L-2-hydroxyglutarate impairs neuronal differentiation through epigenetic activation of MYC expression

Abstract

High levels of L- and D-2-hydroxyglutarate, the reduced forms of α-ketoglutarate (αKG), are implicated in human neurodevelopmental disorders and cancer. Both enantiomers exert effects on epigenetics by modulating a family of αKG-dependent dioxygenases involved in histone, DNA and RNA demethylation. L-2HG dehydrogenase (L2HGDH) converts L-2HG to αKG. Its deficiency is a rare, autosomal recessive inborn error of metabolism (IEM) characterized by systemic elevations of L-2HG, progressive neurological disability and a high risk of malignancy in the brain. The mechanisms behind these aberrations are unknown. Here we used an isogenic, patient-derived induced pluripotent stem cell (iPSC) system to study the impact of L2HGDH deficiency on neural progenitor cell (NPC) function and neuronal differentiation. We demonstrate that L2HGDH deficiency causes accumulation of L-2HG, NPC hyperproliferation, increased clonogenicity, excessive growth, and defective neuronal differentiation in 2D cultures and cortical spheroids. Editing the L2HGDH locus to wild-type reverses these effects. Blocking L-2HG accumulation in NPCs with a glutaminase inhibitor also induces neuronal differentiation. L-2HG-dependent inhibition of KDM5 histone demethylases leads to widespread retention of H3K4me2 and H3K4me3, markers of active gene expression. These marks are prominently elevated at the MYC locus in L2HGDH-deficient cells, and consequently cells express high MYC both in 2D culture and in many distinct cell types within cortical spheroids. Although thousands of loci display altered histone methylation, genetically or pharmacologically normalizing MYC is sufficient to completely reverse defective neuronal differentiation. These data indicate that the primary metabolic disturbance in an iPSC IEM model activates the MYC oncogene, favoring stem cell self-renewal and suppressing lineage commitment to neurons.

Figure 1.. An isogenic, patient-derived iPSC system to study L2HGDH deficiency.

(A) L2HGDH genotypes of two L-2HGA patients and their parents. (B) Relative 2HG abudnace in plasma from L-2HGA patients, parents, and unrelated subjects. (C) Volcano plot of plasma metabolites comparing Patient 1 to unrelated subjects (n = 427; sibling excluded). A linear mixed-effects model was used to account for repeated measures per individual, with subject ID as a random effect. P-values were adjusted for multiple testing (Benjamini-Hochberg), with significance thresholds of adjusted p < 0.01 and fold-change > 2 or < 0.5. (D) Volcano plot of fibroblast metabolites comparing Patient 1 to 29 unrelated lines (sibling excluded), each profiled in quadruplicate. A mixed-effects model was used as in (C), with cell line as the random effect. (E) L-2HG concentrations in neonatal control, adult control, and patient fibroblasts. (F) Alignment showing the pathogenic L2HGDH variant (c.829C>T, red), and edits (blue) introduced by the single-stranded oligonucleotide (ssODN) repair template. Electropherograms show unedited and biallelically corrected iPSC alleles. (G and H) L-2HG levels in PSCs (G) and NPCs (H) from control and patient lines with unedited or corrected L2HGDH alleles. For (B), (E), (G), and (H), data are shown as box-and-whisker plots with jittered points (n = 4 biological replicates for E–H); boxes represent the 25th–75th percentile, horizontal lines indicate medians, and whiskers extend to 1.5× the interquartile range. Statistical significance was assessed using one-way ANOVA with Tukey’s HSD test.

Figure 2.. L-2HG accumulation enhances NPC proliferation and self-renewal while inhibiting neuronal differentiation.

(A) Growth curves of NPCs derived from H9 control, Patient 1 unedited, and Patient 1 corrected iPSCs. (B) Stereoscopic images of cortical spheroids at day 30 of differentiation. Scale bar, 1000 μm. (C) Quantification of spheroid surface area at day 30. (D) Images of colonies formed from single NPCs (top: stereoscope; bottom: phase contrast). Scale bars, 2 mm and 100 μm. (E) Frequency of colony formation in single NPCs. (F) Immunostaining for MAP2 and βIII-Tubulin in Patient 1 unedited, corrected, and corrected + 30 μM octyl-L-2HG (o-L-2HG) neurons at day 14. DAPI marks nuclei. Scale bar, 100 μm. (G) Quantification of neurite length from images in (F) using the SNT plugin in ImageJ. (H) GSEA plot showing enrichment of the Reactome_Neuronal_System gene set in corrected versus unedited neurons. (I) Fura-2 ratio recordings of Ca²⁺ dynamics during 45 mM KCl application in neurons differentiated for 47 days (unedited: n = 29; corrected: n = 24). (J) Quantification of PAX6 signal intensity normalized to DAPI in unedited and corrected neurons. (K) UMAP of single-cell transcriptomes from day 45 cortical spheroids (unedited and corrected; 12,660 cells each). Cell type annotations include: astrocyte progenitor (Astro. prog.), cycling gliogenic radial glia (Cycl. glio. rad. glia), cycling intermediate progenitor (Cycl. inter. prog.), early gliogenic precursor (Early glio. prec.), gliogenic intermediate progenitor (Glio. inter. prog.), immature neuron (Immature n.), late neurogenic precursor (Late neuro. prec.), excitatory neuron (Excit. n.), interneuron, maturing inhibitory neuron (Maturing inhib. n.), neurogenic radial glia (Neuro. rad. glia), and oligodendrocyte progenitor cell (OPC). (L and M) Proportions of immature neurons (L) and excitatory neurons (M) among cell types shown in (K). For (A), (C), (E), (G), (J), (L), and (M), data are presented as mean ± 1 SEM. For (A), differences in cell growth were analyzed using two-way ANOVA (cell line × time), followed by Tukey’s HSD post hoc test. Statistical significance was assessed using Student’s t-test for (E) (n = 8), (J) (n = 679 unedited; n = 430 corrected), and for (L) and (M) (n = 3 each). One-way ANOVA followed by Tukey’s HSD test was used for (C) (n = 10 H9, n = 15 unedited, n = 10 corrected) and (G) (n = 96 unedited, n = 86 corrected, n = 108 corrected + o-L-2HG).

Figure 3.. Suppressing L-2HG synthesis improves neurite formation in L2HGDH-deficient NPCs.

(A) Schematic of ¹³C-labeling routes to L-2HG from [U-¹³C]glucose (blue) or [U-¹³C]glutamine (green) in NPCs. Abbreviations: Glc, glucose; Pyr, pyruvate; Ac-CoA, acetyl-CoA; OAA, oxaloacetate; Cit, citrate; αKG, α-ketoglutarate; Gln, glutamine; GLS, glutaminase; GLSi, glutaminase inhibitor (CB839); Glu, glutamate. (B) Ion counts of ¹³C-labeled 2HG isotopologues in NPCs derived from H9, control hiPSC, and Patients 1 and 2 after 4 h of labeling with [U-¹³C]glucose or [U-¹³C]glutamine. (C) Ion counts of ¹³C-labeled 2HG isotopologues in control hiPSC-derived and Patient 1 NPCs cultured in [U-¹³C]glutamine and treated with DMSO or 1 μM CB839 (GLSi) for 4 h. (D) Immunofluorescence for MAP2 and βIII-Tubulin in Patient 1 NPCs after 14 days of neuronal differentiation, treated with DMSO, 30 μM octyl-L-2HG (o-L-2HG), 1 μM CB839 (GLSi), or both. DAPI marks nuclei. Scale bar, 100 μm. (E) Quantification of total neurite length from cells in (D), measured with the SNT plugin in ImageJ. For (B) and (C), ion counts were normalized to total ion current (TIC). Data were non-normally distributed (Shapiro–Wilk test), so statistical significance was assessed by Kruskal–Wallis test with Dunn’s post hoc correction (n = 4 per group). For (E), one-way ANOVA with Tukey’s HSD test was used (n = 90 for DMSO + DMSO; n = 116 for DMSO + o-L-2HG; n = 89 for GLSi + DMSO; n = 101 for GLSi + o-L-2HG). Data shown as mean ± 1 SEM.

Figure 4.. Increased MYC expression in L2HGDH-deficient NPCs and cortical spheroids.

(A) Gene set enrichment analysis (GSEA) mountain plot showing enrichment of Dang_MYC_Targets_Up gene set in unedited versus corrected Patient 1 NPCs. (B) Quantification of MYC mRNA in control H9 NPCs, unedited Patient 1 NPCs, and corrected Patient 1 NPCs. mRNA levels were obtained from RNA sequencing of biological triplicates and expressed as transcripts per kilobase million (TPM). (C) Immunoblot analysis of nuclear c-MYC and Neurogenin-2 in unedited and corrected Patient 1 NPCs. TBP was used as a loading control for nuclear lysates. (D) UMAP heatmaps displaying MYC expression in single-cell RNA-seq of unedited and corrected day 45 cortical spheroids. Color intensity (purple) indicates MYC transcript levels in individual cells. (E) Quantification of NEUROG2 mRNA in control H9 NPCs, unedited Patient 1 NPCs, and corrected Patient 1 NPCs. mRNA levels were obtained from RNA sequencing of biological triplicates and expressed as transcripts per kilobase million (TPM). (F) Immunoblot analysis of nuclear c-MYC and Neurogenin-2 in Patient 1 NPCs treated with DMSO or 1 μM CB839 (GLSi) for 4 or 7 days. TBP served as a loading control for nuclear lysates. For (B) and (E), statistical significance was assessed on log-transformed TPM values using one-way ANOVA followed by Tukey’s HSD test. Error bars represent ±1 SEM of three biological replicates.

Figure 5.. Increased activating histone methylation marks H3K4me2 and H3K4me3 at the MYC locus in L2HGDH-Deficient NPCs.

(A) DNA dot blot analysis of 5mC levels in NPCs, including control H9 NPCs, control iPSC-derived NPCs, unedited Patient 1 NPCs, unedited Patient 2 NPCs, corrected Patient 1 NPCs, and corrected Patient 2 NPCs. (B) m6A dot blot analysis of total RNA in NPCs, including control H9 NPCs, control NPCs, unedited Patient 1 NPCs, unedited Patient 2 NPCs, corrected Patient 1 NPCs, and corrected Patient 2 NPCs. (C) Immunoblot analysis of activating histone markers H3K4me2 and H3K4me3 levels in unedited and corrected Patient 1 NPCs. Histone H3 was used as a loading control for histone lysates. (D) and (E) ChIP–Seq profiles of H3K4me2 (D) and H3K4me3 (E) in unedited and corrected Patient 1 NPCs. ChIP–Seq signals were plotted over center peaks (± 5 kb from peak center) identified in corrected Patient 1 NPCs. Sites were sorted by the ChIP–Seq signal intensity from corrected Patient 1 NPCs. (F) and (G) Representative ChIP-seq tracks showing H3K4me2 (F) and H3K4me3 (G) enrichment at the MYC locus in unedited and corrected Patient 1 NPCs.

Figure 6.. c-MYC depletion restores neuronal differentiation in L2HGDH-deficient NPCs.

(A) Immunoblot of nuclear c-MYC and Neurogenin-2 in unedited Patient 1 NPCs treated with DMSO or 20 μM nocodazole (Noc) for 7 days. TBP was used as a loading control. (B) Quantification of neurite lengths in Patient 1 NPCs treated with DMSO or 20 μM Noc at the NPC stage and throughout 14-day neuronal differentiation. (C) Immunoblot of nuclear c-MYC and Neurogenin-2 in Patient 1 NPCs transduced with lentivirus expressing control shRNA (shSCR) or MYC-targeting shRNA (shMYC). (D) Quantification of neurite lengths in Patient 1 NPCs transduced with shSCR or shMYC. The black dashed line denotes the mean neurite length in corrected neurons. (E) Representative recordings of intracellular Ca²⁺ dynamics in neurons transduced with shSCR (n = 29) or shMYC (n = 24), captured every 10 s using Fura-2 ratio imaging during 45 mM KCl application. Neurons were differentiated for 47 days. (F) Immunoblot of nuclear c-MYC and Neurogenin-2 in Patient 1 NPCs treated with DMSO or 50 μM EN4 for the indicated durations. (G) Quantification of neurite lengths in Patient 1 NPCs treated with: DMSO; EN4 throughout differentiation; EN4 at the NPC stage and throughout differentiation; or EN4 only at the NPC stage. The dashed line indicates the corrected neuron mean. (H) Representative Ca²⁺ recordings in neurons treated with DMSO (n = 28) or 50 μM EN4 (n = 26), acquired as in (E). For (B), (D), and (G), neurite lengths were measured using the SNT plugin in ImageJ. Data are mean ± 1 SEM. Statistical significance was determined by Student’s t test (B, D) or one-way ANOVA with Tukey’s HSD test (G).