Single-cell transcriptomic analysis of somatosensory neurons uncovers temporal development of neuropathic pain

Abstract

Peripheral nerve injury could lead to chronic neuropathic pain. Understanding transcriptional changes induced by nerve injury could provide fundamental insights into the complex pathogenesis of neuropathic pain. Gene expression profiles of dorsal root ganglia (DRG) in neuropathic pain condition have been studied. However, little is known about transcriptomic changes in individual DRG neurons after peripheral nerve injury. Here we performed single-cell RNA sequencing on dissociated mouse DRG cells after spared nerve injury (SNI). In addition to DRG neuron types that are found under physiological conditions, we identified three SNI-induced neuronal clusters (SNIICs) characterized by the expression of Atf3/Gfra3/Gal (SNIIC1), Atf3/Mrgprd (SNIIC2) and Atf3/S100b/Gal (SNIIC3). These SNIICs originated from Cldn9⁺/Gal⁺, Mrgprd⁺ and Trappc3l⁺ DRG neurons, respectively. Interestingly, SNIIC2 switched to SNIIC1 by increasing Gal and reducing Mrgprd expression 2 days after nerve injury. Inferring the gene regulatory networks after nerve injury, we revealed that activated transcription factors Atf3 and Egr1 in SNIICs could enhance Gal expression while activated Cpeb1 in SNIIC2 might suppress Mrgprd expression within 2 days after SNI. Furthermore, we mined the transcriptomic changes in the development of neuropathic pain to identify potential analgesic targets. We revealed that cardiotrophin-like cytokine factor 1, which activates astrocytes in the dorsal horn of spinal cord, was upregulated in SNIIC1 neurons and contributed to SNI-induced mechanical allodynia. Therefore, our results provide a new landscape to understand the dynamic course of neuron type changes and their underlying molecular mechanisms during the development of neuropathic pain.

Fig. 1. The heterogeneity of DRG cells in SNI model.

a Experimental design. L4 and L5 DRGs are dissected after SNI, and the dissociated cell suspension is processed with 10× Genomics scRNA-seq. b t-SNE plot of 36,810 DRG cells shows the cellular heterogeneity of DRG cells. Dots, individual cells; colors, cell types. c Feature heatmap shows the cell type-specific marker genes in each DRG cell type. The color represents the Log₂ normalized transcript counts. d t-SNE plot of 6935 DRG neurons shows the neuronal clusters. Dots, individual cells; colors, neuron clusters. e Dot plot shows the expression of marker genes in each DRG neuron clusters. The size of the dot indicates the percentage of cells expressing the gene in each cluster, and the color depicts the average normalized transcript counts in the neurons.

Fig. 2. Integrated analysis of DRG neuron sequenced by 10× Genomics and Smart-seq2.

a t-SNE plots of 3227 (479 for Smart-seq2 and 2748 for 10× Genomics) L4 and L5 DRG neurons sequenced by 10× Genomics and Smart-seq2 after alignment. Dots, individual cells. b Summary of neuron classification based on 10× Genomics and the previously reported classification. c ISH shows the Prokr2- and Smr2-expressing neurons in mouse L5 DRG (n = 3). Scale bar, 50 µm. d ISH shows Dcn expression in both DRG neurons (arrow) and fibroblast (arrowhead). Scale bar, 50 µm. e RNAscope result shows the Rxfp1-expressing DRG neuron (arrow). Scale bar, 50 µm. f Statistical results show the percentages of Dcn⁺ and Rxfp1⁺ DRG neurons (n = 3). g RNAscope assay shows that the Zcchc12 (arrowhead) and Cldn9 (arrow) are not co-expressed in DRG neurons. Scale bar, 20 µm. h Dual fluorescent ISH shows that Dcn⁺ neurons are a subset of tdTomato⁺ neurons in Zcchc12-CreER::Ai9 mouse L5 DRG (arrow). Scale bar, 20 µm. i RNAscope assay shows that Rxfp1⁺ neurons are a subset of tdTomato⁺ neurons in Zcchc12-CreER::Ai9 mouse L5 DRG (arrows). Scale bar, 20 µm. The data are shown as means ± SEM.

Fig. 3. Three emerging neuron clusters appeared at different time points after SNI.

a t-SNE plot shows somatosensory neuron clusters in control and at different time points after SNI. Dots, individual cells; colors, neuron clusters. b A heatmap shows the DEGs in the three emerging neuron clusters. The color represents the scaled transcript counts. c, d Dual fluorescent ISH confirms the co-expression of Atf3 with Mrgprd, Gfra3, or S100b at different time points after SNI (n = 3). Scale bar, 20 µm. The data are shown as means ± SEM. *P < 0.05, ***P < 0.001, n.s. no significance vs the control group.

Fig. 4. The origins of emerging neuron clusters induced by SNI.

a, b RNAscope results show that the percentage of Cldn9⁺ neurons was decreased after SNI (n = 3). Scale bar, 20 µm. c A heatmap shows the Pearson correlation among different neuron clusters based on whole transcript counts. d A proposed model for the neuron type switches induced by SNI. e, f Dual fluorescent ISH combined with immunostaining shows the co-expression of tdTomato (red), Mrgprd (green) and Gal (blue) in DRG neurons of Mrgprd-CreER::Ai9 mice. A part of Mrgprd⁺ neurons expressed Gal but express much less Mrgprd (arrowheads) 2 days after SNI (n = 3). Scale bar, 20 µm. The data are shown as means ± SEM. ***P < 0.001 vs the control group.

Fig. 5. Gene regulatory network underlying peripheral nerve injury.

a t-SNE plot shows the major neuron clusters based on the regulon activity matrices. Dots, individual cells; colors, neuron clusters. b Gene regulatory networks of Gal and Mrgprd were inferred by SCENIC. Colors indicated Normalized Enrichment Score (NES). c t-SNE plots show the activity of each TF and the enriched motif. The activated cells were labeled by blue. d Flowchart shows the siRNA transfection in the cultured DRG neurons filtered by FACS. e, f qPCR results show the effects of TF knockdown on expression of Gal or Mrgprd. The data are normalized to negative control (n ≥ 3). The data are shown as means ± SEM. *P < 0.05, **P < 0.01, n.s. no significance vs the control group.

Fig. 6. Pseudo-time analysis reveals the gene expression changes and functional adaptation during Cldn9⁺ neuron type-switch.

a Single-cell trajectory shows the fates of Cldn9⁺ neurons after SNI. Two branches were identified, one led to Atf3⁺/Gfra3⁺/Gal⁺ neurons from Cldn9⁺ neurons (Fate1, F1), the other remained as it was (F2). The color indicates the time points after SNI (top) or represents neuron clusters (bottom). b Gene signatures during Cldn9⁺ neuron switch. The DEGs (rows) are shown using heatmap (left) with neurons (columns) in pseudo-time from Root to F1. Gene-expression trends in each group (middle). GO terms associated with DEGs in the four kinetic clusters (right). c Heatmaps show the DEGs identified by BEAM (rows), and cells (columns) are shown in pseudo-time from Root to F1.

Fig. 7. Clcf1 is involved in the temporal development of neuropathic pain.

a RNAscope results show that the percentage of Clcf1⁺ neurons was increased after SNI (n = 3). Scale bar, 20 μm. b RNAscope results show the co-expression of Clcf1 and Crlf1 after SNI (n = 3). Scale bar, 20 μm. c ELISA results showing CLC levels in cultured DRG neurons from control and SNI mice (n = 11 for each group). d Intrathecal injection (i.t.) of CLC/CLF dose-dependently decreased the mechanical thresholds in C57 mice (n = 14 for PBS group; n = 12, 11, 12 for CLC/CLF group of 6, 30, or 150 ng injection, respectively). e Immunostaining results show that the astrocytes are activated 4 h after CLC/CLF complex injection (30 ng per mouse). Scale bar, 50 μm. The enlarged ROI and outline are shown on the right. Scale bar, 20 μm. f, g The signal of activated astrocytes and their distribution in spinal cord area were both increased after CLC/CLF injection. The data are shown as means ± SEM (n = 3). h A flowchart shows that  a basal threshold was tested before the injection, and the mice were injected with the siRNA for two consecutive days (4 μg each time). The von Frey test was performed on SNI 4d, and the knockdown efficiency of siClcf1 was detected from the mouse L4, L5 DRGs. i qPCR results show that the expression of Clcf1 was reduced in the DRGs after siClcf1 injection. j Knockdown of Clcf1 partially alleviated the mechanical allodynia induced by SNI (n = 18 for control and n = 18 for siClcf1). The data are shown as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, n.s. no significance vs the control group.