An Integrated Multi-Tissue Single-Cell Transcriptomic Analysis Highlights the CCL2-CCR2 Signaling Pathway as a Potential Therapeutic Target in Neuropathic Pain

Abstract

Background

Neuropathic pain involves complex neuroimmune interactions across the somatosensory nervous system, yet a comprehensive multi-tissue cellular landscape remains lacking.

Methods

We integrated single-cell RNA sequencing data from the sciatic nerve (SN), dorsal root ganglion (DRG), and spinal cord (SC) of spared nerve injury (SNI) mice. The therapeutic potential of the selective CCR2 antagonist RS504393 was evaluated in vivo.

Results

Macrophages dominated the injured SN (40.5%), exhibiting an inferred transcriptional shift from a resident-like Mrc1⁺ phenotype toward proinflammatory Ccl2⁺/Spp1⁺ subtypes. Injury elevated glycolytic activity in SN macrophages and spinal microglia, but suppressed metabolism in DRG neurons. Cell-cell communication analysis predicted enhanced CCL2-CCR2 signaling from peripheral macrophages to sensory neurons in the Pain group. In vivo, RS504393 administration significantly alleviated mechanical and thermal hypersensitivity, reduced spinal inflammatory infiltration, and suppressed microglial activation.

Conclusion

This study provides an integrated multi-tissue single-cell analysis of neuropathic pain and highlights the CCL2-CCR2 signaling pathway as a potential non-opioid therapeutic target at the peripheral neuroimmune interface, with pharmacological evidence suggesting downstream central effects.

Introduction

Neuropathic pain (NP) is a complex chronic pain condition caused by somatosensory nervous system lesions, exhibiting marked phenotypic and genetic heterogeneity.^1,2 Current treatments often fail to achieve satisfactory relief owing to complex pathological mechanisms. The development of NP involves neuroimmune interactions and chronic neuroinflammation within the peripheral and central nervous systems.^3,4 The spared nerve injury (SNI) model is widely used for dissecting these fundamental mechanisms.^5,6

During the pathogenesis of neuropathic pain, the sciatic nerve (SN), dorsal root ganglion (DRG), and spinal cord (SC) constitute critical anatomical nodes for nociceptive signal transmission. Following injury, localized Wallerian degeneration and substantial immune cell infiltration occur in the SN.^7,8 The DRG undergoes profound transcriptional reprogramming, contributing to peripheral sensitization.^9,10 Meanwhile, central sensitization is induced in the SC dorsal horn through glial activation.^11,12 These three nodes are anatomically and functionally connected along the nociceptive axis, providing a biological basis for integrated cross-tissue analysis. However, a comprehensive cellular landscape linking the peripheral injury site to the central projection area remains lacking.

Single-cell RNA sequencing (scRNA-seq) enables the precise dissection of tissue heterogeneity.^13,14 While macrophage infiltration is known to regulate the inflammatory microenvironment after nerve injury,^15,16 how macrophage metabolic shifts drive pain maintenance remains unresolved.¹⁷ Furthermore, the specific networks mediating cross-tissue signaling from peripheral nerves to the central nervous system require further characterization. Prior scRNA-seq studies have individually characterized immune responses in the injured SN,¹⁸ transcriptional changes in DRG satellite glial cells,¹⁹ and spinal cord glial dynamics under chronic pain;²⁰ however, none examined cross-tissue neuroimmune interactions along the SN–DRG–SC signaling pathway.

This study aimed to investigate the value of cross-tissue cellular communication and macrophage metabolic shifts in NP. By integrating multi-tissue scRNA-seq data from the SN,¹⁸ DRG,¹⁹ and SC,²⁰ all of which share the same peripheral nerve injury context and were processed with batch-effect correction to mitigate technical variation, this study systematically characterized macrophage developmental trajectories and core transcription factor networks, providing the first cross-tissue single-cell dissection of the SN–DRG–SC neuroimmune landscape in neuropathic pain. We note that cross-tissue communication inferred from unmatched public datasets is computational and hypothesis-generating; the identified signaling interactions therefore require experimental verification. Building on bioinformatics predictions, the functional role of the CCL2-CCR2 signaling pathway was validated in attenuating spinal microglial activation at the in vivo pharmacological level, aiming to provide novel cross-tissue therapeutic targets for NP.

Materials and Methods

scRNA-Seq Data Acquisition and Preprocessing

The scRNA-seq datasets used in this study were retrieved from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database, including three independent datasets: the SN dataset GSE153762 (Pain group only, n = 3 biological replicates, 3 days after SNI; no naïve Control sciatic nerve samples were available in this dataset),¹⁸ the DRG dataset GSE174430 (Control group, n = 1; Pain group, n = 2),¹⁹ and the SC dataset GSE208766 (Control group, n = 2; Pain group, n = 2).²⁰ Complete sample-level metadata for all 10 samples, including GEO sample IDs, tissue, condition, time point, sex, sequencing platform, and retained cell number, are provided in Supplementary Table S1. Raw FASTQ files were aligned to the mm10 mouse genome using Cell Ranger (v6.1.2; default parameters). Seurat (v4.3.0) was used for quality control, retaining 37,642 cells expressing 1000–6000 genes with <10% mitochondrial transcripts. Data were normalized (NormalizeData), scaled (ScaleData), and 2000 highly variable genes (HVGs) were identified (FindVariableFeatures) for principal component analysis (PCA). The top 30 principal components were retained and corrected for batch effects using Harmony (v1.0; theta = 2, batch covariate = sample ID).

Cell Clustering and Dimensionality Reduction Visualization

The batch-corrected Harmony embeddings were used to construct a k-nearest-neighbor (k-NN) graph using the FindNeighbors function (dims = 1:30, k.param = 20). Subsequently, unsupervised clustering was performed on the k-NN graph using the Leiden algorithm via the FindClusters function with a resolution of 0.5. For two-dimensional visualization, the Uniform Manifold Approximation and Projection (UMAP) algorithm was applied using the RunUMAP function (dims = 1:30, min.dist = 0.3, spread = 1.0). The cells were then visualized in UMAP space, with color-coding according to tissue origin, treatment condition, and cell type.

Cell Type Annotation

Preliminary cell types were assigned using SingleR (v2.0.0; MouseRNAseqData reference) and manually curated via scMayoMap. Initial annotation identified 14 broad cell classes based on canonical lineage markers. Within the DRG neuronal population, neurons were further subdivided into general neurons and nociceptor-enriched neurons based on nociceptive marker expression (Calca, Trpv1). Within the myeloid lineage, dendritic cells were subdivided into classical dendritic cells and monocyte-derived dendritic cells based on differential expression of monocytic markers (Ly6c2, Ccr2). Low-abundance clusters that could not be confidently assigned to major lineages were grouped as “Other”. The final annotation yielded 18 displayed categories representing major cell types and biologically relevant subtypes. Annotations were validated via cluster-specific marker genes (FindAllMarkers: Wilcoxon test, P < 0.05, |log₂ fold change (FC)|> 0.5) and visualized using dot plots reflecting expression proportion (size) and average level (color).

Differential Expression Analysis

Differential expression analysis between Control and Pain groups within the DRG neuronal subpopulation to identify differentially expressed genes (DEGs) was performed using FindMarkers (Wilcoxon test). DEGs (Benjamini-Hochberg P < 0.05, |log₂FC|>0.5) were visualized via volcano plots. To ensure robustness, genes with mean normalized expression below 0.01 in both groups were excluded, and log₂ fold changes were calculated using a pseudocount of 0.01. Additionally, to prevent artifacts from sparse single-cell data, genes were required to be expressed (counts > 0) in at least 10% of cells in either group.

Gene Set Activity Scoring

Metabolic pathway activities (glycolysis mmu00010, oxidative phosphorylation mmu00190 from Kyoto Encyclopedia of Genes and Genomes [KEGG]) were scored using AUCell (v1.20.0; AUCell_buildRankings, AUCell_calcAUC) and visualized via heatmaps. HK2 expression was plotted as grouped bar charts (mean ± standard error of the mean [SEM]) to corroborate metabolic scores. Pain-related gene signatures in DRG neurons (eg, Trpv1, Trpa1, Scn9a, Calca) were scored using UCell (v2.2.0; ScoreSignatures_UCell).^10,21 Scores were compared (Mann–Whitney U-test) and visualized via violin plots (median with interquartile range [IQR]). To address pseudoreplication concerns, pathway activity scores were additionally summarized at the biological-sample level by computing per-sample mean AUCell scores; sample-level statistics are provided in Supplementary Table S3 and Supplementary Figure S2. Given the small sample sizes (n = 2 vs n = 2 for SC; n = 1 vs n = 2 for DRG), sample-level comparisons did not reach statistical significance, and cell-level analyses are retained as exploratory summaries with consistency across replicates examined as a supplementary criterion.

Pseudotime Trajectory Analysis

SN macrophage trajectory inference utilized Monocle3 (v1.3.1) via preprocess_cds (num_dim=30), reduce_dimension (method=“UMAP”), and learn_graph. Resident-like macrophages (Mrc1⁺) were designated as the root based on prior biological knowledge of their homeostatic identity in uninjured nerve tissue. Pseudotime (0–1.0) was calculated (order_cells). Four macrophage states (Ccl2⁺, proinflammatory, resident-like, transitional) and pseudotime-dependent gene expression dynamics were visualized via UMAP, scatter, and line plots. As the SN dataset represents a single post-injury time point without multi-timepoint sampling, RNA velocity, or lineage tracing data, the resulting trajectory reflects an inferred transcriptional ordering consistent with a resident-to-inflammatory axis, rather than a temporally demonstrated cellular transition.

Weighted Gene Co-Expression Network Analysis

Weighted gene co-expression network analysis (WGCNA, v1.72) was applied to the DRG neuronal matrix. After HVG filtering, a scale-free network (R² > 0.85) was constructed (pickSoftThreshold, blockwiseModules: min.module.size=30, mergeCutHeight=0.25). Pearson correlations between module eigengenes and pain phenotypes were analyzed. Genes from pain-correlated modules (adjusted P < 0.05) underwent Gene Ontology (GO) biological process enrichment via clusterProfiler (v4.6.0), visualized as dot plots.

Transcription Factor Activity Inference

Transcription factor (TF) activities were inferred using decoupleR (v2.4.0) via VIPER and DoRothEA regulons (confidence A-C). DRG neuronal scores (Control vs Pain) were compared (Wilcoxon test, P < 0.001).

Cell-Cell Communication Analysis

CellChat (v1.6.1) was used to analyze cross-tissue communication (computeCommunProb: population.size=TRUE; filterCommunication: min.cells=10), prioritizing SN macrophage-to-DRG neuron signaling.²² Fold changes (Pain/Control) for 12 predicted ligand-receptor pairs were plotted. Ccl2 (SN macrophages) and Ccr2 (DRG neurons) expressions were evaluated using histograms and scatter plots (Mann–Whitney U-test). Because the SN dataset contains Pain-group samples only, SN macrophages contributed exclusively as Pain-condition sender cells in the CellChat analysis. The Control-condition communication scores are derived from the DRG and SC datasets and reflect predicted signaling in the absence of nerve injury; comparisons involving SN macrophages should be interpreted in light of this asymmetric group composition. To assess the robustness of the CCL2–CCR2 interaction, a permutation-based specificity analysis was performed by comparing the observed CCL2–CCR2 communication score against a null distribution generated from 10,000 randomly sampled ligand–receptor pairs from the CellChatDB. Biological replicate consistency of CCL2 and CCR2 expression was assessed by quantifying the proportion of CCL2⁺ SN macrophages across three Pain replicates and CCR2⁺ DRG neurons across available Control and Pain samples (Supplementary Figure S3).

Animal Experiments

All animal experiments were approved by the Animal Research Ethics Committee of Guangxi University of Chinese Medicine (Approval No. DW20240319-057). All experimental procedures were conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals: Eighth Edition (National Academies Press), the American Veterinary Medical Association (AVMA) Guidelines, the principles of the Basel Declaration, and the ethical guidelines by the International Council for Laboratory Animal Science (ICLAS). Furthermore, the study design and reporting strictly adhered to the ARRIVE guidelines 2.0.

Adult male C57BL/6J mice (8–10 weeks of age, 20–25 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). All animals were housed in a specific pathogen-free (SPF) facility maintained at 22–24°C and 45–55% humidity, under a 12 h light/dark cycle with free access to food and water. Sample size was determined a priori using G*Power 3.1. Assuming a large effect size (Cohen’s f = 0.40), with α = 0.05, power = 0.80, and k = 3 groups (one-way ANOVA at day 14), a minimum of 7 animals per group was required. Ten animals per group were enrolled to accommodate potential attrition. Animals were allocated to three groups using a randomized block design based on body weight at the time of surgery: Sham group (sham surgery + vehicle, n = 10), SNI group (SNI surgery + vehicle, n = 10), and RS504393 group (SNI surgery + RS504393, n = 10). All animals in each group were used for behavioral assessment. A subset of five animals per group was randomly selected for histological analyses (HE staining and immunofluorescence). Behavioral testing was performed by an investigator blinded to group allocation. Histological image acquisition and quantification were likewise conducted by a blinded investigator. Exclusion criteria were defined a priori as failure to develop mechanical allodynia (withdrawal threshold > 4 g at day 3 post-SNI). No animals were excluded during the study, and no attrition occurred beyond the planned endpoints.

The SNI model was established according to a reported method.⁵ Under ketamine/xylazine anesthesia (100 mg/kg ketamine and 15 mg/kg xylazine, intraperitoneally), the left sciatic nerve branches were exposed. The common peroneal and tibial nerves were ligated (8–0 nylon) and axotomized (2–4 mm resected), while the sural nerve was preserved. Sham mice underwent identical exposure without nerve injury. To reduce animal suffering, mice were placed on a heating pad during the surgery and recovery period until fully awake, and post-operative health was monitored daily. In accordance with NC3Rs guidelines, humane endpoints were defined as rapid weight loss (>20% of baseline), severe lethargy, or an inability to access food and water. Fortunately, no animals reached these endpoints prior to the scheduled sacrifice.

RS504393 (Tocris, Cat. No. 2391), a selective CCR2 antagonist, was administered subcutaneously at a dose of 3 mg/kg (dissolved in 5% dimethyl sulfoxide (DMSO) in phosphate-buffered saline (PBS)). The 3 mg/kg subcutaneous dose was adopted from Sadler et al,²³ which demonstrated effective CCR2 blockade and pain behavior reversal at this dose in rodents. Dosing began on day 3 post-SNI, when mechanical allodynia was confirmed, and continued daily until day 14 to cover the early maintenance phase of neuropathic pain following SNI. On day 3, behavioral assessment was performed prior to the first drug administration to confirm successful SNI induction; on days 7 and 14, behavioral assessments were conducted 2 hours after RS504393 or vehicle administration. Controls received vehicle equivalents. On day 14, mice were deeply anesthetized via an intraperitoneal injection of sodium pentobarbital (200 mg/kg) in accordance with the AVMA Guidelines for the Euthanasia of Animals, and subsequently transcardially perfused with cold 0.9% saline and 4% paraformaldehyde (PFA).

Behavioral Assessment

Pain behaviors were tested at baseline and on days 3, 7, and 14 post-surgery. Mice were habituated to testing chambers for 30 min. Mechanical allodynia was evaluated using von Frey filaments (Stoelting, 58011) applied perpendicularly to the left hindpaw for 8 s using the up-down method. The median of five measurements represented the withdrawal threshold (g). Thermal hyperalgesia was measured using a radiant heat apparatus (IITC, Model 390). The latency from heat onset to paw withdrawal was recorded (3 measurements/paw, 5 min intervals, averaged). A 30 s cutoff prevented tissue damage.

Hematoxylin and Eosin (HE) Staining

Post-perfusion, the L4–L6 spinal cord segments were post-fixed in 4% PFA for 24 h, dehydrated, and paraffin-embedded. Serial transverse sections (4 μm) were mounted, deparaffinized, and stained with hematoxylin (Solarbio, G1140) for 8 min. After 1% HCl-ethanol differentiation and bluing, sections were counterstained with eosin (Solarbio, G1100) for 2 min. Slides were dehydrated, cleared, mounted, and imaged via light microscopy (Olympus, BX53; scale bar: 50 μm). Inflammatory cell density in the spinal dorsal horn was quantified using ImageJ by counting cells with condensed, irregular, deeply hematoxylin-stained nuclei in three non-overlapping fields per section (field area = 0.1 mm²); the mean count per animal was used for statistical comparison.

Immunofluorescence Double-Labeling

Lumbar spinal cord cryosections (20 μm) were washed in 0.01 M PBS and blocked (5% normal donkey serum [Solarbio, SL050], 0.3% Triton X-100 [Sigma, T8787]) for 1 h at room temperature. Sections were incubated overnight at 4°C with primary antibodies: rabbit anti-CCR2 (Abcam, ab273050, 1:200) and mouse anti-IBA1 (Wako, 019–19,741, 1:500). After PBS washes, sections were incubated for 1 h with Alexa Fluor 488-conjugated donkey anti-rabbit (Thermo Fisher, A21206, 1:500) and Alexa Fluor 594-conjugated donkey anti-mouse (Thermo Fisher, A21203, 1:500) secondary antibodies. Nuclei were stained with DAPI (Sigma, D9542, 1 μg/mL). Images were acquired via fluorescence microscopy (Olympus, BX53; scale bar: 100 μm) and quantified using ImageJ. Mean fluorescence intensity of CCR2, IBA1, and CCR2/IBA1 co-localization was measured separately in three non-overlapping fields per section in the spinal dorsal horn; the mean value per animal was used for statistical comparison.

Statistical Analysis

For animal behavioral and histological data, results are expressed as mean ± SEM. Group comparisons across multiple time points were performed using two-way repeated measures analysis of variance (ANOVA) followed by Bonferroni post hoc correction for multiple comparisons, using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). For bioinformatic analyses, all statistical procedures were conducted within the R (v4.3.1) and Python (v3.9, scanpy v1.9.3) environments. Inter-group differences were evaluated using the Wilcoxon rank-sum test. For multiple comparisons, P values were adjusted using the Benjamini–Hochberg method. Pearson correlation coefficients were calculated to assess the correlations between modules and phenotypes. All data visualizations were generated using ggplot2 (v3.4.2) and the built-in plotting functions within scanpy. Statistical significance was defined as P < 0.05 for all analyses. Single-cell comparisons treat cells as observations; consistency across biological replicates was examined to support interpretation.

Results

Macrophages Dominate the Cellular Landscape of the Sciatic Nerve Following Injury

UMAP visualization showed distinct separation of cell populations among the SN, DRG, and SC tissues, as well as between the Pain and Control groups (Figure 1A). In the DRG, satellite glial cells (SGCs) were the predominant population (66.2%). In the SC, astrocytes (33.8%), microglia (31.0%), and oligodendrocytes (25.6%) were the major cell types. The SN displayed the most diverse cellular composition, predominantly comprising macrophages (40.5% of 10,987 total cells), fibroblasts (18.4%), and Schwann cells (15.1%) (Figure 1B and C). Cell-type annotations aligned with the expression of specific marker genes: Mpz and Mbp in Schwann cells; Cd68 and Lyz2 in macrophages; Col1a1 in fibroblasts; Flt1 in endothelial cells; Cx3cr1 and Ctss in microglia; Gfap and S100b in astrocytes; Mog in oligodendrocytes; Snap25 in DRG neurons; and Calca in nociceptor-enriched neurons (Figure 2).

Figure 1.

Multi-tissue single-cell transcriptomic atlas and cellular landscape of the sciatic nerve following peripheral nerve injury. (A) UMAP plot of 37,642 single cells from three tissues (SN, DRG, and SC), colored by tissue origin (left) and treatment condition (right). (B) Heatmap of proportional cell-type composition (%) across the three tissues. (C) Horizontal bar chart of cell subpopulation proportions within the SN (n = 10,987 cells).

Abbreviations: SN, sciatic nerve; DRG, dorsal root ganglion; SC, spinal cord; UMAP, Uniform Manifold Approximation and Projection; SGCs, satellite glial cells.

Figure 2.

Marker gene expression profiles for cell-type annotations across the multi-tissue single-cell transcriptomic atlas. Dot plot of representative marker gene expression across annotated cell types (dot size: proportion of expressing cells; color intensity: mean expression level).

Abbreviations: SN, sciatic nerve; DRG, dorsal root ganglion; SC, spinal cord; NK cells, natural killer cells; DC, dendritic cells.

Injury Induces Transcriptional Reprogramming in DRG Neurons and Metabolic Shifts in Local Immune and Glial Cells

In DRG neurons from the Pain group, 3 genes (Neat1, Gm242418, Nfib) were upregulated and 71 genes (including Nptx2, Dnajb1, Trpc3) were downregulated compared to the Control group (adjusted P < 0.05, |log₂FC| > 0.5; Figure 3A). Glycolytic and oxidative phosphorylation (OXPHOS) pathway activities were elevated in SN macrophages, SN Schwann cells, and SC microglia in the Pain group. For SN macrophages, metabolic scores are available for the Pain group only, as the SN dataset lacks a matched naïve Control condition. To address this limitation, we re-analyzed an independent longitudinal SN dataset (GSE198582) containing both naïve and day-3 post-injury samples;²⁴ SN macrophages at day 3 post-injury showed significantly elevated glycolytic activity compared to naïve controls (Mann–Whitney U-test, P = 2.56 × 10⁻²⁵³; Supplementary Figure S1 and Supplementary Table S2), supporting the descriptive metabolic scores observed in the primary dataset. Conversely, DRG neurons in the Pain group showed reduced glycolysis (from 0.259 to 0.188) and OXPHOS (from 0.344 to 0.244) activity scores (Figure 3B). Nociceptive gene set activity scores in DRG neurons displayed a subset of high-activity cells in the Pain group, though the overall between-group difference was not statistically significant (P = 0.145; Figure 3C). Given the single Control biological replicate in the DRG dataset, these comparisons are exploratory and should be interpreted with caution. HK2 expression was highest in SC microglia, moderate in SN macrophages (predominantly in the Pain group), and negligible in DRG neurons (Figure 3D).

Figure 3.

Transcriptional profiles in DRG neurons and metabolic pathway activities in immune and glial cells following peripheral nerve injury. (A) Volcano plot of differentially expressed genes (DEGs) in DRG neurons between Pain and Control groups (red: upregulated, n = 3; blue: downregulated, n = 71; gray: not significant, n = 1696; Wilcoxon rank-sum test, adjusted P < 0.05, |log₂FC| > 0.5). Genes expressed in <10% of cells in both groups were excluded to prevent fold-change artifacts from sparse expression. Statistical comparisons are performed at the cell level (Control n = 1 biological replicate, Pain n = 2 biological replicates). (B) Heatmap of glycolytic and OXPHOS pathway activity scores across major cell populations (numerical values: mean AUCell scores; color intensity: activity level; SN macrophage scores are available for the Pain condition only, as the SN dataset lacks a matched naïve Control). (C) Violin plot of nociceptive gene set activity scores in DRG neurons (Control, n = 69 cells from 1 biological replicate; Pain, n = 229 cells from 2 biological replicates; P = 0.145, Mann–Whitney U-test; median with interquartile range; comparison is exploratory due to limited biological replication). (D) Bar chart of Hk2 expression levels across distinct cell populations (mean ± SEM; SN macrophage values reflect the Pain condition only).

Abbreviations: DRG, dorsal root ganglion; SN, sciatic nerve; SC, spinal cord; Mac, macrophages; Sat Glia, satellite glial cells; OXPHOS, oxidative phosphorylation; HK2, hexokinase 2; FC, fold change; SEM, standard error of the mean; AUCell, area under the curve cell scoring.

Sciatic Nerve Macrophages Show an Inferred Transcriptional Ordering Toward Proinflammatory Subtypes After Injury

SN macrophages formed a continuous inferred trajectory in UMAP space, with pseudotime values increasing from the main cell cluster to an isolated upper-left cluster (Figure 4A). Four macrophage subclusters were identified: Ccl2⁺, proinflammatory, resident-like, and transitional. Proinflammatory and resident-like macrophages populated the main early-pseudotime cluster, Ccl2⁺ macrophages localized to the high-pseudotime terminal cluster, and transitional macrophages distributed between them (Figure 4B). Ccl2 expression increased along the pseudotime axis, peaking in the terminal branch. Spp1 maintained high expression from low-to-moderate pseudotime through the terminal stage. Ccl3 showed localized enrichment in the high-pseudotime terminal region (Figure 4C). The resident-like marker genes Cx3cr1 and Mrc1 were highly expressed early and downregulated along the pseudotime axis. Conversely, the proinflammatory markers Ccl2 and Il1b were upregulated in the mid-to-late stages, with Ccl2 peaking at the trajectory terminus (Figure 4D). This inferred ordering is consistent with a transcriptional shift toward inflammatory states but does not establish that resident-like macrophages undergo direct conversion into Ccl2⁺ cells.

Figure 4.

Pseudotime trajectory and subpopulation dynamics of sciatic nerve macrophages following peripheral nerve injury. (A) UMAP plot of the SN macrophage pseudotime trajectory (color gradient: pseudotime values 0 to 1.0). (B) UMAP plot of SN macrophage subpopulations based on subtype annotations. (C) Scatter plots of signature gene (Ccl2, Spp1, and Ccl3) expression along the pseudotime axis (color intensity: expression level). (D) Line plots of mean expression dynamics of representative marker genes (Ccl2, Il1b, Cx3cr1, and Mrc1) along the pseudotime trajectory.

Abbreviations: SN, sciatic nerve; UMAP, Uniform Manifold Approximation and Projection; Ccl2⁺, Ccl2-expressing; Il1b⁺, Il1b-expressing; Mrc1⁺, Mrc1-expressing; Spp1, secreted phosphoprotein 1; Cx3cr1, CX3C chemokine receptor 1.

Upregulated ATF3 and Jun Activities Correlate with Inflammatory Gene Modules in Injured DRG Neurons

The M7 module exhibited the strongest positive correlation with the pain-related phenotype (Pearson r = 0.17, P < 0.001), while the M5 module showed the largest absolute and significant negative correlation (P < 0.001) (Figure 5A). Although statistically significant, the modest magnitude of these correlations reflects the complexity of pain-related transcriptional programs and the limited biological replication in the DRG dataset (Control n = 1). The pain-associated modules were primarily enriched in GO biological processes including cell junction disassembly, microglial cell activation (highest GeneRatio values), negative regulation of T-cell-mediated cytotoxicity, synapse pruning, and antigen presentation (Figure 5B). These immune-related terms reflect neuronal expression of genes canonically associated with glial or immune functions, likely representing neuron-intrinsic inflammatory programs or neuron-glia signaling pathways activated by injury. Activity scores for the transcription factors ATF3, Jun, STAT3, and NF-κB were higher in the Pain group compared to the Control group (all P < 0.001). STAT3 exhibited the highest absolute activity scores in both groups. ATF3 activity increased from near zero in the Control group to elevated levels in the Pain group (Figure 5C). In UMAP space, DRG neurons with high ATF3 activity (scores ~0.0 to 0.45) clustered within specific upper and central subpopulations (Figure 5D).

Figure 5.

Transcription factor activities and gene co-expression modules in DRG neurons following peripheral nerve injury. (A) Bar chart of correlation coefficients between gene co-expression modules and the pain phenotype. (B) Dot plot of GO biological process enrichment for pain-associated modules (X-axis: GeneRatio; Y-axis: GO terms; dot size: gene count; color intensity: adjusted P value). (C) Grouped bar chart comparing the activity scores of four transcription factors (ATF3, Jun, STAT3, and NF-κB) in DRG neurons between Control and Pain groups (***, P < 0.001 at cell level; Control n = 1 biological replicate, Pain n = 2 biological replicates; findings are exploratory due to limited biological replication). (D) UMAP plot of the spatial distribution of ATF3 activity in DRG neurons (color gradient: activity values 0.0 to 0.45). ***, P < 0.001; *, P < 0.05.

Abbreviations: DRG, dorsal root ganglion; GO, Gene Ontology; UMAP, Uniform Manifold Approximation and Projection; TF, transcription factor; ATF3, activating transcription factor 3; Jun, Jun proto-oncogene; STAT3, signal transducer and activator of transcription 3; NF-κB, nuclear factor kappa B.

Predicted Macrophage-to-DRG Neuron Communication via the CCL2-CCR2 Signaling Pathway is Elevated Following Peripheral Injury

Predicted communication scores from SN macrophages to DRG neurons for proinflammatory ligand-receptor pairs (CCL2–CCR2, CCL3–CCR1, CSF1–CSF1R, and IL1B–IL1R1) were higher in the Pain group than in the Control group, with CCL2–CCR2 showing the largest increase. Conversely, the SPP1–ITGAV score was higher in the Control group and decreased in the Pain group (Figure 6A). The CCL2–CCR2 pair exhibited the highest Pain/Control FC. Proinflammatory pairs (CCL3–CCR1, CSF1–CSF1R, CCL3–CCR5, IL1B–IL1R1) showed FC > 1; TGFB1–TGFBR1 and TNF–TNFRSF1A had modest FC > 1; and SPP1–ITGAV, SPP1–CD44, CXCL10–CXCR3, NGF–NTRK1, and BDNF–NTRK2 had FC < 1 (Figure 6B). In the Pain group, approximately half of the SN macrophages expressed Ccl2, displaying a right-skewed distribution (Figure 6C). In DRG neurons, the proportion of CCR2⁺ cells showed a non-significant trend toward elevation in the Pain group compared to the Control group (P = 0.057; Figure 6D), a result that should be interpreted with caution given the single Control biological replicate. Across the three Pain group biological replicates, the proportion of Ccl2⁺ SN macrophages ranged from 38.8% to 55.2% (Figure 7A). In DRG neurons, the proportion of Ccr2⁺ cells was 1.4% in the single Control replicate, compared to 6.2% and 3.4% in the two Pain replicates (Figure 7B). The single Control replicate precludes estimation of Control variability or statistical comparison. The relatively low proportion of CCR2-expressing DRG neurons and the non-significant between-group comparison (P = 0.057) indicate that neuronal CCR2 expression in this dataset remains tentative and requires independent validation through spatial transcriptomics or protein-level confirmation. To assess the specificity of the CCL2–CCR2 prediction, a permutation-based analysis confirmed that the CCL2–CCR2 communication score ranked at the 96.15th percentile relative to 10,000 randomly sampled ligand–receptor pairs (permutation P = 0.039), indicating that this interaction exceeds chance expectation. CCL2 was expressed in 48.9% of SN macrophages, ranking in the top 2.7% of all expressed genes, and CCR2 showed consistent elevation across Pain DRG replicates, further supporting the robustness of this prediction (Supplementary Figure S3).

Figure 6.

Predicted cell-cell communication from SN macrophages to DRG neurons via the CCL2–CCR2 signaling pathway following peripheral nerve injury. (A) Grouped bar chart comparing representative ligand–receptor communication scores from SN macrophages to DRG neurons (Control vs Pain). Because the SN dataset contains Pain-group samples only, SN macrophages contributed as Pain-condition sender cells; the Control-condition scores reflect the predicted signaling baseline derived from DRG and SC Control samples in the integrated analysis. (B) Horizontal bar chart of fold changes (Pain/Control) of ligand–receptor communication pairs (dashed reference line: FC = 1). (C) Histogram of single-cell Ccl2 expression in Pain group SN macrophages (dashed line: mean = 1.77; parentheses: number and proportion of Ccl2⁺ cells). (D) Scatter plot of Ccr2 expression in DRG neurons between Control and Pain groups (Mann–Whitney U-test at cell level; parentheses: proportion of Ccr2⁺ cells); P = 0.057; Control n = 1 biological replicate, Pain n = 2 biological replicates; this comparison is exploratory due to limited biological replication.

Abbreviations: SN, sciatic nerve; DRG, dorsal root ganglion; CCL2, C-C motif chemokine ligand 2; CCR2, C-C motif chemokine receptor 2; CSF1, colony-stimulating factor 1; IL1B, interleukin-1 beta; TGFB1, transforming growth factor beta 1; TNF, tumor necrosis factor; SPP1, secreted phosphoprotein 1; BDNF, brain-derived neurotrophic factor; NGF, nerve growth factor; FC, fold change.

Figure 7.

Biological replicate validation of Ccl2 expression in SN macrophages and Ccr2 expression in DRG neurons. (A) Bar chart of mean Ccl2 expression and the proportion of Ccl2⁺ SN macrophages across three biological replicates (rep1–3) in the Pain group. (B) Bar chart of Ccr2 expression and the proportion of Ccr2⁺ DRG neurons across biological replicates (Control, n = 1 shown as individual value; Pain, n = 2 shown as mean ± SEM). Control variability cannot be estimated from a single replicate.

Abbreviations: CCL2, C-C motif chemokine ligand 2; CCR2, C-C motif chemokine receptor 2; SEM, standard error of the mean; rep, biological replicate.

CCR2 Antagonism Alleviates Neuropathic Pain and Reduces Spinal CCR2⁺/IBA1⁺ Cell Activation in SNI Mice

Baseline (BL) mechanical withdrawal thresholds and thermal withdrawal latencies were comparable across all groups. From day 3 through day 14 post-surgery, the SNI group exhibited significantly decreased mechanical withdrawal thresholds and thermal withdrawal latencies compared to the Sham group. RS504393 administration did not produce significant behavioral improvement at day 3, but significantly restored both mechanical withdrawal thresholds and thermal withdrawal latencies at days 7 and 14 post-surgery compared to the SNI group (Figure 8A and B). Spinal cords from the SNI group displayed disrupted cellular arrangement and significantly increased inflammatory cell infiltration compared to the Sham group; quantification confirmed elevated inflammatory cell density in the SNI group relative to Sham, which was significantly attenuated by RS504393 treatment (one-way ANOVA with Bonferroni correction, all P < 0.001; Figure 8C and E). Quantification of fluorescence intensity in the spinal dorsal horn revealed that CCR2, IBA1, and CCR2/IBA1 co-localization signals were all significantly elevated in the SNI group compared to Sham and significantly reduced following RS504393 treatment (one-way ANOVA with Bonferroni correction, all P < 0.001; n = 5 per group; Figure 8D and F). IBA1 labels both resident microglia and infiltrating monocyte-derived macrophages; therefore, the CCR2⁺/IBA1⁺ population represents a mixed myeloid cell population rather than microglia alone.

Figure 8.

Effects of RS504393 on neuropathic pain and CCR2-mediated microglial activation in SNI mice. (A) Mechanical withdrawal thresholds (g) at baseline (BL) and days 3, 7, and 14 post-surgery (n = 10 per group; von Frey test). (B) Thermal withdrawal latencies (s) at corresponding time points (n = 10 per group; Hargreaves method). (C) Representative hematoxylin and eosin-stained spinal cord sections (n = 5 per group; scale bar = 50 μm). (D) Representative immunofluorescence images of CCR2 (green), IBA1 (red), and their co-localization in the spinal dorsal horn, with DAPI counterstain (n = 5 per group; scale bar = 100 μm); IBA1 labels both resident microglia and infiltrating monocyte-derived macrophages. (E) Quantification of inflammatory cell density in the spinal dorsal horn (n = 5 per group; individual animal values shown; mean ± SEM; one-way ANOVA with Bonferroni correction; *** P < 0.001). (F) Quantification of CCR2, IBA1, and CCR2/IBA1 co-localization mean fluorescence intensity in the spinal dorsal horn (n = 5 per group; individual animal values shown; mean ± SEM; one-way ANOVA with Bonferroni correction; *** P < 0.001). For (A and B), two-way repeated measures ANOVA with Bonferroni post hoc correction was used; *** P < 0.001 compared to Sham; ### P < 0.001 compared to SNI.

Abbreviation: ns, not significant.

Discussion

Single-cell transcriptomic analysis reveals that macrophages constitute 40.5% of the cellular population in the sciatic nerve following injury, indicating substantial local immune infiltration.⁷ Pseudotime trajectory analysis reveals a transcriptional ordering consistent with a shift from an early Mrc1⁺ resident-like phenotype toward proinflammatory terminal subtypes, though, given the single post-injury time point of the dataset, this represents a computational inference rather than a demonstrated temporal transition. The pronounced enrichment of Ccl2 and Spp1 at the trajectory terminus coincides with the formation of a localized inflammatory state. Proinflammatory macrophage polarization exacerbates tissue damage through the release of inflammasome-associated factors, whereas targeted inhibition of this process is associated with enhanced nerve repair.^15,16 This inferred macrophage state transition may represent an upstream event associated with subsequent neuroimmune alterations, though lineage tracing would be required to demonstrate this relationship.

Within the dorsal root ganglion, sensory neurons display a highly selective transcriptional response characterized by the marked upregulation of transcription factors including ATF3, Jun, and STAT3. ATF3 functions as a classic marker of nerve stress and plays a central role in neuropathic pain progression by orchestrating inflammatory signaling cascades between neurons and local macrophages.²⁵ While ATF3 activation promotes neuronal survival and axonal regeneration,²⁶ its persistent overactivation observed in specific DRG subpopulations is strongly associated with the activation of pain-related gene modules encompassing neuro-glial interactions. This indicates a complex intersection between regenerative responses and nociceptive hypersensitivity in primary sensory neurons.

Cross-tissue metabolic evaluations demonstrate that peripherally infiltrating macrophages and centrally activated microglia exhibit concurrent elevations in glycolytic and oxidative phosphorylation activities. The high expression of key glycolytic enzymes, notably HK2, in microglia and macrophages underscores the essential role of glucose metabolism in driving pain-related neuroinflammation.¹⁷ This elevated metabolic profile meets the immense energy demands of activated immune cells to synthesize and secrete proinflammatory cytokines.²⁷ Conversely, DRG neurons display a marked reduction in overall metabolic activity. Impaired neuronal mitochondrial function and subsequent oxidative stress are recognized contributors to neuropathic pain.²⁸ Organisms may counteract this neuronal metabolic deficit through compensatory mechanisms, such as glia-to-neuron mitochondrial transfer, which helps restore energy homeostasis and protects neurons from progressive peripheral neuropathy.²⁹ The independent validation using GSE198582 and the cross-dataset consistency of sample-level metabolic scores provide convergent support for these metabolic patterns, while acknowledging that definitive sample-level statistical confirmation awaits studies with larger biological replication. Recent integrative analyses of peripheral nerve vascular and metabolic microenvironments further corroborate the importance of cell-type-specific metabolic profiling in nerve injury contexts.^30,31

CellChat analysis identifies compatible CCL2–CCR2 co-expression patterns across SN macrophages and DRG neurons, with predicted communication scores elevated in the Pain group; these predictions reflect co-expression patterns across independently generated datasets and do not demonstrate that CCL2 released in the sciatic nerve physically reaches and activates DRG neurons in vivo. These computational findings are consistent with prior evidence that peripheral CCL2 promotes immune cell recruitment and modulates sensory pathway excitability. Pharmacological targeting of chemokine axes has demonstrated considerable analgesic potential in preclinical models.^32–34 Subcutaneous administration of the CCR2 antagonist RS504393 ameliorated neuropathic pain behaviors and reduced CCR2⁺/IBA1⁺ myeloid cell activation in the spinal dorsal horn, providing indirect evidence that this peripheral signaling axis is associated with central effects. These in vivo findings support a CCR2-dependent neuroimmune mechanism in neuropathic pain maintenance. However, systemic CCR2 antagonism affects circulating monocytes, peripheral macrophages, spinal infiltrating cells, and potentially resident microglia, preventing attribution of the analgesic effect to a specific cellular pathway. Furthermore, IBA1 co-localization does not distinguish resident microglia from infiltrating monocyte-derived cells without lineage-specific markers such as TMEM119 or P2RY12. The pharmacological validation therefore corroborates CCR2 as a functionally relevant therapeutic target while leaving the precise cellular mechanism as a hypothesis requiring orthogonal validation. To address this, we plan to perform RNAscope co-localization with nociceptor markers (eg, CGRP) on DRG sections in future work to confirm CCR2 protein expression in sensory neurons. Furthermore, while the scRNA-seq data suggest CCR2 expression in a small subset of DRG neurons, the low proportion of CCR2⁺ neurons (Figure 7B) and non-significant group-level comparison (P = 0.057) highlight that neuronal CCR2 remains a candidate mechanism awaiting confirmation.

This study integrated independently generated public datasets with unequal biological replication across tissues (DRG Control n = 1, Pain n = 2; SC Control n = 2, Pain n = 2). The single Control replicate in the DRG dataset limits statistical inference for group-level comparisons; DRG-related findings (eg, CCR2 expression P = 0.057) should be interpreted as hypothesis-generating observations. In particular, the relatively low proportion of CCR2-expressing DRG neurons (3.4–6.2% in Pain replicates vs 1.4% in Control) and the lack of protein-level or spatial validation leave neuronal CCR2 expression as a candidate mechanism rather than an established finding. To address this computational limitation, this study implemented downstream in vivo pharmacological validation to provide robust empirical confirmation of the bioinformatics hypotheses. Additionally, discrepancies in sensory neuron subtyping algorithms across current analysis pipelines may introduce variability in cellular annotation. Furthermore, translating the rodent-derived mechanisms identified in this study to human clinical applications faces an inherent species gap. Future research should prioritize establishing unified molecular classification standards and constructing high-resolution human transcriptomic atlases to facilitate precise clinical translation.

Integration of multi-tissue single-cell transcriptomics and in vivo validation delineates a systemic neuropathic pain mechanism driven by macrophage state transitions, selective neuronal transcriptional reprogramming, and divergent metabolic adaptations. The convergent computational and pharmacological evidence positions the CCL2–CCR2 signaling pathway as a candidate mechanism and therapeutic hypothesis warranting further experimental investigation for the development of non-opioid analgesic strategies.

Funding Statement

This work was supported by the National Natural Science Foundation of China (82505721, 82260976); Guangxi Clinical Research Center for Traditional Chinese Medicine Rehabilitation Construction Project, the seventh batch of Guangxi Clinical Medical Research Centers (Guikefa [2025] No. 293); Traditional Chinese Medicine Rehabilitation Center Construction Unit Project of the National Administration of Traditional Chinese Medicine (Guozhongyi Yizheng Han [2023] No. 255); Guangxi Science and Technology Program Project (Guike LT2600640073); Natural Science Foundation of Guangxi University of Chinese Medicine (2024QN028, 2025QN023).

Abbreviations

ANOVA, analysis of variance; AVMA, American Veterinary Medical Association; BL, baseline; DEGs, differentially expressed genes; DMSO, dimethyl sulfoxide; DRG, dorsal root ganglion; FC, fold change; GEO, Gene Expression Omnibus; GO, Gene Ontology; HE, hematoxylin and eosin; HVGs, highly variable genes; IQR, interquartile range; KEGG, Kyoto Encyclopedia of Genes and Genomes; k-NN, k-nearest-neighbor; NF-κB, nuclear factor kappa B; NCBI, National Center for Biotechnology Information; NP, neuropathic pain; OXPHOS, oxidative phosphorylation; PBS, phosphate-buffered saline; PCA, principal component analysis; PFA, paraformaldehyde; SC, spinal cord; scRNA-seq, single-cell RNA sequencing; SEM, standard error of the mean; SGCs, satellite glial cells; SN, sciatic nerve; SNI, spared nerve injury; TF, transcription factor; UMAP, Uniform Manifold Approximation and Projection; WGCNA, weighted gene co-expression network analysis.

Data Sharing Statement

The scRNA-seq datasets analyzed in this study are publicly available in the NCBI Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE153762, GSE174430, GSE198582, and GSE208766. All analysis scripts (R code for quality control, batch correction, trajectory analysis, WGCNA, transcription factor inference, and Python code for cell-cell communication analysis) are provided as Supplementary Raw Data. All other original data are included in this article and available from the corresponding author, Liangyuan Tan, upon request.

Ethics Statement

The animal study was approved by the Animal Research Ethics Committee of Guangxi University of Chinese Medicine (Approval No. DW20240319-057). The study was conducted in accordance with the local legislation and institutional requirements.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.