RNA profiling of human dorsal root ganglia reveals sex differences in mechanisms promoting neuropathic pain

Abstract

Neuropathic pain is a leading cause of high-impact pain, is often disabling and is poorly managed by current therapeutics. Here we focused on a unique group of neuropathic pain patients undergoing thoracic vertebrectomy where the dorsal root ganglia is removed as part of the surgery allowing for molecular characterization and identification of mechanistic drivers of neuropathic pain independently of preclinical models. Our goal was to quantify whole transcriptome RNA abundances using RNA-seq in pain-associated human dorsal root ganglia from these patients, allowing comprehensive identification of molecular changes in these samples by contrasting them with non-pain-associated dorsal root ganglia. We sequenced 70 human dorsal root ganglia, and among these 50 met inclusion criteria for sufficient neuronal mRNA signal for downstream analysis. Our expression analysis revealed profound sex differences in differentially expressed genes including increase of IL1B, TNF, CXCL14 and OSM in male and CCL1, CCL21, PENK and TLR3 in female dorsal root ganglia associated with neuropathic pain. Coexpression modules revealed enrichment in members of JUN-FOS signalling in males and centromere protein coding genes in females. Neuro-immune signalling pathways revealed distinct cytokine signalling pathways associated with neuropathic pain in males (OSM, LIF, SOCS1) and females (CCL1, CCL19, CCL21). We validated cellular expression profiles of a subset of these findings using RNAscope in situ hybridization. Our findings give direct support for sex differences in underlying mechanisms of neuropathic pain in patient populations.

Keywords: dorsal root ganglia; neuropathy; pain; pain transcriptomics; transcriptome-wide association study.

Figure 1

Top 25 pain-associated genes in the male cohort (increased in pain). Pain-associated genes in male samples (Supplementary Table 3A) show systematic increases in pain. Quantile plots (quantile versus value) for gene relative abundances (in TPMs) for the top 25 genes are shown in male pain samples (in red, upper line) and male no-pain samples (in blue, lower line). These include multiple members of AP-1 signalling (EGR3, FOSL1), pro-inflammatory cytokines (IL1B, CCL3, CCL4), TNF signalling (TNF, IL1B) and other transcriptional regulators (NR4A2, FOXS1, HBEGF) relevant to the peripheral nervous system and pain.

Figure 2

Top 25 pain-associated genes in the female cohort (increased in pain). Pain-associated genes in female samples (Supplementary Table 3C) show systematic increases in pain. Quantile plots (quantile versus value) for gene relative abundances (in TPMs) in the top 25 genes are shown in female pain samples (in magenta, upper line) and female no-pain samples (in green, lower line). These include multiple members of receptor genes (ADORA2B, IL1RAPL2, GPR160), pro-inflammatory and proliferation-related genes (HAMP, FREM1), vesicular trafficking genes (LYG2, RASEF) and interferon-response genes (USP6, TTC12).

Figure 3

Sex differential aspects of the pain-associated transcriptome. (A) The top 25 pain-associated genes (Supplementary Table 3A and C) increased in pain (with a median fold change of 2-fold or higher) for each sex show a remarkably sex-differential signal, with only HAMP showing 2-fold or greater change in the median in both sexes among the top 25. Log2 fold changes in the median between pain and no-pain subcohorts are shown for males and females as a scatter plot, with male pain-associated genes shown in blue (right cluster) and female pain-associated genes shown in pink (left cluster). LY96 is present in both male and female lists. (B) Pain-associated ligands in each sex signal to hDRG-expressed receptors that are enriched in sensory neuronal subpopulations, but have little overlap across sexes. Based on our interactome analysis, we show such ligand–receptor pairs, alongside receptor expression in human DRG neuronal subtypes as a heatmap. Overlap with relevant signalling pathways are also shown. Male signalling is enriched in TNF-alpha pathway, while female signalling is enriched in interferon signalling. Although ICAM3 (in males) and TGFB2 or NOV (in females) are not in the pain-associated gene lists, they are increased in pain for the corresponding sex at the median or upper quartile levels and are thus shown here.

Figure 4

OSM coexpression module. Coexpression of individual genes with OSM in male samples was quantified using Pearson’s correlation (Pearson’s R). (A) Empirical cumulative distribution function (CDF) of Pearson’s R (with OSM expression) for transcription factor genes with Pearson’s R > 0.55, overlaid with Pearson’s R for genes of key signalling molecules in the TNF-alpha pathway (TNF, TP53 and OSM) and IL32 (another pro-inflammatory cytokine) and members of enriched gene sets that overlap with the coexpressed TF genes are also shown. (B) Heatmap of correlation matrix (Pearson’s R) between members of OSM coexpression module in enriched gene sets (TNF signalling via NFKB, p53 signalling) that overlap with it. Rows and columns of the correlation matrix have the same genes, with the diagonal showing perfect correlation (R = 1).

Figure 5

IFIT1 coexpression module. Coexpression of individual genes with IFIT1 in female samples was quantified using Pearson’s correlation (Pearson’s R). (A) Empirical cumulative distribution function (CDF) of Pearson’s R (with IFIT1 expression) for transcription factor genes with Pearson’s R > 0.55, overlaid with Pearson’s R for genes of key signalling molecules in the interferon signalling pathways and members of enriched gene sets that overlap with the coexpressed TF genes are also shown. (B) Heatmap of correlation matrix (Pearson’s R) between members of IFIT1 coexpression module in enriched gene sets (interferon alpha and gamma response, respectively; PI3K/AKT/mTOR signalling) that overlap with it. Rows and columns of the matrix have the same genes, with the diagonal showing perfect correlation (Pearson’s R = 1).

Figure 6

RNAscope for male pain-associated genes TNF, OSM, IL1B and IL32. (A) RNAscope TNF expression (red) overlaid with TRPV1 expression (green), AIF1 (white) and DAPI (blue) in pseudocolour. (B) Schema showing how TNF-alpha signalling cascades overlap with other signalling pathways, based on existing literature (icons from BioRender). (C) TNF, OSM, IL1B and IL32 quantile plots show the shift in abundance between male pain and non-pain subcohorts, while the bar plots show the RNA-seq abundance for the patient DRGs that were queried by RNAscope assay. (D) RNAscope OSM, IL1B and IL32 expression (red) overlaid with TRPV1 expression (green), AIF1 (white) and DAPI (blue) in pseudocolour. Scale bar = 20 μm, large globular structures are considered to be lipofuscin. Yellow arrows point out cells with overlap of expression between the gene of interest (red) and AIF1 (white), cyan arrows point out overlap of expression between the gene of interest (red) and TRPV1 (green), respectively, in the zoomed-in images, and pink arrows point to lipofuscin. (A and D) In all micrographs, wide field views and zoomed-in views on single neurons and surrounding cells are shown to display overall signal distribution, and colocalization of signal with specific neuronal and macrophage cell markers for each RNAscope probe.

Figure 7

RNAscope for female pain-associated genes IFIT1 and HLA-DQB1. (A) RNAscope IFIT1 expression (red) overlaid with TRPV1 expression (green), AIF1 (white) and DAPI (blue) in pseudocolour. (B) IFIT1 and HLA-DQB1 quantile plots show the shift in abundance between female pain and non-pain subcohorts, while the bar plots show the RNA-seq abundance for the patient DRGs that were queried by RNAscope assay. (C) RNAscope HLA-DQB1 expression (red) overlaid with TRPV1 expression (green), AIF1 (white) and DAPI (blue) in pseudocolour. (D) Schema showing how interferons are typically activated and drive a signalling programme, based on existing literature (icons from BioRender). (E) Quantile plots in the female cohort for other IFIT1-correlated interferon signalling pathway genes—IFIT2, TLR1, NMI—show increase in gene abundance in pain. Scale bar = 20 μm, large globular structures are considered to be lipofuscin. Yellow arrows point out cells with overlap of expression between the gene of interest (red) and AIF1 (white), cyan arrows point out overlap of expression between the gene of interest (red) and TRPV1 (green), respectively, in the zoomed in images, and pink arrows point to lipofuscin (A and C). In all micrographs, wide field views and zoomed-in views on single neurons and surrounding cells are shown to display overall signal distribution, and colocalization of signal with specific neuronal and macrophage cell markers for each RNAscope probe.