Vitamin C epigenetically controls osteogenesis and bone mineralization

Abstract

Vitamin C deficiency disrupts the integrity of connective tissues including bone. For decades this function has been primarily attributed to Vitamin C as a cofactor for collagen maturation. Here, we demonstrate that Vitamin C epigenetically orchestrates osteogenic differentiation and function by modulating chromatin accessibility and priming transcriptional activity. Vitamin C regulates histone demethylation (H3K9me3 and H3K27me3) and promotes TET-mediated 5hmC DNA hydroxymethylation at promoters, enhancers and super-enhancers near bone-specific genes. This epigenetic circuit licenses osteoblastogenesis by permitting the expression of all major pro-osteogenic genes. Osteogenic cell differentiation is strictly and continuously dependent on Vitamin C, whereas Vitamin C is dispensable for adipogenesis. Importantly, deletion of 5hmC-writers, Tet1 and Tet2, in Vitamin C-sufficient murine bone causes severe skeletal defects which mimic bone phenotypes of Vitamin C-insufficient Gulo knockout mice, a model of Vitamin C deficiency and scurvy. Thus, Vitamin C's epigenetic functions are central to osteoblastogenesis and bone formation and may be leveraged to prevent common bone-degenerating conditions.

Fig. 1. Vitamin C deficiency causes vast transcriptional changes in bone.

a Experimental setup; wk, week. b RNA-seq analysis of femoral bone using principal component analysis and hierarchical clustering. c Venn diagrams depicting differentially expressed genes (p-adj < 0.05). d Gene expression heatmap displaying differentially expressed genes between VitC-positive and VitC-negative mice (p-adj < 0.05). e–h Volcano plot and GSEA analysis of the Bone60 gene-set in VitC ± versus VitC+ RNA-seq data, NES normalized enrichment score, FDR false discovery rate. g Heatmap and h gene expression bar charts depict leading-edge genes. Bar graphs represent mean ± SD. *p < 0.05; **p < 0.01; FDR adjusted Wald test (e), FDR-adjusted two-tailed, unpaired t tests; N = 3 per group from biologically independent animals (h). Source data are provided as a Source Data File.

Fig. 2. Histone and DNA demethylation signatures in bone are highly sensitive to Vitamin C.

a H3K9me3 and H3K27me3 western blots in indicated tissues from 20 week old Gulo^–/– mice supplemented with VitC (VitC+) or without VitC from WK 15 to WK 20 (VitC ±). b Relative western blot quantitation to (a) shown as fold change to VitC+. c 5hmC dot blot in indicated tissues from same experimental setup as in a and relative quantitation (d). e Immunofluorescence of 5hmC in calvarial sections; Gulo^−/− mice were crossed with transgenic Dmp1-EGFP_Topaz mice to visualize osteocytes; DIC, differential interference contrast, scale bar represents 50 µm. Bar graph represent mean ± SD; *p < 0.05; **p < 0.01; one-way ANOVA with Dunnett’s multiple comparison tests (d); N = 5 per tissue and group from biologically independent animals (b, d). Source data are provided as a Source Data File.

Fig. 3. Vitamin C-sensitive DNA hydroxymethylation levels distal and proximal to bone-specific genes correlate with their expression.

a hMeDIP-seq analysis of femoral bone showing average 5hmC peak number around transcriptional start sites (TSS), average 5hmC signal around TSS or in gene bodies for all genes or for the Bone60 gene-set; TES, transcriptional end site. b Overlaid 5hmC peak occupancy comparing VitC+ and VitC ± groups near Bone60 genes in femurs; [value]=max peak scale. c Correlation between gene expression and 5hmC occupancy (TSS + /−30 kb) in VitC +/− vs VitC+ femurs. LE, leading edge as per GSEA, FC, fold change. d Super-enhancer (SE) based clustering of 5hmC peaks with Runx2 as the highest ranking Bone60 gene and number of Bone60 genes associated with 5hmC-SE and 5hmC-TE as well as example of a 5hmC-SE (e); TE, typical enhancers; NLE, non-leading edge. f Ranked decrease in 5hmC signal at 5hmC-SE after VitC withdrawal. Two-sided Fishers exact test (c). Source data are provided as a Source Data File.

Fig. 4. Vitamin C governs all differentiation stages of the osteogenic lineage.

a Colony formation unit assay (CFU) of BMSCs isolated from VitC treated (VitC+) or VitC depleted (VitC ±) Gulo^−/− mice from week 15 to week 20. b Extracellular matrix (ECM) deposition and ECM mineralization in BMSCs isolated from VitC-deficient Gulo^−/− mice. c Quantitation of ECM deposition, alkaline phosphatase activity (ALPL) and ECM mineralization in VitC treated and untreated BMSCs. d Western blot for H3K9me3 and H3K27me3 and dot blot for 5hmC during osteogenic lineage progression. e 5hmC immunofluorescence and alizarin red stain in mature osteoblasts within mineralizing nodules (*). f Quantification of Dmp1 promoter activity (EGFP fluorescent signal) and ECM mineralization in cultures in which VitC was removed after D35 and representative images from experiments quantified in the left panel. (*) mineralized nodules; BMSCs from Gulo^−/− mice with Dmp1-EGFP_Topaz reporter were used for this experiment; RFU, relative fluorescence units. g mRNA expression of osteoblast and osteocyte markers in BMSCs with or without VitC, or cultures in which VitC was removed starting at D35 (VitC +/−). Scale bars in (e) and (f) represent 50 µm. Graphs represent mean ± SD, *p < 0.05; **p < 0.01; ***p < 0.001. Unpaired two-tailed t test in (a), paired two-tailed t test in (c), two-way ANOVA with Tukey’s (EGFP) and Sidak’s (Mineral) multiple comparison tests in (f), two-way ANOVA with Tukey’s multiple comparison tests in (g); significances in (g) represent D50 VitC+ vs D50 VitC ±. N = 3 (a, f mineral, g), n = 4 (f, EGFP), n = 5 (c, ECM) n = 6 (c, ALPL & Mineral) per group from cells derived from biologically independent animals. Source data as well as exact p = values for all comparisons in (g) are provided in the Source Data File.

Fig. 5. Vitamin C-dependent collagenous matrix formation is dispensable for Vitamin C-mediated osteoblastic gene expression.

a Gene expression of collagen hydroxylases and cross-linkers in BMCSs from 5-week VitC-deficient Gulo^–/– mice cultured in the presence or absence of VitC, or treated with VitC until day 35 after which VitC was removed. b Inhibition of VitC-dependent extracellular matrix (ECM) deposition by 1,4-DPCA. c Experimental setup using 1,4-DPCA during VitC-induced osteogenic differentiation. d–f Effects of 1,4-DPCA on ECM deposition (RFU, relative fluorescence units) (d), ECM mineralization (e) and osteogenic gene expression (f); VitC-untreated BMSCs at day 0 were used as an additional control. g Effects of 1,4-DPCA on ECM deposition and ECM mineralization in human BMSCs during osteogenic differentiation. h Osteogenic gene expression in hBMSCs treated with or without 1,4-DPCA. Graphs represent mean ± SD, *p < 0.05; **p < 0.01; ***p < 0.001 comparing VitC+ and VitC- groups at corresponding time points (a) test against VitC- D0 (f) and test against VitC-/DPCA- if not otherwise noted (g, h). #p < 0.05; ##p < 0.01; ###p < 0.001 comparing VitC+ groups with vs without 1,4-DPCA at D35 (f, h); in h for IBSP # p = 0.0447. Two-way ANOVA (a) and one-way ANOVA (d, f, g, h) with Tukey’s multiple comparison tests. N = 3 (a, d, f–h) per group from cells derived from biologically independent animals/donors. Source data are provided as a Source Data File.

Fig. 6. Vitamin C controls the activity of epigenetic αKGDDs during osteogenic differentiation.

a Inhibition of VitC dependent α-ketoglutarate dehydrogenases by L-2-hydroxyglutarate (L-2-HG) during osteogenic differentiation. b Inhibition of VitC dependent 5hmC synthesis and H3K9me3 as well as H3K27me3 demethylation by L-2-HG during osteogenic differentiation of primary mBMSCs and quantification of blots (c). d Effects of L-2-HG in differentiating calvarial MC3T3-E1 pre-osteoblasts on 5hmC deposition and H3K9me3 as well as H3K27me3 demethylation and quantification of the blots (e). Bar graphs represent mean ± SD, *p < 0.05; **p < 0.01; ***p < 0.001. One-way ANOVA with Tukey’s multiple comparison test for dot blot quantifications and two-way ANOVA analysis with Tukey’s multiple comparison tests for western blot quantifications in c, e. In (c), for H3K9me3 ***p = 0.0008, in (e), for H3K9me3 *p = 0.0155. N = 3 (c, e) per group from cells derived from biologically independent animals (c) or independent biological experiments (e). Source data are provided as a Source Data File.

Fig. 7. Loss of Histone demethylases and DNA hydroxymethylases, particularly TET2, impairs osteogenic differentiation.

a Gene expression of VitC-dependent epigenetic hydroxylases in MC3T3-E1 cells and mouse femur. b shRNA mediated hydroxylases knockdown efficiencies in MC3T3-E1 cells. c Ranked relative extracellular matrix (ECM) deposition, collagen cross links measured via FTIR, alkaline phosphatase activity (ALPL) and ECM mineralization after suppression of hydroxylases in MC3T3-E1. d Two-way ranked osteoblastic gene expressions after hydroxylases knockdowns in MC3T3-E1 cells; FC, fold change; EV, empty vector. e 5hmC dot blot after Tet knockdowns in MC3T3-E1 cells. Graphs represent mean ± SD, ***p < 0.0001. One-way ANOVA with Dunnett’s multiple comparison test (b), unpaired, two tailed t test (d). Data points in (b–d) were compared to EVCtrl. N = 2 (a, MC3T3-E1), n = 6 (a, bone), n = 3/tested shRNA (b), n = 3 (c) from biologically independent experiments (a, MC3T3-E1, b,c) or from biologically independent animals (a, bone). Source data are provided as a Source Data File.

Fig. 8. TET2 overexpression potentiates osteoblastic differentiation.

a TET2 western blot after CRISPR-dCas9-mediated TET2 o/e (overexpression) and b 5hmC dot blot under the same conditions. c ECM mineralization after TET2 o/e at day 21 of differentiation. d Osteoblastic gene expression during TET2 o/e. Bar graphs represent mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001. Unpaired two-tailed t test in (c); two-way ANOVA analysis with Tukey’s multiple comparison tests, groups compared to D0 Ctrl or as marked in d. N = 3 (c, d) per group, from biologically independent experiments. Source data as well as exact p = values for all comparisons in d are provided in the Source Data File.

Fig. 9. Vitamin C-induced 5hmC is a stable epigenetic mark that is required for osteogenic differentiation.

a Schematic representation of the active DNA demethylation cycle and experimental design for the shown experiments in differentiating osteoblasts. b Dot blots at day 14 with and without VitC and/or the DNMT-inhibitor RG108, which decreases global CpG methylation levels (5mC). c ECM deposition, alkaline phosphatase activity (ALPL) and ECM mineralization in the presence or absence of VitC and/or RG108. d Dot blots for 5hmC and 5fC (formyl-cytosine) in osteoblasts at indicated time points in the presence or absence of VitC. e Osteoblastic gene expression after VitC and/or RG108 administration. Bar graphs represent mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001. Two-way ANOVA analysis with Sidak’s multiple comparison tests (c and e), groups compared to day 0 Ctrl or as marked in e. In c, for ECM RG108- VitC- vs RG108 + VitC- **p = 0.0062 and RG108- VitC+ vs RG108 + VitC + **p = 0.0029. N = 3 (c, e) per group, from biologically independent experiments. Source data as well as exact p = values for all comparisons in (e) are provided in the Source Data File.

Fig. 10. Conditional Tet1 and Tet2 deletion impairs DNA hydroxymethylation at bone related loci and causes bone loss in mice.

a Schematic representation of the mouse models used. Bone tissue was collected at 11 weeks; cKO, conditional knockout. b Prrx1 driven TET1 and TET2 knockdown efficiencies in femurs of Tet1/2 double knockout mice. c 5hmC dot blot and relative quantification. d–f hMeDIP-Seq of femurs from Tet1/2 double knockout mice. d Principal component (PC) analysis and 5hmC peaks around selected bone specific genes (e). f Correlation between 5hmC occupancy (TSS + /−30 kb) and gene expression as measured by RNA-Seq; FC, fold change. g Expression of Bone60 genes in conditional Tet1, Tet2 or Tet1/2 femurs versus Ctrl femurs. h Wordcloud representing phenotypes associated with distal intergenic loci with the strongest 5hmC loss in double knockout vs Ctrl femurs as analyzed by GREAT analysis; FDR, false discovery rate. i µCT analysis of L5 spine and relative quantification; TV total volume, BV bone volume, Tb.N. trabecular number, Tb.Th. trabecular thickness, Tb.Sp. trabecular separation (j). Bar graphs represent mean ± SD; *p < 0.05; **p < 0.01; ***p < 0.001. One-way ANOVA analysis with Tukey’s (c) and Dunnett’s (j) multiple comparison tests. FDR-adjusted two-tailed, unpaired t tests in g, two-sided Fishers exact test (f). In c, Ctrl vs Tet1/2 cKO ***p = 0.0009, Tet1 cKO vs Tet1/2 cKO ***p = < 0.0001, Tet1 cKO vs Tet2 cKO **p = 0.005, Tet2 cKO vs Tet1/2 cKO *p = 0.0118. N = 5 (c) & n = 6 per group in j from biologically independent animals. Source data are provided as a Source Data File.