Human genetic variation shapes the response of neurons to interferons

Abstract

Inflammation is increasingly recognized as important to neuropathology, including more classic neuroimmune disease as well as neurodegenerative and neuropsychiatric disorders. Interferons (IFN) are important mediators of central nervous system inflammation. Individuals appear to vary in susceptibility to neuroinflammatory pathology, suggesting that identifying human genetic modifiers of the neuronal IFN response might provide insight into disease pathophysiology. To identify potential modifiers, we stimulated neuronal "cellular villages" of iPSC-derived neurons from over one hundred donors with IFN-alpha (IFNa) or IFN-gamma (IFNg). We then correlated allele states of common variable SNPs to gene expression to identify hundreds of expression quantitative trait loci (eQTLs), many of which emerged specifically upon IFN treatment. We characterized the distinct but overlapping neuronal transcriptional responses to IFNa and IFNg, and identified specific response QTLs. Functional annotation of STAT1 binding to the genome in response to IFN stimulus identified STAT1 binding sites as enriched for response-regulating human genetic variation and also enabled identification of loci with IFN-dependent allele-specific binding of STAT1. These results demonstrate how human genetic variation can influence IFN-dependent mechanisms in neurons in disease-relevant ways.

Figure 1:. Interindividual variation in the transcriptomic response to IFNs.

a) Experimental overview. iPSC-derived neurons from 102 donors (genotyped separately before analysis) were cultured together and treated with IFNa, IFNg or PBS (control). Cells were analyzed by single-cell RNA-seq, then assigned to the individual donors based on combinations of many transcribed SNPs. Later, in some analyses, genetic sequence variation and gene expression are correlated to identify eQTLs, and this is compared to transcription factor binding. b) Volcano plots show overall response to 7 hours of IFNa (left) and 24 hours of IFNg (right) treatment with robust gene upregulation. Data are from all donors, all cells (pseudo-bulked together). c) Differential gene-expression responses to IFNg and IFNa treatment. d) Gene set enrichment analysis comparing differentially expressed genes in IFNg versus IFNa treatment. Top five enriched by normalized enrichment score (NES) biological pathways shown (highest and lowest). e) IFN-response metagenes were generated by calculating mean expression of top induced genes with >4-fold induction (n=94 for IFNa, n=109 for IFNg). IFNa-response and IFNg-response metagene expression per donor in both control (gray) and IFN conditions is shown for IFNa (red) and IFNg (blue). f) IFN-response metagene expression for each donor in response to either IFNa (x-axis) or IFNg (y-axis). g) Relative expression (visualized as Z-score of the donor-specific measurements) of gene expression (log counts per million) of top induced genes (fold change >4) in response to IFNa (top) or IFNg (bottom). Each column represents a donor; each row represents a gene. Hierarchical clustering was performed for both donors and genes in the IFNa group, and this order was maintained to display IFNg data. h) “mini” metagene magnitude shown per donor in both control (gray) and IFNa-stimulated (red) conditions.

Figure 2:. Genetic determinants of IFN response and cis-eQTL discovery.

a) Comparison of normalized eQTL effect sizes in IFN versus control. Significant eQTLs in either IFN conditions are displayed, comparing their effect size normalized to respective gene expression in either control or IFN conditions. Off-diagonal effects are response-specific. b) All eQTLs and their corresponding genes with q-value <0.05 were considered, and diagram displays overlap in genes with eQTLs in each condition. c) Example eQTLs including those considered shared (top) and IFN-responsive (middle, bottom) d). Schematic of significant rQTLs found in IFNa (red) and IFNg (blue) response and their area of function in the classical MHC class I peptide presentation pathway.

Figure 3:. STAT1 binding underlies much interindividual variation in response to interferons.

a) Distribution of all eQTLs found (in any condition) in the genome. b) STAT1 binding across the genome in response to either IFNg (right) or IFNa (left). c) Enriched motifs in STAT1 peaks relevant to STAT binding. d) average mean read density of STAT1 binding (measured using Cut&Tag) at SNPs and eQTLs identified in control and IFN-response conditions. e) Overlap of eQTL categories with STAT1 binding sites. Chi-squared test performed on IFN-shared eQTLs and IFN-responsive eQTLs.

Figure 4:. Allele-specific and interferon-dependent function at a CASP7 eQTL.

a) SNPs within 500kb of the lead CASP7 eQTL are shown colored by R² value with respect to the lead CASP7 eQTL (green). b) Top, example western blot for CASP7 with TUBB3 as loading control. Bottom, protein quantification normalized to TUBB3 expression (n=3 independent pairs) *p<0.01 by student’s t-test. c) STAT1 binding at the CASP7 promoter in cells carrying two copies of the major (CTT;CTT) or minor (C;C) allele. Signal is normalized to total number of reads. d) Relative STAT1 Cut&Tag read density across CASP7 eQTL in cells heterozygous at the allele (CTT; C) (n=4). **p<0.0001 by chi-squared test (for each experiment). H3K4me3 read density across the same site as STAT1 binding, shown as control (n=2; p-value for each chi-squared test = 0.04 and 0.52).