Single-cell and single-nucleus transcriptomics of the dorsal root ganglion in neuropathic pain: cell-state remodeling and translational prospects Authors: Xu-Ran Liao, Huan Zheng, Jiang-Yuan Luo, Bi-Xin Zheng Published: 2026 ABSTRACT Neuropathic pain remains a major unmet clinical challenge, as current therapies provide limited efficacy and poor tolerability. A major obstacle to developing effective analgesics is the pronounced cellular and molecular heterogeneity of the dorsal root ganglion (DRG), which integrates neuronal, glial, immune, and stromal responses to injury. Recent advances in single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq) have enabled cell-type-resolved analyses of these responses and revealed disease-associated cellular states that are obscured by bulk tissue profiling, thereby offering new opportunities to redefine disease mechanisms and therapeutic priorities. This review synthesizes evidence from single-cell studies of peripheral nerve injury, diabetic painful neuropathy, and chemotherapy-induced neuropathy. Although these conditions share common biological processes, including neuronal state remodeling, disrupted glial homeostatic support, context-dependent immune responses, and extracellular matrix reorganization, they do not converge on a single conserved molecular program. Peripheral nerve injury is characterized by neuronal injury and repair states, reactive and metabolic glial remodeling, and neuroimmune recruitment and crosstalk. Diabetic neuropathy is associated with altered sensory coding, impaired satellite glial lipid support, and neurodegenerative remodeling. Chemotherapy-induced neuropathy exhibits substantial agent-specific heterogeneity, including subtype-selective neuronal vulnerability and metalloproteinase-dysregulated satellite glial states following paclitaxel treatment, as well as sex-associated fibrotic remodeling following bortezomib treatment. These cellular states may serve adaptive, maladaptive, or degenerative functions. However, because most available studies rely on single-time-point or repeated cross-sectional sampling, they cannot directly establish temporal progression or causal relationships. We further examine how human DRG atlases bridge experimental models and human disease by determining whether candidate molecular targets and cell states are conserved and by identifying their cellular localization. These datasets also highlight species-specific differences in neuronal organization and non-neuronal transcriptional programs that may limit the direct translation of findings from rodent models. Consequently, the primary translational value of DRG single-cell studies lies in elucidating disease mechanisms and prioritizing candidate therapeutic targets for further investigation. Advancing these discoveries toward clinical application will require complementary evidence from human genetics, human DRG transcriptomic datasets, functional validation studies, and clinical pharmacology. Collectively, these complementary approaches may facilitate the development of mechanism-based and cell-type-informed analgesic strategies. Introduction Neuropathic pain is a chronic condition resulting from lesions or diseases affecting the somatosensory nervous system and clinically manifests as spontaneous pain, hyperalgesia, and mechanical allodynia. With etiologies ranging from diabetic peripheral neuropathy and chemotherapy-induced peripheral neuropathy to trigeminal neuralgia, neuropathic pain poses a substantial global health burden that is compounded by psychiatric comorbidities, such as anxiety and depression [1, 2]. Existing pharmacological management remains limited, with guideline-supported first-line treatments, including gabapentinoids, tricyclic antidepressants, and serotonin–noradrenaline reuptake inhibitors, providing incomplete relief for many patients, whereas opioid use is constrained by safety concerns and inconsistent long-term benefits [3–5]. This therapeutic inadequacy is largely attributable to the translational gap between preclinical models and human pathology, together with an incomplete understanding of the molecular heterogeneity underlying pain chronification. The dorsal root ganglion (DRG), which harbors the somata of pseudounipolar primary sensory neurons, functions as a key first relay station for somatosensory transmission. Owing to its relatively permeable blood–nerve barrier, the DRG is uniquely vulnerable to systemic metabolic insults and neurotoxic agents [6]. Under pathological conditions, the DRG becomes an active site of pain-related remodeling. Changes in ion-channel expression, neuronal excitability, and ectopic discharge can increase afferent input to the spinal dorsal horn and contribute to central sensitization [7–9]. This remodeling extends beyond neurons to satellite glial cells, Schwann cells, immune cells, vascular cells, and stromal populations. Depending on the disease context and stage, these responses may contribute to tissue repair, sensory sensitization, or neurodegeneration [10–12]. Conventional bulk RNA sequencing averages gene expression across whole tissues and can therefore obscure cell-type-specific changes and signals from less abundant populations, including nociceptor subtypes and infiltrating immune cells [13–15]. The advent of single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq) has enabled cell-type-resolved profiling of DRG heterogeneity [16]. Compared with scRNA-seq, snRNA-seq can improve the representation of large sensory-neuron nuclei and permits the analysis of frozen human tissues, although it captures nuclear rather than whole-cell transcriptomes [17]. When combined with spatial transcriptomics and computational approaches, these technologies can identify disease-associated cell states, suggest possible temporal trajectories, and nominate candidate intercellular communication pathways [18]. This review organizes the emerging single-cell evidence around the concept that neuropathic pain involves coordinated cell-state remodeling across the neuronal, glial, immune, and stromal compartments of the DRG. A cell state is defined here as a reproducible transcriptomic configuration within a specified cell population. We first summarize the technological evolution from scRNA-seq to snRNA-seq and spatially informed approaches, and then describe the cellular taxonomy of the DRG across species. We next compare disease-associated remodeling across peripheral nerve injury, diabetic neuropathy, and chemotherapy-induced neuropathy. Finally, we examine the correspondence between experimental findings and emerging human DRG atlases and discuss how these data may support translational interpretation, including target prioritization, mechanistic refinement, biomarker concepts, and analgesic development. This narrative review was informed by searches of PubMed, Web of Science, and Scopus for studies published between January 2018 and January 2026. Search terms included combinations of “dorsal root ganglion,” “neuropathic pain,” “single-cell RNA sequencing,” “single-nucleus RNA sequencing,” “spatial transcriptomics,” “human DRG,” and “translational.” We prioritized studies that applied sc/snRNA-seq, spatial or multi-omic techniques, human DRG datasets, or functional validation relevant to translational pain research. Methodological considerations for single-cell profiling of the DRG Single-cell profiling of the DRG presents several technical challenges that can affect cell recovery, transcript detection, and biological interpretation. These challenges are particularly important because sensory neurons exhibit substantial size variability and are closely associated with satellite glial cells and other non-neuronal cell populations [19, 20]. In a typical workflow, freshly isolated DRG tissue is dissociated into intact cells for scRNA-seq, whereas nuclei are isolated from fresh or frozen tissue for snRNA-seq. Following barcoding, library preparation, and sequencing, the resulting data undergo quality control, normalization, dimensionality reduction, clustering, cell-type annotation, and differential expression analysis [21]. The choice between scRNA-seq and snRNA-seq involves important trade-offs. scRNA-seq captures whole-cell transcripts but requires enzymatic and mechanical dissociation, which may preferentially exclude large or fragile myelinated neurons and induce stress-response genes such as Fos and Jun [17, 22]. By contrast, snRNA-seq avoids extensive tissue dissociation, improves the recovery of large neuronal populations, and can be applied to frozen human tissue [23, 24]. However, snRNA-seq captures only nuclear transcripts and may underrepresent cytoplasmic and locally translated mRNAs, particularly those enriched in distal axonal compartments. Recent efforts to address these trade-offs include a single-soma deep-sequencing strategy that combines laser capture microdissection with Smart-seq2 to generate full-length transcriptomic profiles and detect, on average, more than 9,000 unique genes per human DRG neuron, substantially exceeding the sensitivity of standard snRNA-seq approaches [17, 25, 26]. Both scRNA-seq and snRNA-seq largely lose the spatial organization of the DRG. Spatial transcriptomics preserves anatomical context and localizes neuronal–glial units, immune niches, and disease-associated cellular states [27–31]. Integrating spatial transcriptomic and single-cell datasets therefore enhances the interpretation of cell-type-specific changes, although spatial transcriptomic platforms vary in spatial resolution and transcript detection sensitivity. Cross-study comparisons are also influenced by batch effects arising from differences in sequencing depth, transcript capture efficiency, cell or nucleus recovery, quality-control thresholds, normalization methods, and data-integration strategies. These factors can affect cluster resolution, estimated cell-type proportions, and differential-expression results, particularly when comparing studies that use different pain models, sampling time points, sexes, or species [32–34]. Greater standardization of experimental design, data processing, and cell-type annotation would improve reproducibility and facilitate comparisons across studies. Cellular taxonomy of the DRG The DRG comprises a heterogeneous assembly of neuronal and non-neuronal cell populations that function as an integrated unit to maintain somatosensory homeostasis. Recent scRNA-seq and snRNA-seq studies have significantly refined the cellular taxonomy of this structure, revealing greater diversity than was previously recognized using histological or electrophysiological approaches [13, 32, 35]. Fig. 1 presents a schematic overview of this multicellular landscape and its key marker genes, and Table 1 provides a more detailed comparison of marker profiles, functional roles, and human–mouse correspondence.Fig. 1Major cell populations of the dorsal root ganglion and representative marker genesTable 1Representative DRG cell populations, marker genes, functional roles, and human–mouse correspondence in sc/snRna-seq studiesCell TypePopulation / programKey Markers (Mouse)Key Markers (Human)Main roleHuman–mouse correspondenceReferencesSensory NeuronsPeptidergic nociceptors (PEP)Calca, Tac1, Trpv1, Ntrk1CALCA, TAC1, TRPV1, NTRK1Thermal/chemical nociception; CGRP/SP releaseModerate–high; broader TRPV1 distribution in humans[13, 30, 35, 36]Non-peptidergic nociceptors (NP)Mrgprd, P2rx3, RetMRGPRD (low/diffuse), P2RX3Mechanical nociception and pruriceptionModerate; diffuse MRGPRD expression in humans[13, 35–37]C-low-threshold mechanoreceptors (C-LTMRs)Th, Slc17a8 (Vglut3)TH (low), SLC17A8/VGLUT3 (variable)Affective touch; gentle mechanosensationModerate; one-to-one physiological correspondence incomplete[17, 38, 39]Myelinated LTMR neurons (NF/Aβ)Nefh, Ntrk2, Ntrk3, RetNEFH, NTRK2, NTRK3Low-threshold mechanosensation; touch and vibrationHigh[13, 30, 35]ProprioceptorsPvalb, Ntrk3, Etv1PVALB, NTRK3Proprioception and muscle spindle innervationHigh[13, 30, 35]Aδ-like myelinated nociceptors / HTMRsNefh, Scn10a, Trpv1NEFH, SCN10A, TRPV1 (partial overlap)High-threshold mechano-/thermo-nociceptionModerate; correspondence is supported but not one-to-one[13, 35, 36]Human-enriched PIEZO2-high PEP neuronsNo clear counterpartPIEZO2 (high), CALCAPutative mechanosensory functionLow/absent; no clear murine counterpart[17]Human-enriched NTRK3/DCC nociceptor-related neuronsNo clear counterpartNTRK3, DCC, opioid-related transcriptsHuman-enriched sensory program; function uncertainLow/absent; persists into adulthood in human DRG[40]Satellite Glial CellsHomeostatic SGCsKcnj10, Fabp7, Glul, Gja1KCNJ10, FABP7, GLUL, GJA1Perineuronal ensheathment; K+ buffering; glial homeostasisModerate; broad identity conserved, fine subtype unresolved[18, 41, 42]Spatially biased SGC programsFabp7, Kcnj10Not establishedPreferential association with selected neuronal classesUncertain; supported mainly by mouse spatial data[43]UGT8+ metabolic SGCsUgt8, GalCer biosynthesis genesNot establishedGalCer biosynthesis; lipid support at the neuron–glia interfaceUncertain; defined mainly in rodent DPN[44]TIMP3+ SGCsTimp3TIMP3ECM homeostasis; metalloproteinase restraintModerate; human functional validation is limited[45]Schwann CellsMyelinating Schwann cellsMpz, Mbp, PrxMPZ, MBPAxonal myelination and saltatory conductionHigh[46, 47]Non-myelinating Schwann cellsNcam1, L1cam, Scn7aNCAM1, L1CAMEnsheathment of unmyelinated C-fibersHigh[46, 47]Stress response-related Schwann cells (SRSCs)Npy, Spp1Not establishedInjury-associated remodelingUnknown; described in mouse[48]Repair Schwann cellsJun, Ngf, BdnfJUN, NGF, BDNFInjury-induced repair; trophic supportModerate[46, 49]MacrophagesResident macrophages (sNAMs)Cx3cr1, Csf1r, Aif1, Cd68CX3CR1, CSF1R, AIF1Tissue surveillance; debris clearance; immune homeostasisHigh; broad population conserved[50–52]CCR2+ recruited monocyte/macrophage statesCcr2, Trem2, Lgals3CCR2, TREM2, LGALS3Context-dependent immune modulationModerate/uncertain; mainly mouse peripheral-nerve evidence[53, 54]Mast CellsMrgprb2+ mast cellsKit, Mrgprb2, Tpsb2KIT, MRGPRX2Neuropeptide- triggered degranulation; neurogenic inflammationModerate; MRGPRX2 is a functional counterpart, not a direct ortholog[55]Endothelial CellsFenestrated capillary endothelial cellsPecam1, Cldn5, VwfPECAM1, CLDN5, VWFPermeable neurovascular unit; nutrient exchange; barrier functionHigh[56, 57]PericytesDRG pericytesRgs5, Pdgfrb, Acta2RGS5, PDGFRB, ACTA2Capillary support; vascular tone and barrier stabilityHigh[56–58]Fibroblast-like stromal cellsStromal / fibroblast-like cellsCol1a1, Dcn, Col3a1COL1A1, DCN, COL3A1ECM synthesis; structural support and signalingModerate-high; broad identity conserved[59, 60]Activated myofibroblast-like cellsActa2, Col1a1ACTA2, COL1A1Fibrotic ECM deposition after injuryModerate; direct human DRG evidence limited[59, 60]Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; HTMRs, high-threshold mechanoreceptors; LTMRs, low-threshold mechanoreceptors; sNAMs, sensory neuron-associated macrophages; SGCs, satellite glial cells Major cell populations of the dorsal root ganglion and representative marker genes Table 1Representative DRG cell populations, marker genes, functional roles, and human–mouse correspondence in sc/snRna-seq studiesCell TypePopulation / programKey Markers (Mouse)Key Markers (Human)Main roleHuman–mouse correspondenceReferencesSensory NeuronsPeptidergic nociceptors (PEP)Calca, Tac1, Trpv1, Ntrk1CALCA, TAC1, TRPV1, NTRK1Thermal/chemical nociception; CGRP/SP releaseModerate–high; broader TRPV1 distribution in humans[13, 30, 35, 36]Non-peptidergic nociceptors (NP)Mrgprd, P2rx3, RetMRGPRD (low/diffuse), P2RX3Mechanical nociception and pruriceptionModerate; diffuse MRGPRD expression in humans[13, 35–37]C-low-threshold mechanoreceptors (C-LTMRs)Th, Slc17a8 (Vglut3)TH (low), SLC17A8/VGLUT3 (variable)Affective touch; gentle mechanosensationModerate; one-to-one physiological correspondence incomplete[17, 38, 39]Myelinated LTMR neurons (NF/Aβ)Nefh, Ntrk2, Ntrk3, RetNEFH, NTRK2, NTRK3Low-threshold mechanosensation; touch and vibrationHigh[13, 30, 35]ProprioceptorsPvalb, Ntrk3, Etv1PVALB, NTRK3Proprioception and muscle spindle innervationHigh[13, 30, 35]Aδ-like myelinated nociceptors / HTMRsNefh, Scn10a, Trpv1NEFH, SCN10A, TRPV1 (partial overlap)High-threshold mechano-/thermo-nociceptionModerate; correspondence is supported but not one-to-one[13, 35, 36]Human-enriched PIEZO2-high PEP neuronsNo clear counterpartPIEZO2 (high), CALCAPutative mechanosensory functionLow/absent; no clear murine counterpart[17]Human-enriched NTRK3/DCC nociceptor-related neuronsNo clear counterpartNTRK3, DCC, opioid-related transcriptsHuman-enriched sensory program; function uncertainLow/absent; persists into adulthood in human DRG[40]Satellite Glial CellsHomeostatic SGCsKcnj10, Fabp7, Glul, Gja1KCNJ10, FABP7, GLUL, GJA1Perineuronal ensheathment; K+ buffering; glial homeostasisModerate; broad identity conserved, fine subtype unresolved[18, 41, 42]Spatially biased SGC programsFabp7, Kcnj10Not establishedPreferential association with selected neuronal classesUncertain; supported mainly by mouse spatial data[43]UGT8+ metabolic SGCsUgt8, GalCer biosynthesis genesNot establishedGalCer biosynthesis; lipid support at the neuron–glia interfaceUncertain; defined mainly in rodent DPN[44]TIMP3+ SGCsTimp3TIMP3ECM homeostasis; metalloproteinase restraintModerate; human functional validation is limited[45]Schwann CellsMyelinating Schwann cellsMpz, Mbp, PrxMPZ, MBPAxonal myelination and saltatory conductionHigh[46, 47]Non-myelinating Schwann cellsNcam1, L1cam, Scn7aNCAM1, L1CAMEnsheathment of unmyelinated C-fibersHigh[46, 47]Stress response-related Schwann cells (SRSCs)Npy, Spp1Not establishedInjury-associated remodelingUnknown; described in mouse[48]Repair Schwann cellsJun, Ngf, BdnfJUN, NGF, BDNFInjury-induced repair; trophic supportModerate[46, 49]MacrophagesResident macrophages (sNAMs)Cx3cr1, Csf1r, Aif1, Cd68CX3CR1, CSF1R, AIF1Tissue surveillance; debris clearance; immune homeostasisHigh; broad population conserved[50–52]CCR2+ recruited monocyte/macrophage statesCcr2, Trem2, Lgals3CCR2, TREM2, LGALS3Context-dependent immune modulationModerate/uncertain; mainly mouse peripheral-nerve evidence[53, 54]Mast CellsMrgprb2+ mast cellsKit, Mrgprb2, Tpsb2KIT, MRGPRX2Neuropeptide- triggered degranulation; neurogenic inflammationModerate; MRGPRX2 is a functional counterpart, not a direct ortholog[55]Endothelial CellsFenestrated capillary endothelial cellsPecam1, Cldn5, VwfPECAM1, CLDN5, VWFPermeable neurovascular unit; nutrient exchange; barrier functionHigh[56, 57]PericytesDRG pericytesRgs5, Pdgfrb, Acta2RGS5, PDGFRB, ACTA2Capillary support; vascular tone and barrier stabilityHigh[56–58]Fibroblast-like stromal cellsStromal / fibroblast-like cellsCol1a1, Dcn, Col3a1COL1A1, DCN, COL3A1ECM synthesis; structural support and signalingModerate-high; broad identity conserved[59, 60]Activated myofibroblast-like cellsActa2, Col1a1ACTA2, COL1A1Fibrotic ECM deposition after injuryModerate; direct human DRG evidence limited[59, 60]Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; HTMRs, high-threshold mechanoreceptors; LTMRs, low-threshold mechanoreceptors; sNAMs, sensory neuron-associated macrophages; SGCs, satellite glial cells Representative DRG cell populations, marker genes, functional roles, and human–mouse correspondence in sc/snRna-seq studies Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; HTMRs, high-threshold mechanoreceptors; LTMRs, low-threshold mechanoreceptors; sNAMs, sensory neuron-associated macrophages; SGCs, satellite glial cells Sensory neurons Somatosensory neurons in the DRG convert mechanical, thermal, and chemical stimuli into electrical signals transmitted to the spinal dorsal horn. Transcriptomic profiling has resolved these neurons into approximately 14–18 conserved subtypes across species, depending on the sequencing platform and data integration strategy. These subtypes can be broadly grouped into peptidergic nociceptors (PEP), characterized by Calca and Tac1 expressions; non-peptidergic nociceptors (NP), marked by Mrgprd and P2rx3; and NF/Aβ myelinated low-threshold mechanoreceptor neurons, defined by Nefh and Ntrk2/3 [32, 34, 35, 61, 62]. Additional functional specialization has been identified within these broad categories. C-low-threshold mechanoreceptors (C-LTMRs), defined by Th and Vglut3 co-expression, represent a distinct lineage implicated in affective touch and temperature sensation [38, 39]. At the subcellular level, Landeros et al. described a neuronal membrane proteasome complex on nociceptor surfaces that modulates excitability through extracellular peptide degradation, adding a previously unrecognized layer of regulation [63]. An important consideration for translational research is the incomplete conservation of neuronal subtypes across species. In mice, PEP and NP populations form transcriptionally discrete clusters, whereas human nociceptors exhibit a more continuous transcriptional spectrum [30, 35, 36]. Notably, Yu et al. identified the hPEP.PIEZOh cluster, a human-enriched subtype with high PIEZO2 expression that lacks a clear murine counterpart and may contribute to visceral mechanosensation [17]. Satellite glial cells SGCs form a continuous perineuronal sheath around each sensory neuron soma, creating a restricted microenvironment that supports metabolic coupling, potassium buffering, and neurotransmitter recycling [41, 42, 64]. Although SGCs were traditionally considered as a homogeneous population, recent transcriptomic and spatial analyses have revealed considerable molecular heterogeneity among SGCs. Different SGC subtypes appear to preferentially associate with specific neuronal classes; for example, Fabp7-enriched SGCs are predominantly found around large mechanoreceptors, whereas Kir4.1-expressing SGCs preferentially ensheathe nociceptors [43]. In addition, SGCs actively participate in lipid metabolism. Xu et al. characterized a subpopulation enriched in enzymes involved in galactosylceramide biosynthesis, including Ugt8, indicating that SGCs contribute to the structural integrity of the neuron–glia interface through lipid homeostasis [44]. Under homeostatic conditions, SGCs express high levels of Timp3, which helps maintain extracellular matrix stability [45]. Schwann cells SCs in the DRG and peripheral nerves are broadly categorized into myelinating (Mpz+) and nonmyelinating (Ncam1+) subtypes. After nerve injury, SCs can revert to a progenitor-like repair phenotype characterized by the upregulation of Jun and neurotrophic factors, such as Ngf and Bdnf [46, 47, 49]. Recent single-cell studies have further expanded this classification. Lan et al. identified a subtype termed stress response–related Schwann cells, characterized by high neuropeptide Y (Npy) expression, that functions as a mechanical sensor coordinating tissue remodeling through osteopontin (Spp1) signaling [48]. Furthermore, Li et al. provided evidence that specific SC subpopulations regulate the local inflammatory environment through interactions with macrophages, indicating a more active role in immune regulation than was previously recognized [65]. Immune cells The DRG maintains a resident immune niche composed primarily of macrophages, often referred to as sensory neuron–associated macrophages. Single-cell profiling has shown that these macrophages exist along a spectrum of activation states rather than conforming to the classical M1/M2 dichotomy [50–52]. Their functions may depending on the type of injury, disease stage, and local microenvironment. Other immune cell populations may also contribute to DRG remodeling. For example, mast-cell activation through Mrgprb2 promotes neurogenic inflammation and pain in selected nerve-injury models [55]. Collectively, these findings underscore the functional heterogeneity of immune responses associated with neuropathic pain. Vascular and stromal cells The vascular compartment of the DRG consists of fenestrated endothelial cells (Pecam1+) and pericytes (Rgs5+, Pdgfrb+), which collectively form a neurovascular unit that is more permeable than the blood–brain barrier [56]. Under metabolic stress, such as hyperglycemia, reduced mural cell support and disruption of endothelial tight junctions may impair barrier function and facilitate immune cell infiltration [57, 58]. Fibroblasts within the DRG and peripheral nerve synthesize extracellular matrix components, and transcriptomic studies have distinguished epineurial, endoneurial, and related stromal subpopulations, some of which can adopt a pro-fibrotic myofibroblast-like state characterized by Acta2 expression under chronic pain or injury conditions [59, 60]. Etiology-specific DRG cell-state reprogramming Building on the cellular taxonomy described above, this section examines how distinct neuropathic pain etiologies reshape cellular states within the DRG. Single-cell and single-nucleus studies show that peripheral nerve injury, diabetic neuropathy, and chemotherapy-induced neuropathy share several injury-response programs, but also display distinct neuronal, glial, immune, and stromal signatures. Table 2 summarizes shared and etiology-associated DRG cell-state programs and their supporting evidence, and Figure 2 provides a schematic comparison of homeostatic and disease-associated remodeling across these etiologies.Table 2Shared and etiology-associated DRG cell-state programs in neuropathic painEtiology/modelMain compartmentPatternCell-state program or processKey featuresEvidenceReferencesPNIInjured sensory neuronsShared; PNI-associatedInjury/repair neuronal stateAtf3, Sprr1a, Gap43 ↑; subtype markers such as Tac1, Mrgprd, or Nefh ↓Transcriptomic; selected functional support[66, 67]PNISatellite glial cellsShared glial remodeling; PNI-associatedReactive/metabolic SGC remodelingGfap, Apoe ↑; Kcnj10, Hmgcs1, Fdft1 ↓Transcriptomic; selected functional support[68–70]PNIInjured neurons, Schwann cells, macrophagesShared neuroimmune process; PNI-associatedNeuroimmune recruitment and crosstalkJun/Npy-associated Schwann-cell states; Il34–Csf1r, Ccl2–Ccr2, and Csf1–Csf1rsignalingTranscriptomic + functional (animal)[48, 49, 53, 65, 71]PNIMast cellsPNI-associatedNeuropeptide-responsive mast-cell activationMrgprb2Functional (animal; selected models)[55]DPNSensory neuronsShared neuronal remodeling; DPN-associatedSensory-coding remodelingFxyd7, Atp1b1, and Mrgprd ↑in selected neuronsTranscriptomic + functional (animal)[37, 72]DPNSatellite glial cellsShared loss of glial support; DPN-associatedImpaired SGC lipid supportUgt8 expression and GalCer biosynthesis ↓Transcriptomic + functional (animal)[44]DPNSatellite glial cells and sensory neuronsDPN-associatedPutative neuron–SGC signalingPtn–Ncl and Mdk–NclComputational inference[73]Human diabetic peripheral neuropathySatellite glial cells and non-myelinating SchwanncellsHuman disease-associatedGlial remodeling with neurodegenerationSPP1+ Nageotte nodules associated with dystrophic or sprouting axonsHuman spatial + histopathological[31]Paclitaxel-induced CINPSensory neurons, including C-LTMR-like neuronsPaclitaxel-associatedSubtype-selective neuronal vulnerabilityCamk1d ↑ across several neuronal subtypes; prominent remodeling of Th/Slc17a8-defined neuronsTranscriptomic + functional (animal)[74]Paclitaxel-induced CINPSatellite glial cellsPaclitaxel-associatedMetalloproteinase-related SGC remodelingTimp3 ↓; altered Mmp14 and Adam17 signalingTranscriptomic + functional (animal)[45]Bortezomib-induced CINPSatellite glial cellsBortezomib- and sex-associatedProfibrotic SGC remodelingAtf3-associated fibrotic and ECM programsTranscriptomic (animal; sex-associated)[75]Bortezomib-induced CINPSatellite glial cells, macrophages, stromal cellsBortezomib-associated fibrotic remodelingMacrophage-stromal fibrotic nicheMacrophage expansion, collagen deposition, and ECM remodelingTranscriptomic + depletion evidence (animal)[75]Abbreviations: CINP, chemotherapy-induced neuropathic pain; DPN, diabetic painful neuropathy; DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; PNI, peripheral nerve injury; SGC, satellite glial cellFig. 2Shared and etiology-associated DRG cell-state remodeling in neuropathic pain Table 2Shared and etiology-associated DRG cell-state programs in neuropathic painEtiology/modelMain compartmentPatternCell-state program or processKey featuresEvidenceReferencesPNIInjured sensory neuronsShared; PNI-associatedInjury/repair neuronal stateAtf3, Sprr1a, Gap43 ↑; subtype markers such as Tac1, Mrgprd, or Nefh ↓Transcriptomic; selected functional support[66, 67]PNISatellite glial cellsShared glial remodeling; PNI-associatedReactive/metabolic SGC remodelingGfap, Apoe ↑; Kcnj10, Hmgcs1, Fdft1 ↓Transcriptomic; selected functional support[68–70]PNIInjured neurons, Schwann cells, macrophagesShared neuroimmune process; PNI-associatedNeuroimmune recruitment and crosstalkJun/Npy-associated Schwann-cell states; Il34–Csf1r, Ccl2–Ccr2, and Csf1–Csf1rsignalingTranscriptomic + functional (animal)[48, 49, 53, 65, 71]PNIMast cellsPNI-associatedNeuropeptide-responsive mast-cell activationMrgprb2Functional (animal; selected models)[55]DPNSensory neuronsShared neuronal remodeling; DPN-associatedSensory-coding remodelingFxyd7, Atp1b1, and Mrgprd ↑in selected neuronsTranscriptomic + functional (animal)[37, 72]DPNSatellite glial cellsShared loss of glial support; DPN-associatedImpaired SGC lipid supportUgt8 expression and GalCer biosynthesis ↓Transcriptomic + functional (animal)[44]DPNSatellite glial cells and sensory neuronsDPN-associatedPutative neuron–SGC signalingPtn–Ncl and Mdk–NclComputational inference[73]Human diabetic peripheral neuropathySatellite glial cells and non-myelinating SchwanncellsHuman disease-associatedGlial remodeling with neurodegenerationSPP1+ Nageotte nodules associated with dystrophic or sprouting axonsHuman spatial + histopathological[31]Paclitaxel-induced CINPSensory neurons, including C-LTMR-like neuronsPaclitaxel-associatedSubtype-selective neuronal vulnerabilityCamk1d ↑ across several neuronal subtypes; prominent remodeling of Th/Slc17a8-defined neuronsTranscriptomic + functional (animal)[74]Paclitaxel-induced CINPSatellite glial cellsPaclitaxel-associatedMetalloproteinase-related SGC remodelingTimp3 ↓; altered Mmp14 and Adam17 signalingTranscriptomic + functional (animal)[45]Bortezomib-induced CINPSatellite glial cellsBortezomib- and sex-associatedProfibrotic SGC remodelingAtf3-associated fibrotic and ECM programsTranscriptomic (animal; sex-associated)[75]Bortezomib-induced CINPSatellite glial cells, macrophages, stromal cellsBortezomib-associated fibrotic remodelingMacrophage-stromal fibrotic nicheMacrophage expansion, collagen deposition, and ECM remodelingTranscriptomic + depletion evidence (animal)[75]Abbreviations: CINP, chemotherapy-induced neuropathic pain; DPN, diabetic painful neuropathy; DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; PNI, peripheral nerve injury; SGC, satellite glial cell Shared and etiology-associated DRG cell-state programs in neuropathic pain Abbreviations: CINP, chemotherapy-induced neuropathic pain; DPN, diabetic painful neuropathy; DRG, dorsal root ganglion; ECM, extracellular matrix; GalCer, galactosylceramide; PNI, peripheral nerve injury; SGC, satellite glial cell Shared and etiology-associated DRG cell-state remodeling in neuropathic pain Peripheral nerve injury Peripheral nerve injury (PNI) is a leading cause of neuropathic pain, frequently resulting from trauma, compression, or surgical complications [76–78]. Preclinical models, including chronic constriction injury, spared nerve injury, and spinal nerve ligation, have been instrumental in elucidating the mechanisms underlying the transition from acute trauma to chronic pain [79, 80]. Recent sc/snRNA-seq studies indicate that PNI induces coordinated transcriptional remodeling across neurons, satellite glial cells, Schwann cells, and immune cell populations within the DRG [81]. At the neuronal level, PNI is characterized by a partial loss of subtype identity and activation of injury-response programs. Injured sensory neurons exhibit reduced expression of several subtype-defining genes accompanied by robust induction of Atf3, although the specific transcriptional changes vary across neuronal subtypes, injury models, and sampling time points [66]. Atf3-positive neurons also express regeneration-associated genes, including Sprr1a and Gap43, suggesting activation of an axonal repair program [67]. Other injury-induced signals may link repair and pain modulation. Adcyap1, which encodes PACAP, has been proposed to promote axon regeneration while limiting nociceptive signaling after PNI [82]. Clcf1 has also been identified in injury-induced neuronal clusters and may connect peripheral neuronal reprogramming with spinal astrocyte activation and central sensitization [16]. In addition, the iPain atlas identified a senescence-like nociceptor state in selected chronic pain models, and senolytic treatment reduced pain-related behaviors in mice [81]. However, behavioral improvement alone does not demonstrate that the underlying transcriptional state itself is reversible or remains stable across disease stages. These findings suggest that neuronal injury states comprise both adaptive repair responses and maladaptive programs that contribute to persistent pain. Non-neuronal cells also undergo extensive remodeling following PNI. Satellite glial cells exhibit increased expression of reactive genes, including Gfap and Apoe, accompanied by reduced expression of homeostatic genes such as Kcnj10 [68, 69]. Downregulation of cholesterol biosynthesis genes, including Hmgcs1 and Fdft1, may further impair lipid support for neuronal repair [69, 70]. Similarly, Schwann cells display injury-induced plasticity, including repair-associated and stress-related states [83]. Recent evidence suggests that Il34/Csf1r-mediated Schwann cell–macrophage communication may sustain neuroinflammation and chronic pain after nerve injury [65]. The immune response in PNI follows a sequential pattern. In the acute phase, injured neurons secrete chemokines, including Ccl2 and Csf1, promoting the infiltration of Ccr2+ monocyte–derived macrophages [53, 71]. These macrophages may sustain a pro-inflammatory DRG milieu, although their functions are likely to be context- and time-dependent. Mast-cell activation through Mrgprb2 has also been implicated in neurogenic inflammation and pain after nerve injury [55]. Together, these findings suggest that PNI is not only a disorder of damaged axons, but also a failure of multicellular homeostasis within the DRG. Diabetic painful neuropathy Diabetic painful neuropathy (DPN) is the most common chronic complication of diabetes, affecting up to half of all people with diabetes, of whom a substantial subset develops neuropathic pain. [84, 85]. Unlike the acute mechanical insult of PNI, DPN develops under chronic metabolic stress caused by hyperglycemia, dyslipidemia, oxidative stress, and microvascular dysfunction [86]. Single-cell studies suggest that this metabolic context produces a pattern of DRG remodeling that differs from that induced by traumatic nerve injury. At the neuronal level, single-cell profiling has identified disease-specific subpopulations associated with DPN. Zhou et al. described mechanical allodynia-associated clusters defined by Fxyd7 and Atp1b1 upregulation, implicating ionic dysregulation in the maintenance of mechanical hypersensitivity [72]. George et al. reported aberrant Mrgprd upregulation in nonpeptidergic neurons in diabetic neuropathy, suggesting a role for ectopic activation of cutaneous sensory afferents in painful phenotypes [37]. Altered expression of mechanosensitive and voltage-gated ion channels, including Piezo2, Scn9a, and Scn3a, may further contribute to sensory abnormalities in DPN [86]. These findings suggest that DPN-related neuronal remodeling is driven primarily by metabolic stress rather than by a classic axotomy-like injury program alone. A distinguishing feature of DPN at the single-cell level is the metabolic impairment of SGCs. Reduced Ugt8 expression and impaired galactosylceramide biosynthesis disrupt lipid raft integrity and weaken SGC–neuron metabolic coupling, thereby contributing to nociceptor hyperactivity [44]. Inferred ligand–receptor analyses have also suggested altered Ptn–Ncl and Mdk–Ncl signaling between glial cells and neurons [72, 73]. This metabolic uncoupling contrasts with the more inflammatory gliosis typically observed after PNI. Human DPN tissue provides additional evidence of advanced DRG remodeling. Spatial analysis of Nageotte nodules showed that these structures are mainly composed of SPP1-expressing satellite glial cells and non-myelinating Schwann cells associated with dystrophic or sprouting sensory axons [31]. This finding provides a human pathological correlate of multicellular DRG degeneration and likely reflects relatively advanced neurodegenerative remodeling in DPN. The immune response in DPN also appears to be context-dependent. Ccr2-positive macrophages in prediabetic mice can express Trem2 and Lgals3 and may delay terminal sensory axon degeneration through galectin 3 signaling [54]. This contrasts with the predominantly pain-promoting macrophage response described in some PNI models and suggests that indiscriminate anti-inflammatory strategies may not be appropriate across all neuropathic pain etiologies [53, 54]. Chemotherapy-induced neuropathic pain Chemotherapy-induced neuropathic pain (CINP) is a common, dose-limiting adverse effect of several antineoplastic agents [87]. Unlike PNI and DPN, CINP results from drug-induced neurotoxicity, including mitochondrial dysfunction, impaired axonal transport, and altered neuroimmune signaling [88]. Single-cell studies further suggest that CINP is not a uniform syndrome but instead comprises agent-specific patterns of neuronal and non-neuronal vulnerability. In paclitaxel-induced neuropathy, Sun et al. identified widespread Camk1d upregulation across several sensory-neuron subtypes, together with prominent transcriptomic remodeling of Th/Slc17a8-defined C-LTMR-like neurons. Functional knockdown experiments implicated Camk1d in thermal and cold hypersensitivity, but did not demonstrate that its expression or function was restricted to the C-LTMR-like population [74]. Concurrently, Tonello et al. showed that paclitaxel reduces Timp3 expression in satellite glial cells, leading to disinhibition of metalloproteinase-related signaling involving Mmp14 and Adam17 and contributing to a pro-inflammatory microenvironment [45]. These findings suggest that paclitaxel-induced neuropathy involves both neuronal subtype vulnerability and glial microenvironmental remodeling. Bortezomib-induced neuropathy appears to involve a distinct glial and stromal response. Thomsen et al. reported that male mice develop a specific Atf3-driven profibrotic program in SGCs after bortezomib treatment, resulting in pathological collagen deposition that physically constricts sensory neurons [75]. This response was linked in part to macrophage–SGC crosstalk and was less prominent in female mice, which showed relative enrichment of lipid metabolic programs [75]. These findings highlight sex as an important biological variable in CINP and suggest that different chemotherapeutic agents may engage distinct DRG pathological programs within the DRG. Single-cell studies reveal distinct, treatment-associated DRG responses to different chemotherapeutic agents. Paclitaxel-induced neuropathy is associated with C-LTMR-like neuronal vulnerability and SGC metalloproteinase signaling, whereas bortezomib-induced neuropathy appears to involve sex-dependent neurofibrotic remodeling [45, 74, 75]. These differences support the view that CINP is not a uniform neurotoxic syndrome, but rather a group of agent-specific pathobiological states that may require mechanism-tailored therapeutic strategies. Shared and etiology-specific DRG remodeling Comparisons across PNI, DPN, and CINP indicate partially shared biological processes rather than a single conserved molecular program. A cell state refers here to a reproducible transcriptomic configuration within a defined cell population, whereas a cell-state transition requires evidence of temporal progression. Because most current studies rely on single-time-point or repeated cross-sectional sampling, they identify disease-associated cell states and potential temporal patterns rather than directly tracking transitions within the same cells [44, 45, 66, 72, 75]. The principal overlap across etiologies occurs at the level of broad biological processes, including neuronal state remodeling, altered glial homeostatic support, context-dependent immune responses, and extracellular matrix remodeling [45, 53, 89]. These cell states may serve adaptive, maladaptive, or degenerative functions depending on the cell type, disease stage, and experimental context. For example, injury-associated neuronal programs may comprise both regenerative and pain- associated components [66], whereas Nageotte nodules in human diabetic peripheral neuropathy are more consistent with advanced neurodegenerative remodeling than with an early or pain-specific mechanism [31]. Accordingly, similar transcriptomic features across etiologies should not be assumed to produce equivalent functional outcomes. Etiology-specific patterns remain prominent. PNI is characterized by neuronal injury and repair states, reactive and metabolic SGC remodeling, and neuroimmune cell recruitment. DPN is associated with altered sensory coding, impaired lipid-support functions in SGCs, and inferred neuron–SGC signaling [66, 90, 91]. CINP exhibits substantial agent-associated variation, with subtype-selective neuronal vulnerability and metalloproteinase-dysregulated SGC states following paclitaxel treatment, as well as male-biased profibrotic remodeling following bortezomib treatment. However, the temporal coordination of these responses remains incompletely understood. Available evidence suggests that PNI begins with early neuronal injury responses and chemokine induction, followed by glial and immune remodeling, whereas metabolic, neuronal, and glial abnormalities in DPN may evolve in parallel [44, 72]. In CINP, the timing and persistence of these responses vary according to the chemotherapeutic agent and biological context. Collectively, current evidence supports etiology-specific temporal models rather than a unified sequence of DRG remodeling [45, 74]. Large-diameter myelinated neurons, including NF mechanoreceptors and proprioceptors, remain less well characterized than nociceptors in disease-focused datasets. Altered Nefh-expressing populations have been reported following PNI; however, their contribution to persistent pain remains uncertain [66]. Comparable analyses in DPN and CINP remain limited. The C-LTMR-like population implicated in paclitaxel-induced neuropathy represents a distinct unmyelinated mechanoreceptive lineage and should not be interpreted as evidence for the involvement of large myelinated neurons [38, 74]. Thus, the improved recovery of large neuronal nuclei by snRNA-seq has not yet translated into a clear understanding of their disease-specific roles. Single-cell profiling of the human DRG The limited translational success of analgesic candidates identified in rodent models has underscored the need to define the molecular architecture of the human DRG at cellular resolution [92]. Recent single-nucleus RNA sequencing, spatial transcriptomics, single-soma deep RNA sequencing, and cross-species integration studies have generated increasingly comprehensive human DRG atlases [30, 35]. These datasets demonstrate broad conservation of the major sensory neuron classes while also revealing species-specific differences in neuronal subtype organization, glial programs, and the distribution of candidate therapeutic targets. Table 3 summarizes key human DRG evidence, relevant rodent comparisons, translational implications, and major caveats.Table 3Human DRG evidence and translational implicationsHuman evidenceRepresentative findingsRodent comparison / contextTranslational implicationKey caveatReferencesNociceptor architecture and marker distributionHuman nociceptors show a more continuous transcriptional spectrum, broader TRPV1 distribution and lower or more diffuse MRGPRD expressionMouse peptidergic and nonpeptidergic nociceptors form more discrete clustersRodent subtype maps and target distributions require cautious human extrapolationDonor/postmortem atlases do not necessarily reflect protein function, excitability, or treatment response[30, 34–36]Human-enriched neuronal populationsPIEZO2-high peptidergic neurons and NTRK3/DCC-expressing nociceptor-related neurons have been identified in human DRGNo clear one-to-one adult mouse counterpartHuman-enriched sensory programs for target discoveryFunctional roles and pain-state relevance remain unresolved[17, 40]Human glial programsHuman SGCs and Schwann cells show species-divergent neurotrophic, ECM, and inflammatory expression programs, including NGF-related patternsDirect correspondence with rodent glial programs remains uncertainHuman cellular context for glial mechanisms and candidate targetsExpression patterns alone do not establish clinical causality or therapeutic relevance[18]Human disease pathologyNageotte nodules in diabetic peripheral neuropathy contain SPP1-expressing SGCs and non-myelinating Schwann cells associated with dystrophic or sprouting axonsNot fully captured by standard rodent diabetic neuropathy modelsDirect human tissue evidence of multicellular DRG remodeling and neurodegenerationMay represent advanced pathology rather than an early, reversible, or pain-specific state[31]Integration with genetic evidenceGWAS-supported pain candidates, such as SCN11A and EFNB2 can be mapped to human DRG cell populationsGenetic signals can be integrated with cell-type-resolved human DRG expressionPrioritization of genetically anchored human pain candidatesGenetic and cellular anchoring does not establish functional validation or therapeutic tractability[93]Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GWAS, genome-wide association study; SGCs, satellite glial cells Table 3Human DRG evidence and translational implicationsHuman evidenceRepresentative findingsRodent comparison / contextTranslational implicationKey caveatReferencesNociceptor architecture and marker distributionHuman nociceptors show a more continuous transcriptional spectrum, broader TRPV1 distribution and lower or more diffuse MRGPRD expressionMouse peptidergic and nonpeptidergic nociceptors form more discrete clustersRodent subtype maps and target distributions require cautious human extrapolationDonor/postmortem atlases do not necessarily reflect protein function, excitability, or treatment response[30, 34–36]Human-enriched neuronal populationsPIEZO2-high peptidergic neurons and NTRK3/DCC-expressing nociceptor-related neurons have been identified in human DRGNo clear one-to-one adult mouse counterpartHuman-enriched sensory programs for target discoveryFunctional roles and pain-state relevance remain unresolved[17, 40]Human glial programsHuman SGCs and Schwann cells show species-divergent neurotrophic, ECM, and inflammatory expression programs, including NGF-related patternsDirect correspondence with rodent glial programs remains uncertainHuman cellular context for glial mechanisms and candidate targetsExpression patterns alone do not establish clinical causality or therapeutic relevance[18]Human disease pathologyNageotte nodules in diabetic peripheral neuropathy contain SPP1-expressing SGCs and non-myelinating Schwann cells associated with dystrophic or sprouting axonsNot fully captured by standard rodent diabetic neuropathy modelsDirect human tissue evidence of multicellular DRG remodeling and neurodegenerationMay represent advanced pathology rather than an early, reversible, or pain-specific state[31]Integration with genetic evidenceGWAS-supported pain candidates, such as SCN11A and EFNB2 can be mapped to human DRG cell populationsGenetic signals can be integrated with cell-type-resolved human DRG expressionPrioritization of genetically anchored human pain candidatesGenetic and cellular anchoring does not establish functional validation or therapeutic tractability[93]Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GWAS, genome-wide association study; SGCs, satellite glial cells Human DRG evidence and translational implications Abbreviations: DRG, dorsal root ganglion; ECM, extracellular matrix; GWAS, genome-wide association study; SGCs, satellite glial cells Human DRG atlases Human DRG profiling has evolved from neuronal taxonomy toward integrated and spatially informed reference atlases. Nguyen et al. used snRNA-seq to classify human DRG neurons into major sensory subtypes, providing an initial framework for comparisons with rodent sensory neurons [35]. Tavares-Ferreira et al. complemented this approach by combining snRNA-seq with spatial transcriptomics, thereby preserving anatomical context and identifying molecular signatures of human nociceptors within intact DRG sections [30]. Yu et al. applied single-soma deep RNA sequencing to individual human DRG neurons, improving transcript detection and revealing neuronal diversity that may be underrepresented by nuclear profiling alone [17]. Cross-species integration efforts, such as the harmonized trigeminal and DRG atlas generated by Bhuiyan et al., further provide standardized annotations for comparing neuronal and non-neuronal cell populations across species [34]. The primary value of these atlases is that they provide a human molecular framework for interpreting findings from rodent models. They help determine whether candidate molecular targets or cell-state programs are present in the human DRG, whether they are enriched in nociceptors or non-neuronal cell populations, and whether their cellular distribution is conserved across species. Cross-species differences in DRG cell populations Species differences are evident not only in neuronal populations but also in glial and immune cell populations. At the neuronal level, human nociceptor populations appear more transcriptionally continuous than their murine counterparts. Markers such as TRPV1 and MRGPRD, which help distinguish peptidergic and non-peptidergic nociceptor populations in rodents, exhibit broader and less discrete expression patterns in the human DRG [30, 32]. Some human-enriched neuronal populations, including PIEZO2-high peptidergic neurons and NTRK3/DCC-expressing nociceptor-related neurons [35, 40], also lack clear one-to-one correspondence with mouse subtypes. These differences have important implications for analgesic development, because molecular targets that appear restricted to discrete nociceptor populations in mice may be more broadly expressed or differently distributed in humans. Cross-species differences also extend to satellite glial cells and Schwann cells [94]. Human glial populations exhibit distinct neurotrophic, extracellular matrix, and inflammatory gene expression profiles, including prominent NGF-related signaling in satellite glial cells [18]. These findings help define the human cellular context of candidate molecular targets but do not establish their roles in clinical pain. Consequently, glial and immune mechanisms identified in rodent models require validation in human DRG tissue. Human DRG evidence in neuropathic pain Disease- focused human DRG studies remain limited. Spatial profiling of Nageotte nodules in diabetic peripheral neuropathy identified SPP1-expressing satellite glial cells and non-myelinating Schwann cells associated with neuronal loss and dystrophic or sprouting sensory axons [31]. These findings provide direct evidence of multicellular DRG remodeling in human disease, but likely reflect advanced neurodegenerative pathology rather than an early or pain-specific mechanism. Bulk transcriptomic analyses of clinically phenotyped human DRG samples have also identified sex-associated neuroimmune signatures [33]. However, the cellular origins of these signatures require confirmation using single-cell or spatial transcriptomic approaches. Most human DRG single-cell datasets are derived from organ donors or postmortem specimens with incomplete pain phenotyping. Donor age, sex, comorbidities, medication exposure, postmortem interval, tissue preservation, and agonal state may all influence transcriptomic profiles [95, 96]. Moreover, transcript abundance does not necessarily predict protein expression, neuronal excitability, or clinical pain phenotypes [97]. The cross-sectional design of current datasets also limits the ability to distinguish among causal, compensatory, and degenerative cell states. Translational prospects Translation of DRG single-cell discoveries requires a stepwise framework that moves from transcriptomic association to translational interpretation, human-relevant validation, and clinically tractable opportunities. Figure 3 summarizes this translational framework.Fig. 3Translational framework for DRG single-cell discoveries in neuropathic pain Translational framework for DRG single-cell discoveries in neuropathic pain Target prioritization Therapeutic candidates identified through DRG single-cell studies vary considerably in the strength of supporting evidence. Target prioritization should integrate cell-type-resolved human DRG expression with human genetic evidence, functional studies in human sensory neurons, experimental perturbation, and clinical pharmacology. Integration of GWAS data can link candidate genes and cell populations to human pain susceptibility [98, 99], whereas studies using native or cultured human sensory neurons can determine whether transcriptomically nominated targets alter neuronal excitability or other pain-relevant phenotypes. Candidates supported only by differential expression or computational inference should be regarded as hypothesis-generating. Nav1.8, encoded by SCN10A, illustrates a relatively advanced example of target prioritization. Nav1.8 is enriched in nociceptor populations and is preferentially expressed in peripheral sensory neurons, providing a biological rationale for the development of peripherally acting analgesics [100]. This rationale is supported by randomized clinical trials of the selective Nav1.8 inhibitor suzetrigine in acute postoperative pain [101], leading to regulatory approval for the treatment of moderate-to-severe acute pain in adults. However, these findings should not be extrapolated to neuropathic pain, for which the clinical efficacy of Nav1.8 inhibition remains under investigation. Accordingly, Nav1.8 should be regarded as a clinically validated target for acute pain and as a promising, but not yet established, target for peripheral neuropathic pain. Nav1.7, encoded by SCN9A, represents a more complex translational example. Human loss-of-function variants in SCN9A cause congenital insensitivity to pain [102], providing strong genetic evidence supporting Nav1.7 as a pain-related target. Nevertheless, some selective Nav1.7 inhibitors have demonstrated limited clinical efficacy [103]. Human DRG transcriptomic studies indicate that SCN9A expression is enriched in selected nociceptor populations rather than being uniformly distributed across all sensory neurons [30]. This observation raises the possibility that molecular heterogeneity across patients and pain conditions may have diluted treatment effects in previous clinical trials. However, this interpretation remains hypothetical and requires prospective validation in clinically defined cohorts using target-relevant biomarkers. Integrating GWAS data with human DRG single-cell datasets provides an additional strategy for to target identification. SCN11A and EFNB2 are enriched in pain-associated human DRG neuronal populations and can therefore be considered genetically anchored candidate genes [93]. However, their causal roles and therapeutic tractability remain uncertain. ERO1 has stronger functional support. Its inhibition reduces excitability in cultured human DRG neurons and attenuates acute pain-like behavior in mice [104]. Whether this mechanism contributes to chronic neuropathic pain remains unresolved. Several additional pathways remain at the preclinical stage, including Camk1d-associated neuronal vulnerability in paclitaxel-induced neuropathy [74], Ugt8/GalCer-related SGC–neuron metabolic uncoupling in diabetic neuropathy [44], Timp3-associated extracellular matrix remodeling [45], and Il34/Csf1r-mediated Schwann cell–macrophage communication following nerve injury [11]. These findings provide mechanistic insights, however, their translational potential will depend on conservation in the human DRG, a demonstrated causal contribution to pain maintenance, and therapeutic tractability in defined patient subgroups. Translational challenges and validation Candidate cell-state signatures may eventually support mechanism-based patient stratification; however, their current clinical utility remains limited. ATF3/Atf3-associated injury-state neurons, senescence-like nociceptor states, and CCR2/Ccr2- and LGALS3/Lgals3-associated macrophage responses may define biologically informative cell states in experimental models [53, 66, 105]. However, the human DRG cannot be routinely sampled in living patients, and no validated surrogate assay currently captures DRG cell states with sufficient reliability for clinical stratification. Potential surrogates, including skin-biopsy-derived molecular profiles, circulating extracellular vesicles, serum proteins, imaging correlates, and organoid-informed response predictors, should therefore be regarded as future research directions rather than as near-term clinical biomarkers [106, 107]. Emerging experimental platforms may help address part of this validation gap; however, they also require cautious interpretation. Human sensory organoids provide a useful system for testing human-relevant sensory neuron programs and candidate targets [108]. However, most current organoid models resemble developmental sensory states and do not fully recapitulate the architecture of the mature adult DRG, including neuron–satellite glial organization, the vascular niche, immune components, and the pathological features of chronic neuropathic pain. Similarly, nociceptor-selective AAV strategies and promoter-guided gene delivery offer promising approaches for cell-type-selective intervention; however, delivery specificity, tissue tropism, durability of expression, immunogenicity, off-target effects, and long-term safety remain unresolved [109]. Limitations and interpretive challenges Technical variation in tissue processing, transcript capture, sequencing depth, and data integration can affect cell recovery, cluster definition, and cross-study comparability. These effects are particularly relevant when comparing datasets generated using different platforms, species, pain models, and sampling time points. A more fundamental limitation is that current single-cell studies cannot directly establish temporal progression or causal mechanisms. Clustering and differential-expression analyses identify disease-associated transcriptional states but do not establish causality [110]. Pseudotime and trajectory inference methods order cells according to transcriptional states and can generate hypotheses regarding state progression. However, pseudotime is not equivalent to biological time, and most DRG datasets do not track the same cells longitudinally [111]. Even studies that include multiple sampling time points generally provide repeated cross-sectional snapshots. Accordingly, demonstrated cell-state transitions require support from time-resolved sampling, lineage tracing, live-cell imaging, perturbation experiments, or other forms of direct temporal evidence. Likewise, ligand–receptor inference methods predict potential cell–cell communication based on transcript co-expression and curated interaction databases; however, these predictions do not establish spatial proximity, receptor activation, or functional signaling [112]. Furthermore, transcript abundance also does not necessarily predict protein expression, ion-channel activity, neuronal excitability, or pain-related behavior. Human DRG studies provide important translational context but remain constrained by limited tissue availability, donor heterogeneity, postmortem effects, and incomplete clinical pain phenotyping [30, 35]. Consequently, spatial localization, protein-level characterization, functional perturbation, electrophysiological validation, and clinically annotated human datasets are required before transcriptomic findings can be considered causal or therapeutically relevant. Conclusions and future directions Single-cell and single-nucleus transcriptomics have revealed that neuropathic pain is associated with coordinated remodeling across neuronal, glial, immune, and stromal cell populations within the DRG. Although PNI, DPN, and CINP share broad biological processes, including neuronal state remodeling, altered glial homeostatic support, context-dependent immune responses, and extracellular matrix remodeling, these processes are mediated through distinct cell types, molecular programs, and disease contexts. Human DRG studies provide an essential framework for evaluating the conservation and cellular context of findings from experimental models. Future studies should integrate longitudinal and spatial transcriptomic profiling with multi-omic, protein-level, and functional validation in clinically annotated human tissues and more mature human-relevant models. Together, these complementary approaches may help distinguish disease-associated transcriptional states from causal mechanisms and identify therapeutically tractable targets for mechanism-based, cell-type-informed analgesic development.