Cross-species transcriptomic atlas of dorsal root ganglia reveals species-specific programs for sensory function

Abstract

Sensory neurons of the dorsal root ganglion (DRG) are critical for maintaining tissue homeostasis by sensing and initiating responses to stimuli. While most preclinical studies of DRGs are conducted in rodents, much less is known about the mechanisms of sensory perception in primates. We generated a transcriptome atlas of mouse, guinea pig, cynomolgus monkey, and human DRGs by implementing a common laboratory workflow and multiple data-integration approaches to generate high-resolution cross-species mappings of sensory neuron subtypes. Using our atlas, we identified conserved core modules highlighting subtype-specific biological processes related to inflammatory response. We also identified divergent expression of key genes involved in DRG function, suggesting species-specific adaptations specifically in nociceptors that likely point to divergent function of nociceptors. Among these, we validated that TAFA4, a member of the druggable genome, was expressed in distinct populations of DRG neurons across species, highlighting species-specific programs that are critical for therapeutic development.

Fig. 1. Comparison of nuclei isolation methods and tissue types in dorsal root ganglia single-nucleus RNA-seq.

a Overview of laboratory and computational workflow. The graphic was created with BioRender.com b Representative micrographs of nuclei from FACS and density-gradient (DG) protocols. Blue arrows indicate neuronal nuclei (PI+ and NeuN+) and yellow arrows indicate non-neuronal nuclei (PI+ and NeuN−). Distribution of nuclei size by capture method. Dotted lines represent smoothed distributions of binned data. The experiment was repeated 4 times independently with similar results. c–e Summary metrics computed for different nuclei isolation methods and tissue types in mouse data. c, d Relative cell-type composition for the nuclei isolation methods or tissue types. Black dashed line was drawn at 50%. e DRG neuron signature score for different nuclei isolation methods and tissue types. n = 37,384 nuclei examined over 10 independent experiments. The lower and upper hinges of the boxes correspond to the 25th–75th percentile with the line in the middle depicting the median. The whiskers are set at the minimum and maximum values of the dataset. f Volcano plot showing results from differential expression analysis of transcripts from FACS versus DG nuclei. Statistically significant genes are indicated by red or blue colors. Gray dotted lines represent log-fold change thresholds of 2 and −2.

Fig. 2. Single-nuclei transcriptome atlas of mouse dorsal root ganglia (DRG) sensory neurons.

a UMAP of 37,384 mice DRG nuclei colored by cell types and annotated by marker genes as indicated in the main text. b UMAP of 5656 mice DRG neurons colored by subtypes and annotated by marker genes as indicated in the main text. NP denotes non-peptidergic; PEP denotes peptidergic for a and b. c Fraction of nuclei (dot size) in each subset expressing canonical marker genes (columns) and their scaled average expression level in expressing cells (dot color) within subtypes (rows). d Heatmap comparing mean expression levels (color bars) of top differentially expressed genes (rows) between C-LTMR1 and C-LTMR2 subtypes. e Representative images for RNAScope validation of Rgs5, Th, and Slc17a8 expression in mouse DRGs. Neurons are outlined in turquoise. White arrowheads indicate Slc17a8-positive cells expressing both high levels of Th and Rgs5 and yellow arrowheads indicate cells that express Slc17a8-positive cells but low amounts of Th and Rgs5 transcript. f Quantification of RNA transcript punctate dots (representing expression level) normalized by slide area for each DRG level and C-LTMR subtype. Rgs5 area-normalized puncta per cell differences analyzed using a mixed-effects model (main effect of C-LTMR type: F_{(1, 8)} = 175.5, p < 0.0001; C-LTMR type and spinal level interaction: F_{(2, 8)} = 9.565, p = 0.0076) with a Bonferonni’s multiple comparisons post-test C-LTMR1 vs. 2 (cervical & thoracic: adjusted p <0.0001; lumbar: adjusted p = 0.0131). N = 4 mice. Asterisks represent statistical significance (**** for <0.0001 and * for <0.05). Black dashed line represents the manually determined threshold for Th high and low cells. The lower and upper hinges of the boxes correspond to the 25th to 75th percentile with the line in the middle depicting the median. The whiskers are set at the minimum and maximum values of the dataset. g Distribution of the two C-LTMRs subpopulations by DRG level. Source data are provided as a Source Data file. f and g share the same color legend.

Fig. 3. DRG sensory neuron homology consensus across mouse, guinea pig, cynomolgus monkey and human.

a Summary metrics for mouse, guinea pig, cynomolgus monkey, and human DRG single-nuclei RNA-seq data. Evolutionary distances (Million Years Ago; MYA) are shown in the dendrogram to the left. b UMAP of Seurat-integrated DRG neurons across mouse, guinea pig, cynomolgus monkey, and human, faceted and colored by species, Seurat clusters, DRG subclasses, and subtypes. c Cross-species dendrogram and heatmap of MetaNeighbor mean area under the receiver operator characteristic (AUROC) scores (y-axis in plot) colored by DRG sensory neuron subclasses. d Fraction of nuclei (dot size) in each neuron subtype expressing selected canonical markers (columns) and their scaled average expression level in expressing nuclei (dot color) in mouse, guinea pig, cynomolgus monkey, and human data, by sensory neuron subtypes (rows). Colored asterisks indicate markers used for annotating clusters in the integrated cross-species DRG data. NP denotes non-peptidergic; PEP denotes peptidergic for b–d. Species icons in a and d were created with BioRender.com.

Fig. 4. Conserved and species-specific transcriptional programs in DRG sensory neurons.

a Relative DRG subtype proportions across all species (computed in lumbar levels only). b Heatmap of conserved genes and species-specific genes by DRG subtypes from panel (c). c Customized UpSet plot showing a number of conserved and species-specific genes across datasets for each of the major sensory neuron subtypes, as indicated above each plot. Total number of genes that are unique to one species (one colored box per row), or shared across multiple species (multiple colored boxes per row), are indicated to the right of each row. d Plot indicating evolutionary distance (x-axis; icons at top) and transcriptional correlation (y-axis) for each sensory neuron subtype (indicated by colors) between humans and other preclinical species. Correlation values are indicated by colored circles; the values across all species are connected with a solid line. e Heatmap indicating normalized expression levels of genes associated with pain perception in each neuronal subtype. f Number of genes with a divergent or non-divergent expression between mouse and human DRGs indicated by gene ontology(GO) annotation category. Numeric numbers on each bar represent the total number of expressed genes in a given GO category. g Number of divergently expressed genes between mouse and human samples from panel (f), colored by DRG class (left) or subtype (right). NP denotes non-peptidergic; PEP denotes peptidergic for (a–g). M denotes mouse, G denotes guinea pig, C denotes cynomolgus monkey, and H denotes human for (a–c). Species icons in b and d were created with BioRender.com.

Fig. 5. Divergent expression of TAFA4 across species.

a Dot plots showing the fraction of nuclei (size) and their scaled average expression level (color) in each neuron subtype (columns) across mouse, guinea pig, cynomolgus monkey, and human for genes (row names). NP denotes non-peptidergic; PEP denotes peptidergic. b Representative images of RNAScope validation of FAM19A4/TAFA4, NTRK2, and TUBB3 (mouse, guinea pig, human) or SNAP25 (cynomolgus monkey) in DRGs. Neurons are outlined in turquoise. White arrowheads indicate neurons (TUBB3+ or SNAP25+) expressing both TAFA4 and NTRK2 in human samples. Asterisks in the human images represent lipofuscin aggregates. Scale bar represents 50 µm. GP denotes guinea pig; Cyno denotes cynomolgus monkey. c Percentage of FAM19A4/TAFA4+, NTRK2+, and TAFA4+/NTRK2+ (‘double +’) in mouse, guinea pig, cynomolgus monkey, and human. d Size distribution of FAM19A4/TAFA4+, NTRK2+, and TUBB3+ or SNAP25+ cells in each species. Colored solid lines represent smoothed distributions of binned data. Gray dashed lines represent the 33% and 67% percentiles of the TUBB3+ or SNAP25+ distribution as a surrogate for small, medium, and large diameter neurons. Source data are provided as a Source Data file.