration ± SEM. No significant changes were observed in the electrophysiology profiles (CMAP or duration) for the 6-wk-old mice.

Furthermore, western blot analyses of sciatic nerves from 6-wk-old Shmt1^−/− mice did not exhibit changes in levels of MAG, MBP, or NF155 proteins in either sex (SI Appendix, Fig. S2). In summary, these findings indicate PN associated with impairment of Shmt1 appears congenital and related to functional deficits.

Sex-Specific Transcriptomic Alterations in Dorsal Root Ganglia (L3-L5) of Shmt1^−/− Mice.

The effect of diet, sex, and Shmt1 genotype on PN-associated changes in gene expression was determined by RNA sequencing of L3-L5 dorsal root ganglia (DRGs) from Shmt1^+/+ and Shmt1^−/− male and female mice fed the control or folic acid–supplemented diets for 6 mo (Fig. 3A). In silico deconvolution of the input RNA samples against the published reference single-cell dataset (27) revealed no significant differences in DRG cell type composition among the diet, sex, or genotype groups. Thus, confirming consistency in the dissection and harvest of the DRGs (SI Appendix, Fig. S3). Comparative datasets are provided in the SI Appendix, Tables S1–S7.

Fig. 3.

RNA-seq analysis of bulk RNA extracted from 6-mo-old Shmt1^+/+ and Shmt1^−/− mice DRGs (L3-L5). (A) Schematic overview of the workflow showing Shmt1^+/+ and Shmt1^−/− mice weaned onto AIN93G diets for 6 mo. Shmt1^+/+ and Shmt1^−/− male mice were weaned onto the AIN93G diet. Shmt1^−/− female mice were weaned to either the AIN93G or the AIN93G+FA diet for 6 mo. Mice were on the respective diets for 6 mo before DRGs were harvested from L3-L5 positions of lumbar spinal cord followed by total RNA extraction and purification. Samples pooled from (L3-L5; Left and Right sides) DRGs (N = 5 mice per condition) were used. (B–D) Heatmaps for differentially expressed genes (DEGs) are represented for the following comparisons: (B) Female Shmt1^−/− mice v/s Shmt1^+/+mice, both on AIN93G diet; (C) Female Shmt1^−/− mice on the AIN93G+FA diet v/s the AIN93G diet; (D) Male Shmt1^−/− mice v/s Shmt1^+/+ mice, both on AIN93G diet. (E and F) Pathway enrichment analysis of DEGs for the three respective comparisons using GO terms. Biological processes (BP) downregulated in Shmt1^−/− female mice (E) and upregulated BP in Shmt1^−/− male mice (F) on AIN93G diet are shown. (G) Sex-specific heatmap representing unique and overlapping genes in Shmt1^−/− female and male mice on AIN93G diet compared to control group. All represented genes for heatmaps were chosen based on their expression in neuronal and Schwann cells observed in neurodegenerative diseases. |Log₂fold change| cutoff was set at 0.2 and p_cutoff <0.01 for female Shmt1^−/− mice v/s Shmt1^+/+ mice on AIN93G diet comparison. A p_cutoff<0.05 was used when comparing Shmt1^−/− mice on the AIN93G + FA to Shmt1^−/− mice on the AIN93G diet, and when comparing male Shmt1^−/− mice to Shmt1^+/+ mice on AIN93G diet.

Three groups were assigned for analysis to test the effect of Shmt1 loss-of-function, sex and diet on mRNA expression in the DRGs: 1) Shmt1^−/− female mice vs. Shmt1^+/+ female mice fed the control diet 2) Shmt1^−/− female mice fed the excess FA diet vs. Shmt1^−/− female mice fed the control diet; and 3) Shmt1^−/− male mice vs. Shmt1^+/+ male mice fed the control diet. To identify genes sensitive to Shmt1 expression, DRGs from Shmt1^−/− and Shmt1^+/+ female mice all fed the control diet were compared. 218 genes were differentially expressed (52 upregulated, 166 downregulated; |log₂F_c|> 0.2 and p_cutoff <0.01) in the Shmt1^−/− female mice fed the control diet. Key downregulated transcripts included genes involved in Schwann cell function and myelin maintenance (Mbp, Mag, Egr2, Mal, Prx, Pmp22, and Bcas1) (Fig. 3B and SI Appendix, Table S1). These changes are consistent with the reduced MAG protein levels observed in the sciatic nerve (Fig. 1D), implicating altered DRG Schwann cell transcription impacting motor NCV. Further, in the Shmt1^−/− DRG sensory neurons, downregulation of multiple members of solute carriers (SLCs, including Slc6a1, Slc6a20a/b, Slc9a2, Slc26a2, Slc26a7, Slc22a8, and Slco1c1) was observed (SI Appendix, Fig. S4). Reduced expression of Slc6a1 (GABA transporter-1) and Slc6a20a/b (proline transporters) impair inhibitory neurotransmission and amino acid homeostasis (28, 29). In summary, these findings indicate functional convergence of excitability dysregulation, metabolic vulnerability, and impaired structural support, synergistically contributing to the pathogenesis of PN in Shmt1^−/− mice (SI Appendix, Fig. S4 and Table S3).

To identify diet-sensitive genes, DRGs from Shmt1^−/− female mice fed either the control diet or the excess folic acid diet were compared. Excess FA supplementation altered expression of 158 genes (113 downregulated, 45 upregulated; |log₂F_C| > 0.2, p_cutoff < 0.01). Few neurologic/solute career genes met this stringent threshold. Using a relaxed cutoff (P < 0.05), upregulated gene clusters comprising ion transporters, nociceptors, and neuronal regulators with a potential role in peripheral nerve function and repair were identified in Shmt1^−/− mice DRGs fed the excess FA diet (Fig. 3C and SI Appendix, Fig. S4 and Table S4). Key genes that promote neurotransmission (Slc6a11), myelination (Ncmap), neuronal differentiation (Neurod1), signaling receptors (Ptgfr, P2ry10) and metabolite transport (Slc22a1, Slco1a6) were upregulated, collectively indicating a coordinated enhancement of neuronal integrity and peripheral nerve maintenance (SI Appendix, Fig. S4). Further, elevated expression of solute carriers (Slc9a2/3/9, Slco1c1/5, Slco22a1), suggested altered ionic homeostasis, zinc uptake, and hormones/drug transport, all which influence neuronal resilience under neuropathic stress (30). Upregulation of voltage-gated sodium and calcium channel activities (Cacna1e, Scn1b, Scn4b, and Tmem37) and Slc6a11 in the excess FA group indicates changes in GABAnergic signaling favoring nociception (SI Appendix, Fig. S4). These groups of genes tightly regulate sensory neuronal excitability and pain sensitivity with a potential role in nerve regeneration (31–33). Together, these data provide insights into transcriptional responses in the DRG sensory neurons that integrate excitability control with regenerative support in the context of PN, influenced by dietary folic acid.

To identify genes sensitive to sex, DRGs from Shmt1^−/− male mice were compared to Shmt1^+/+ male mice all fed the control diet. 652 DEGs (422 upregulated, 229 downregulated; |log₂FC| > 0.2, P < 0.05) were identified in Shmt1^−/− male mice compared to the wild type. Few Schwann cell-related signatures were present. The predominant DEGs upregulated in the Shmt1^−/− male mice clustered within the neuropeptide and neurohormone category, such as, Npy, Nts, Neurog1, Nptxr, and Nlgn2 (Fig. 3D and SI Appendix, Tables S5 and S6). These patterns suggest a potential compensatory neuroprotective signaling that may mitigate PN severity in males, as reflected in the motor NCV. Ferroptosis-regulator genes, including Alox12b, Steap3, and Ptgir were also upregulated in males (Fig. 3D and SI Appendix, Table S7), consistent with findings in females, though the exact role of ferroptosis in PN progression remains to be clarified.

Multi-DEG analyses were performed to evaluate genotype by sex comparisons for Shmt1^−/− v/s Shmt1^+/+ male and female mice all fed the control diet. Both overlapping and distinct PN-related transcriptional changes were observed (Fig. 3G). DEGs exhibiting common upregulation in both female and male Shmt1^−/− mice included Alox12b, Steap3, Kcnn3, and Kcne4 (SI Appendix, Tables S2 and S7), whereas Mpz, Lox, and Pmp2 were downregulated in both sexes. Further, female-specific downregulation of Mbp, Mag, and Slc7a11 and male-specific upregulation of Npy, Nts, and Lrrc15 in Shmt1^−/−mice indicated sex-specific differences in PN progression.

Gene Ontology (GO) analyses of DEGs sensitive to both Shmt1 genotype and sex were performed. The most downregulated biological processes (BP) observed in female Shmt1^−/−mice compared to female Shmt1^+/+ mice were myelination, wound healing, and magnesium ion transport (Fig. 3E). When comparing male Shmt1^−/−mice, upregulated BPs included neuropeptide signaling, axon regeneration, and negative regulation of apoptosis (Fig. 3F). These findings further suggest that Shmt1^−/−female mice exhibit structural myelin-related vulnerabilities, whereas males have elevated expression of adaptive neuroprotective signaling pathways.

Dietary Folic Acid Prevents Peripheral Neuropathy in Diabetic (Lepr^db/db) Mice.

In concurrence with the literature, Lepr^db/db male and female mice fed the control diet exhibited significant reduction in motor NCV at 6 mo age (34). NCV was reduced by ~23% in both male (p_adj = 0.013) and female (p_adj = 0.0007) Lepr^db/db mice compared to Lepr^+/+ mice (Fig. 4 A and B). Lepr^db/db mice fed the excess FA diet exhibited significantly restored NCV in both males (137.7% rescue; p_adj = 0.003) and females (126.3% rescue; p_adj = 0.003) compared to Lepr^+/+ mice. Lepr^db/db mice fed AIN93G containing 0.1% deoxyuridine exhibited similar reductions in NCV as observed for Lepr^db/db mice fed the AIN93G diet (Fig. 4B), indicating a lack of dietary rescue by deoxyuridine.

Fig. 4.

The effect of folic acid on diabetic peripheral neuropathy characterized by measuring motor NCV and semiquantitative immunoblots of the sciatic nerves from wild-type and diabetic Lepr^db/db mice. (A) Work flow showing wild-type (Lepr^+/+) mice weaned to the AIN93G diet as the control group, and diabetic (Lepr^db/db) mice randomly weaned onto either the control diet (AIN93G, 2 mg/kg folic acid) or excess FA diet (AIN93G, 8 mg/kg folic acid) or a dU supplemented diet (AIN93G, 2 mg/kg folic acid, 0.1% deoxyuridine) for 6 mo followed by electrophysical and biochemical tests. Both female and male mice were used for the study. (B) Sciatic motor NCV comparison between female (Left) vs. male (Right) mice on different diets are shown. The Y-axis represents relative NCV (m/s) calculated for proximal and distal stimulations as described previously. Each data point represents mean NCV for one mouse, calculated from both left and right limbs using the formula: $\mathrm{NCV} \left(\frac{\mathrm{m}}{\mathrm{s}}\right)=\frac{\mathrm{Distance} \mathrm{between} \mathrm{proximal} \mathrm{and} \mathrm{distal} \mathrm{stimulation}}{\mathrm{Latency} \mathrm{of} \mathrm{response} \mathrm{for} \mathrm{proximal} \mathrm{stimulus}-\mathrm{Latency} \mathrm{of} \mathrm{response} \mathrm{for} \mathrm{distal} \mathrm{stimulus}}$. Every proximal and distal response was recorded for three independent stimulations (24). The X-axis shows diet/genotype groups. Data indicate mean NCV ± SEM (n = 14 to 16 mice/group). (C) Electrophysiology profiles consisting of maximal amplitude (mV) of the response curve (compound motor action potential, CMAP) and duration of the response (ms) are shown. (D) Relative quantification of the immunoblots were done on total proteins extracted from the sciatic nerves harvested from male and female Lepr^+/+ fed the AIN93G diet and Lepr^db/db mice fed either AIN93G or the AIN93G + FA diets for 6 mo. Nerve lysate proteins were quantified and probed with antibodies against periaxin, NF155, MAG, MBP, and normalized against β3 tubulin, as described previously. Left panel D shows the immunoblots whereas densitometric analysis of the immunoblots using ImageJ are represented in panel D, Right. For the densitometric plots, data are shown for mean relative expression of corresponding proteins ± SEM for N = 3 biological replicates.

The impact of the excess FA diet on other measures of nerve function was investigated in Lepr^db/db mice. Distal CMAP amplitude was reduced in both female (58% reduction; p_adj < 0.00001) and male (29.6% reduction; p_adj = 0.06) Lepr^db/db mice compared to Lepr^+/+ mice fed the control diet (Fig. 4C). No significant effects of diet or genotype were observed in the proximal CMAP for male mice. However, proximal CMAP amplitude was reduced in female Lepr^db/db mice fed with control diet compared to wild-type mice fed the control diet (49.8% reduction, p_adj < 0.00001). These findings confirmed impaired axonal conduction in mice affected by diabetic PN (35). Excess dietary FA did not restore CMAP amplitude (distal or proximal) for either sex (Fig. 4 C, Top and Bottom).

Duration of the response time was reduced in both male (23.7% reduction; p_adj < 0.0001) and female (37.4% reduction; p_adj < 0.00001) Lepr^db/db mice compared to Lepr^+/+ mice on the control diet. This indicated a reduction in the number of conducting nerve fibers in the sciatic nerve of diabetic mice. Excess FA diet partially rescued the duration of response time in Lepr^db/db female mice (22.6%; p_adj = 0.028) compared to Lepr^db/dbfed with control diet, but not in male mice (Fig. 4C). Deoxyuridine supplementation had no impact on CMAP or response duration in either sex.

Excess FA Restored Myelin Basic Protein in Lepr^db/db Mice.

Sciatic nerve lysates exhibited significantly reduced MBP protein levels in Lepr^db/dbmale mice (45.1%; P = 0.006) and a similar trend was observed in females (23%; P = 0.2) compared to Lepr^+/+ mice, all fed the control diet for 6 mo (Fig. 4D). Feeding the excess FA diet for 6 mo restored MBP expression in both male (28.5%; P = 0.003) and female (259.1%; P = 0.05) Lepr^db/db mice compared to Lepr^+/+ mice fed the control diet for 6 mo (Fig. 4 D, Top and Bottom).

MAG and periaxin protein levels remained unchanged in both sexes and genotypes. However, NF155, a nodal protein critical for ion channel organization, was selectively reduced in diabetic male mice (49.4%; P = 0.001), but unaffected in females compared to wild-type mice fed the control diet (Fig. 4 D, Top). This sex-specific nodal disruption likely arises from hyperglycemia and may contribute to the differential functional outcomes observed for PN in diabetic mice (36, 37). In summary, unlike the Shmt1^−/− mice, Lepr^db/dbmice developed DPN associated with reduced MBP and NF155 expression. Dietary folic acid prevented these changes, suggesting its protective effects extend across mechanistically distinct forms of PN.

Dietary Uridine Exacerbates NCV in Wild-Type Mice Independent of Folate Status.

The effect of dietary supplementation with uridine for 6 mo on PN was investigated in wild-type male and female mice (Fig. 5A). Both male and female mice fed the uridine-containing diet exhibited reduction of motor NCV in sciatic nerve by ~15% (P = 0.03) and ~21% (P = 0.003), respectively, compared to wild-type mice fed the control diet (Fig. 5B). Both distal and proximal CMAP were increased by 34.5% (P = 0.005) and 23.4% (P = 0.037), respectively, in the female uridine diet groups, compared to control diet (Fig. 5B). However, no change was observed in CMAP for male mice (Fig. 5C). Duration of response time was reduced by 15.04% (P = 0.003) in female mice on the uridine diet compared to the control group, whereas there was no change for male mice (Fig. 5C). This indicated lack of conducting fibers in the sciatic nerves causing a demyelinating neuropathy in the female mice fed the uridine diet.

Fig. 5.

Assessment of motor NCV and sciatic nerve health in wild-type mice fed the uridine supplemented diet for 6 mo. (A) Workflow showing wild-type male and female mice randomly weaned to either the control diet (AIN93G containing 2 mg/kg FA) or the uridine diet (AIN93G containing 2 mg/kg folic acid and 0.6% uridine) for 6 mo followed by measurement of motor NCV on the sciatic nerve. (B) Sciatic motor NCV comparison between female (Left) vs. male (Right) mice on different diets. The Y-axis relative motor NCV (m/s). Each data point represents mean NCV for one mouse, calculated from both the left and right limbs using the same method and formula described previously. The X-axis represents the diet/genotype group. (C) Additional electrophysical parameters consisting of the compound motor action potential (CMAP, mV) and duration of the response curve (ms) are shown. (D) Semiquantitative immunoblot of sciatic nerve lysates and the corresponding densitometric analysis are shown. Data indicate relative expression ± SEM (n = 3 mice/group). Biological replicates (N = 3) from each group have been shown and labeled appropriately.

Semiquantitative immunoblot of the sciatic nerve revealed expression of MBP was reduced in female mice (42%; P = 0.03) fed the uridine diet compared to female mice fed the control diet, confirming the PN to be demyelinating in nature (Fig. 5D). In contrast, MBP expression showed a reduction of 24.6% (P = 0.1) in male mice fed the uridine diet compared to the control diet. Subsequently, MAG, NF155, and periaxin expression remained unchanged across sex and diet groups (Fig. 5D). Dose-dependent (0 to 100 µM) uridine exposure had no effect on embryonic fibroblast proliferation, irrespective of folate concentration in the growth medium (SI Appendix, Fig. S5). Together, these findings demonstrate that chronic uridine intake induces a sex-biased, demyelinating form of PN in wild-type mice, highlighting uridine as a potential dietary risk for nerve dysfunction.

Discussion

This study reveals a shared etiology between folic acid–responsive NTDs and PN, which are both linked to reduced Shmt1 expression. Shmt1 heterozygosity is sufficient to induce both PN and NTDs, highlighting the sensitivity of these conditions to subtle metabolic perturbations in nucleotide metabolism. Furthermore, both pathologies can be prevented with elevated intakes of FA, and both are congenital in nature, as PN manifested early at 6 wk of age. Dietary uridine induced PN in wild-type mice in this study and has previously been shown to induce NTDs in wild-type mice independent of folate status (19), indicating another shared connection between NTDs and PN in mice. While this study indicates a shared genetic and nutritional etiology for PN and NTDs in the Shmt1 mouse model, it is not clear whether they share common pathophysiological mechanisms.

There are also differences in the expression of the two pathological phenotypes. The Shmt1^−/− mouse model exhibited sexual dimorphism in PN severity with female Shmt1^−/− mice exhibiting more severe neuropathic deficits. Sexual dimorphism has not been observed for the NTDs phenotype (12, 17, 38). Furthermore, while the NTD phenotype exhibits low penetrance (17), the PN phenotype exhibits full penetrance.

This study provides insights into the pathways and mechanisms underlying PN in the Shmt1 mouse model and its prevention by excess dietary FA. Reductions in NCV were not associated with gross anatomical abnormalities in the sciatic nerve (SI Appendix, Fig. S1) and with early presentation (Fig. 2 B and C) indicating functional deficits. To examine function, the expression of three candidate proteins associated with nerve function and health: Myelin-associated glycoprotein (MAG), myelin basic protein (MBP), and neurofascin 155 (NF155) were investigated. Shmt1^−/− mice exhibited lower expression of MAG in the sciatic nerve (Fig. 1D) and lower Krox20 mRNA levels (Fig. 1E), which is a transcription factor that regulates MAG and MBP (26) expression, and Schwann cell differentiation. MAG is essential for myelin stability and tethering to axons (39–41). MAG has been clinically associated with a rare human PN in a number of studies (42, 43). Mag^−/− mice exhibit characteristics of unmyelinated axons, highlighting the importance of MAG in maintenance of myelin structure (44–46). The impaired expression of MAG in Shmt1 deficient mice is consistent with transcriptional dysregulation in Schwann cells and compromised myelin–axon interactions and consequently reduced motor NCV. Expression of NF155, which is required for proper organization of ion channel clusters among neurons (37), was not affected in Shmt1^−/− mice, consistent with proper functioning of neuronal junctions in the sciatic nerve and normal motor action potential (CMAP) in Shmt1^−/− mice (Fig. 1 C and D). These findings indicate that the neuropathological changes in Shmt1-deficient mice are likely due to defects in myelin organization and axonal support. Furthermore, no changes were observed among inflammatory cytokines in the plasma when comparing Shmt1^+/+ vs. Shmt1^−/− female mice on the control diet, indicating that the PN was not due to systemic inflammation (SI Appendix, Fig. S6). Exposure to excess FA was associated with improved NCV in Shmt1^−/− female mice, without affecting MAG, MBP, or NF155 levels in the sciatic nerve. Therefore, the neuroprotective effect of FA in this context may involve mechanisms beyond the regulation of Schwann cell proliferation or the expression of myelin proteins. In contrast, MBP expression was reduced in the sciatic nerve of wild-type female mice fed the uridine-supplemented diet (Fig. 5D). Homeostatic disorders of uridine metabolism and its carcinogenicity have been linked to p53 activation, DNA damage, and apoptosis (47, 48).

Signatures of altered nociception and pain sensitivity were observed in the RNA profiles of DRGs in human and mice models of sciatic nerve injury (49–53), prompting studies of the Shmt1^−/− transcriptome. Functional enrichment showed the myelination pathway was downregulated in the female Shmt1^−/−mice L3-L5 DRG transcriptome compared to Shmt1^+/+female mice when both were fed the control diet (Fig. 3E). In contrast, the DEGs in Shmt1^−/− male mice revealed upregulation of neuropeptide Npy signaling and inhibition of neuronal apoptosis pathways (Fig. 3F). A systematic review on DRG transcriptomics of rodent models of peripheral nerve injury revealed neuropeptide (Npy) signature to be consistently upregulated in multiple studies, mitigating the neuropathic burden (54). Hence, the sexual dimorphism observed for the PN phenotype is also consistent with pathways involved in sensory neuronal functions in DRGs, providing insights into sex-specific mechanisms underlying the PN phenotype in this mouse model. Increased expression of DEGs (Alox12b, Kcnn3, Steap3, Lox) that regulate the ferroptosis pathway was common to both male and female Shmt1^−/−mice compared to Shmt1^+/+mice when both fed the control diet (Fig. 3G). This implicates a mechanistic link between Shmt1 impairment and ferroptosis cell death in the DRG sensory neurons. Ferroptosis and its connection to PN has been established recently in the literature (55, 56).

Analysis of DEGs responsive to excess FA consumption did not reveal any relevant insights into its neuroprotective effects of FA related to myelination pathways. However, with a relaxed p-value cutoff, excess FA consumption upregulated certain solute carriers, neurotransmitters, and hormone/drug transporter activities that were otherwise downregulated in Shmt1^−/− female mice fed the control diet (SI Appendix, Fig. S5). These restorations indicate the presence of enhanced sensory neuronal regenerative support and excitability control that was provided by folic acid. Changes in the transcriptome could be mediated directly by the folate receptor (57), changes in cellular methylation, or other mechanisms.

The PN phenotype in Lepr^db/db mice was distinct from Shmt1^−/− mice, although the PN phenotypes in both mouse models were improved by excess folic acid consumption. In female Lepr^db/dbmice fed the excess FA diet, improved NCV as well as duration of response time, a measure that reflects total nerve fibers recruited for electrical signal conduction, was observed. Elevated MBP expression was also observed in Lepr^db/db mice fed the excess FA diet (Fig. 4 D, Top and Bottom), indicating that folic acid promotes proliferation of Schwann cells in the sciatic nerve. However, unlike in Shmt1^−/− mice, MAG expression levels were unaffected in Lepr^db/dbmice. Previous studies have established DPN etiology in the Lepr^db/dbmouse model is associated with glucose-mediated oxidative stress and neuronal apoptosis (22).

In a pilot study, Shmt1 heterozygosity was sufficient to induce PN in mice. However, no interactions were observed for PN between Shmt1 and Lepr genotypes upon crossing Shmt1^+/−and Lepr^db/db mice (SI Appendix, Fig. S7 A–D). Shmt1 heterozygosity did not improve or exacerbate deficits seen in motor NCV in Lepr^db/db mice both fed the AIN93G diet. Mice fed the AIN93G diet lacking both B12 and folic acid also did not exacerbate the PN in Shmt1^+/− or Lepr^db/db mice, indicating neither folic acid nor vitamin B12 dietary deficiency exacerbated the PN phenotype in these models. Lepr^db/db genotype and diet groups had significantly lower concentrations of plasma homocysteine (P < 0.0001), methionine (P = 0.0075), serine (P < 0.0001), and glycine (P = 0.0006) compared to Lepr^+/+ mice on the control diet (SI Appendix, Table S8). The changes in these folate-related biomarkers indicate that Lepr deficiency alters folate mediated one carbon metabolism.

These studies provide additional evidence for the concept of “special nutritional requirements” (58, 59), where changes in a physiological state, induced by genetics, chronic disease (e.g., diabetes), environmental exposures, or other causes, modify nutritional requirements. Not meeting these new nutritional requirements result in the onset of comorbidities (e.g., PN in diabetes). Two randomized placebo controlled trials showed minor improvement in NCV following administration of 1 mg/d FA for 16 wk in patients with diabetic polyneuropathy in Iran (60, 61), but dose–response data were lacking. It is not known whether administration of reduced folates, including 5-methyltetrahydrofolate or folinic acid, would be efficacious. The prevention of PN with excess folic acid in female Shmt1^−/− mice as well as Lepr^db mice offers the potential for a simple dietary prevention of PN, including DPN.

Materials and Methods

Detailed methods are available in SI Appendix.

Mouse Models.

The generation of Shmt1^+/− mice on C57BL/6J background has been described previously (62). Mice were maintained as heterozygous breeding colonies on C57BL/6J background. Wild-type (Shmt1^+/+) and homozygous knockout (Shmt1^−/−) mice were generated by crossing Shmt1^+/− breeding pairs.

Diabetic heterozygous (Lepr^db/+) mice on C57BL6/J background were purchased from Jackson laboratories (strain # 000697, B6.BKS(D)-Lepr^db/J). Wild-type (Lepr^+/+) and diabetic homozygous (Lepr^db/db) mice were generated by crossing Lepr^db/+ pairs.

All pups were genotyped 2 wk after birth and housed in pathogen-free, temperature and humidity-controlled environment with access to ad libitum chow or specialized diet and water. Lepr^db and Shmt1 genotyping was performed using the primers listed in SI Appendix, Table S9, recommended by the Jackson Laboratories and as previously described (16, 17). Mice colonies were housed and genotyped at the Texas A&M Institute of Genomic Medicine. All animal experiments were approved by the Texas A&M University Institutional Animal Care and Use Committee (IACUC) and conducted in accordance with the Animal Welfare Act and applicable federal and state regulations.

Diets.

To investigate whether loss of Shmt1 expression caused diet-dependent PN, 19 to 21 d old male and female Shmt1^+/+and Shmt1^−/− mice were weaned onto the AIN93G diet (Dyets, Inc, Bethlehem, PA) containing 2 mg/kg folic acid (control diet). To assess the dietary effects on PN, Shmt1^−/− male and female mice were weaned onto either the AIN93G diet supplemented with 8 mg/kg FA (excess FA diet; AIN93G+FA) or the AIN93G diet containing 2 mg/kg folic acid and 0.1% 2’-deoxyuridine (dU diet; AIN93G+dU). All mice remained on respective diets for 6 mo prior to NCV tests and biochemical assays.

In parallel, male and female wild-type and Lepr^db/db mice were weaned onto the AIN93G diet to examine the effect of Lepr genotype on DPN. To assess the effect of diet on DPN, Lepr^db/db male and female mice were weaned onto either the excess FA diet (AIN93G + FA) or the dU diet (AIN93G+dU). Mice were maintained on respective diets for 6 mo prior to electrophysical and biochemical tests.

In a separate study, Shmt1^+/+ and Shmt1^−/− female and male mice were weaned onto the AIN93G diet for 3 wk to determine whether the PN was congenital. Finally, to evaluate the effect of dietary uridine on PN, wild-type male and female mice were weaned onto either the AIN93G diet or the AIN93G diet supplemented with 0.6% uridine (U diet) for 6 mo.

Nerve Conduction Studies.

Nerve conduction velocity (NCV) was measured on the sciatic nerve of mice in the motor direction following protocols established by Schulz et al. (24, 63). Anesthesia in mice was induced with 2% isoflurane and 98% oxygen for 2 to 3 min. Complete anesthesia was confirmed by absence of withdrawal reflex upon gently pinching the hind paws for reaction. Anesthesia was maintained throughout the procedure using 1.5% isoflurane and 98.5% oxygen. For the Lepr^db/db mice, anesthesia was induced with 3% isoflurane and 97% oxygen, followed by 2% isoflurane and 98% oxygen for maintenance. Throughout the procedure, mice body temperature was maintained at 32 to 34 °C with heating pads and commercial eye-cream was applied on the pupils for lubrication. The hind limbs were shaved and depilated with commercial hair-removal cream. A sensing ring electrode was placed over the thickest part of the gastrocnemius muscle on the dorsal thigh followed by the reference/grounding electrode positioned ~10 mm apart. Stimulation sites were located 4.5 mm (distal stimulation) and 15 mm (proximal stimulation), respectively, from the sensing electrode. Single square-wave pulses (0.1 ms duration, 10 Hz frequency) were delivered to the sciatic nerve, using monopolar disposable 28G needle electrodes. Prior to actual data collection, all stimulation parameters were optimized for maximal amplitude (mV) of the compound motor action potential (CMAP) and peak latency (ms) for the corresponding neuromuscular functional response curve. The response curve was recorded for at least three independent stimulations (distal and proximal) per mouse from the left and right limbs. Relative NCV (m/s) was calculated as the distance between the proximal and distal stimulations divided by the difference between the corresponding peak latencies.

Additional electrophysical parameters, such as the amplitude of CMAP (mV) for the distal and proximal stimulation sites, and duration time (msec) of each response curve were also noted. The CMAP amplitude reflects axonal conduction capacity, whereas the duration of response is functionally dependent on the total number of nerve fibers recruited for the electrical conduction (24). Together, these measures provide functional assessment of sciatic nerve conduction and structural integrity.

Quantitative Immunoblot.

Sciatic nerves were collected from Shmt1^+/+ and Shmt1^−/−male and female mice (n = 3 to 4) on respective diet groups. Following homogenization, total protein was extracted using RIPA buffer (Pierce) and quantified by Lowry-Bensadoun assay. Equal amounts of denatured proteins were separated on 4 to 15% SDS-PAGE gels (Biorad) at 100 V, 400 mA, and transferred onto a 0.22 µm Immunblot PVDF membrane (Biorad) for 2 h (65 V, 400 mA) using Mini TransBlot apparatus (Biorad) under ice-cold conditions. The membranes were blocked with blotting grade casein (Biorad) and incubated at 4 °C overnight with primary antibodies against myelin basic protein, MBP (Cell Signaling D8X4Q, 1:1,000), myelin associated glycoprotein, MAG (Cell Signaling D4G3, 1:15,000), neurofascin-155, NF155 (Cell Signaling D7B6O, 1:400), periaxin (SC-515672, 1:250), and loading control β3-tubulin (Cell Signaling D71G9, 1:15,000). Protein bands were visualized using Clarity ECL substrate (Bio-Rad) and imaged with a FluorChem E system (Protein Simple). Densitometric analysis was performed with IMageJ (version 1.53a).

RNA Sequencing of L3-L5 DRG Neurons.

Shmt1^+/+ and Shmt1^−/−male and female mice (n = 5, per group) fed either the control or excess FA diet for 24 wk were sacrificed using CO₂. Lumbar (L3-L5) DRGs corresponding to the sciatic nerve were rapidly dissected from left and right sides of the spinal cord on ice, rinsed in ice-cold PBS and directly placed into 500 μL RNAlater, flash frozen, and stored at −80°C.

The DRGs were homogenized using a Mililys personal homogenizer (Bertin Instruments) with three 1-min cycles at 4 °C. Total RNA was extracted using QIAzol and purified with the Rneasy Mini kit (Qiagen Inc.) following the manufacturer’s protocol. RNA integrity and concentration were measured using a Qubit fluorimeter prior to library preparation. Messenger RNA was isolated from 150 ng total RNA using a Nextflex Poly-A Selection kit (Perkin Elmer, Waltham, MA, USA). cDNA libraries were prepared using a Nextflex Rapid Directional RNA 2.0 kit, miniaturized to ²/₅ reaction volume with 12 PCR cycles and automated on a Sciclone NGSx liquid handler. A total of 25 libraries were pooled in equimolar amounts and sequenced on Illumina NovaSeq SP flowcell (2 × 100 bp), yielding 940 million raw clusters, with an average of 37.6 million clusters per sample.

Raw sequencing data were processed using the SUSHI framework (64) from Functional Genomics Center Zurich (FGCZ). Adapter trimming and quality filtering were performed with fastp v0.20 (65). Trimmed reads were pseudoaligned to the GRCm39 mouse reference genome assembly (GENCODE release M31) and gene expression levels were quantified using Kallisto v0.46.1 (66). Differential gene expression (DEG) analysis was performed using the R package edgeR v3.42.2 (67) to compare different conditions. Genes showing |log₂FC| > 0.2 (P < 0.01) were considered differentially expressed. Functional enrichment analysis of the up- and down-regulated genes was performed using the R package clusterProfiler v4.8.1 using the GO biological processes (BP) terms (68). Differential expression results were visualized using the Shiny web application exploreDE (69). In silico cell type deconvolution of the bulk RNAseq samples was performed against the published reference single-cell dataset (27). For each sample, cell type abundance was computed by using CIBERSORTx (70).

Reverse Transcription qPCR on Sciatic Nerve mRNA.

Sciatic nerves from Shmt1^+/+ and Shmt1^−/− male and female mice fed the control diet for 6 mo were dissected and flash frozen in 500 µL RNAlater (Sigma). Total RNA was extracted and purified using the RNeasy kit (Qiagen) according to the manufacturer’s instructions. Complementary DNA was synthesized from RNA samples with the SuperScript IV VILO Master Mix (Invitrogen). Gene expression of Sox10, Sox2, Krox20, Pou3f1, Mbp, and Mpz was analyzed by qPCR using the QuantStudio 5 Real-Time PCR Detection System and PowerUp™ SYBR™ Green Master Mix (Thermo Fisher Scientific, Darmstadt, Germany). Expression levels were normalized to 18S. Primer sequences are listed in SI Appendix, Table S10.

Statistical Analysis.

Comparisons of NCV, CMAP, and response duration between diet and genotype groups were performed using unpaired Student’s t tests (JMP Pro 17). Protein quantification from immunoblotting and RT-qPCR data were compared using Welch’s t tests. Differences between groups were considered significant when P < 0.05. All graphical values are shown as mean parameter values with SE. A Bonferroni correction was applied for multiple comparisons where appropriate.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The RNA-seq data reported in this article have been deposited in NCBI’s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE275111 (71).