Common transcriptional signatures of neuropathic pain

Abstract

The dorsal root ganglia (DRG) are key structures in nociception and chronic pain disorders. Several gene expression studies of DRG in preclinical pain models have been performed, but it is unclear if consistent gene changes are identifiable. We, therefore, compared several recent RNA-Seq data sets on the whole DRG in rodent models of nerve injury. Contrary to previous findings, we show hundreds of common differentially expressed genes and high positive correlation between studies, despite model and species differences. We also find, in contrast to previous studies, that 60% of the common rodent gene response after injury is likely to occur in nociceptors of the DRG. Substantial expression changes are observed at a 1-week time-point, with smaller changes in the same genes at a later 3- to 4-week time-point. However, a subset of genes shows a similar magnitude of changes at both early and late time-points, suggesting their potential involvement in the maintenance of chronic pain. These genes are centred around suppression of endogenous opioid signalling. Reversal of this suppression could allow endogenous and exogenous opioids to exert their analgesic functions and may be an important strategy for treating chronic pain disorders. Currently used drugs, such as amitriptyline and duloxetine, do not seem to appropriately modulate many of the critical pain genes and indeed may transcriptionally suppress endogenous opioid signalling further.

Figure 1.

Comparison on the common DEGs in rodent DRG data sets on the injury pain. (A) Principle component analysis plot of variance-stabilized transformed counts for mouse: Cobos (C), day 7, and Baskozos (B), day 28. (D) Principle component analysis plot of variance-stabilized transformed counts for rat: Perkins (P), day 7, and Baskozos (Br), day 21. To illustrate the correction of technical variations between the data sets, which was performed internally by Deseq2, the PCAs were generated after applying the removeBatchEffect function from the limma package. (B and E) Venn diagrams of DEG numbers in mouse (B) and rat (E) data sets, showing the numbers of overlapping genes. (C and F) Linear correlation between log2fc in expression levels relative to control noninjured DRG for the common genes (adjusted P ≤ 0.05) in mouse (C) and rat (F) data sets. (G) Linear correlation of the expression changes in common genes between at least one mouse and rat data set (|log2fc| ≥ 0.5; adjusted P ≤ 0.05). The sets of the common genes were split into correlated and anticorrelated groups as indicated in the legend. DEG, differentially expressed gene; DRG, dorsal root ganglia; PCA, principle component analysis.

Figure 2.

The GO enrichment analysis of the common rat/mouse DEGs in nerve injury models (Table S5, http://links.lww.com/PAIN/A972). (A and B) The ClusterProfiler analysis of significantly enriched GO terms within the cell component (A) and biological processes (B) categories; FDR-adjusted P ≤ 0.05. (C) TopGO analysis of the common genes with Elim method. The top categories with FDR-adjusted P < 0.01 are shown in bright colours. DEG, differentially expressed gene.

Figure 3.

Sensory neurons genes are differentially expressed in mouse DRG during nerve injury. The set of 233 common mouse DEGs from Table S2 (http://links.lww.com/PAIN/A972) was analysed for nociception-related genes using reported lists of mouse DEGs in DRG sensory neurons. (A) Venn diagram of the DEG numbers related to NP, PEP, and LM sensory neurons, shown in blue, magenta, and green, respectively. (B) Linear correlation of the whole DRG expression changes in the nociception genes with those in individual sensory neurons (NP, PEP, and LM) during nerve injury. (C–E) Protein–protein interaction analysis of biological interactions within common mouse DEGs related to NP (C), PEP (D), or LM (E) neurons. The analysis was performed with STRING, using high confidence of interactions (score 0.7) and not more than 5 interactors in the first shell. The line colour indicates the type of interaction evidence. The nodes were clustered using the MCL algorithm. Those genes that did not have a direct partner in the network are not shown. DEG, differentially expressed gene; DRG, dorsal root ganglia; NP, nonpeptidergic nociceptor; PEP, peptidergic nociceptor.

Figure 4.

Enrichment analysis of the persistent rat DEGs with pronounced expression changes, which are not present in mouse data set of sensory neurons. ClusterProfiler analysis of significantly enriched GO terms among DEGs with |log2fc| at 28 days (data set Br) ≥ than at 7 days (data set P). Only highly DEGs (|log2fc| ≥ 1; 304 genes total) were used. GO terms are shown within the CC (A) and BP (B) categories; FDR-adjusted P < 0.05. BP, biological process; CC, cell component; DEG, differentially expressed gene; FDR, false discovery rate.

Figure 5.

The pain-related PPI network of the persistent rat DEGs. The network was built based on the pain interactome, using rat DEGs with persistent expression changes, together with neighbours from the pain interactome network (DEGs and non-DEGs). The coloured arrows mark the type and direction of interactions; coloured nodes mark the expression changes in rat data sets (up/down/persistent/nonpersistent/non-DEG partner), as explained in the legend. DEG, differentially expressed gene; PPI, protein–protein interaction.

Figure 6.

Cell-line responses to 10-μm amitriptyline and duloxetine for 6 hours relative to the differential gene expression of rat. For each rat gene, the number of anticorrelated and correlated cell-line responses from the drugs amitriptyline (A) and duloxetine (B) were collated. Genes in the rodent model whose changes are expected to have antinociceptive effects are indicated with an asterisk.