Shared patterns of glial transcriptional dysregulation link Huntington's disease and schizophrenia

Abstract

Huntington's disease and juvenile-onset schizophrenia have long been regarded as distinct disorders. However, both manifest cell-intrinsic abnormalities in glial differentiation, with resultant astrocytic dysfunction and hypomyelination. To assess whether a common mechanism might underlie the similar glial pathology of these otherwise disparate conditions, we used comparative correlation network approaches to analyse RNA-sequencing data from human glial progenitor cells (hGPCs) produced from disease-derived pluripotent stem cells. We identified gene sets preserved between Huntington's disease and schizophrenia hGPCs yet distinct from normal controls that included 174 highly connected genes in the shared disease-associated network, focusing on genes involved in synaptic signalling. These synaptic genes were largely suppressed in both schizophrenia and Huntington's disease hGPCs, and gene regulatory network analysis identified a core set of upstream regulators of this network, of which OLIG2 and TCF7L2 were prominent. Among their downstream targets, ADGRL3, a modulator of glutamatergic synapses, was notably suppressed in both schizophrenia and Huntington's disease hGPCs. Chromatin immunoprecipitation sequencing confirmed that OLIG2 and TCF7L2 each bound to the regulatory region of ADGRL3, whose expression was then rescued by lentiviral overexpression of these transcription factors. These data suggest that the disease-associated suppression of OLIG2 and TCF7L2-dependent transcription of glutamate signalling regulators may impair glial receptivity to neuronal glutamate. The consequent loss of activity-dependent mobilization of hGPCs may yield deficient oligodendrocyte production, and hence the hypomyelination noted in these disorders, as well as the disrupted astrocytic differentiation and attendant synaptic dysfunction associated with each. Together, these data highlight the importance of convergent glial molecular pathology in both the pathogenesis and phenotypic similarities of two otherwise unrelated disorders, Huntington's disease and schizophrenia.

Keywords: Huntington disease; astrocyte; glial progenitor cell; neurogenetics; oligodendrocyte; schizophrenia; synapse.

Figure 1

Huntington’s disease and schizophrenia-derived glia share common mechanisms of disrupted glial differentiation. (A–E) Assessment of differential and shared gene expression by Huntington’s disease (HD) and schizophrenia (SCZ) glial progenitor cells (GPCs). (A) Venn diagram depicting the number of differentially expressed genes (DEGs) either unique to or shared by HD and SCZ GPCs. The overlap between the two sets of dysregulated genes was statistically significant (Fisher’s exact test, P = 4 × 10⁻⁵). (B) Functional annotation dot plots representing significant gene ontology (GO) terms for genes dysregulated by both HD and SCZ human GPCs (hGPCs) versus controls (CTR). DEGs in HD- and SCZ-derived GPCs shared implicated functions relating to synapse structure and synaptic transmission. Dot sizes represent gene ratios with respect to the gene set. Dot colours indicate the significance of association to the GO term. (C) Scatter plots showing the correlation in log₂-transformed fold-changes obtained from comparisons of disease versus control HD (x-axis) and SCZ (y-axis) hGPCs. The shaded area indicates 95% confidence interval for linear fit line in blue. (D) Network representation of functional annotation for dysregulated genes in SCZ-astrocytes. (E) Expression heat maps of top DEGs identified in M1–3. (F–H) Differential and shared gene expression by HD and SCZ astrocytes. (F) Venn diagram of differentially expressed genes either unique to or shared by HD and SCZ astrocytes. The overlap between the two sets of dysregulated genes was not statistically significant (Fisher’s exact test, P > 5 × 10⁻²). (G) DEGs in HD- and SCZ-derived astrocytes shared common annotation with functions relating to extracellular matrix organization. Overlapping genes in significant annotation were emphasized to be concordantly downregulated in both HD- and SCZ-GPCs (C) or astrocytes (H). (H) Scatter plots showing the correlation in log₂-transformed fold-changes obtained from comparisons of disease versus control HD (x-axis) and SCZ (y-axis) astrocytes. The shaded area indicates 95% confidence interval. fc = fold-change.

Figure 2

Prioritized modules and their structural preservation across networks. (A) Nine modules were identified as significantly associated with the condition phenotype in glial progenitor cells (GPCs). These modules were tested for network preservation using NetRep, and the maximal P-value was reported. (B) Four of nine modules were preserved between Huntington’s disease (HD) and schizophrenia (SCZ), exhibiting conserved directionality of dysregulation across diseases. (C) Similar to the GPC networks, eight modules were identified as significantly associated with the condition phenotype in astrocytes. The eight modules were tested for network preservation, and the maximal P-value was reported. (D) Only one of eight modules were found to be preserved, representing downregulated genes that were conserved in both HD and SCZ diseases. (E) Gene set enrichment analysis (GSEA) indicates that disease-associated human GPC and astrocyte dysregulation was significantly enriched for modules in bold (adjusted-P < 10⁻²) and with conserved directionality across diseases. Normalized enrichment score (NES) for the modular gene sets, with positive scores indicating enrichment for upregulated genes in the disease condition, and negative scores indicating enrichment for downregulated genes. ***P < 10⁻³, **P < 10⁻², *P < 5 × 10⁻². CTR = control; fc = fold-change.

Figure 3

Gene set intersection highlights related clusters with distinct functions. (A) The module gene set network was compiled from the significant gene set intersection between prioritized disease-associated modules. The nodes represent the modules and the edges represent the significant pairwise gene set intersection (Fisher’s exact test, P < 10⁻¹). The node and edge colours represent the four closely interconnected module clusters identified by community detection, designated as C1 through C4. (B) Gene set enrichment analysis (GSEA) indicated glial progenitor cell (GPC)- or astrocyte- dysregulations to be specifically enriched for a specific cluster. (C) Clusters displayed distinct annotation trends. (D) Clusters were regulated by distinct collections of transcription factors (TFs). CTR = control; DEG = differentially expressed gene; HD = Huntington’s disease; SCZ = schizophrenia.

Figure 4

Synapse-regulatory pathways were dysregulated in both Huntington’s disease and schizophrenia human glial progenitor cells. (A) The multilayered gene correlation network was composed from 1218 genes shared among prioritized modules in Fig. 3A. Node colour indicates which cluster each gene was assigned to, and edge colour indicates which original network a correlation was observed. The network clusters were defined by Gephi’s community detection algorithm and visualized using Cytoscape (Supplementary material, ‘Extended Methods’ section). (B) Clusters C1 and C2 were spread among all four original networks, indicating their crucial contribution to the aggregated network structure. (C) Distribution of node strength and eigenvector centrality across clusters. Cluster C2 was composed of genes with the highest node strength and second-highest eigenvector centrality. (D) Prioritized genes were functionally concordant to known Huntington’s disease (HD)- and schizophrenia (SCZ)-implicated genes. The 174 top-ranked genes were compared against the set of externally acquired genes. Shared members among the top-ranked, SCZ-, and HD- associated genes are highlighted in text. (E and F) Pearson’s R coefficients between the four gene sets based on all tested gene ontology (GO) terms or significant terms (adjusted-P < 10⁻²). Hierarchical clustering indicated a close connection between SCZ, HD and top-ranked genes, but not rheumatoid arthritis (RA). (G) Heat map depicting the adjusted P-values for the top five significant annotations per gene set. Similar to E and F, hierarchical clustering indicates similarity between HD, SCZ and top-ranked genes but not to RA.

Figure 5

Suppression of OLIG2 and TCF7L2-dependent modulators of glutamatergic synapses. (A) Transcription factors enriched for top-ranked genes. (B) OLIG2 expression is high in glial progenitor cells (GPCs) and low in astrocytes. (C) OLIG2 occupancy at the transcription start sites of prioritized genes in the multilayered network of Fig. 3. (D) Tracks showing OLIG2 peaks for OMG, and for TCF7L2 in GPCs. (E) Heat map depicting dysregulated downstream targets of OLIG2 that were also top-ranked genes. (F) Tracks showing OLIG2 peaks for genes related to glutamatergic receptors. (G) Expression of glutamatergic synapse-related genes. (H) Downregulation of glutamatergic synapse-regulators by quantitative-reverse transcription PCR (n = 6, Student’s t-test). (I) Downregulation of ADGRL3 at the protein level, as measured by ELISA. (n = 4, Student’s t-test). (J) Venn diagram indicating a high degree of overlap between OLIG2 and TCF7L2 peaks in GPCs but not in HepG2 hepatocarcinoma cells. (K) Occupancy intensities for peaks identified in Tcf7l2 chromatin immunoprecipitation sequencing (ChIP-seq) in rat at different stages of oligodendrocyte maturation, indicating that Tcf7l2 binding is stage-specific. (L) Tracks showing Tcf7l2 peak for Adgrl3—a glutamatergic synapse regulator. (M) Rescue of ADGRL3 gene expression by OLIG2 or TCF7L2 (n = 4, Dunnett’s test). (N) Rescue of ADGRL3 protein expression by OLIG2, as measured by ELISA. (n = 4, Student’s t-test). ***P < 10⁻³; **P < 10⁻²; *P < 5 × 10⁻²; ⁺P < 10⁻¹. CTR = control; LV = lentivirus.

Figure 6

Our model: dysregulated glial support of glutamatergic synapses reciprocally impacts oligodendrocytic and astrocytic maturation in both Huntington’s disease and schizophrenia. (A) At the glial progenitor cell (GPC) stage, ADGRL3 is regulated by OLIG2. (B) ADGRL3 expression is regulated by TCF7L2 at the oligodendrocyte stage. (C) Diseased GPCs fail to respond to synaptic cues due to decreased glutamate receptor expression, leading to cell-intrinsic impediment to oligodendrocyte (OL) maturation. (D) The cyclic AMP (cAMP) level is modulated by ADGRL3, and as a result so is CREB transcriptional activation. (E) cAMP-dependent CREB activation in response to neuronal activity, resulting in the expression of synapse regulatory genes, as well as astrocytic genes. (F) Immature oligodendrocytes fail to properly myelinate, leading to synaptic loss and functional deficits. (G) Immature astrocytes fail to clear extracellular glutamate, resulting in blockade of spine remodeling, transmitter ‘spillover’, and ultimately postsynaptic excitotoxicity. Solid line: confirmed gene regulatory network from this manuscript; dashed line: proposed regulation, based upon previous studies.