Identification of a reactive astrocyte subpopulation during HIV-associated pain pathogenesis in mouse models Authors: Junying Zheng, Michael Spurgat, Mark Maurelli, Zachary Watson, Sakib Hossain, Adriana Dibua, Subo Yuan, Xin Liu, Benjamin B. Gelman, Shao-Jun Tang Published: 2025 ABSTRACT Pain is a common complication in individuals with HIV, yet the underlying pathogenic mechanisms remain poorly understood. Reactive astrocytes are selectively activated in the spinal dorsal horn of HIV patients experiencing pathological pain, implicating their potential role in pain pathogenesis. However, the specific pathogenic subtype of these reactive astrocytes has not been defined. Here, we identify a distinct subpopulation of pain-associated reactive astrocytes in the lumbar spinal cord of HIV-1 gp120 transgenic mice using single-nucleus RNA sequencing. This astrocyte subpopulation displays transcriptomic features associated with phagocytosis and inflammation and depends on galectin-3 for its formation. Both global and astrocyte-specific heterozygous knockout of Lgals3 (galectin-3) markedly reduce mechanical allodynia in gp120 transgenic mice. These findings reveal a subset of reactive astrocytes that play a role in remodeling the spinal pain circuitry underlying HIV-associated pain. Introduction Pathological pain is a frequent comorbidity in HIV patients, but its underlying mechanism remains unclear1,2. Among the neuropathologies observed in the pain neural circuits of HIV patients, reactive astrocytes in the spinal cord dorsal horn (SDH) are particularly evident3. Importantly, reactive astrocytes were found in the SDH of HIV patients who developed neuropathic pain but not in those without pain disorders3. These findings on HIV patient spinal specimens suggest that reactive astrocytes play a crucial role in the pathogenesis of HIV-associated pain. Under normal conditions, spinal astrocytes maintain circuitry homeostasis by modulating pain processing in neuronal circuits via their various biological activities, such as neurotransmitter uptake and gliotransmitter release4. When the homeostatic function is disrupted in reactive astrocytes, their associated pain neuronal circuits may become sensitized to activation. Reactive astrocytes in the SDH have been reported in animal models of various pathological pain types4,5, including HIV-associated pain models6. Supporting their critical role in pain pathogenesis, reactive astrocyte ablation blocks mechanical allodynia induced by HIV-1 glycoprotein 120 (gp120)7. Concomitantly, astrocytic ablation also reverses gp120-induced neural circuit polarization (NCP) implicated in HIV-associated pain pathogenesis7. Similarly, ablation of reactive astrocytes blocks opioid-induced hyperalgesia and its associated NCP in the SDH8. Astrocytes are heterogenous in function, morphology, and spatiotemporal distribution across CNS regions9,10. Astrocytes undergo a reactive transformation in various disease conditions. Recent data from single-cell RNA sequencing studies found reactive astrocytes can give rise to a spectrum of subtypes with diverse functions11. These astrocyte subtypes have characteristic gene expression signatures implicated in specific cellular pathways, such as extracellular matrix remodeling12, plaque degradation13, inflammation, and anti-inflammation14, and may contribute to relevant pathogenesis of specific diseases. However, despite the emerging consensus about the critical role of reactive astrocytes in pain pathogenesis4,5, little is known about pain-pathogenic subtypes of astrocytes and their contribution to underlying pathogenesis. In this study, we performed snRNA-seq analysis of the lumbar spinal cord of HIV-1 gp120 transgenic (gp120Tg) mice15, a model of HIV-associated pain, and identified a subpopulation of reactive astrocytes, named HIV pain-associated astrocytes (HIPAs). This astrocytic population has transcriptomic signatures of phagocytosis and inflammation and is a phagocyte of neurons. In addition, global knockout (KO) of Gal3 leads to depletion of this astrocytic population and blockage of the development of mechanical and thermal allodynia in gp120Tg mice. Conditional KO of one copy of Gal3 in astrocytes also attenuated mechanical allodynia. These findings suggest that HIPAs play an important role in the pathogenesis of HIV-associated pain. Results Identification of an astrocytic subpopulation specific to the spinal cord of gp120 transgenic mice by snRNA-seq analysis To investigate the cellular mechanism underlying pain pathogenesis induced by HIV-1 gp120, we performed snRNA-seq analysis of spinal cords from Gfap-gp120 transgenic (gp120Tg) mice at 4–5 months of age. Single nuclei were extracted from the lumbar regions of the cord and subjected to cDNA synthesis using 10X Genomics (10X Chromium v3 chemistry). Barcoded cDNA libraries were sequenced using next-generation sequencing, and the snRNA-seq data were processed and analyzed for quality control using Seruat16. Using UMAP, we performed clustering analysis of snRNA-seq data from 20,284 nuclei that passed the quality control filter. The data was derived from wild-type (WT) mice (a pool of spinal lumbar sections from three mice) and gp120Tg mice (a pool of lumbar sections from three mice). This analysis generated a detailed atlas of transcriptomically distinct spinal cell clusters, including various types of neurons and glia (Fig. 1a).Fig. 1Single-nucleus RNA-seq (snRNA-seq) identified an astrocytic subpopulation in the lumbar spinal cord of gp120Tg mice.a An overview of cell clusters from WT (male) and gp120Tg (male) lumbar spinal cords. b, c Identification of astrocytes (red circle) by expression of astrocyte marker genes. d Subsetting of astrocytic cluster (a) revealed eight sub-clusters. e, f Splitting the merged astrocytic clusters (d) by genotypes revealed differences in specific clusters from WT and gp120Tg mice; note the marked decrease of clusters 1 and 2 and increase of cluster 5 in gp120Tg mice. Cluster 7 was a population specifically observed in gp120Tg mice. Clusters with a low expression level of Gfap are circled by blue lines; clusters with a high expression level of Gfap are circled by red lines; g Relative contribution of WT or gp120Tg cells to the individual astrocytic clusters in (d). h The expression of top 10 cluster-specific marker genes in astrocyte clusters from WT mice. Each column represented a nucleus, each row represented a gene. The markers in clusters 1 and 2 (green rectangles) were exemplified in the list on left. Note cluster 1 and 2 express a high level of astrocyte transporter genes Slc6a11 (red arrow) and Slc7a10 (blue arrow) and a very low level of Gfap (black arrow). On the contrary, clusters 3 and 4 express a very low level of Slc6a11 and Slc7a10 but a high level of Gfap. i Expression of WT cluster-specific markers (h) in the gp120Tg astrocyte clusters. j The heatmap showed the expression of top10 cluster-specific markers in astrocytes from the gp120Tg mice. Note cluster 7 expressed a unique set of gene markers (red rectangle). k Expression of selected genes in astrocytic subclusters shown by a dot plot. Cluster 7 has low levels of expression of genes related to astrocytic homeostatic function (orange rectangle), but high levels of expression of genes related to inflammation (red rectangle) and phagocytosis (blue rectangle). Scale bar: scaled average log2 value. Red arrow points to the expression profiles of Lgals3 (gene encoding Galectin 3 (Gal3)). Note the selective expression of Lgals3 in cluster 7. Single-nucleus RNA-seq (snRNA-seq) identified an astrocytic subpopulation in the lumbar spinal cord of gp120Tg mice. a An overview of cell clusters from WT (male) and gp120Tg (male) lumbar spinal cords. b, c Identification of astrocytes (red circle) by expression of astrocyte marker genes. d Subsetting of astrocytic cluster (a) revealed eight sub-clusters. e, f Splitting the merged astrocytic clusters (d) by genotypes revealed differences in specific clusters from WT and gp120Tg mice; note the marked decrease of clusters 1 and 2 and increase of cluster 5 in gp120Tg mice. Cluster 7 was a population specifically observed in gp120Tg mice. Clusters with a low expression level of Gfap are circled by blue lines; clusters with a high expression level of Gfap are circled by red lines; g Relative contribution of WT or gp120Tg cells to the individual astrocytic clusters in (d). h The expression of top 10 cluster-specific marker genes in astrocyte clusters from WT mice. Each column represented a nucleus, each row represented a gene. The markers in clusters 1 and 2 (green rectangles) were exemplified in the list on left. Note cluster 1 and 2 express a high level of astrocyte transporter genes Slc6a11 (red arrow) and Slc7a10 (blue arrow) and a very low level of Gfap (black arrow). On the contrary, clusters 3 and 4 express a very low level of Slc6a11 and Slc7a10 but a high level of Gfap. i Expression of WT cluster-specific markers (h) in the gp120Tg astrocyte clusters. j The heatmap showed the expression of top10 cluster-specific markers in astrocytes from the gp120Tg mice. Note cluster 7 expressed a unique set of gene markers (red rectangle). k Expression of selected genes in astrocytic subclusters shown by a dot plot. Cluster 7 has low levels of expression of genes related to astrocytic homeostatic function (orange rectangle), but high levels of expression of genes related to inflammation (red rectangle) and phagocytosis (blue rectangle). Scale bar: scaled average log2 value. Red arrow points to the expression profiles of Lgals3 (gene encoding Galectin 3 (Gal3)). Note the selective expression of Lgals3 in cluster 7. Previous studies have suggested an association of reactive astrocytes in the spinal dorsal horn (SDH) with the development of neuropathic pain in HIV patients3. Using a gp120 model mouse, we confirmed the role of reactive astrocytes in the pathogenesis of HIV-associated pain7. To identify gp120-induced specific subtypes of astrocytes contributing to pain pathogenesis, we subset astrocytic clusters revealed by feature plots of marker genes Aqp4 and Gja1 (Fig. 1b, c). These pooled astrocytes were subjected to further dimension reduction analysis17. From this analysis, we identified eight astrocytic subclusters from the combined WT and gp120Tg astrocytes (Fig. 1d). Splitting the clusters based on their genotypes (WT vs. gp120Tg) revealed marked differences between WT and gp120Tg astrocytic clusters (Fig. 1e, f). Supplementary Data 1 lists the number of astrocytes in each cluster in the WT and gp120Tg mice. Compared to their WT counterparts, the gp120Tg astrocytes in clusters 1 and 2 were significantly decreased, while those in cluster 5 were increased. Cluster 7 was specifically identified in gp120Tg mice (Fig. 1e–g). To gain insights into the biological functions of individual astrocytic clusters in the WT and gp120Tg datasets, we used Seurat FindAllMarkers to identify specific marker genes for each cluster in the WT and gp120Tg mice (Supplementary Data 2; Supplementary Data 3). These markers represent genes that are significantly upregulated or downregulated (adjusted p-value < 0.05) in a given astrocytic cluster compared to all other astrocytic clusters. As shown in Fig. 1h, the top 10 cluster-specific marker genes in the WT dataset exemplified the expression of the different clusters. We found that the WT astrocytic clusters could be separated into two major categories based on their Gfap expression levels, with clusters 1, 2, 5, and 6 showing low Gfap expression (Gfap-low) and clusters 3 and 4 showing high Gfap expression (Gfap-high) (Fig.1h; Supplementary Fig. 1a), consistent with previous observations13. Clusters 1 and 2 were the two major subpopulations expressing Slc6a11 and Slc7a10 (Fig.1h). Slc6a11 encodes an astrocyte neurotransmitter transporter that plays a key role in GABA uptake18, while Slc7a10 encodes a sodium-independent amino acid transporter with a primary role related to modulation of excitatory glutamatergic neurotransmission19. Clusters 3 and 4 expressed high Gfap levels but low levels of Slc6a11 and Slc7a10 (Fig.1h). In the gp120Tg mice, there was a clear shift from Gfap-low to Gfap-high populations in the astrocytic clusters (Supplementary Fig. 1b). To understand how the WT astrocytic clusters transformed into the gp120Tg astrocytic clusters, we examined the expression of the top 10 WT cluster-specific marker genes in the gp120Tg astrocytes (Fig. 1i). In the gp120Tg mice, there are eight clusters. Since clusters 6 and 8 are very small subpopulations, we did not pursue them in this study. From the gp120Tg heatmap (Fig. 1i), we can see that the marker genes of WT clusters 1 and 2 are highly expressed in gp120Tg clusters 1, 2, and 5. The WT marker genes of clusters 3 and 4 are highly expressed in gp120Tg clusters 3 and 4. These data suggest that, in gp120Tg mice, clusters 1, 2, and 5 originated from WT clusters 1 and 2, whereas clusters 3 and 4 were derived from WT clusters 3 and 4. Notably, cluster 7 did not express any WT cluster-specific marker genes, suggesting that cluster 7 is a de novo astrocyte cluster formed only in gp120Tg mice. To further determine the transcriptomic signature of Cluster 7, we generated gp120Tg cluster-specific marker gene lists (Supplementary Data 3). Figure 1j is a heatmap demonstrating the expression of the top 10 upregulated cluster-specific marker genes in gp120Tg astrocytes. The results clearly show that Cluster 7 expresses a unique set of genes. We named this cluster HIV-pain associated astrocytes (HIPA) based on their potential pathogenic contributions to pain associated with HIV infection (see below). HIPAs comprised 20% of total gp120Tg spinal astrocytes. To gain insight into their potential biological function, we imported the 112 HIPA (cluster 7)-specific marker genes from the gp120Tg dataset (Supplementary Data 3, cluster 7) into Ingenuity Pathway Analysis (IPA). IPA identified 33 well-characterized inflammatory genes (Supplementary Fig. 2) and 7 well-characterized phagocytic genes (Supplementary Fig. 3). Hence, HIPAs have enriched expression of genes related to inflammation and phagocytosis. To visualize the expression profiles of the top IPA-selected genes related to astrocytic homeostasis, inflammation, and phagocytosis, we generated a dot plot using the combined WT and gp120Tg dataset (Fig. 1k). The plot showed that HIPAs (cluster 7) expressed low levels of homeostatic genes (Gja1, Aqp4, Slc1a2, etc.) but high levels of inflammatory genes (Gfap, Anxa3, Nox4, Cdh13, Aldh1a1, Bcl2, Adamts5, Adra1a, Col4a5) and phagocytic genes (Lgals3, Palld, Ptprj, Rsas2, C3) (Fig. 1k). These data indicate that HIPAs are both inflammatory cells and phagocytes, which may contribute to the development of pain associated with HIV infection. Gal3 is a specific marker of HIPAs Among the up-regulated genes related to phagocytosis or inflammation, the expression of gene coding for Gal3 (Lgals3) was high and selective in HIPAs (cluster 7) (Fig. 1k). Hence, we used Gal3 as a transcriptomic marker for HIPAs. snRNA-seq data revealed that Gal3+ astrocytes compose ~6% of the entire spinal cord astrocytes (Supplementary Data 4) and predominantly located in cluster 7 (HIPA). Gal3 was reported to play a role in neuropathic pain in a peripheral nerve injury model20,21, suggesting that HIPAs are a potentially important contributor to HIV-associated pain pathogenesis. To further validate Gal3 as a marker for HIPAs, we performed double fluorescent immunostaining of GFAP and Gal3, followed by confocal imaging (Fig. 2a, d, e). As Gal3 was reported to express in microglia in the brain22, we performed a control experiment using Iba1, a microglial marker (Fig. 2b, d). We found that Gal3 was almost exclusively expressed in astrocytes but not in microglia in the gp120Tg spinal cords (Fig. 2a, b, d, e). Approximately 95% of the Gal3+ cells were GFAP+, while ~5% of GFAP+ cells were Gal3+ microglia (Fig. 2c), indicating Gal3+ cells are predominantly astrocytes. In some cases, Gal3+ cells were observed in clumps containing both GFAP+ and Iba1+ cells (Fig. 2e). However, optical sectioning image analysis showed that Gal3 signals originated from astrocytes rather than microglia (Fig. 2f–h). These results confirm that Gal3 is a specific marker for HIPAs in the gp120Tg spinal cords, as suggested by snRNA-seq data.Fig. 2Gal3 was predominantly expressed in spinal astrocytes in gp120Tg mice.a–d Double immunofluorescent staining (IF) of Gal3 (red) with astrocytic marker (GFAP) (white) and microglial marker Iba1 (green) respectively showed most Gal3+ cells were also GFAP+ (white open arrows) (a) in the gp120Tg lumbar spinal cord, although Gal3+ microglia were occasionally observed (white filled arrow) (b). SDH spinal dorsal horn. c Quantitative summary of Gal3+ microglia and astrocytes. n = 6 (male); unpaired two-tail t test; Data are presented as mean ± SEM. d Co-staining of Iba1, GFAP and Gal3 revealed Gal3 expressed in astrocyte (open arrows) but not microglia (filled arrows) in the gp120Tg spinal cord. e A 3D confocal image of a cell clump of microglia (filled arrow) and an astrocyte (open arrow) with Gal3 signals (red). f–h Different optical sections of the clump showed that Gal3 was exclusively in the astrocyte (open arrow) but not in the nearby microglia (filled arrow). Gal3 was predominantly expressed in spinal astrocytes in gp120Tg mice. a–d Double immunofluorescent staining (IF) of Gal3 (red) with astrocytic marker (GFAP) (white) and microglial marker Iba1 (green) respectively showed most Gal3+ cells were also GFAP+ (white open arrows) (a) in the gp120Tg lumbar spinal cord, although Gal3+ microglia were occasionally observed (white filled arrow) (b). SDH spinal dorsal horn. c Quantitative summary of Gal3+ microglia and astrocytes. n = 6 (male); unpaired two-tail t test; Data are presented as mean ± SEM. d Co-staining of Iba1, GFAP and Gal3 revealed Gal3 expressed in astrocyte (open arrows) but not microglia (filled arrows) in the gp120Tg spinal cord. e A 3D confocal image of a cell clump of microglia (filled arrow) and an astrocyte (open arrow) with Gal3 signals (red). f–h Different optical sections of the clump showed that Gal3 was exclusively in the astrocyte (open arrow) but not in the nearby microglia (filled arrow). Differentiation of HIPAs in the spinal cord To gain insights into the induction of HIPAs in gp120Tg spinal cords, we compared Gal3 protein levels in the spinal cord of WT and gp120Tg mice by western blotting (Fig. 3a) and found that Gal3 expression was low in WT but significantly upregulated in gp120Tg spinal cords (Fig. 3a). Immunostaining analysis confirmed Gal3 upregulation in the spinal cord of gp120Tg mice, with the induction of HIPAs in gp120Tg spinal cords (Fig. 3b–e). The increase of Gal3+ astrocytes in the gp120Tg spinal cord was further validated by the co-staining of Gal3 with SOX9 and GFAP (Supplementary Fig. 4). To demonstrate the Gal3 is specifically induced by gp120, we compared the effect of native and heat-denatured forms of purified recombinant gp120 spinally injected (i.t.), and observed that only native gp120, but not heat-denatured gp120, induced Gal3 expression (Supplementary Fig. 5), aligning with their activity in inducing pain6.Fig. 3Induction of HIPAs in the spinal cord of gp120Tg mice.a HIPA marker Gal3 was significantly upregulated in the spinal cords of gp120Tg mice. Wt: n = 4 (male); gp120Tg: n = 6 (male); unpaired two-tail t test; Data are presented as mean ± SEM. b IF revealed Gal3+ cells (red) were predominantly at the central canal (CC) region (arrow) in WT mice. c The spatial distribution of HIPAs (Gal3+ cells) in the gp120Tg spinal cord shown in a low power image. HIPAs were scattered (open arrowheads) in the gray matter (outlined by dashed lines in (b) and (c) from the CC regions (white arrow) toward to the dorsal horn and ventral horn. Compared with WT (b), Gal3 signals around the central canal (arrow) were diminished. d The relative Gal3 immunostaining signal in the entire spinal cord in (b) and (c) was quantified. Wt: n = 3 (male); gp120Tg: n = 3 (male); paired one-tail t-test; Box-and-whisker plots were generated in Microsoft Excel; boxes represent the interquartile range (25th–75th percentiles), the center line indicates the median, whiskers extend to data points within 1.5× the interquartile range. e A co-staining of Gal3 with GFAP and Vimentin demonstrated HIPAs (white arrows) forming a trajectory originating from the CC region to the outer margin of the gray matter in the gp120Tg mice. The inset is an enlarged selected area near the CC, demonstrating the colocalization of Gal3 with GFAP and Vimentin and the migration of HIPAs outside the CC. The experiment (e) was independently repeated at least three times. Induction of HIPAs in the spinal cord of gp120Tg mice. a HIPA marker Gal3 was significantly upregulated in the spinal cords of gp120Tg mice. Wt: n = 4 (male); gp120Tg: n = 6 (male); unpaired two-tail t test; Data are presented as mean ± SEM. b IF revealed Gal3+ cells (red) were predominantly at the central canal (CC) region (arrow) in WT mice. c The spatial distribution of HIPAs (Gal3+ cells) in the gp120Tg spinal cord shown in a low power image. HIPAs were scattered (open arrowheads) in the gray matter (outlined by dashed lines in (b) and (c) from the CC regions (white arrow) toward to the dorsal horn and ventral horn. Compared with WT (b), Gal3 signals around the central canal (arrow) were diminished. d The relative Gal3 immunostaining signal in the entire spinal cord in (b) and (c) was quantified. Wt: n = 3 (male); gp120Tg: n = 3 (male); paired one-tail t-test; Box-and-whisker plots were generated in Microsoft Excel; boxes represent the interquartile range (25th–75th percentiles), the center line indicates the median, whiskers extend to data points within 1.5× the interquartile range. e A co-staining of Gal3 with GFAP and Vimentin demonstrated HIPAs (white arrows) forming a trajectory originating from the CC region to the outer margin of the gray matter in the gp120Tg mice. The inset is an enlarged selected area near the CC, demonstrating the colocalization of Gal3 with GFAP and Vimentin and the migration of HIPAs outside the CC. The experiment (e) was independently repeated at least three times. In WT spinal cords, Gal3+ cells were largely restricted to the central canal area (Fig. 3b), with only a few sparse Gal3+ cells in dorsal horns and marginal areas (Fig. 3b). In contrast, in gp120Tg spinal cords, Gal3+ cells markedly increased, especially in the gray matter (Fig. 3c, e). Co-staining of Gal3 with GFAP and Vimentin revealed that HIPAs appeared to form a trajectory originating from the central canal to the outer margin of the gray matter in the gp120Tg mice (Fig. 3e). Concomitantly, Gal3+ cells were diminished in the central canal region of gp120Tg mice (Fig. 3c), giving an apparent impression that Gal3+ HIPAs were derived from Gal3+ precursors migrating out from the central canal region. In the central canal of WT mice, Gal3 signals were confined to ependymal cells lining the canal (Fig. 4a). Consistent with reported findings23, Gal3+ spinal ependymal cells expressed vimentin (Fig. 3e) but not GFAP (Fig. 4b–d). In contrast, in the gp120Tg mice, Gal3 signals were completely absent from the ependymal cell layer (Fig. 4e, h). Concomitantly, many Gal3+ cells were observed outside the ependymal cell layer (Figs. 3c,e and 4e,h,i); these expressed GFAP (Figs. 3e and 4h,i) and appeared to form a trajectory toward spinal dorsal and ventral horns (Fig. 3c,e). Unlike ependymal cells in the central canal of WT mice that were GFAP-negative (Fig. 4b, c), some spinal ependymal cells in the central canal of gp120Tg mice expressed GFAP (Fig. 4f, g), signaling their commitment to an astrocytic lineage. These data are consistent with the idea that Gal3+ cells in the central canal ependymal layer may be induced to migrate out and differentiate into HIPAs in the spinal dorsal and ventral horns in gp120Tg mice.Fig. 4Distribution of Gal3 in the central canal and the morphological features of HIPAs in the spinal cord of gp120Tg mice.a–h Distribution of Gal3 (red) protein in ependymal cells of spinal CCs (encircled by dash lines) in WT and gp120Tg mice. a In the WT mice, Gal3 protein was specifically expressed in ependymal cells around the CC (wide filled arrow); b, c Gal3+ cells in the WT CC did not express astrocytic marker GFAP; d no Gal3+GFAP+ cells were detected in the ependymal cell layer. e In the gp120Tg mice, the expression of Gal3 in ependymal cells was markedly diminished (wide open arrow); instead, clear Gal3 expression was detected in cells outside the ependymal layer (wide filled arrows); f, g in contrast to ependymal cells in WT mice (b, c), GFAP+ cells were observed in the ependyma layer in gp120Tg mice (thin arrows). h HIPAs (wide filled arrows) were GFAP+ cells outside the gp120Tg CC. i–k Confocal images showing morphological heterogeneity of HIPAs at different locations outside the ependymal layer in the gp120Tg spinal cord. i Most HIPAs (wide filled arrows) close to the CC region were smaller and Gal3 protein in these cells was largely restricted to the nucleus. Insert is a high-power image of the selected region demonstrating that Gal3 is expressed largely in the nucleus area of a HIPA close to CC. Experiments (a–i) were independently repeated at least three times. j A 3D Z-stack image demonstrates a HIPA (wide arrow) in the SDH with large cellular volume and more process with Gal3 distributed throughout the whole astrocyte including the processes. Gal3-negative astrocytes (thin arrow) were smaller with fewer short processes. k A 3D Z-stacked confocal image to show the heterogenous morphologies of HIPAs in SDH. Filled arrow pointed to a HIPA with a relatively smaller volume and a few long processes. Open arrow pointed to a HIPA with bigger cell volume body and many short processes. Note Gal3 signals (red) across the whole cells in these HIPAs. Distribution of Gal3 in the central canal and the morphological features of HIPAs in the spinal cord of gp120Tg mice. a–h Distribution of Gal3 (red) protein in ependymal cells of spinal CCs (encircled by dash lines) in WT and gp120Tg mice. a In the WT mice, Gal3 protein was specifically expressed in ependymal cells around the CC (wide filled arrow); b, c Gal3+ cells in the WT CC did not express astrocytic marker GFAP; d no Gal3+GFAP+ cells were detected in the ependymal cell layer. e In the gp120Tg mice, the expression of Gal3 in ependymal cells was markedly diminished (wide open arrow); instead, clear Gal3 expression was detected in cells outside the ependymal layer (wide filled arrows); f, g in contrast to ependymal cells in WT mice (b, c), GFAP+ cells were observed in the ependyma layer in gp120Tg mice (thin arrows). h HIPAs (wide filled arrows) were GFAP+ cells outside the gp120Tg CC. i–k Confocal images showing morphological heterogeneity of HIPAs at different locations outside the ependymal layer in the gp120Tg spinal cord. i Most HIPAs (wide filled arrows) close to the CC region were smaller and Gal3 protein in these cells was largely restricted to the nucleus. Insert is a high-power image of the selected region demonstrating that Gal3 is expressed largely in the nucleus area of a HIPA close to CC. Experiments (a–i) were independently repeated at least three times. j A 3D Z-stack image demonstrates a HIPA (wide arrow) in the SDH with large cellular volume and more process with Gal3 distributed throughout the whole astrocyte including the processes. Gal3-negative astrocytes (thin arrow) were smaller with fewer short processes. k A 3D Z-stacked confocal image to show the heterogenous morphologies of HIPAs in SDH. Filled arrow pointed to a HIPA with a relatively smaller volume and a few long processes. Open arrow pointed to a HIPA with bigger cell volume body and many short processes. Note Gal3 signals (red) across the whole cells in these HIPAs. We observed evident morphological changes of HIPAs at different positions in relation to the central canal. HIPAs in areas adjacent to the central canal ependymal layer were relatively small with shorter processes, and Gal3 protein was mainly restricted to nuclei (Fig. 4i). In contrast, HIPAs in regions farther away from the central canal, particularly in the dorsal horn, showed an increase in morphological complexity, with larger size compared to non-HIPA astrocytes (Fig. 4j) and more processes (Fig. 4j, k). In these HIPAs, Gal3 protein was observed throughout the entire cell, including the cytoplasm and processes (Fig. 4j, k). HIPAs in the spinal dorsal horn exhibited more heterogeneous in morphology, some with relatively smaller cell bodies and a few long processes and others with larger bodies and many short processes (Fig. 4k). By using Imaris AI software to create surface of DAPI staining, we also identified HIPAs with multiple nuclei (Supplementary Fig. 6). To assess the relevance of HIPAs in the gp120Tg model, we examined spinal autopsies from HIV patients. We found that Gal3 protein was significantly induced in the spinal cord of HIV patients compared to control patients who died from non-HIV causes (Supplementary Fig. 7a). Additionally, immunofluorescent staining showed a drastic increase in Gal3+ astrocytes in the spinal cords of HIV patients (Supplementary Fig. 7b, c), indicating that HIPAs are also induced following HIV infection, like in the gp120Tg model. HIPAs are phagocytes of neurons Previous studies have shown that cells expressing Gal3, such as Gal3+ macrophages, act as phagocytes and release Gal3 as an opsonin to label target cells for phagocytosis24,25. Our discovery of upregulated phagocytosis-related genes in HIPAs (Fig. 1k) suggests that this population of Gal3+ astrocytes have the potential to function as phagocytes to clear damaged neurons induced by HIV neurotoxins, such as gp120. Given that our previous work has revealed neurodegeneration in the SDH of HIV patients with neuropathic pain6 and in a mouse model of HIV-associated pain6,26, we hypothesized that HIPAs could contribute to neurodegeneration as phagocytes. To evaluate the capability of HIPAs to phagocytose neurons, we performed double fluorescent immunostaining experiments to label NeuN (a marker of neuronal soma) and Gal3. Then we used ZEISS Airyscan confocal super-resolution microscopy and took 3D Z-stack confocal images demonstrating the interaction of HIPA with neuronal soma (Fig. 5a, Supplementary Video 1). By using Imaris AI software to create solid surfaces for NeuN and Gal3 signals, we can clearly view from different angles the neuronal somas enwrapped/engulfed by a HIPA (Fig. 5b–d).Fig. 5HIPAs engulfed neuronal soma in the gp120Tg spinal cord.a A 3D Z-stack confocal image taken by ZEISS Airyscan confocal super-resolution microscopy to demonstrate the interaction of a HIPA (wide open arrow) with neuronal somas (filled arrow). b Imaris AI software removed the Gal3 signals and created solid green surfaces of NeuN signals to demonstrate the neuronal somas. c Imaris created solid surfaces of Gal3 signals (pink) and NeuN signals (green) to demonstrate the neuronal somas (open arrows) enwrapped by the processes of a HIPA (pink). d A view of image (c) from a different angle. e Shown here were two representative HIPAs (red) (open wide arrows) in close interactions with neurons (green) (thin arrows). f–k Imaris 3D image segmentation of the selected area in (e). This image analysis removed all the signals that were not interact with HIPAs to visualize NeuN signals engulfed by HIPAs. f–h Imaris image segmentations of one region in (e) to show intact neuronal somas (thin white arrows) engulfed by a HIPA. i–k Imaris image segmentations of another region in (e) to show neuronal debris (thin red arrows) inside a HIPA. The engulfed neuronal somas revealed by creating the solid opaque surfaces (open arrow) (f and i), transparent surfaces (open arrow) (g and j), and by removing HIPA components (h and k). l Fraction of Gal3+ astrocytes with phagocyted neuronal components. Wt: n = 3 (male); gp120Tg: n = 3 (male); paired two-tail t test; Box-and-whisker plots were generated in Microsoft Excel; boxes represent the interquartile range (25th–75th percentiles), the center line indicates the median, whiskers extend to data points within 1.5× the interquartile range. m Combination of Fluoro-Jade C (FJC) staining and Gal3 IF staining revealed extensive overlap (arrows) of degenerative neurons revealed by FJC staining and HIPAs revealed by Gal3 IF staining. The experiment was independently repeated 3 times. n Quantification of FJC staining signals in the WT and gp120Tg spinal cords. Wt: n = 4 (male); gp120Tg: n = 5 (male); unpaired t two-tail test with Welch’s correction; Data are presented as mean ± SEM. HIPAs engulfed neuronal soma in the gp120Tg spinal cord. a A 3D Z-stack confocal image taken by ZEISS Airyscan confocal super-resolution microscopy to demonstrate the interaction of a HIPA (wide open arrow) with neuronal somas (filled arrow). b Imaris AI software removed the Gal3 signals and created solid green surfaces of NeuN signals to demonstrate the neuronal somas. c Imaris created solid surfaces of Gal3 signals (pink) and NeuN signals (green) to demonstrate the neuronal somas (open arrows) enwrapped by the processes of a HIPA (pink). d A view of image (c) from a different angle. e Shown here were two representative HIPAs (red) (open wide arrows) in close interactions with neurons (green) (thin arrows). f–k Imaris 3D image segmentation of the selected area in (e). This image analysis removed all the signals that were not interact with HIPAs to visualize NeuN signals engulfed by HIPAs. f–h Imaris image segmentations of one region in (e) to show intact neuronal somas (thin white arrows) engulfed by a HIPA. i–k Imaris image segmentations of another region in (e) to show neuronal debris (thin red arrows) inside a HIPA. The engulfed neuronal somas revealed by creating the solid opaque surfaces (open arrow) (f and i), transparent surfaces (open arrow) (g and j), and by removing HIPA components (h and k). l Fraction of Gal3+ astrocytes with phagocyted neuronal components. Wt: n = 3 (male); gp120Tg: n = 3 (male); paired two-tail t test; Box-and-whisker plots were generated in Microsoft Excel; boxes represent the interquartile range (25th–75th percentiles), the center line indicates the median, whiskers extend to data points within 1.5× the interquartile range. m Combination of Fluoro-Jade C (FJC) staining and Gal3 IF staining revealed extensive overlap (arrows) of degenerative neurons revealed by FJC staining and HIPAs revealed by Gal3 IF staining. The experiment was independently repeated 3 times. n Quantification of FJC staining signals in the WT and gp120Tg spinal cords. Wt: n = 4 (male); gp120Tg: n = 5 (male); unpaired t two-tail test with Welch’s correction; Data are presented as mean ± SEM. To visualize neuronal components inside HIPAs, we performed Imaris image segmentation analysis on representative Gal3+ cells (Fig. 5e). This analysis removed NeuN signals not interacting with HIPAs, allowing us to visualize the NeuN signals inside HIPAs (Fig. 5f–k). By creating solid opaque HIPA surfaces (Fig. 5f, i), transparent HIPA surfaces (Fig. 5g, j), or by removing all HIPA components (Fig. 5h, k), we were able to reveal the neuronal components engulfed by HIPAs. These image analyses showed intact neuronal somas (Fig. 5h) or debris (Fig. 5k) inside HIPAs, demonstrating that HIPAs are astrocytic phagocytes capable of phagocytosing neuronal components in the SDH of gp120Tg mice. Moreover, we made 3D videos using Imaris AI software to further confirm HIPAs phagocytosis of neuronal somas in the gp120Tg spinal cords (Supplementary Video 1, 2). Approximately 15% of Gal3+ astrocytes in the gp120Tg spinal cord demonstrated phagocyted neuronal components (Fig. 5l). To examine the association of HIPAs with neuronal degeneration, we performed Fluoro-Jade C (FJC) staining combined with Gal3 immunofluorescent staining. We observed a significant increase of FJC signals in the spinal cords of the gp120Tg mice compared to the WT mice (Fig. 5m, n), indicating an increase of degenerated neurons. Furthermore, we found that FJC-labeled degenerated neurons co-localized with Gal3 signals (Fig. 5m), suggesting an association of HIPAs with degenerated neurons. Gal3 knockout diminished HIPAs in gp120Tg spinal cords Given the proposed roles of Gal3 in regulation of cell growth and differentiation27,28, we hypothesized that Gal3 is involved in the formation of HIPAs. To test this, we generated gp120Tg mice with Gal3 KO (gp120Tg/Gal3−/−) by crossing gp120Tg mice with mice carrying a Gal3 targeting mutation29. Western blotting analysis confirmed that Gal3 KO significantly attenuated the upregulation of Vim and GFAP protein expression in the gp120Tg spinal cord (Supplementary Fig. 8a–c), indicating that Gal3 is important for the astrocytic reaction induced by gp120. We then performed snRNA-seq analysis to investigate if HIPAs could still form in gp120Tg/Gal3−/− mice. Integrated UMAP clustering analysis of pooled 48,935 lumbar spinal nuclei from wild-type (n = 3), gp120Tg (n = 3), gp120Tg/Gal3−/− (n = 3), and Gal3−/− (n = 3) mice showed all major spinal cell types, including different types of neurons and glia (Fig. 6a). Astrocytes were subset for a second-dimension reduction analysis (Fig. 6b, c) and the astrocytic clusters were displayed in an integrated UMAP plot (Fig. 6d). Splitting the astrocytic clusters based on their genotypes revealed that, similar to the results shown in Fig. 1e, f, cluster 7 (HIPAs) was specifically observed in the gp120Tg mice and largely disappeared in WT (Fig. 6e–f). Importantly, this group of astrocytes induced in gp120Tg mice was largely absent in gp120Tg/Gal3−/− mice (Fig. 6g, i), suggesting that Gal3 KO prevented HIPA formation induced by gp120. Moreover, expression of the top 50 upregulated genes in HIPAs, including genes involved in inflammatory and phagocytic pathways, were also dramatically attenuated in the gp120Tg/Gal3−/− mice (Fig. 6j–l). Astrocytes expressing inflammatory genes (e.g., Anxa3) and phagocytotic gene (e.g., C3), which were increased in the gp120Tg mice (Fig. 6o, s) compared to WT (Fig. 6n, r) and primarily in cluster 7 (HIPAs), were diminished in the gp120Tg/Gal3−/− mice (Fig. 6p, t). In contrast, Gal3 KO in WT mice did not induce evident changes in astrocytic clusters (Fig. 6e,h,n,q,r,u) and the expression profiles of HIPA-associated genes (Fig. 6m). To confirm the effect of Gal3 KO on HIPA formation, we performed RNAscope analysis of C3, whose expression is enriched in HIPAs (Fig. 1k). The results showed that C3+ astrocytes increased in gp120Tg mice and the increase was blocked in gp120Tg/Gal3−/− mice (Supplementary Fig. 9). These data strongly suggest that Gal3 induced by gp120 is important for the formation of HIPAs.Fig. 6Gal3 KO in gp120Tg mice diminished HIPAs (cluster 7) and attenuated the activation of other subpopulations of astrocytes.a A UMAP cluster plot of the integrated snRNA-seq datasets from WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− mice (male) revealed clusters of the main cell types in lumbar spinal cords. Feature plots of astrocyte marker genes Aqp4 (b) and Gfap (c) to show the astrocytic cluster. d A subset analysis of combined astrocytes from WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− datasets. e–h Splitting the overlayed astrocytic clusters (d) by genotypes. h The cluster pattern of Gal3 KO astrocytes (h) was similar to that of WT (e). i Quantitative analysis of the relative population sizes of nuclei of individual clusters from different genotypes. Note that Gal3 KO depleted HIPAs (cluster 7) in gp120Tg mice, prevented the decrease of the two homeostatic astrocytic subpopulations (cluster 1, 2), and attenuated the increase of cluster 5. j–m Comparison of the expression of top 50 upregulated genes in HIPA (cluster 7 in f) in gp120Tg (k), WT (j), gp120Tg/Gal3 KO (l), or Gal3 KO (m) astrocytic subclusters. Genes included those implicated in inflammatory and phagocytotic pathways. The color bars on top of individual heatmaps were used to represent individual clusters, and the color scale bar on the right represented the expression levels (log2 values). The red arrows point to cluster 7 in individual heatmaps. Note there is no cluster 7 in Gal3 KO mice. Compared with WT (j), these genes were significantly upregulated in cluster 7 of gp120Tg astrocytes (k), but the upregulation was blocked in gp120Tg/Gal3 KO astrocytes (l). Compared with the heatmap of WT astrocytes (j), Gal3 KO (m) did not evidently alter the expression of these genes. n–q Astrocytes expressing inflammatory gene Anxa3 (enclosed by the red rectangles) in the WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− spinal cords. r–u Astrocytes expressing phagocytotic gene C3 (enclosed by the red rectangles) in the WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− spinal cords. Scale bar: gene expression (log2 value). Gal3 KO in gp120Tg mice diminished HIPAs (cluster 7) and attenuated the activation of other subpopulations of astrocytes. a A UMAP cluster plot of the integrated snRNA-seq datasets from WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− mice (male) revealed clusters of the main cell types in lumbar spinal cords. Feature plots of astrocyte marker genes Aqp4 (b) and Gfap (c) to show the astrocytic cluster. d A subset analysis of combined astrocytes from WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− datasets. e–h Splitting the overlayed astrocytic clusters (d) by genotypes. h The cluster pattern of Gal3 KO astrocytes (h) was similar to that of WT (e). i Quantitative analysis of the relative population sizes of nuclei of individual clusters from different genotypes. Note that Gal3 KO depleted HIPAs (cluster 7) in gp120Tg mice, prevented the decrease of the two homeostatic astrocytic subpopulations (cluster 1, 2), and attenuated the increase of cluster 5. j–m Comparison of the expression of top 50 upregulated genes in HIPA (cluster 7 in f) in gp120Tg (k), WT (j), gp120Tg/Gal3 KO (l), or Gal3 KO (m) astrocytic subclusters. Genes included those implicated in inflammatory and phagocytotic pathways. The color bars on top of individual heatmaps were used to represent individual clusters, and the color scale bar on the right represented the expression levels (log2 values). The red arrows point to cluster 7 in individual heatmaps. Note there is no cluster 7 in Gal3 KO mice. Compared with WT (j), these genes were significantly upregulated in cluster 7 of gp120Tg astrocytes (k), but the upregulation was blocked in gp120Tg/Gal3 KO astrocytes (l). Compared with the heatmap of WT astrocytes (j), Gal3 KO (m) did not evidently alter the expression of these genes. n–q Astrocytes expressing inflammatory gene Anxa3 (enclosed by the red rectangles) in the WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− spinal cords. r–u Astrocytes expressing phagocytotic gene C3 (enclosed by the red rectangles) in the WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− spinal cords. Scale bar: gene expression (log2 value). Gal3 KO attenuates pathogenic processes of gp120-induced pain Next, we sought to understand the potential contribution of the HIPA population of reactive astrocytes to pain pathogenesis. Firstly, we determined the effect of Gal3 KO on mechanical allodynia in gp120Tg mice. As previous studies have shown, young adult gp120Tg mice (2 months) do not manifest mechanical allodynia measured by von Frey testing30. However, we observed that mechanical allodynia began to manifest at later ages (4–5 months) (Fig. 7a). Importantly, expression of mechanical allodynia in gp120Tg mice (4-5 months) was blocked by global KO of Gal3 (gp120Tg/Gal3−/−) or heterozygous KO of astrocytic Gal3 in the gp120Tg mice (gp120Tg/Gal3CKO+/−) (Fig. 7a), indicating that Gal3 in HIPA contributes to the development of mechanical allodynia in gp120Tg mice. In addition, we performed Hargreave tests to assess the role of Gal3 in thermal pain (Supplementary Fig. 10). Compared with WT mice, gp120Tg mice manifested significant increase of thermal pain sensitivity, and the thermal pain hypersensitivity was significantly attenuated by global Gal3 KO in the gp120Tg mice. In contrast to its effect on mechanical pain hypersensitivity, heterozygous KO of astrocytic Gal3 in gp120Tg mice did not significantly attenuate thermal pain, indicating astrocytic Gal3’s differential role in mediating the development of distinct pain modalities.Fig. 7Gal3 KO attenuated mechanical hyperalgesia, neuronal degeneration, and pain related signaling in gp120Tg mice.a Von Frey test of mechanical hyperalgesia. Gp120Tg mice (5 months) developed mechanical hyperalgesia. Global KO of Gal3 or heterozygous KO of astrocyte specific Gal3 blocked the expression of mechanical hyperalgesia of gp120Tg mice. Comparisons between groups were performed using unpaired t-test. Wt: n = 13; gp120Tg: n = 13; gp10Tg/Gal3KO: n = 10; gp120Tg/Gal3CKO: n = 6. Mixed males and females were used for the behavior test. Data are presented as mean ± SEM. b–e FJC staining of lumbar spinal dorsal horns from WT (b), gp120Tg (c), gp120Tg/Gal3−/− (d) mice. FJC staining signals (arrows) were markedly increased in gp120Tg mice (c), which were attenuated by Gal3 KO (d). Inserts are the enlarged images of the selected areas to show the FJC signals. Asterisk: central canal. e Quantification of FJC staining signals. Wt: n = 4; gp120Tg: n = 5; gp120Tg/Gal3−/−: n = 4. Data are presented as mean ± SEM. f Immunoblotting comparison of spinal protein levels of selected molecular markers of neuroinflammation in WT, gp120Tg, gp120Tg/Gal3−/− mice. Quantification of active Il-1β (g), pro-IL-1β (h), active-caspase1 (i), Wnt5a (j), and Iba1 (k). g–k Wt: n = 4 (male); gp120Tg: n = 6 (male); gp120/Gal3−/−: n = 3 (male); Data was analyzed by One-way ANOVA and unpaired t test. g–j Data are presented as mean ± SD. k Data are presented as mean ± SEM. Gal3 KO attenuated mechanical hyperalgesia, neuronal degeneration, and pain related signaling in gp120Tg mice. a Von Frey test of mechanical hyperalgesia. Gp120Tg mice (5 months) developed mechanical hyperalgesia. Global KO of Gal3 or heterozygous KO of astrocyte specific Gal3 blocked the expression of mechanical hyperalgesia of gp120Tg mice. Comparisons between groups were performed using unpaired t-test. Wt: n = 13; gp120Tg: n = 13; gp10Tg/Gal3KO: n = 10; gp120Tg/Gal3CKO: n = 6. Mixed males and females were used for the behavior test. Data are presented as mean ± SEM. b–e FJC staining of lumbar spinal dorsal horns from WT (b), gp120Tg (c), gp120Tg/Gal3−/− (d) mice. FJC staining signals (arrows) were markedly increased in gp120Tg mice (c), which were attenuated by Gal3 KO (d). Inserts are the enlarged images of the selected areas to show the FJC signals. Asterisk: central canal. e Quantification of FJC staining signals. Wt: n = 4; gp120Tg: n = 5; gp120Tg/Gal3−/−: n = 4. Data are presented as mean ± SEM. f Immunoblotting comparison of spinal protein levels of selected molecular markers of neuroinflammation in WT, gp120Tg, gp120Tg/Gal3−/− mice. Quantification of active Il-1β (g), pro-IL-1β (h), active-caspase1 (i), Wnt5a (j), and Iba1 (k). g–k Wt: n = 4 (male); gp120Tg: n = 6 (male); gp120/Gal3−/−: n = 3 (male); Data was analyzed by One-way ANOVA and unpaired t test. g–j Data are presented as mean ± SD. k Data are presented as mean ± SEM. Since Gal3 is required for the formation of HIPAs (Fig.6e–g, i), it is probable that Gal3 controls the expression of gp120-induced allodynia via HIPAs. How might HIPAs contribute to the pain pathogenesis initiated by gp120? Previous studies on HIV patient autopsies and animal models suggest that neurodegeneration, indicated by synaptic degeneration, is associated with HIV pain pathogenesis6,26,31. As HIPAs are neuronal phagocytes (Fig. 5), we determined the effect of Gal3 KO on neurodegeneration in the SDH of gp120Tg mice (5 months), by FJC staining to detect the degenerative neurons in WT, gp120Tg, and gp120Tg/Gal3−/− mice (5 months). Compared with WT spinal cords, gp120Tg spinal cords had significantly more FJC staining signals (Fig. 7b, c, e), indicating an increase of neurodegeneration. Gal3 KO blocked the increase of FJC signals in spinal cords of gp120Tg mice (Fig. 7d, e), suggesting Gal3 was involved in gp120-induced neurodegeneration in the SDH pain circuits. Neuroinflammation in the SDH is a biomarker for pain development in HIV patients6, with activation of astrocytes being a hallmark specifically in HIV patients with chronic pain3. These findings suggest that HIPAs are probably among the reactive astrocytes that contribute to the pathogenic neuroinflammation underlying HIV-associated pain. Aligning with this hypothesis, snRNA-seq analysis revealed upregulated expression of inflammatory genes in HIPAs (Fig. 1k, Supplementary Fig. 2). To test this hypothesis, we compared protein levels of neuroinflammation markers in the spinal cords of WT, gp120Tg, and gp120Tg/Gal3−/− mice (5 months) using immunoblotting analysis (Fig. 7f). Specifically, we focused on neuroinflammation markers IL-1β, Wnt5a, and reactive microglia due to their confirmed contribution to the development of gp120-induced pain7. Our results showed that cleaved mature IL-1β was upregulated in gp120Tg spinal cords, and this upregulation was abolished in gp120Tg/Gal3−/− mice (Fig. 7g). However, pro-IL-1β levels were not significantly different among WT, gp120Tg, and gp120Tg/Gal3−/− mice (Fig. 7h). These data suggest that Gal3 regulates IL-1β at the stage of its maturation rather than at transcription or translation stages. Since reactive astrocytes are the source of gp120-induced IL-1β7 and Gal3 KO diminished HIPA (Fig. 6g, i), we reasoned that Gal3 likely controls IL-1β maturation in HIPAs. To gain insight into the potential Gal3-regulated IL-1β activation mechanism, we tested the involvement of inflammasomes. We measured the levels of active-caspase 1, which is the inflammasome-associated pro-IL-1β processing proteinase and observed that active-caspase1 was significantly upregulated in gp120Tg mice (Fig. 7f, i). Strikingly, Gal3 KO not only abolished the upregulation but also brought the active-caspase1 level drastically below the basal level (Fig. 7i). These results indicate that Gal3 regulates gp120-induced IL-1β via controlling the activation of inflammasomes in HIPAs. Wnt5a is a regulator of neuroinflammation that is specifically elevated in the SDH of HIV-positive patients experiencing pain32 and plays a significant role in the pathogenesis of gp120-induced pain33. We observed that spinal Wnt5a was markedly upregulated in gp120Tg mice and that Gal3 KO significantly mitigated this increase in gp120Tg/Gal3−/− mice (Fig. 7j). Reactive microglia are an important source of inflammatory mediators and critical for the expression of the early phase of gp120-induced pain7. In line with previous research26, we observed upregulation of the microglia/macrophage marker Iba1 in the spinal cord of gp120Tg mice (Fig. 7k), suggesting microglia activation. Importantly, we discovered that Gal3 KO attenuated the Iba1 upregulation (Fig. 7k), indicating an inhibitory effect of Gal3 KO on microglial activation. Discussion By snRNA-seq analysis, we identified a population of reactive astrocytes (termed HIPAs) in the lumbar spinal cord of HIV gp120Tg mice. We also elucidated a Gal3-mediated mechanism for HIPA formation. We further showed that HIPAs played a role in controlling the development of gp120-induced pain. Reactive astrocytes emerge as key contributors to the development of pathological pain4,5. In the context of HIV-associated pain, the observations of astrogliosis specifically in the SDH of patients with pain3 and blockage of pain development after ablation of reactive astrocytes in mouse models7 suggest the involvement of reactive astrocytes in the pathogenesis of this condition. However, the specific subtype of astrocytes responsible for HIV-associated pain was unknown. Our studies reported here show that HIPAs are a subtype of reactive astrocytes induced in the spinal cord of gp120Tg mice. Depletion of HIPAs by Gal3 KO was associated with diminished allodynia, demonstrating their contribution to HIV-related pain pathogenesis in animal models. Furthermore, the finding of Gal3+ astrocytes in the spinal cord of HIV patients (Supplementary Fig. 7) suggests that HIPAs may also contribute to pain pathogenesis in patients. HIPAs express multiple upregulated inflammatory genes, including Annexin A3 (Anxa3), Nox4, and Adamts5. Upregulations of these genes have been implicated in pain pathogenesis in neuropathic pain models34–36. For instance, Anxa3 was upregulated in the spinal cords of a rat neuropathic pain model induced by chronic constriction injury (CCI), and its inhibition alleviated CCI-induced mechanical allodynia and thermal hyperalgesia34. Nox4 is one of the major sources of intracellular reactive oxygen species (ROS)37 and promotes caspase 1 production and IL-1β activation38, both of which are known to play important roles in pain pathogenesis39,40. ADAMTSs are in the family of zinc metalloendopeptidases that participate in diverse biological processes, including extracellular matrix remodeling and inflammation41. Adamts5 is a protease of Reelin, the extracellular matrix glycoprotein in the spinal cord dorsal horn42. Reduction of Reelin resulted in mechanical and thermal hyperalgesia36,43. The upregulated expression of these inflammatory genes in HIPAs suggests that they are inflammatory cells that modulate neuroinflammation during pain pathogenesis. Consistent with this notion, HIPA depletion induced by Gal3 KO significantly attenuated the expression of these genes (Fig.6j–l, n–p) and neuroinflammation in the spinal cords of gp120Tg/Gal3−/− mice, as shown by the reversal of upregulated proinflammatory mediator IL-1β (Fig. 7g) and proinflammatory regulator Wnt5a (Fig. 7j), both of which are critical for gp120-induced pain pathogenesis7,33. HIPAs express high levels of phagocytosis-related genes, including Lgals3, C3, and Palld. Gal3 (Lgals3) and C3 act as phagocytosis ligands, where their deposition on the surface of target cells facilitates phagocytosis44,45. Anxa3, an inflammation-related gene expressed in HIPAs, also plays a specialized role in phagocytosis by mediating calcium-dependent granules-phagosome fusion46. The expression of phagocytosis-related genes suggests that HIPAs are phagocytes and may promote neurodegeneration in the pain circuits observed in HIV patients and mouse models of HIV-associated pain6. In support of this idea, we observed internalized neuronal components within HIPAs (Fig. 5). Furthermore, depletion of HIPAs is associated with a significant decrease in neuronal degeneration in gp120Tg/Gal3−/− mice (Fig. 7e). It will be interesting for future studies to determine if HIPAs show any pruning selectivity between excitatory and inhibitory neurons. The viral tools that have been developed for imaging analysis of astrocytic phagocytosis of excitatory and inhibitory synapses will be useful for this purpose47. It will also be important for future investigation to determine the role of phagocytic receptors of Gal3+ astrocytes (e.g., MerTK and MEGF10)48 in the gp120-induced phagocytosis. Previous studies have shown that microglia also play a critical role in gp120-induced synaptic degeneration in the SDH26. It will be interesting for future studies to investigate how HIPAs and microglia coordinate gp120-induced synapse degeneration in the spinal neural circuits. The relevance of synaptic degeneration in the pathogenesis of HIV-associated pain is supported by the finding that synapse loss is specifically observed in the SDH of HIV patients who developed pathological pain but not in HIV patients without pain complication6. Gal3+ reactive astrocytes, named Mac2 astrocytes, were identified in the spinal cords of various amyotrophic lateral sclerosis (ALS) mouse models49. Like HIPAs, Mac2 cells also expressed other astrocytic markers, such as GFAP, ALDH1L1, and S100β, but not microglial markers. Mac2 cells also manifested morphologies similar to HIPAs, including large round cell bodies and relatively short processes. As ALS patients develop pain50, it will be significant to test if Mac2 cells also contribute to ALS-related pain pathogenesis. The observation of almost complete depletion of HIPAs in the spinal cords of gp120Tg/Gal3−/− mice suggests that Gal3 plays a key role in HIPA genesis (Fig. 6g, i). In WT mice, Gal3 was expressed at a basal level in cells restricted to the ependymal layer around the spinal central canal (Figs. 3b and 4a, d). However, in gp120Tg mice, Gal3 protein expression was markedly upregulated (Fig. 3a, d), and Gal3+ cells in the central canal were reduced (Figs. 3c and 4e, h). Instead, most Gal3+ cells were observed out of the central canal region (Figs. 3c and 4e, h, i), forming an apparent trajectory from the central canal to the spinal marginal area (Fig. 3e). These observations raised the possibility that HIPAs originate from Gal3+ ependymal cells of the central canal, which are a pool of neural stem cells with the potential to differentiate into discrete multilineages51. Previous studies have shown that Gal3 is implicated in glial fate choices in stem cell differentiation in the postnatal subventricular zone of mouse brains52. Gal3 overexpression promotes astrocytic differentiation while simultaneously suppressing oligodendrogenesis52. Indeed, in the gp120Tg central canal, GFAP expression was detected in ependymal cells that were Gal3+ (Fig. 4h), suggesting that Gal3+ ependymal cells adopt an astrocytic lineage. These data seem to align with the possibility stated above. However, at this stage we cannot rule out alternative possibilities, such as HIPA differentiation from other astrocytes in response to gp120-induced signals. Further direct evidence is needed to validate the idea. Gal3 has been shown to regulate astrocyte differentiation through multiple mechanisms, both during normal development in neurogenic niches such as the subventricular zone (SVZ) and in disease conditions such brain injury53. In the SVZ of postnatal mice, Gal3 overexpression increases gliogenesis, favoring astrocyte generation over oligodendrocytes, by enhancing BMP signaling which biases neural stem cells toward astrocyte lineage commitment, whereas Gal3 knockdown or KO attenuate gliogenesis52. Gal3 may also be involved in modulating astrocyte maturation and maintaining astrocyte structural integrity during differentiation54–56. In brains with intracerebral hemorrhages, Gal3 marks proliferating astrocytes and probably regulates reactive astrogliosis28. It remains to be tested if HIPA genesis utilizes the Gal3-regulated processes reported during development and brain disease complications or mechanism. Methods Animal All animal procedures were performed in compliance with protocol 0904031B approved by the Institutional Animal Care and Use Committee at the University of Texas Medical Branch, and in accordance with all relevant guidelines and regulations. Gp120 Tg mice used in this study (provided by Dr. Marcus Kaul) express HIV-1 LAV gp120 under the control of the glial fibrillary acidic protein (Gfap) promoter15. Lgals3 Homozygotes KO (Gal3−/−) mice were obtained from the Jackson Laboratory (Strain: #006338). We received Lgals3-Flox mice57 from Dr. Danial Saban. Gfap-Cre mice were purchased from purchased from Jackson Laboratory (Strain: #024098). Mice were housed in a specific pathogen-free (SPF) facility at UTMB under controlled environmental conditions, including a 12-h light/dark cycle (lights on at 7:00 AM and off at 7:00 PM), ambient temperature maintained at 22 ± 2 °C, and relative humidity of 40–60%. Standard rodent chow and water were provided. Mice were anesthetized with 14% urethane or solfuorane. All primer sequences used for genotyping are included in the Supplementary Data 5. Male mice were used for snRNA-seq, western blot, injections, and immunofluorescent staining experiments. Mixed males and females were used for behavioral testing and RNAscope validation of the snRNA-seq data. Lumbar spinal cord dissection We performed spinal nuclei dissociation using the same tissue dissociation procedures for all experimental groups. To isolate nuclei for snRNA-seq, we euthanized WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− mice with 14% urethane followed by decapitation. We rapidly dissected out the lumbar spinal cords and temporarily stored them in 6 ml of cold Hibernate A/B27 (HABG) medium for further processing. HABG was prepared by adding 30 ml of Hibernate A (BrainBits, cat. no. HA), 600 µl of B27 (ThermoFisher, cat. no.17504044), 88 µl of 0.5 mM Glutamax (Invitrogen, cat. no. 35050-061), penicillin-streptomycin (ThermoFisher, cat. no.15070063) to a final concentration of 2%, and DNase I (ThermoFisher, cat. no. AM2224) to a final concentration of 80 U/ml. Nucleus dissociation from the lumbar spinal cord We collected three lumbar spinal sections from each group and pooled them together for nuclei dissociation. To isolate nuclei for snRNA-seq, we used the detergent-mechanical cell lysis method58. In brief, the lumbar spinal sections were placed in a pre-chilled Dounce homogenizer, and 500 μl of pre-chilled detergent lysis buffer was added. The detergent lysis buffer was prepared by adding 3 µl of 20% Triton-X to 600 µl of low sucrose buffer (sucrose: 0.32 M; HEPES (pH = 8.0):10 mM; CaCl2: 5 mM; MgAc: 3 mM; EDTA: 0.1 mM; DTT: 1 mM). We homogenized the tissue first by 5 strokes with pestle A (‘loose’ pestle), then 5–10 strokes with pestle B (‘tight’ pestle) to prepare crude nuclei. We did not lift pestles out of the lysis buffer between strokes to minimize bubble creation during homogenization. We added 1 ml of low sucrose buffer to the Dounce homogenizer containing the crude nuclei and mixed gently by pipetting 2–3 times. We passed the crude nuclei preparation through a 40 µm strainer (Fisher scientific, cat. no. 08-771-1) into a pre-chilled 50 ml conical tube containing 1 ml of low sucrose buffer. Then, an additional 1 ml of low sucrose buffer was passed through the strainer, which brought the volume of the nucleus preparation to 3.5 ml (3 ml of low sucrose buffer and 500 μl of lysis buffer). We centrifuged the samples for 10 min (3200 × g; 4 °C), and resuspended the pellets using 3 ml of low sucrose buffer by gently swirling to wash the pellet off the wall. The suspension was put on ice for 2 min and then transferred to an Oak Ridge tube. We homogenized the suspended nuclei for 30 s on ice using a homogenizer (IKA, Ultra Turrax T8; at setting 1). We carefully added 12.5 ml of density sucrose buffer (sucrose: 1 M; HEPES (pH = 8.0): 10 mM; MgAc: 3 mM; DTT: 1 mM) underneath the low sucrose buffer homogenate using a serological pipette to prevent creating any bubbles that could disrupt the density layers. We then centrifuged the Oak Ridge tube at 3200 × g for 20 min at 4 °C. Once the centrifugation was complete, we immediately decanted the supernatant in a flicking motion. The nuclei remaining on the wall were resuspended with 500 µl of resuspension solution (BSA: 0.4 mg/ml; RNAse Inhibitor: 0.2 U/μl). The sample was then passed through a 40 μm Flowmi pipette tip filter (Sigma-Aldrich, cat no: H13680-0040) before counting and loading to remove clumped nuclei and large debris that can clog the microfluidics of Chromium chips. We collected the filtered nuclei in a pre-chilled tube and determined the nucleus yield using a hemocytometer. The nuclei were immediately processed for the next step. Due to high myeline contents in spinal cords, the yield of nuclei was relatively low59. 3’ single-nucleus RNA library construction and sequencing Gel Bead-In Emulsions (GEMs) were generated using the Chromium Platform from 10x Genomics (cat. no. PN-1000075) following the Chromium protocol CG000185 Rev. The Chromium Single Cell 3′ Library & Gel Bead Kit was used for GEM and library preparation. Briefly, 10,000 to 20,000 nuclei were loaded onto the Chromium controller to recover 7000–15,000 nuclei for library preparation and sequencing. GEMs were captured and underwent reverse transcription by incubating at 53 °C for 45 min in a thermal cycler. Next, GEMs were broken, and the sample cDNA was recovered and cleaned according to the instructions. A 35 µl sample of recovered cDNA was used for cDNA amplification using the reagents provided in the kit, and 12 PCR cycles were used. The amplified cDNA was cleaned and quantified using an Agilent Bioanalyzer High Sensitivity chip. For generating the 3ʹ Gene Expression library, 10 µl of purified cDNA sample was enzymatically fragmented and size-selected using SPRIselect magnetic beads. Illumina sequencing adapters were then ligated to the size-selected fragments and cleaned up using SPRIselect magnetic beads. Sample indices were selected and amplified, followed by a double-sided size selection using SPRIselect magnetic beads. The final library quality was assessed using an Agilent Bioanalyzer High Sensitivity chip. The libraries were sequenced on the Illumina NovaSeq to target a minimum of ~15,000 reads per nucleus with a total of 300 million paired-end reads per sample. The library sequencing was performed at the Single Cell Genomics Core, Baylor College of Medicine, USA. Seurat snRNA-seq integration analysis The raw sequencing data were pre-processed using Cell Ranger (v.3.1.0) (10X Genomics). The resulting filtered single-cell expression matrices files from the WT, gp120Tg, gp120Tg/Gal3−/−, and Gal3−/− mice were imported into R for Seurat snRNA-seq integration analysis. Seurat R package (v.4.0.5) was utilized for dimension reduction and cluster analysis. Seurat implements a global-scaling normalization method “LogNormalize” by default that normalizes gene expression measurements for each nucleus by total expression, multiplies this by a scale factor (10,000 by default), and log-transforms the result. To remove low-quality nuclei and doublets, we set the filter criteria of nFeature_RNA > 200 and nFeature_RNA < 5000 as a QC control for snRNA-seq data. This was followed by normalization and selection of all highly variable genes falling within a selected cut-off window for PCA analysis. Statistically significant PCs (p-value < 0.001) were utilized in cluster determination. UMAP plots were produced at a resolution of 0.6, and clusters were annotated based on the expression level of canonical marker genes, and gene expression was visualized using feature maps. The astrocyte clusters were identified using the classic marker genes Gfap, Aqp4, and Gja1. The astrocyte clusters from the integrated UMAP plot were subset for astrocyte subset analysis. UMAP plots of astrocytes were produced at a resolution of 0.5. The integrated UMAP plot was split by genotype to view the astrocyte cluster at each condition. Immunofluorescent (IF) staining Mice were anesthetized with 14% urethane and transcardially perfused with 30 ml of cold 1×PBS. The L4–L5 lumbar spinal cord segments were collected, fixed in freshly prepared 4% paraformaldehyde for 24 h at 4 °C, and then immersed in 30% sucrose in 1×PBS at 4 °C until they sank to the bottom. The tissues were embedded in OCT mounting medium, stored in an air-tight container at –80 °C, and equilibrated to –20 °C in a cryostat for ~1 h prior to sectioning. Cryo-sections ranging from 10 to 35 μm were cut, mounted onto SuperFrost® Plus slides, or stored in Hito floating section storage solution (Hitobiotec) at −20 °C until they were stained for immunocytochemistry. For immunostaining, sections were rinsed with 1×PBS twice to remove storage solution and blocked with 5% BSA and 0.3% Triton X-100 in 1×PBS for 2 h at room temperature. This was followed by 48 h of incubation with primary antibodies, including anti-Gal3 antibody (1:500, Santa Cruz, cat. no. sc-23938), anti-GFAP antibody (1:500, Millipore, cat. no. AB5541), anti-Iba1 antibody (1:500, 1:1000; Wako, cat. no. 016–20001), ant-Vimentin antibody (1:500, Abcam, cat. no. ab137321), anti-NeuN antibody (1:500, Millipore, cat. no. ABN90), anti-Sox9 antibody (Sigma-Aldrich, cat. No. ab5535) and anti-vGlut1 antibody (Synaptic systems, cat. no. 35 303). After five washes with PBS, the sections were incubated with the corresponding second antibodies at room temperature for 2 h. The stained sections were mounted with ProLong Gold Antifade Mountant (Invitrogen, cat. no. P36930) and stored overnight at 4 °C. Images were obtained using Zeiss LSM 880 with Airyscan (Carl Zeiss Microscopy GmbH) and Nikon A1R MP (Nikon Instruments Inc.), and the 3-D images were processed using Imaris 9.6 software (Oxford Instruments). Fluoro-Jade C (FJC) staining 10-μm cryo-sections of lumbar spinal cord were prepared for Fluoro-Jade C (FJC) staining to identify degenerating neurons. FJC staining was performed using the Fluoro-Jade C Ready-to-Dilute Staining Kit (Biosensis, cat. no. TR-100-FJ) following the manufacturer’s instructions. After FJC staining, immunofluorescence (IF) staining was performed on the same sections using the anti-Gal3 antibody (1:500, Santa Cruz, cat. no. sc-23938) according to the protocol described above. The stained sections were mounted with ProLong Gold Antifade Mountant (Invitrogen, cat. no. P36930) and stored overnight at 4 °C. Images were visualized using Zeiss LSM 880 with Airyscan (Carl Zeiss Microscopy GmbH) and Nikon A1R MP (Nikon Instruments Inc.). RNAscope in situ hybridization (ISH) and quantification of C3+ astrocytes Mice were anesthetized with isolfuorane then euthanized by decapitation. Spinal cords were collected and first fixed in fresh 4% PFA for 24 h at 4 °C. The spinal cords were then immersed in 15% sucrose solution at 4 °C for 12 h, then 30% sucrose solution at 4 °C for 12 h. Each spinal cord was then embedded in OCT mounting medium and stored at −80 °C prior to sectioning. The mounted tissue blocks were equilibrated to −20 °C in a cryostat, then 15 μm sections were cut and mounted onto SuperFrost® Plus slides and used immediately. Probes for the gene C3 (417841) were purchased from Advanced Cell Diagnostics (Newark, CA). RNAscope procedures, including tissue pre-treatment and probe hybridization, were performed according to the protocol of RNAscope® Multiplex Fluorescent V2 Assay provided by the manufacturer (Advanced Cell Diagnostics). After the last step of RNAscope, the sections were immunostained for Gfap and Lgals3 as described (Dual ISH-IHC, Advanced Cell Diagnostics). The stained sections were mounted with VECTASHIELD HardSet Antifade Mounting Medium with DAPI (Vector Laboratories, cat. no. H-1500-10) and stored overnight at 4 °C before imaging using a Nikon AX confocal microscope. Anti-Gfap antibody (abcam, cat. ab134436) and Anti-Lgals3 Antibody (Santa Cruz Biotechnology, catalog no. sc-23938) were both used at 1:200 dilution. The images were acquired on a Nikon confocal microscope and analyzed using ImageJ (v1.54p). We quantified C3 signals within HIPAs (Gal3+) in gp120Tg mice and non-HIPA astrocytes (Gal3−) in wild-type, gp120Tg, and gp120Tg/gal3-KO mice and found that 3 pixels of C3 signals were an effective threshold to differentiate HIPAs from non-HIPA astrocytes. Using this cutoff value, we designed a computerized procedure to count C3+ astrocytes (GFAP+). To validate the data from this automatic procedure, we also asked two individuals without access to group information to count C3+/GFAP+ cells. Postmortem human spinal tissues Spinal cord specimens of HIV+ patients and non-infected controls were procured from the Texas NeuroAIDS Research Center (TNRC) of the National NeuroAIDS Tissue Consortium (NNTC). The research received oversight from the Institutional Review Board (IRB) of the University of Texas Medical Branch (UTMB) (Federal Wide Assurance FWA#00002729). The UTMB IRB approval number is 98-402. The authorization for autopsy was obtained from next-of-kin in accordance with Texas statutory law. Informed consent was obtained from decedents who were participants in a clinical research protocol when alive. Consent forms were reviewed by the IRB, date-stamped, and approved annually. Informed consent was obtained annually. Western blotting analysis To prepare mouse tissue, mice were euthanized using excess anesthesia (isoflurane), and L4–L5 lumbar spinal cord segments were collected. Spinal tissues were homogenized in RIPA lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, 10% glycerol, 1 mM EDTA, pH 8.0) containing phenylmethylsulfonyl fluoride (PMSF) and protease inhibitor mixture (Sigma-Aldrich, cat. no. P8340). The protein concentration in the homogenates was determined using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, cat. no. 23227). The protein concentration was adjusted to 2 μg/μl for immunoblotting analysis. Equal amounts of protein (30 μg/lane) were loaded and separated by SDS-PAGE, followed by transferring to polyvinylidene difluoride membranes (Millipore, cat. no. IPVH00010). For immunoblotting analysis, the blots were blocked by 5% nonfat milk in Tris-buffered saline–Tween-20 (20 mM Tris-HCl, 150 mM NaCl, pH 7.5, 0.1% Tween-20) for 2 h at room temperature, then sequentially incubated with primary and secondary antibodies. The primary antibodies used were anti-Gal3 (1:500, Santa Cruz, cat. no. sc-23938), anti-cleaved Il1β (1:500, Millipore, cat. no. ab1413-I), anti-pro-Il1β (1:500, Cell Signaling Technology, cat. no. 12242), anti-cleaved caspase-1 (1:1000, Cell Signaling Technology, cat. no. 4199), anti-wnt5a (1:1000, Abcam, cat. no. ab72583), anti-Iba1 (1:1000, Wako, cat. no. AB_839506), and anti-actin (1:1000, Santa Cruz, cat. no. sc-1616). The protein bands were visualized using the Pierce Enhanced Chemiluminescence Kit (Thermo Fisher Scientific, cat. no. 32106), with β-actin blotted as a loading control. The intensity of the bands was quantified using densitometry analysis with NIH ImageJ. Uncropped and unprocessed scans are provided in the source files. Gp120 injection Recombinant gp120 protein (Cat # HRP20082) was obtained through BEI Resources - a division of the National Institute of Allergy and Infectious Diseases. A 2 µL aliquot of the stock solution was diluted with 0.1% bovine serum albumin in 0.1 M phosphate-buffered saline for a final concentration of 15 ng/µL. Mice were initially anesthetized by inhaling 3% isoflurane. Once fully anesthetized, a 30-gauge needle was inserted at a 45° angle into the intervertebral space between L5 and L6. Correct needle placement was confirmed by a tail twitch from the anesthetized mouse. Finally, 7 µL of the gp120 solution was expelled at an approximate rate of 1 µL per 3 s. After delivery, the needle was held at the injection position for 15 s to ensure no spillage of solution. Subjects were injected every other day for a total of 3 injections per mouse. Mechanical pain tests To measure paw withdrawal thresholds (PWTs), von Frey testing was conducted on the plantar surface of the hind paw. Mice were habituated to the testing surroundings for 2 h per day for three days prior to the testing day. On the day of testing, calibrated von Frey filaments (ranging from 0.1 to 2.0 g) were applied perpendicularly to the central area of the hind paw. The Dixon “up and down paradigm” method was used to calculate the PWT values. To minimize subjective effects and bias, behavioral tests were performed in a double-blind manner, with the genotype of individual animals not revealed to the experimenters. Thermal hypersensitivity tests Thermal withdrawal latencies were measured using the Hargreaves test. For this, mice were placed in clear plastic chambers (7 × 9 × 10 cm) on an elevated surface and allowed to acclimatise to their environment before testing. The radiant heat source was aimed at the plantar surface of each hind paw in the area immediately proximal to the toes. The paw withdrawal latency (PWL) was defined as the time from the onset of radiant heat until a withdrawal of the hind paw occurred. A 20-s cutoff time was used to avoid tissue injury and sensitisation. Three measures of PWL were taken and averaged for each hind paw. Statistical analysis Based on our results from previous experiments, we calculated a cohen’s d of 2.08 between groups. To achieve an alpha of 0.05 and power of 80%, we use G*Power (v3.1) to determine that a minimum five animals were required. Further statistical analysis was conducted using Prism 5 software (GraphPad). Data are presented as mean ± SEM or mean ± SD. Immunoblotting data were analyzed using either Student’s t test or nonparametric Mann–Whitney test or One-way ANOVA. Immunofluorescent staining, FJ-C staining, and pain behavioral data were analyzed using a One-way ANOVA or Student’s t test. A p-value of less than 0.05 was considered statistically significant for all tests. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary video 1 Supplementary video 2 Reporting Summary Transparent Peer Review file Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary video 1 Supplementary video 2 Reporting Summary Transparent Peer Review file Source data Source data Source data