RBP-Driven RNA sorting and local translation establish the molecular identities of individual sensory axons

Abstract

Neurons extend long axons that traverse distinct microenvironments, yet how these subcellular compartments acquire and maintain specialized identities remains unclear. Here, we use spatial translatomics to define the local translatomes of somatosensory dorsal root ganglion (DRG) neurons. Translating Ribosome Affinity Purification and RNA sequencing (TRAP-seq) reveal thousands of mRNAs preferentially translated within central axons, peripheral axons, or DRG somata, establishing compartment-specific translational programs. Many of these transcripts encode ion channels and neurotransmitter receptors that may confer distinct electrophysiological and regenerative properties to each axon. Integration of the TRAP-seq data with published RNA-seq identify locally translated components that change following neuropathic injury and could thereby adjust neuronal activity. We identify RNA regulons coordinated by RNA-binding proteins (RBPs) SFPQ and SRSF10, which preferentially bind and traffic mRNAs to peripheral or central axons, respectively. These findings indicate that RBP-guided RNA sorting and local translation enable the establishment and dynamic local modulation of somatosensory function.

Here authors define compartment-specific translatomes in somatosensory dorsal root ganglion neurons and identify distinct local translational programs across central axons, peripheral axons and neuronal somata. They show that the RNA-binding proteins SFPQ and SRSF10 coordinate the selective trafficking and local translation of mRNAs in peripheral and central axons.

Introduction

Neurons possess several distinctive and essential features, including the ability to form long-range connections, efficiently propagate electrical signals, and exhibit plasticity in response to multiple stimuli. These capabilities rely on dynamic regulation of gene transcription and translation, which together shape the cellular proteome. Transcriptional programs adapt rapidly to changes in the overall neuronal state, including alterations in membrane polarization, metabolic activity, and the inflammatory environment^1–3. In contrast to transcription, translation can be regulated at the subcellular level, allowing for localized responses to changes in the microenvironment, such as those occurring at synapses and along the axons and dendrites^4–10. As a result, local translation offers an additional layer of spatiotemporal regulation, fine-tuning the neuronal proteome in response to both global and microenvironmental signals.

It was not long ago that scientists believed local translation could not occur in axons. The first evidence of RNA in axons was detected in the 1960s^11,12, however, it took another 30 years of technological advances in microscopy for researchers to show that ribosomes were present in axons and capable of translating mRNA^13–15. Since 2000, multiple studies have revealed that local protein synthesis is necessary for axon development and regeneration, and proteins transported from the soma are not sufficient to maintain local proteomes^5,16–18. Despite this progress, it is not understood how mRNA transport and translation establish local proteomes and govern the specificity of axonal functions. Analysis of local translation in the axons of retinal ganglion cells in the central nervous system (CNS) suggests that translatomes are consistent across distinct axon branches¹⁹, which would preclude dynamic regulation of individual axonal proteomes in response to changes in the local microenvironment. However, it is not known whether it is possible for neurons to differentially regulate mRNA transport and translation to establish and modulate distinctive features of individual axons.

In the peripheral nervous system (PNS), sensory neurons of the dorsal root ganglia (DRG), trigeminal ganglia, and nodose ganglia are distinguished by their pseudounipolar morphology, where a single axon stem bifurcates into central and peripheral projections that innervate the spinal cord and peripheral tissues, respectively^20,21. In humans, each of these axons can extend up to 1 meter from the cell soma in the ganglion. The peripheral axons of DRG neurons are responsible for dynamically transducing environmental stimuli into electrical signals, while the central axons transmit these signals to the CNS. Both peripheral and central axons generate and propagate action potentials, are myelinated in many DRG subtypes, and exhibit the same microtubule orientation with plus-ends towards the terminals^21,22. Studies of DRG axonal function and composition have historically focused on the peripheral axon as it is more accessible for surgical isolation, injury models, and electrophysiological recordings, while investigations of the central axon are underrepresented²³. Therefore, it is not known how the individualized proteomes of central and peripheral axons are decided and established²⁴. The spatially and functionally distinct central and peripheral axon branches make DRG neurons an ideal system for addressing whether subcellular axonal translatomes differ, how any such differences arise, and how they contribute to the compartmentalized demands of encoding, transmitting, and filtering sensory information.

In this study, we identify and compare the spatial translatomes of the two axons of somatosensory neurons. The sodium channel, Nav1.8, is expressed in approximately 60% of DRG neurons, including most subtypes involved in pain sensation^20,25,26. Using Nav1.8^cre; L10a-GFP^fl/fl mice and a combination of Translating Ribosome Affinity Purification and highly sensitive RNA-sequencing (TRAP-seq)²⁷, we profile the central and peripheral axonal translatomes of somatosensory neurons from microdissected mouse tissues—specifically, (1) lumbar DRG cell bodies, (2) sciatic nerves (peripheral axons), and (3) lumbar dorsal roots (central axons). Our analysis identifies thousands of axon-selective transcripts, including specialized ion channel and receptor isoforms with preferential translation in peripheral and central axons. These findings indicate that translation is an indispensable mechanism for locally modulating axonal physiology within distinct branches of the same neuron. We show that many of the locally translated mRNAs are selectively localized to the central versus peripheral axons, indicating that there is disparate transport and/or stabilization of individual mRNAs in these two axons. Notably, we identify distinct RNA-binding proteins (RBPs) that predominantly bind locally translated mRNAs of the peripheral or central axons, suggesting RBPs enable selective localization of functional RNAs cargoes to individual axons. Together, our findings offer a comprehensive understanding of how selective RNA transport and translation enable highly specialized subcellular proteomes that establish the spatial organization of neurons in normal physiology and disease states. This resource (trap.painseq.com) provides a wealth of data to investigate how cellular functions are selectively regulated at the local level and highlights how dysregulation of axonal translatomes may contribute to neuropathic diseases and disorders.

Results

Defining subcellular translatomes in somatosensory dorsal root ganglion neurons

To generate a comprehensive resource of the axonally transported and translated mRNAs in DRG sensory neurons in vivo, we performed spatial translatomics using Translating Ribosome Affinity Purification (TRAP) paired with ultra-low methods for RNA-sequencing to account for the technical difficulties of extracting high-integrity RNA from axons in vivo. The lumbar ganglia (soma), sciatic nerves (peripheral axons), and lumbar dorsal roots and spinal cords (central axons) were microdissected from ~ 15 perinatal mice (P3-7) containing GFP-tagged ribosomal subunits in Nav1.8+ neuronal cell types (Nav1.8^cre;L10a-GFP^fl/fl) (Fig. 1A). Ribosome-associated mRNAs were isolated from tissue lysates using GFP-antibody affinity purification and ultra-low concentration RNA extraction, yielding ~10-100 pg/μL RNA (Supplementary Fig. 1A, B). We carried out the same procedure on littermate negative control mice (L10a-GFP^fl/fl) that lacked GFP-tagged ribosomes. Libraries were prepared from low-input total RNA samples with SMART-Seq, and subsequently sequenced and aligned to elucidate the distinct central and peripheral axonal translatomes (n = 4 sets/condition, each set consisted of 13–17 individual animals) (Supplementary Data S1)²⁸.

Fig. 1. Defining subcellular translatomes in somatosensory dorsal root ganglia neurons.

A Schematic depicting the TRAP-sequencing protocol. Nav1.8^Cre/+ mice were crossed to L10a-EGFP^fl/fl mice to generate Nav1.8^Cre;L10a-EGFP^fl/fl mice and L10a-EGFP^fl/fl negative control mice. In these mice, Nav1.8^Cre drives expression of an EGFP-tagged ribosomal subunit, L10a, selectively in a set of DRG neurons. Tissue microdissections were performed in mice aged P3-7 to collect lumbar dorsal root ganglia (soma compartment), lumbar dorsal roots/ spinal cord (central axon compartment), and sciatic nerves (peripheral axon compartment). Using the Translating Ribosomal Affinity Purification (TRAP) protocol, ribosomal-associated mRNAs were isolated from tissue lysates using an anti-EGFP antibody immunoprecipitation followed by RNA extraction using filter columns. Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B Principal component analysis of DRG soma, peripheral axon, and central axon samples (n = 4 peripheral and central axon samples, n = 8 DRG soma samples). C Heatmap showing normalized gene expression within the DRG soma, peripheral axon, and central axon TRAP datasets for all genes with expression ≥ 5 counts. D–F Volcano plots demonstrating differential gene expression between the TRAP datasets comparing peripheral axons vs. DRG soma (D), central axons vs. DRG soma (E), and central vs. peripheral axons (F). Differential gene expression analysis was conducted using the RStudio DESeq2 package, using a two-sided Wald test and Benjamini-Hochberg FDR-adjusted p-values (p.adj ≤ 0.05; fold-change ≥ 1.5). G Bar graphs depicting the normalized gene expression of select mRNA transcripts (with known expression in axons) within the DRG soma, peripheral axon, and central axon compartments. Statistical significance determined by two-sided, unpaired t tests for a priori peripheral vs. central axon comparisons (n = 4 peripheral and central axon samples, n = 8 DRG soma samples) (p-value ≤ 0.05). Data are represented as mean ± SEM. Source data are provided as a Source Data file.

Principal component analysis and pairwise comparisons demonstrate that the DRG soma samples cluster together across the multiple samples, while central and peripheral axon samples cluster distinctly from the soma and from each other (Fig. 1B). This distinct clustering indicates that the transcripts were highly consistent and specific for each compartment. Heatmap visualization with hierarchical clustering of ~ 18,000 normalized genes revealed a clear segregation of samples by condition and distinct translational profiles across DRG soma, peripheral axon, and central axon samples (Fig. 1C). Volcano plots demonstrating the significance and fold-change between compartments for all genes indicated that both peripheral and central axons expressed many genes that were enriched (padj. ≤ 0.05; F.C. ≥ 1.5) over the DRG compartment, demonstrating substantial axonal translation in DRG sensory neurons in vivo (Fig. 1D, E). Moreover, the volcano plot comparing central and peripheral axon compartments indicates that large numbers of distinct transcripts are translated in the peripheral versus central axons, demonstrating the subcellular specificity with which DRG neurons control local translation to regulate spatially distinct neuronal functions in DRG soma and peripheral and central axons (Fig. 1F).

Interestingly, previously characterized axonally translated genes such as Actb (β-actin), Bcl2l2 (Bclw), Rpl29 (60 s ribosomal protein L29), Lamb2 (Laminin β2), Vim (Vimentin), and Kpnb1 (Importin β1) all have higher axonal mRNA levels in the peripheral axon compartment than in the central axon (Fig. 1G)^3,29,30. This suggests that the characterization of axonal translatomes in scientific literature mirrors DRG peripheral axons, while the translational landscape of DRG central axons may differ from the prior data. Thus, defining what is locally translated in both peripheral and central DRG axons is critical for understanding how compartment-specific protein synthesis contributes to somatosensory information transfer.

Central and peripheral DRG axons have distinct translatomes

Our analysis reveals that the peripheral and central axonal translatomes are remarkably different from one another (Fig. 2A and Supplementary Data S2). In total, the DRG axonal translatome contains roughly 13,000 genes, of which 5960 genes are shared between the two axons, 3379 are enriched in central axons and 3420 are enriched in peripheral axons (Fig. 2B). Gene ontological (GO) functional enrichment analysis of the genes shared between the axonal compartments reveals local translation of transcripts related to autophagy, RBP complex formation, ubiquitin-dependent proteasomal mechanisms, mitochondrial function, and ribosome biogenesis in both axons (Fig. 2C and Supplementary Data S3). Many of these transcripts are also translated in the cell soma; ~ 60% of the transcripts translated in peripheral axons were also translated in soma, and 72% of the transcripts identified in the TRAP-Seq of central axons were translated in the soma as well (Supplementary Fig. 2A–C). These ubiquitous components are also related to ubiquitin-dependent proteasomal degradation, autophagy, ribonucleoprotein complexes, and organelle function (e.g., mitochondria, Golgi, endoplasmic reticulum, and ribosomes) (Supplementary Fig. 2D, E). Functional annotation analysis of the axon-specific transcripts that are not shared with the DRG soma identified enrichment of gene categories associated with synaptic specialization in the central axon samples, while peripheral axon-specific transcripts were enriched for categories related to cytoskeletal structure and contractility (Supplementary Fig. 2F, G and Supplementary Data S4, 5).

Fig. 2. Central and peripheral DRG axons have distinct translatomes.

A Schematic depicting the translatomic gene sets compared in Fig. 2 (i.e., central axon vs. peripheral axon). Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B Heatmap highlighting clusters of genes with (1) similar expression across central and peripheral axon compartments, (2) higher expression in the central axon compartment, and (3) higher expression in the peripheral axon compartment (n = 4 samples/group). C–E Gene ontology (GO) enrichment analysis visualized as a dot plot, where each dot represents a significantly enriched biological process (BP), cellular component (CC), or molecular function (MF). Gene Ontology enrichment was performed using clusterProfiler enrichGO based on a one-sided hypergeometric test. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR). C Dot plot showing enriched GO terms from the translatomic geneset shared between both axons (p.adj ≥ 0.05; abs[F.C] ≤ 1.5). D Dot plot showing GO terms enriched in the central axon gene set (p.adj ≤ 0.05; F.C ≥ 1.5). E Dot plot showing GO terms enriched in the peripheral axon gene set. (p.adj ≤ 0.05; F.C ≤ −1.5).

The 3379 genes comprising the central axon translatome were enriched for transcripts associated with neuronal excitability, synaptic organization, and neurotransmission (Fig. 2D and Supplementary Data S6). The top enriched terms included gene sets annotated to both presynaptic and postsynaptic processes. Genes implicated in presynaptic assembly included Nrxn1, Cbln1/2, Pclo, and Efnb3, consistent with the presence of transcripts encoding adhesion molecules and active-zone scaffolds. Furthermore, axonal translation of voltage-gated calcium channel components and SNARE proteins (Cacng2/4/8, Stx1b, Vamp2) is consistent with enrichment of transcripts involved in presynaptic transmission. In addition to these presynaptic components, transcripts encoding postsynaptic scaffolding and receptor-associated proteins such as Grip1/2, Dlg2/3, Gphn, Nptxr, and Arc were also detected. CA-enriched transcripts, therefore, include multiple genes implicated in pre- and postsynaptic signaling pathways. Based on the classes of locally translated receptors, these data are consistent with the possibility that presynaptic DRG axon terminals engage in local translational regulation related to glutamatergic (Grin/Gria/Grm/Grik), GABA-ergic (Gabra/Gabrb), endogenous opioid (Oprk/Oprl) and serotonergic circuits (Htr), which aligns with known mechanisms of pain modulation in the dorsal horn. Given the complex cellular composition of the spinal cord environment, we cannot rule out that these sequences may reflect some contamination from surrounding tissue as well as compartment-specific translation. Nonetheless, collectively, these findings suggest that DRG central axons contain a diverse repertoire of synaptic-related transcripts, rather than functioning solely as passive conduits of sensory information.

In contrast to the central axons, GO analysis of 3420 transcripts locally translated in the peripheral axons of DRG neurons revealed that transcripts were enriched for processes related to translational capacity, cytoskeletal organization, and myelination-associated components (Fig. 2E and Supplementary Data S7). The enriched mRNAs were strongly associated with ribosomal (Rpl/Rps) and cytoplasmic translation (Eif/Etf) categories, indicating enrichment of transcripts encoding ribosomal proteins and translation factors. Furthermore, terms such as actomyosin structure, organization and myofibril assembly were identified among enriched categories, pointing to regulation of cytoskeletal and contractile elements. For example, local translation of the Rho GTPase signaling pathway controlling myosin contractility (Rhoa, Rock1/2, Cdc42bpa), actin cytoskeletal scaffolding proteins (Iqgap1-3), and non-muscle myosins (Myh9/Myh14) is consistent with transcripts that may contribute to local cytoskeletal dynamics. Lastly, transcripts associated with axon–Schwann cell and axon–ECM interactions (Pak2, Itgb1, Sdc4, Phldb2) were enriched, suggesting potential roles in mediating axon–glial and extracellular interactions within the peripheral nerve environment. While some of these transcripts may derive from the tissue context of sciatic nerve preparations, together these findings highlight enrichment of transcripts related to translational machinery and structural organization in peripheral axons, consistent with a compartment specialized for structural maintenance and remodeling.

Cross-dataset integration reveals compartment-specific translational remodeling during neuropathy

Peripheral nerve injury induces a robust reprogramming of the axonal transcriptome and proteome, with increased axonal protein synthesis and dynamic changes in membrane protein composition and often leads to peripheral neuropathy³¹. However, the mechanisms governing these adaptations are not well understood. In chemotherapy-induced peripheral neuropathy (CIPN), impaired axonal transport of SFPQ-containing RNA granules has been linked to reduced expression of key neuronal survival factors, implicating a connection between axonal mRNA regulation and axon degeneration^32,33. To determine whether locally translated mRNAs in distinct neuronal domains overlap with transcripts altered during neuronal injury, we performed a cross-dataset comparison between our compartment-resolved TRAP translatomes and a published single-cell RNA-seq dataset of DRG neurons treated with paclitaxel (Fig. 3A and Supplementary Data S8). In these previous studies, mice were injected intraperitoneally with 4 mg/kg paclitaxel every other day for six days (four injections total), and transcripts altered in each of the DRG cell subtypes following paclitaxel treatment were identified³¹. Across all DRG neuron subtypes, ~ 1–7% of transcripts present in our TRAP-seq datasets correlate with those that also change significantly with paclitaxel treatment (Fig. 3B). The proportion of paclitaxel-affected transcripts was highest for those translated in the soma and those shared among multiple compartments. Individual analysis of paclitaxel-sensitive genes in Nav1.8+ DRG subtypes, including Mrgprd+ non-peptidergic nociceptors (NPs), Tac1+ /Gpx3+ peptidergic nociceptors (PEP1), Tac1+ /Hpca+ peptidergic nociceptors (PEP2), Nefh+ Aβ low-threshold mechanoreceptors (LTMRs) (NF1), and Sst+ pruriceptors (SST) further demonstrate that paclitaxel treatment causes changes across all domains of pain-sensing neurons (Fig. 3C). However, these analyses also suggest that paclitaxel has a greater impact on transcripts in peripheral axon translatome than the central axon.

Fig. 3. Cross-dataset integration reveals compartment-specific translational remodeling during neuropathy.

A Schematic depicting the cross-dataset integration performed between TRAP-seq (all tissues) and Pain-seq (paclitaxel-treated DRG scRNA-seq) datasets. Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B Bioinformatic analysis showing the percentage of locally translated mRNAs from the TRAP-seq datasets (peripheral axon, central axon, DRG soma, and shared soma + axon) that are differentially expressed in the Pain-seq datasets following treatment with paclitaxel. Each datapoint represents a different DRG subtype identified via scRNA-seq. Statistical analyses were conducted using one-way repeated-measures ANOVA with Geisser–Greenhouse correction followed by Tukey’s multiple-comparisons test (p.adj ≤ 0.05). C Percentage of locally translated mRNAs in the TRAP-seq peripheral axon, central axon, soma, and shared datasets that are differentially expressed in Pain-seq datasets following treatment with paclitaxel, specifically in Nav1.8+ DRG subtypes (NP, PEP1, PEP2, NF1, and SST). D GO enrichment analysis showing significantly enriched GO terms and select locally translated genes from each tissue compartment that are differentially expressed following paclitaxel treatment (p.adj ≤ 0.05). Red: mRNAs that are upregulated following paclitaxel treatment. Blue: mRNAs that are downregulated following paclitaxel treatment. Source data are provided as a Source Data file.

To identify the biological programs most strongly impacted by paclitaxel within each subcellular compartment, we performed GO enrichment analysis on mRNAs whose overall expression levels change with paclitaxel and are locally translated in one or another domain (Fig. 3D). Paclitaxel affects multiple mRNAs that are “shared” in soma and axonal translatomes; these locally translated mRNAs are involved in axonal transport and vesicle localization, such as Rab11b, Sybu, and Kif5 family motors, suggesting that paclitaxel perturbs intracellular trafficking pathways operating across neuronal compartments. Furthermore, the somatic compartment upregulated pathways related to cell respiration and apoptotic signaling, including Park7, Sod2, and Cdkn1a. Paclitaxel alters many mRNAs that are predominantly translated in the peripheral axon compartment, including components involved in cytoplasmic translation, environmental sensing, and neuronal projection extension, such as Itgb1, Ctnnb1, and Flrt3, consistent with paclitaxel-induced remodeling of local growth and structural pathways. In contrast, paclitaxel-sensitive mRNAs that are translated in the central axon include transcripts linked to protein-complex assembly and synaptic/ secretory function, such as Celf4, Kcnc3, Sh3gl2, and Gad2, reflecting changes in neurotransmitter release and presynaptic physiology. Together, these analyses reveal that paclitaxel-induced neuropathy evokes extensive remodeling in DRG neurons, highlighting compartment-resolved molecular responses that may contribute to neuropathic phenotypes.

DRG axons locally translate specialized ion channels and membrane-bound receptors

While analysis of the initial axonal translatomes demonstrates that the translatomes of the two axons differ greatly from one another, some RNAs identified in the bulk sciatic nerve, dorsal root, and spinal cord tissues may not be derived from Nav1.8+ sensory axons, but rather from local glia, epineural cells, and neuronal cell bodies in the spinal cord. For example, compared to the DRG compartment, the central translatome exhibited high levels of Olig2, a marker of oligodendrocytes, and the peripheral translatome had high levels of Ncmap, a marker of myelinating Schwann cells. To distinguish which of the thousands of transcripts were definitively derived from Nav1.8+ sensory neurons rather than other cells, we determined the transcripts with significantly higher expression in TRAP-seq of sciatic nerves, DRG or dorsal roots in Nav1.8-GFP mice as compared to negative controls of TRAP-seq from the same locations in mice lacking L10a-GFP (p.adj ≤ 0.05) (Fig. 4A)³⁴. After bioinformatically filtering out the background transcripts with high expression in negative controls, the specific translatomic gene lists were refined to include 549 transcripts from the soma, 304 from the peripheral axons, and 225 from the central axons (Fig. 4B–E and Supplementary Data S9). This method of stringent background filtering established higher confidence cohorts of locally translated mRNAs within DRG soma and peripheral and central axons. While this additional analysis increases confidence that the selected components are translated in axons of Nav1.8 + DRG neurons, it is intentionally conservative. Accordingly, global analyses of compartment-specific translational programs were performed on the full dataset as discussed above, whereas the filtered dataset defines high-confidence Nav1.8-specific transcripts used for downstream validation.

Fig. 4. Identification of higher confidence translatomes through background filtering.

A Schematic depicting the gene sets compared in Fig. 4 (i.e., GFP+ tissue samples vs. GFP- tissue samples). Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B Venn diagram detailing the number of mRNA transcripts in DRG soma, peripheral axon, and central axon translatomic datasets after background filtering to remove transcripts with expression in GFP-negative control samples. C–E Volcano plots demonstrating the number mRNA transcripts with significantly higher expression in GFP+ TRAP samples from the DRG soma (C), peripheral axon (D), and central axon (E) as compared to GFP-negative control samples from the same tissue compartment. Differential gene expression analysis was conducted using the RStudio DESeq2 package, using a two-sided Wald test and Benjamini-Hochberg FDR adjusted p-values (p.adj ≤ 0.05; fold-change ≥ 1). F A gene concept network demonstrates the color-coded linkages between genes from the central axon translatome and significantly enriched GO terms. The size of the dot in the network correlates to the number of mRNAs in the central axon translatome associated with that GO category. The higher-confidence central axon translatome encodes for genes related to sensory perception, pain perception, and serotonin signaling. G A gene concept network demonstrates the color-coded linkages between genes from the peripheral axon translatome and significantly enriched GO terms. The higher confidence peripheral axon translatome encodes for genes related to synaptic specialization, cilia, and cytosolic ribosomal proteins.

Our background filtered list of locally translated mRNAs confirmed that central and peripheral axons locally translate markedly distinct proteins, including many specialized synaptic receptors and ion channels required for sensory processing. For example, local translation of serotonin receptors, Htr3a and Htr5a, in the central axons suggests serotonin binding may be an important component of descending modulation of nociceptive inputs within the primary somatosensory neurons (Fig. 4F). Several genes encoding ion channel proteins were identified in this central axon gene set, including Scn11a, which encodes Nav1.9. The sodium channel Nav1.9 is selectively expressed in DRG neurons, and point mutations in this channel protein can cause either hyperexcitability and chronic pain or chronic insensitivity to pain³⁵. Several distinct ion channels important for action potential propagation were identified in the higher confidence peripheral axon dataset, such as Hcn1 and Scn2a (Nav1.2) (Fig. 4G). We also identified transcripts encoding the calcium channel, Catsper2, which was initially identified as a voltage-dependent, Ca²⁺ selective, pH-sensitive ion channel highly expressed in sperm, as well as two ancillary components of this channel (Catsperd and Catsperζ)³⁶. Thus, several components that are locally translated in the peripheral axons also play critical roles in membrane physiology and nociception and indicate that regulated axonal translation in both central and peripheral axons may modify pain responses.

To verify our findings of local and selective translation in the central or peripheral axons, we focused on several channels and receptors identified in our high-confidence, background-filtered screen (Supplementary Data S10). Puromycin was intraperitoneally injected into mice to label nascent protein chains in the dorsal root ganglia, sciatic nerves, and dorsal roots (Fig. 5A). Proximity labeling (Puro-PLA) enabled the visualization of specific nascent proteins via proximity of antibodies bound to puromycin and the translating protein of interest (Supplementary Fig. 3). To make sure that we are analyzing nascent peptides within the axons targeted by our TRAP experiments, we focused on PLA puncta that are within the Nav1.8 labeled axons. PLA visualization and quantification verify that individual mRNAs identified by TRAP-seq as being preferentially translated in the peripheral axons of the sciatic nerve (Hcn1 and Scn2a) are indeed translated there, and that we find very little translation in the central axons of the dorsal roots (Fig. 5B, C”). In contrast, our analysis indicates that there is extensive translation of Htr3a and Scn11a in central axons, with little or no translation taking place in the peripheral axons (Fig. 5D, E”). Consistent with developmental single-cell datasets from Sharma et al. showing that Htr3a expression at early postnatal stages is expressed in a broad population of DRG neurons²⁶, HTR3A is translated in both Nav1.8 positive and negative central axons. HTR3A is also translated in perineural cells at these timepoints, as we observed in sciatic nerve tissue.

Fig. 5. Diverse membrane-associated ion channels are locally translated.

A Puromycin proximity ligation assay (Puro-PLA) enables in situ visualization of nascent protein in mouse tissues. Step 1: Puromycin injected into mice incorporates into the C-terminus of translating proteins, labeling nascent peptide chains. Step 2: PLA probes bind two targets of interest (i.e., Puromycin and ion channel) and enable amplification of a fluorescent signal if the targets are within 40 nm. Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B–E Representative 20 × 2 fluorescent images of HCN1 (B), NAV1.2 (C), 5-HTR3A (D), and NAV1.9 (E) PLA puncta and TdTomato-labeled Nav1.8+ (Nav1.8-tdT) axons in dorsal roots and sciatic nerves, representing DRG central and peripheral axons, respectively. Scale bar = 40 μm. B’–E’ Higher magnification inset images (yellow box) demonstrating the location and abundance of HCN1 (B’), NAV1.2 (C’), 5-HTR3A (D’), and NAV1.9 (E’) puncta within Nav1.8-tdT+ axons in the dorsal root and sciatic nerve. Scale bar = 10 μm. B” Quantification of the percent area occupied by HCN1 PLA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 7) and sciatic nerves (n = 7). C” Quantification of the percent area occupied by NAV1.2 PLA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 4) and sciatic nerves (n = 4). D” Quantification of the percent area occupied by 5-HTR3A PLA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 5) and sciatic nerves (n = 5). E” Quantification of the percent area occupied by NAV1.9 PLA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 6) and sciatic nerves (n = 5). Representative images with enhanced brightness shown; raw images used in analysis. Statistical significance determined by two-way ANOVA with Tukey’s multiple comparisons test and multiple two-sided, unpaired t tests (p-value ≤ 0.05). Data are represented as mean ± SD. All statistical analyses were derived from data using 4 mice. Source data are provided as a Source Data file.

Together, the results seen with nascent-chain puromycin labeling (Puro-PLA) validate the specificity and sensitivity of our TRAP-seq approach and demonstrate that the background-filtered translatomes constitute a powerful resource for discerning axonally synthesized proteins. Moreover, our data show that discrete cohorts of ion channels and receptors are synthesized locally within each axonal compartment, supporting the conclusion that peripheral and central branches possess distinct translational identities that may contribute to their specialized physiological roles in nociceptive signaling.

For TRAP-seq experiments, we used postnatal mice, as the higher levels of axonal translation in young animals and the lower levels of surrounding myelination and glial cells facilitate such studies. To determine whether the mRNAs that we have identified as being selectively translated in central and peripheral axons are also differentially translated in adults, we used puromycin labeling to assess local translation of validated ion channels (HCN1, NAV1.2, 5-HTR3A, NAV1.9) in two-month old mice. As expected, the overall level of translation for all components is lower in adult mice than in neonatal animals. However, we find that axonal translation of key ion channels is maintained in mature animals and that there continues to be similar selectivity for central versus peripheral axons (Fig. 6 and Supplementary Fig. 4A–D).

Fig. 6. 5-HTR3A translation is upregulated with paclitaxel-treatment in adult mice.

A 2-mo-old Nav1.8^Cre;TdTomato^fl/+ mice were i.p. injected with 20 mg/kg paclitaxel or vehicle (saline) every other day for four days (Days 1,3,5,7). On Day 16, mice were injected with 22.5 mg/kg puromycin dihydrochloride and incubated for 45 mins to label nascent proteins. Dorsal roots (central axons) and sciatic nerves (peripheral axons) were collected for Puro-PLA. Created in BioRender. Silagi, E. (2026) https://BioRender.com/ra0oidx. B Representative high magnification ( ~120x) inset images demonstrating the location and abundance of 5-HTR3A Puro-PLA puncta (white) within TUJ1+ axons (red) in the dorsal roots and sciatic nerves of mice treated with vehicle or paclitaxel. Scale bar = 10 μm. C Quantification of the percent area occupied by 5-HTR3A Puro-PLA puncta within TUJ1+ axons in individual dorsal roots (n = 13) and sciatic nerves (n = 14). Samples also shown parsed by treatment with either vehicle (n = 6 dorsal roots; n = 6 sciatic nerves) or paclitaxel (n = 7 dorsal roots; n = 8 sciatic nerves). Representative images with enhanced brightness shown; raw images used in analysis. Statistical significance determined by two-way ANOVA with Tukey’s multiple comparisons test and multiple two-sided, unpaired t tests (p-value ≤ 0.05). Data are represented as mean ± SD. All statistical analyses were derived from data using 4 mice. Source data are provided as a Source Data file.

To understand the ability of local translation to respond to environmental cues and so contribute to neuronal plasticity, we focused on the serotonin receptor 5-HTR3A. We asked whether paclitaxel treatment alters the translation of 5-HTR3A, which we identified and verified as being locally translated in the central axons, and whose expression was previously shown to change in response to chemotherapy³¹. As shown, paclitaxel treatment results in selective increases in 5-HTR3A translation in central axons of adult mice (Fig. 6A–C). Notably, these experiments indicate that translation of serotonin receptors in the central axons changes in response to toxic stimuli and so may have a role in neuropathic pain.

mRNAs encoding locally translated proteins are selectively targeted to axons

For the identified proteins to be translated in either axon, the corresponding mRNA must both be transported to that axon and engage with translational machinery at that location. To determine whether the distinct translatomes of the central and peripheral axon reflect differential localization of the mRNA transcripts, or whether these mRNAs are similarly localized to both axons but are selectively translated at one site, we utilized in vivo RNAscope fluorescent in situ hybridization (FISH) to evaluate the subcellular localization of select mRNAs. Single-molecule RNAscope demonstrated the mRNA expression pattern for mRNAs encoding Hcn1, Scn2a, Htr3a, Scn11a, and Catsper2 within the DRG soma, sciatic nerves (peripheral), and dorsal roots (central) (Fig. 7A–D’ and Supplementary Fig. 5). Again, we focused on RNAscope puncta that are within the Nav1.8 labeled axons, or within the Nav1.8 cell bodies in the DRGs. Using particle analysis, we found that Hcn1, Scn2a, and Catsper2 transcripts have significantly more puncta per area in Nav1.8+ peripheral axons compared to central axons, indicating that these transcripts are both selectively localized and selectively translated in peripheral axons (Fig. 7E, F). In contrast, Htr3a and Scn11a have significantly more puncta per area in Nav1.8+ central axons than in peripheral axons (Fig. 7G, H). As expected, all these mRNAs were also localized to the Nav1.8+ cell bodies of DRG neurons. These data indicate that the distinct axonal translatomes in the central and peripheral axons predominantly reflect differential transport and localization of the mRNAs from the cell body to one axon or the other.

Fig. 7. Transcripts encoding locally translated ion channels have axon-selective expression.

A–D Representative 20 × 2 fluorescent images of Hcn1 (A), Scn2a (B), Htr3a (D), and Scn11a (D) RNAscope mRNA puncta in DRG soma, central axons, and peripheral axons, respectively. Scale bar = 100 μm (soma) and 40 μm (axons). Merged DRG images show target RNAscope puncta (white), TdTomato-labeled Nav1.8+ (Nav1.8-tdT) neurons (red), and DAPI-labeled nuclei (blue). A’–D’) Higher magnification inset images (yellow box) demonstrating the location and abundance of Hcn1 (A’), Scn2a (B’), Htr3a (D’), and Scn11a (D’) mRNA puncta within Nav1.8-tdT+ soma in the ganglia (scale bar = 25–35 μm), and axons in the dorsal root and sciatic nerve (scale bar = 15 μm). E Quantification of the percent area occupied by Hcn1 mRNA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 7) and sciatic nerves (n = 5). F Quantification of the percent area occupied by Scn2a mRNA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 6) and sciatic nerves (n = 5). G Quantification of the percent area occupied by Htr3a mRNA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 5) and sciatic nerves (n = 4). H Quantification of the percent area occupied by Scn11a mRNA puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 5) and sciatic nerves (n = 5). Representative images with enhanced brightness shown; raw images used in analysis. Statistical significance determined by two-way ANOVA with Tukey’s multiple comparisons test and multiple two-sided, unpaired t tests (p-value ≤ 0.05). All statistical analyses were derived from data using 4 mice. Source data are provided as a Source Data file.

RBPs are positioned to orchestrate the subcellular localization of axonal translatomes

Our TRAP-seq, PLA and RNAscope findings indicate that many of the channels and receptors critical for initiating and propagating somatosensation are locally produced within either the central or peripheral axon and that the corresponding mRNAs are highly localized to one or the other axon as well. An important step in enabling such highly localized and regulated translation is the sorting and transport of mRNAs to distinct subcellular regions. Sorting of mRNAs to specific locations often depends on structural elements within mRNA sequences; axonal mRNA sorting is usually coordinated through RNA-binding protein (RBP) recognition of select localization motifs within the 3’ untranslated region (3’UTR) of the target mRNA³⁷. RBP-bound mRNAs assemble into ribonucleoprotein (RNP) transport granules, characterized by their ability to undergo liquid-liquid phase separation and be transported directly or indirectly along microtubules by motor proteins^38–40.

To investigate the potential logic for differential RNA sorting to central and peripheral axons, we bioinformatically analyzed the 3’UTRs of axonally translated mRNAs, with the goal of identifying RBPs that selectively bind and traffic mRNAs to one or the other axon⁴¹. RBPs were identified as candidates for mRNA sorting if they fulfilled two criteria: (1) multiple putative binding motifs for that RPB are in the 3’UTR of key axonally translated mRNAs, and (2) the RBP is predicted and/or demonstrated to undergo phase separation and be present in RNA granules^42–44. A heatmap displaying the number of predicted motifs per 3’UTR was used to rank RBPs that were significantly more associated with peripherally or centrally translated mRNAs (Supplementary Fig. 6A). Motif predictions identify potential RBP binding sites and do not demonstrate direct occupancy at specific nucleotide positions. We predict that 16 RBPs are likely to bind the 3’UTRs of peripherally and centrally translated mRNAs, while 84 RBPs have more putative binding sites within peripherally translated mRNA transcripts, and 24 RBPs have more putative binding sites within centrally translated mRNA transcripts (Supplementary Fig. 6A).

We further investigated the three cohorts of RBPs using the STRING database protein-protein interaction functional annotation analysis (Supplementary Fig. 6B–D)⁴⁵. As predicted, all cohorts were enriched for RBPs involved in RNP granules cellular components. The set of RBPs that are predicted to bind equally to mRNAs present in the peripheral and central axons were enriched for nuclear proteins (Supplementary Fig. 6B). Interestingly, peripheral-associated RBPs were instead enriched for proteins with specific neuronal functions in the synapse, dendrite, and distal axon (Supplementary Fig. 6C). RBPs predicted to bind to the centrally located and centrally translated mRNA were enriched for cytoplasmic proteins that are involved in both RNP and stress granules (Supplementary Fig. 6D). These data prompted us to examine the possibility that some of these RBPs may be constituents of cytoplasmic RNP granules involved in selective axonal transport and translation.

If individual RBPs are implicated in mRNA sorting and axon-specific translation, we would expect to see spatially selective expression of these RBPs in the axons in vivo. Therefore, several candidate RBPs were further investigated to determine if in vivo axonal localization patterns correlate with their predicted mRNA binding patterns. We focused on the localization patterns of Splicing factor proline- and glutamine-rich (SFPQ), Fus-interacting protein 1/ serine and arginine rich splicing factor 10 (Fusip1/SRSF10), and Embryonic lethal, abnormal vision, drosophila-like 1 (HuR/ELAVL1). SFPQ contains motif-predicted 3’UTR binding sites on multiple peripherally translated mRNAs required for axonal and synaptic function including Hcn1, Gabrb3, Calb1, and Dnajc6 (Supplementary Fig. 6A). In contrast, SRSF10 contains motif-predicted 3’UTR binding sites on multiple centrally translated mRNAs encoding ion channels and receptors related to pain signaling: Htr3a, Htr5a, Trpv1, P2ry2 and Chrm1 (Supplementary Fig. 6A). ELAVL1 has predicted binding sites on almost all locally translated mRNAs regardless of axon specificity (Supplementary Fig. 6A).

Fluorescence immunohistochemistry paired with particle analysis demonstrates that SFPQ protein is enriched in peripheral axons of the sciatic nerve (Fig. 8A, D), consistent with our prediction that SFPQ preferentially binds to mRNAs in the peripheral axons^32,33. In contrast, we find that the SRSF10 protein is equally present in both central axons of the dorsal root and peripheral axons of the sciatic nerve (Fig. 8B, D). The third RBP tested, ELAVL1, is present in both the dorsal root and the sciatic nerve but is largely restricted to the nuclei of glia and other supporting cells (Fig. 8C, D). Thus, both SFPQ and SRSF10 are appropriately localized to contribute to trafficking and translation of mRNAs in axons, and SFPQ is appropriately located to promote selective sorting towards the peripheral axon. Our data also indicate another difference between the RBP puncta in the central versus peripheral axons: the puncta of SFPQ and SRSF10 were consistently larger within the sciatic nerve than in the dorsal root. This difference in size of the RBP-puncta suggests that there may be distinct features of RNP granules in each axon (Fig. 8E).

Fig. 8. RBPs are positioned to orchestrate the specialized axonal translatomes.

A–C Representative 60x fluorescent immunohistochemistry images of SFPQ (A), SRSF10 (B), and ELAVL1 (C) protein in dorsal roots and sciatic nerves, representing DRG central and peripheral axons, respectively. The solid yellow box shows a 4x magnified image of the area surrounded by yellow dotted lines, demonstrating the location and abundance of protein within axons and nuclei of supporting cells. Scale bars: Full image = 40 μm; inlaid image = 10 μm. Merged images show target IHC (white), TdTomato-labeled Nav1.8+ (Nav1.8-tdT) neurons (red), and DAPI-labeled nuclei (blue). D, E Quantification of the percent area (D) and average size (E) of SFPQ, SRSF10, and ELAVL1-positive puncta within Nav1.8-tdT+ axons in individual dorsal roots (n = 4) and sciatic nerves (n = 4). n = 4 biological replicates; multiple technical replicates averaged per biological replicate. Statistical significance determined by two-way ANOVA with Tukey’s multiple comparisons test and multiple two-sided, unpaired t tests (p-value ≤ 0.05). Data are represented as mean ± SD. All statistical analyses were derived from data using 4 mice. Source data are provided as a Source Data file.

Select RBPs guide the axonal sorting and transport of functional RNA regulons

To determine whether SRSF10 and SFPQ bind and sort mRNAs for axonal translation, we performed RNA Immunoprecipitation paired with sequencing (RIP-seq) in DRG neurons. In our experiments, SRSF10 and SFPQ protein were immunoprecipitated from DRG neurons grown in culture for 5 days, and mRNAs from these immunoprecipitates and from rabbit IgG negative controls were extracted and sequenced (Supplementary Fig. 7A–D and Supplementary Data S11). We identified the mRNAs that are significantly enriched in the immunoprecipitates compared to the inputs. The mRNAs enriched in the SFPQ and SRSF10 immunoprecipitates include transcripts encoding several ion channels and receptors that are locally translated within axons, including Hcn1, Scn2a, Scn11a, and Catsper2 (Fig. 9A, B and Supplementary Data S12, 13).

Fig. 9. Select RBPs bind functional RNA regulons of central or peripheral axons.

A, B Volcano plots from RIP-Seq experiments demonstrating the mRNA transcripts with significantly higher expression in SFPQ IP (A) and SRSF10 IP (B) samples as compared to the input fraction from the same DRG neuron cell culture experiments. Select locally translated mRNAs bound by SFPQ and SRSF10 are indicated on the volcano plots. Differential gene expression analysis was conducted using the RStudio DESeq2 package, using a two-sided Wald test and Benjamini-Hochberg FDR-adjusted p-values (n = 4 sets; p.adj ≤ 0.05; fold-change ≥ 1). C Chi-square analysis comparing RBP-bound transcripts to compartment-resolved TRAP translatomes, using 2 × 2 contingency tables generated by intersecting RIP-seq–identified SFPQ-bound or SRSF10-bound mRNAs with the background-filtered peripheral and central axon gene sets. A pearson chi-square test of independence revealed a significant association between RBP identity and axonal compartment (p ≤ 0.0001), with SFPQ-bound transcripts enriched among peripherally translated mRNAs and SRSF10-bound transcripts enriched among centrally translated mRNAs. C’ Odds-ratio analysis quantifying the direction and magnitude of these biases (from panel C). Points represent odds ratios, and error bars indicate 95% confidence intervals. SFPQ-bound transcripts were 1.26-fold more likely to belong to the peripheral axon translatome (OR = 1.26, 95% CI 1.09-1.49), whereas SRSF10-bound transcripts were preferentially associated with the central axon translatome (OR = 0.62, 95% CI 0.49-0.78) (n = 4 sets). D, E Gene ontology (GO) enrichment analysis visualized as a dot plot, where each dot represents the most significantly enriched biological processes (BP), cellular compartments (CC), or molecular functions (MF) from the SFPQ-associated peripheral axon regulon (D) and SRSF10-associated central axon regulon (E). Gene Ontology enrichment was performed using clusterProfiler enrichGO based on a one-sided hypergeometric test. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) (p.adj ≤ 0.05). Source data are provided as a Source Data file.

To define the functional identity of each RBP’s cargo, we first performed GO enrichment analysis on the full set of SFPQ- and SRSF10-bound transcripts identified by RIP-seq, whether or not these RNAs appeared in the local translatomes (Supplementary Fig. 7E, F’ and Supplementary Data S14, 15). SFPQ-bound mRNAs were enriched for biological processes related to neuronal projection organization and assembly, including axon shaft components, distal axons, growth cones, and synaptic structures, consistent with SFPQ’s established role in long-range axonal RNA trafficking (Supplementary Fig. 7E and Supplementary Data S14). In contrast, SRSF10-bound mRNAs were enriched for pathways related to mRNA processing and RNA splicing, but also contained transcripts encoding multiple structural and signaling components found in central axons—including presynaptic active zone proteins, calcium channel complexes, and microtubule motor complexes (Supplementary Fig. 7F, F’ and Supplementary Data S15). These results indicate that SFPQ and SRSF10 engage functionally distinct RNA regulons, each composed of transcripts that support different aspects of neuronal physiology.

Having established the global functional signatures of these RBP-bound regulons, we next asked whether SFPQ and SRSF10 preferentially associate with mRNAs that are locally translated in peripheral versus central axons. Chi-square analysis comparing RIP-seq targets to the TRAP datasets revealed a significant association between RBP identity and axonal compartment (Fig. 9C). SFPQ-bound transcripts were selectively enriched among peripherally translated mRNAs, whereas SRSF10-bound transcripts were disproportionately enriched among centrally translated mRNAs. Odds-ratio analysis further quantified these biases, showing that SFPQ-bound transcripts were 26% more likely to be peripheral, while SRSF10-bound transcripts were 38% less likely to be peripheral (i.e., 62% more likely to be central) (Fig. 9C′). These findings demonstrate that each RBP binds distinct RNA regulons that preferentially traffic to one or the other axon of sensory neurons.

To further resolve the biological functions of these compartment-directed regulons, we performed GO enrichment analyses on SFPQ- and SRSF10-bound transcripts that overlapped with the peripheral and central axonal translatomes, respectively (Fig. 9D, E). The SFPQ-associated peripheral translatome was enriched for biological processes related to metabolic pathways, axon development, neuron projection morphogenesis, as well as molecular functions linked to ATP binding and enzymatic activity. These categories align with the metabolic and structural demands of the distal sensory terminal and suggest that SFPQ coordinates a peripheral regulon tuned for axon maintenance, remodeling, and localized metabolism (Fig. 9D). In contrast, the SRSF10-associated central translatome was enriched for pathways governing calcium ion transmembrane import and cytoskeletal organization, and molecular functions such as microtubule binding and voltage-gated calcium channel activity (Fig. 9E). These functions reflect the specialized physiology of central presynaptic terminals, where precise control of calcium influx, cytoskeletal scaffolding, and synaptic machinery is essential for neurotransmitter release. Together, these results reveal that SFPQ and SRSF10 enable polarized RNA regulons that contribute to the divergent structural and synaptic identities of peripheral and central DRG axons. By coordinating the selective transport and local translation of distinct gene cohorts into each axonal branch, these RBPs provide a mechanistic framework for how a single sensory neuron maintains functional asymmetry across its long expanse.

Discussion

Spatial translatomics and high-resolution microscopy have revolutionized our ability to identify locally translated mRNAs and understand both the underlying mechanisms and the functional consequences of these subcellular specializations. Here, we show that neuronal axon projections from the same cell can have distinct translatomes to fulfill specialized functional requirements. While molecules that modulate synaptic strength and nociceptive signaling are locally translated in central axons of the somatosensory neurons, peripheral axons invest in sustaining the structural and metabolic demands of long-range projections and regeneration. Our studies also reveal the logic behind axonal sorting of select mRNAs; the RBPs, SFPQ and SRSF10, preferentially bind the 3’UTRs of mRNAs that are axonally translated, with SFPQ implicated in sorting to peripheral axons and SRSF10 emerging as a regulator of centrally translated mRNAs. Furthermore, the set of mRNAs that are locally translated in the peripheral and central axons includes multiple ion channels and receptors related to neuronal firing, nociceptive processing, and regenerative capacity of sensory neurons. Thus, regulated translation has the potential of selectively and dynamically modifying signal transmission within individual axon branches both in normal physiology and in pathophysiology.

Perhaps surprisingly, our studies indicate that many mRNAs that are translated within axons encode plasma membrane-bound ion channels and receptors. It is therefore appropriate to question whether axons are capable of the translational and secretory pathways required for membrane targeting and integration. Early electron microscopy studies did not detect ribosomes, rough endoplasmic reticulum, or Golgi apparatus in axons⁴⁶. More recently, monosomes, small diameter ER tubules, and Golgi satellites have been identified in peripheral axons^17,47–51. It has been suggested that axonal ion channels can be shuttled from axonal Golgi satellites to the plasma membrane through Rab6-positive exocytic vesicles⁵². In addition, atypical membrane proteins that are core-glycosylated rather than N-glycosylated may bypass the Golgi apparatus and be more directly shuttled to the axonal plasma membrane^53,54. Indeed, both the peripheral and central axonal translatomes include glycosylating enzymes (Galnt5 in the peripheral axon dataset and St6galnac2 in the central axon dataset). Several key locally translated mRNAs identified here, including Scn2a, Gabrb3, and Grin2a, encode proteins that have been observed to be core-glycosylated in axonal membranes⁵³. Taken together, these observations confirm that axonal translation, processing, and membrane targeting of ion channels and receptors occurs in distal axons.

The discovery that DRG central axons locally translate transcripts encoding ion channels, neurotransmitter receptors, and synaptic scaffold proteins suggests that presynaptic terminals contain a translatome that can support local neuroplasticity. Local synthesis of glutamatergic and GABAergic receptor components (Grin1/2, Grm1/3/5, Gabra1–5), along with scaffolding proteins such as Nrxn1, Nlgn1, and Dlg2/3 in the central axons, is consistent with the potential for activity-dependent modulation of synaptic function within the dorsal horn. Similarly, translation of voltage-gated calcium channel subunits and presynaptic release machinery (Stx1b, Vamp2) provides a potential mechanism to tune excitability and neurotransmitter release in response to somatosensory stimulation. Together, these findings indicate that DRG central axons are translationally active structures with the potential to influence synaptic signaling, thereby contributing to the habituation and/or sensitization of nociceptive circuits.

Our findings reveal that the neuromodulatory serotonin receptor, 5-HTR3A, is selectively located and translated in central axons, indicating that serotonergic signals in the spinal cord directly modulate primary somatosensory inputs. Indeed, axons from serotonergic neurons of the raphe nuclei are known to descend to the dorsal horn of the spinal cord, and ~ 15% of axo-axonic contacts with DRGs are serotonergic^55,56. While previous models emphasized indirect modulation of somatosensory neurons through GABAergic interneurons that express excitatory 5-HTR3A receptors^57,58, our findings suggest serotonergic stimulation of the central terminals of primary afferents themselves. Adjustments of pain signaling and sensation within the excitatory, glutamatergic central axon terminals may provide an important initial site of pain modulation. Consistent with this hypothesis, treatment with paclitaxel, which causes a painful, neuropathic condition, results in an increase in translation of 5-HTR3A receptors in the central axons³¹. Together, these results support a model in which the local translation of 5-HTR3A equips DRG central axons to integrate descending signals with ongoing sensory activity, shaping early stages of nociceptive information processing^56,59.

Local translation also provides a dynamic mechanism for structural remodeling within axonal compartments. In central DRG axons, the local synthesis of cytoskeletal and scaffolding proteins supports the dynamic modification of presynaptic architecture and receptor alignment, enabling synaptic plasticity within the dorsal horn. In peripheral DRG axons, local translation is poised to support the structural remodeling required for axonal maintenance, growth-cone dynamics, and interactions with the extracellular environment. The capacity for local synthesis of cytoskeletal and translational components may underlie the unique regenerative potential of peripheral, but not central, axons. Peripheral nerve regeneration depends on both neuron-intrinsic components and signaling pathways initiated by glial and immune cells in the local microenvironment^60,61. Dynamic remodeling of the axon cytoskeleton and functional connections to cell-extrinsic ECM proteins and glial cells may allow damaged sensory fibers to restore axonal integrity, rebuild growth cones, and re-establish functional connections after injury. Furthermore, local synthesis of ribosomal components and other protein-synthesis machinery suggests that peripheral axons can replenish or repair their translational apparatus on site, supporting the high biosynthetic demands of regeneration after mechanical injuries or other cell stresses²⁹. Together, localized regulation and synthesis of cytoskeletal outgrowth and translational machinery provide peripheral axons with regenerative capacity in response to nerve injury.

Our data indicate that local translation also can play a role in modulating axonal physiology by dynamically adjusting the complement of ion channels along DRG neurons. Although differential ion channel insertion in axonal membranes has long been known to influence neuroplasticity by altering membrane composition and excitability^62–64, our findings reveal that genes encoding several voltage-gated sodium channels are locally and differentially translated in the axons. There are nine distinct genes that encode different voltage-gated sodium channels, Nav1.1–1.9⁶⁵, and individual neurons of the CNS and PNS each express a subset of these channel genes^35,66,67. Our data indicate that Scn2a, which encodes Nav1.2, is translated in peripheral axons. Nav1.2 is a tetrodotoxin-sensitive (TTX) channel with fast kinetics^35,66–69. In addition, we find that the Hyperpolarization Activated Cyclic Nucleotide Gated Potassium Channel 1 (HCN1) is also selectively translated in the peripheral axons. HCN1 exhibits rapid kinetics, enhances action potential initiation and enables increased action potential frequency^70–72. Peripheral axons resemble dendrites in that they are the site of initial stimulation and signal integration⁵⁹ and we hypothesize that HCN1 and Nav1.2 may enable initiation and propagation of action potentials in both CNS dendrites and DRG peripheral axons. In contrast, we find that Scn11a, which encodes Nav1.9, is preferentially translated in the central axons. Unlike Nav1.2, Nav1.9 is a slowly activating/ inactivating, TTX-resistant channel that amplifies subthreshold signals^35,66–69. Nav1.9 is predominantly expressed in nociceptive neurons in the PNS, and we suggest that this channel may amplify subthreshold pain signals in the central axon and so adjust the electrophysiologic input to the spinal cord³⁵. Mutations or altered expression of many of these axonally translated components have been implicated in human pain syndromes. Gain-of-function mutations of Scn11a, a component of the central axon translatome, result in episodic cold pain sensation and painful peripheral neuropathy in people, while loss of function mutations disrupt inflammatory and cold-triggered pain sensation in mice and humans^35,73,74. Similarly, null mutations of Hcn1, which is translated in peripheral axons, result in reduced neuropathic pain and cold allodynia after nerve injury, when compared to wild-type mice after nerve injury⁷⁵. Therefore, precise control over Scn11a and Hcn1 localization and translation is likely required for appropriate pain sensation and may play a prominent role in neuropathic pain syndromes^73,74. Determining whether axon-restricted translation of these ion channels directly alters sensory excitability or pain behaviors will require future in vivo studies.

A common and often devastating cause of neuropathic pain is chemotherapy-induced peripheral neuropathy. Here, we integrate the analysis of locally translated mRNAs in DRG soma and axons with the genesets that change following paclitaxel treatment, a chemotherapeutic agent known to cause axonal degeneration and peripheral neuropathy. We found that paclitaxel alters the expression of many mRNAs encoding proteins critical for axonal transport and vesicle localization that are locally translated in all of the subcellular domains. In addition, mRNAs with altered expression in response to paclitaxel include mRNAs translated in the peripheral axons that contribute to cytoplasmic translation, neuronal projection extension, and environmental sensing, whereas paclitaxel-altered transcripts that are translated in central axons are associated with synaptic and secretory pathways. These findings highlight domain-specific molecular responses that overlap with neuropathy-associated gene programs and may contribute to chronic pain phenotypes. Although the TRAP experiments were performed during early postnatal development, we find that local translation of multiple identified components continues in adults, albeit at lower levels. Moreover, we find that paclitaxel causes an increase in translation of 5-HTR3A selectively within the central axons. Future studies will be needed to assess whether changes in axonal translation are a prevalent phenomenon in neuropathy models and to determine the physiologic consequences of these changes.

A key question addressed here is how specific RNA regulons, sets of functionally related mRNAs with synchronized expression, are coordinately transported and sorted to a specific axon. RBPs are known to regulate RNA regulons at the level of splicing, nuclear export, transport, localization, and stability². It has been shown that co-regulation of RNA regulons by specific RBPs is required for axon viability^33,76. In this study, we implicate two RBPs in the organization of central and peripheral axonal translatomes. SFPQ and SRSF10 were initially identified as RNA splicing factors containing both RNA recognition motifs (RRMs) and intrinsically disordered regions^77–79. In addition to its role in splicing, SFPQ has been identified as a component of axonal RNP granules and enables axonal transport of mRNA cargoes^33,80–82. While less is known about SRSF10, it has been shown to regulate metabolic processes such as glycolysis⁸³, and has also been functionally linked to neurogenesis and spermatogenesis^84–86. Here, we show that SFPQ and SRSF10 bind multiple locally translated mRNAs and that the subcellular distribution of these two RBPS is strikingly different, with SFPQ enriched in peripheral axons and SRSF10 present in both axons. Defining how this asymmetry is established will be an exciting area of future investigation. RIP-seq and GO analyses demonstrate that the SFPQ-associated regulon was enriched for metabolic pathways, neuronal projection morphogenesis, and distal axon maintenance, while the SRSF10-associated regulon was enriched for calcium channel complexes, cytoskeletal scaffolding, and presynaptic active zone components. These patterns support a model in which polarized RNA regulons reinforce the divergent structural and synaptic identities of peripheral and central axons.

While many RBPs haven been implicated in axonal trafficking, our data indicate that SFPQ is involved in selectively sorting mRNAs to the peripheral, rather than the central axons. This role joins a host of cytoplasmic functions qualifying SFPQ as a master regulator of axonal survival⁸⁰. For example, in peripheral neurons, SFPQ directly binds a kinesin complex containing the motor, KIF5A, and adapter, KLC1, to transport mRNAs encoding axonal survival factors, such as Laminb2 and Bclw^32,33,87. Indeed, both SFPQ knockdown and selective interruption in transport activity led to axon degeneration in sensory neurons. Moreover, cytoplasmic aggregation of SFPQ is implicated in several neurodegenerative disorders including frontotemporal lobar degeneration (FTLD), and Alzheimer’s disease, as well as sensory neuropathies including chemotherapy-induced peripheral neuropathy, while SFPQ mutations can cause motor neuron dysfunction and amyotrophic lateral sclerosis (ALS)^80–82. Our findings now extend our understanding of SFPQ and degeneration, with evidence that SFPQ- dependent RNA granule transport and sorting is likely to contribute to loss of epidermal innervation by peripheral somatosensory axons in neuropathy.

SRSF10 has known roles in regulating RNAs involved in neurogenesis and myelination, but has not previously been implicated in axonal trafficking or neurodegeneration. While SFSF10 is present at similar levels in both the central and peripheral axons, our data suggest that it has a preferential role in transcripts targeted to the central axons, including transcripts encoding calcium channels and regulators. Interestingly, conditional knockouts of SRSF10 in neural progenitor cells result in impaired performance on multiple learning and memory tasks, which involve localized calcium regulation⁸⁴.

The comprehensive comparisons of central and peripheral axon translatomes in DRG neurons assembled here provide critical resources for understanding the basis of compartmentalized neural function. The axonal translatomes indicate that genes associated with synaptic, electrophysiologic, structural, and regenerative functions are locally translated and are subject to modulation in response to environmental cues. Identification of distinct RBPs that coordinate intracellular transport and translation of RNA regulons highlights candidate mechanisms that may contribute to the asymmetric molecular composition of neuronal projections. Overall, the spatial regulation of translation and the selective routing of RNA regulons by SFPQ and SRSF10 provide a conceptual framework for understanding how subcellular specialization is maintained across distinct axonal environments and how neuropathic stress influences compartment-specific molecular programs.

Methods

Mouse lines and animal care

All experimental procedures were conducted in accordance with the National Institutes of Health (NIH) guidelines and were approved by the Dana-Farber Cancer Institute (DFCI) Institutional Animal Care and Use Committee (IACUC) in protocol #01-133.

Mouse Strains: L10a-eGFP^fl/fl (B6;129S4-Gt(ROSA)26Sor^{tm9(EGFP/Rpl10a)Amc}/J) mice were the generous gift of Myriam Heiman (Massachusetts Institute of Technology)²⁷. Nav1.8^cre (B6.129-Scn10a^tm2(cre)Jwo/H) and TdTomato^fl/fl (B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J) mice were kindly provided by David Ginty (Harvard Medical School). For in vivo TRAP studies, Nav1.8^cre/+ and L10a-eGFP^fl/fl mice were crossed to generate Nav1.8^cre; L10a-eGFP^fl/fl mice (GFP + ) and L10a-eGFP^fl/fl (GFP-) littermate controls. For tissue collection and histology, Nav1.8^cre and TdTomato^fl/fl mice were crossed to generate Nav1.8^cre; TdTomato^fl/fl mice with labeled sensory neurons.

Mice were group housed in a temperature-controlled environment (22 ± 2 °C) with a 12 h light/dark cycle and had free access to food and water. Mice aged P3-7 were used for all experiments unless otherwise stated. For TRAP experiments, 13-17 littermate animals of both sexes were pooled by genotype (GFP positive or negative) and used to generate TRAP samples. Pooling of both sexes was required to generate enough tissue material for TRAP-Sequencing. For tissue collection and histology, all TdTomato-expressing littermate animals, presumably including both sexes, were used to generate tissue samples for downstream applications.

Primary rat cell cultures

Timed pregnant Sprague-Dawley rats were purchased from Charles River for primary DRG neuron isolation. Embryos from timed pregnant Sprague-Dawley rats (aged E15) were used for primary cell culture isolation. All embryos, presumably of both sexes, were used. DRGs were dissected, dissociated, and plated in Matrigel (1:45; Thermo Fisher Scientific)-coated p35 dishes. DRG cultures were maintained in NeuroBasal medium supplemented with 2% B27, 1% Glutamax, 1% penicillin and streptomycin, 0.08% glucose, 1–100 ng/ml nerve growth factor (NGF)/brain-derived neurotrophic factor (BDNF; PeproTech), and 0.5 µM Cytarabine (AraC). Cultures were maintained in incubators at 37 °C with 7.5% CO₂ for 5 days.

Translating ribosomal affinity purification (TRAP)

The TRAP protocol was adapted from Myriam Heiman²⁷. Roughly 15 Nav1.8-Cre; L10a-eGFP mice aged P3-7 and L10a-eGP littermate controls were euthanized by CO₂ inhalation and cervical dislocation. Early postnatal (P3–P7) mice were used to reduce myelination and glial abundance in peripheral nerve and spinal cord preparations, thereby increasing neuronal specificity and RNA yield in compartment-specific TRAP-seq experiments. Lumbar (L1-6) dorsal root ganglia (Soma), sciatic nerve (PA), and lumbar dorsal roots and spinal cords (CA) were quickly manually dissected in Dissection Buffer (1X HBSS, 2.5 mM HEPES-KOH [pH7.4], 25 mM Glucose, 4 mM NaHCO₃, 100 µg/mL cycloheximide). Dorsal roots and sciatic nerves were transected at their junction with the DRG soma. The central compartment included dorsal roots extending into the lumbar spinal cord, and the peripheral compartment included sciatic nerve segments extending distally to approximately the knee, representing proximal-to-mid axonal regions.

The pooled Soma and CA compartment tissues were immediately homogenized in ice-cold Polysome Extraction Buffer (20 mH M HEPES-KOH [pH 7.4], 5 mM MgCl₂, 150 mM KCl, 0.5 mM DTT, 100 µg/mL cycloheximide, protease inhibitors (EDTA-free), 40 U/mL RNAsin, 20 U/mL Superasin) using glass Dounce homogenizers. The pooled PA compartment tissue was enzymatically digested (0.1% Collagenase II, 0.2% Hyaluronidase, 40 U/mL RNAsin, 20 U/mL Superasin) in Dissection Buffer for 5 min at room temperature, then resuspended in ice-cold Polysome Extraction Buffer and Dounce homogenized. Homogenates were centrifuged for 10 min at 2000 × g, 4 °C, to pellet large cell debris, and 1% NP-40 (AG Scientific, San Diego, CA) and 30 mM DHPC (Avanti Polar Lipids, Alabaster, AL) were added to the supernatant. After incubation on ice for 5 min, the lysates were centrifuged for 10 min at 20,000 × g, 4 °C. Supernatants were pre-cleared with unbound Pierce Protein L Magnetic Beads at 4 °C for 1 h. Custom anti-GFP antibody (Htz-GFP-19C8; RRID:AB_2716737 and Htz-GFP-19F7; RRID:AB_2716736) was used to coat Pierce Protein L Magnetic Beads at a concentration of 2 μg antibody and 10 μl beads per mouse. The coated beads were added to the supernatant and incubated at 4 °C with end-to-end rotation overnight. Beads were subsequently collected on a magnetic rack, washed three times with high-salt IP Wash Buffer (20 mM HEPES [pH 7.4], 350 mM KCl, 5 mM MgCl₂, 1% NP-40, 0.5 mM dithiothreitol, 100 μg/ml cycloheximide, 3% bovine serum albumin [IgG Protease-free]) and immediately placed in Lysis Buffer with β-ME (Absolutely RNA Nanoprep Kit) (Agilent Technologies, Santa Clara, CA) and incubated for 10 mins at room temperature. Eluted RNA in Lysis Buffer was removed from the beads on a magnetic rack. RNA was further purified using the Absolutely RNA Nanoprep Kit with column purification and DNase digestion. RNA integrity and concentration was measured on an Agilent 2100 Bioanalyzer System with RNA 6000 Pico Kit (Agilent Technologies, Santa Clara, CA); amounts of RNA were estimated as <100 pg for axon samples and < 1 ng for soma samples. Four biological replicates were performed for CA and PA tissues, and eight replicates were performed for somatic tissues.

Low input total RNA sequencing and analysis

Low-input total RNA Sequencing ( >10 pg total RNA or 1–500 cells) was performed at the University of California-San Francisco Functional Genomics Core Facility. RNA libraries were prepared using the SMART-Seq v4 Mouse Kit (Takara Bio Inc.) and sequenced on an Illumina HiSeq 4000 instrument via 50 bp sequencing. Sequences were aligned to the mouse reference genome with STAR_2.7.2b²⁸. RNA-seq analysis of raw counts was performed in R v4.3.2 for sample-level quality control (principal component analysis), Differential Gene Expression testing for pairwise, group comparisons (DESeq2)³⁴, and Functional Annotation (enrichGO)^88,89.

Integrative cross dataset analysis

To identify genes enriched in paclitaxel-treated cells relative to naïve cells within each cell type, we performed differential gene expression analysis using Seurat’s FindAllMarkers function with default settings on data generated by Renthal and colleagues^32,33. P-values were subsequently adjusted for multiple comparisons using the Benjamini-Hochberg procedure (FDR). We then intersected the differentially expressed genes from each cell subtype with those identified in our TRAP-seq data across anatomical compartments. For each resulting intersected gene list, we conducted Gene Ontology (GO) enrichment analysis – covering biological process, molecular function, and cellular component categories – using the normalized gene counts from our TRAP-seq dataset serving as the background set.

Paclitaxel treatment

2-mo-old Nav1.8^cre; TdTomato^fl/fl mice (20–25 g) of both sexes were injected i.p. with 20 mg/kg paclitaxel (#T7402) (Sigma Aldrich, St. Louis, MO) every other day (days 1, 3, 5, and 7, for a total of four injections). Paclitaxel was prepared in one part vehicle (1:1 vol/vol Cremophor EL [EMD Millipore] and 200-proof ethanol) and two parts sterile saline (UPS) and injected at 10 µl/g. Control mice were injected with one part vehicle and two parts saline. Tissues were collected 9 d after the final injection (day 16).

Puromycin injection

P7 Nav1.8^cre;TdTomato^fl/fl mice and 2-mo-old Nav1.8^cre;TdTomato^fl/+ mice were i.p. injected with puromycin dihydrochloride (#P33020) (Research Products International, Mount Prospect, IL) in 1XPBS at a concentration of 22.5 mg/kg and incubated for 25 mins (P7) or 45 mins (2-mo). Mice were euthanized by CO₂ inhalation followed by cervical dislocation. Puromycin incorporation into mouse DRG, nerve, and dorsal root tissues was confirmed by immunohistochemistry with an overnight incubation of 647-conjugated anti-puromycin antibody (Sigma, MABE343-AF647; RRID: AB_2736876; 1:1000) in 5% NGS-PBST at 4 °C (see Fluorescent immunohistochemistry method).

Mouse histological methods

Dorsal root ganglia (Soma), sciatic nerves (PA), and dorsal roots (CA) were rapidly dissected from euthanized mice in ice-cold 1X PBS and placed immediately fixed in 4% PFA overnight. Tissues were washed 3 × 5 min with 1XPBS, then prepared for embedding with 10% to 30% sucrose gradient over 3 days, embedded in NEG-50 medium (VWR, Radnor, Pennsylvania), and sectioned at 16 μm on a cryostat.

Proximity ligation assay

Proximity ligation was performed according to the Duolink In Situ Proximity Ligation (Millipore Sigma, Burlington, MA) protocol. In brief, 20μg of primary anti-Puromycin (Sigma, #MABE343; RRID:AB_2566826) antibody was conjugated to the Duolink PLA Minus Probe using the Duolink In Situ Probemaker Kit (Millipore Sigma, Burlington, MA). Frozen tissue sections were accommodated to room temperature for 30 mins. Slides were then permeabilized with 0.1% PBS-Triton-X-100 (PBST) and blocked with Duolink Blocking Solution. Slides were incubated with Puromycin-conjugated Duolink Probe (1:250) and primary antibodies, anti-5HTR3A (Abcam, #AB38148; RRID:AB_945424; 1:500), anti-SCN11A (Alomone, #ASC-017; RRID:AB_2040200; 1:500), anti-SCN2A (Alomone, #ASC-002; RRID:AB_2040005; 1:500), or anti-HCN1(Alomone, #APC-056; RRID:AB_2039900; 1:500) at 4 °C overnight. Unbound primary antibodies were conjugated to Duolink In Situ PLA Probe Anti-rabbit Plus (#DUO92002, Millipore Sigma) for one hour, followed by ligation and polymerization steps with Duolink In Situ Detection Reagents Far Red (#DUO92013, Millipore Sigma) to amplify the fluorescent signal. DRG axons in 2-mo-old animals were labeled with a 488-conjugated anti-TUBB3/TUJ1 antibody (BioLegend, #803203; RRID:AB_2564757, 1:500) overnight in 4 °C. DRG axons in P7 mice were visualized by endogenous Nav1.8-TdTomato fluorescent signal. Coverslips were mounted with ProLong Gold Antifade Mountant complete with DAPI (#P36931, ThermoFisher).

RNAscope in situ hybridization

In situ hybridization was performed according to the Co-detection Immunohistochemistry with RNAscope Fluorescent Multiplex Reagent (ACDBio, Burlington, MA) protocol for fixed-frozen tissue samples. In brief, sample slides were accommodated to room temperature for 30 mins. Sections were then prepared by washing with 1X PBS before baking for 30 mins at 60 °C. Sections were then post-fixed with 4% PFA for 15 minutes at 4 °C. Samples were dehydrated through serial incubations in 50, 70, and 100% Ethanol. Samples were treated with hydrogen peroxide for 10 min and then underwent antigen retrieval with RNAscope Target Retrieval Solution in a steamer basket for 5 min. Primary anti-RFP antibody (VWR, #RL600-401-379; 1:500) was applied to slides and incubated at 4 °C overnight. Samples then underwent post-primary fixation in 4% PFA for 30 mins, followed by Protease Plus incubation for 30 mins in a HybEZ Oven at 40 °C, probe hybridization for 30 mins at 40 °C, AMP hybridization for 30 mins at 40 °C, HRP signal development for 15 mins at 40 °C, Opal 690 Reagent (Akoya Biosciences) incubation for 30 mins at 40 °C, and HRP blocking for 15 mins at 40 °C. Samples were then incubated with secondary antibody (Invitrogen: AlexaFluor H + L) for 60 mins, followed by two wash steps. Coverslips were mounted with ProLong Gold Antifade Mountant complete with DAPI (#P36931, ThermoFisher).

Fluorescent Immunohistochemistry

Fixed frozen sample slides were accommodated to room temperature for 30 mins. Samples were then permeabilized 3 x 5 min with 0.1% PBS-Triton-X-100 (PBST) and blocked with 5% normal goat serum (NGS) (in 0.1% PBST) for 1 h at room temperature. Primary anti-RFP (VWR, #RL600-401-379), anti-SFPQ (Abcam, #ab38148; 1:500), anti-SRSF10 (Bioss, #BS-13229R; 1:400), and/or anti-HuR (Proteintech, #11910-1-AP; 1:250) antibody were added to slides and incubated at 4 °C overnight. Slides were washed 3 x 3 min with 0.1% PBST and then incubated with Donkey anti-Rabbit Alexa Fluor 568 secondary antibody (ThermoFisher Scientific Cat#A10042; RRID:AB_2556622; 1:1000) for 1 h at room temperature, followed by additional wash steps. Coverslips were mounted with ProLong Gold Antifade Mountant complete with DAPI (#P36931, ThermoFisher).

Confocal microscopy and image analysis

All mouse histological methods were imaged on a Nikon Eclipse Ni C2 Si laser scanning upright confocal with 4 laser lines. Slides were imaged at 20 x, 20 × 2 (digital zoom), and 60x magnification, acquired using NIS Elements software v.4.20.03, and saved in.nd2 format for further processing and analysis. Images were opened in ImageJ v2.16.0 and subjected to basic preprocessing steps to enhance both Nav1.8 + or Tuj1 + axon and particle visibility; this included adjusting brightness and contrast and removing noise from the background by despeckling. A default threshold was first applied to the axon channel, and a mask was generated from the selection and added as a region of interest (ROI). PLA or RNAscope puncta were thresholded using the ImageJ Renyi Entropy default algorithm, and the IHC signal was manually thresholded in ImageJ (using consistent settings between samples) to ensure clear delineation between the particles and the background. Once the appropriate threshold was set, the PLA, RNAscope, or IHC area contained within the previously set axon ROI was analyzed using the Analyze Particles function. For PLA images, particle analysis was performed within a size range of 2 - 50 pixels for P7 samples, and within a 2 - 25 pixel and 0.45 - 1.0 circularity range for 2-mo-old samples to account for increased background in adult samples. For RNAscope images, particle analysis was performed within a size range of 0.2 - 200 pixels. The results were output to a results table, which included measurements such as particle count, percent area, and particle size. Following automatic particle detection using ImageJ macros, the results were visually inspected to ensure accurate particle identification. If necessary, additional particles were manually added or removed from the analysis based on visual verification.

RNA Extraction and qRT-PCR

RNA was extracted from mouse testes, liver, dorsal root ganglia, sciatic nerves, and dorsal roots by basic phenol:chloroform extraction followed by isopropanol RNA precipitation. RNA was converted to cDNA with the High-Capacity RNA-to-cDNA Kit (Applied Biosystems). Template cDNA and Catsper2 primers were incorporated into a SYBR Green master mixture (Applied Biosystems), and mRNA expression was quantified using the EP Gradient s Realplex Mastercycler (Eppendorf). GAPDH was used to normalize gene expression. Each sample was assessed in duplicate and included a template-free control. All primers used were synthesized by Eurofins.

RBP Bioinformatic analysis

3′ UTR sequences were obtained from the UCSC Genome Browser (GRCm39/mm39 assembly). For each analyzed gene, the longest annotated 3′ UTR sequence was extracted and used for motif analysis. Bioinformatic RBP motif analysis of 3’UTRs were performed using the oRNAment and RBPmap databases^42,43. The locally translated mRNAs associated with significant functional annotations were used as the input transcript reference list. Transcripts were limited to 3’UTRs. Predicted motif score cutoff was set to 90%. RBPs found to have predicted binding motifs with the key 3’UTRs were further analyzed for their ability to form ribonucleoprotein granules using the RPS.renlab database⁴⁴. Functional Annotation Analysis for RBP protein-protein interactions was performed using the STRING database⁴⁵. The minimum required interaction score was set to High Confidence (0.700).

RNA Immunoprecipitation (RIP)

RNA Immunoprecipitation was performed according to the EZ- Magna RIP RNA-Binding Protein Immunoprecipitation (Millipore Sigma) protocol. In brief, on Day 5 of primary DRG neuron cell culture, cells were washed with ice-cold PBS and lysed at −80 °C overnight in Complete Lysis Buffer. Lysates were spun down to pellet cellular debris, and the supernatant added to Protein A/G Beads conjugated with 5 μg of primary anti-SFPQ (Abcam, #ab38148), anti-SRSF10 (MBL, #50304712), or anti-rabbit IgG (EZ-Magna Kit). Lysate-bead mixtures were incubated at 4 °C overnight with end-to-end rotation. Protein-RNA complexes were eluted from beads in a Proteinase K Buffer for 30 mins at 55 °C. RNA was further purified using the phenol:chloroform:isoamyl alcohol procedure, followed by absolute ethanol and salt RNA precipitation.

RIP-Sequencing

Sequencing was performed by the DFCI MBCF:Genomics Core Facility using NovaSeq X: Low Input mRNAseq ( > 1 ng RNA). Libraries were prepared with Takara SmartSeq v4 full-length cDNA synthesis using oligo dT priming & 5’ template switching enzyme. Sequencing was performed with 40 M 150 bp read pairs from Illumina NovaSeq X Plus. Sequences were aligned to the rat reference genome with STAR_2.7.2b²⁸. RNA-seq analysis of raw counts was performed in R for sample-level quality control (principal component analysis), Differential Gene Expression testing for pairwise, group comparisons (DESeq2)³⁴, and Functional Annotation (enrichGO)^88,89.

Western blot

RBP immunoprecipitations for RIP experiments were confirmed by Western Blot. Post-IP (EZ-Magna kit protocol), lysates were separated by NuPAGE 4–12% Bis-Tris SDS-Page Protein Gels and probed with anti-SFPQ (1:1000; Abcam, #ab38148), and anti-SRSF10 (1:1000; MBL, #50304712). Bands were visualized with Goat anti-rabbit IgG (H + L)- HRP conjugated secondary antibodies (Bio-Rad; 1706515; RRID:AB_11125143; 1:10,000) and SuperSignal chemiluminescent substrates signal (Thermo Fisher Scientific). Blots were imaged using the AI600 Chemiluminescent Imager (GE Healthcare).

Statistical methods

All statistical analyses were performed using R v4.3.2 and GraphPad Prism v10. Differential gene expression analysis was conducted using the RStudio DESeq2 package, using a two-sided Wald test and Benjamini-Hochberg FDR-adjusted p-values. Gene Ontology enrichment was performed using clusterProfiler enrichGO based on a one-sided hypergeometric test. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR). All histological data were quantified in Prism and are represented as mean ± standard error of the mean (SEM). Grouped data with multiple comparisons were analyzed for statistical significance using two-way ANOVA with Tukey’s multiple comparisons test. Contingency data were analyzed for statistical significance using Chi-Square. A P-value ≤ 0.05 was considered statistically significant. Data distribution was assumed to be normal, but this was not formally tested. Effect size was determined by Cohen’s d measurements, or the standardized difference between the means of two individual groups ((mean 1 – mean 2) / standard deviation). Statistical details of experiments can be found in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.S.S. conceptualized and designed the study, performed the experiments, analyzed the data, and wrote the manuscript. E.N. assisted with tissue processing, performed immunohistochemical experiments, ran bioinformatic analysis, and reviewed and edited the manuscript. J.Z.P. assisted with tissue processing, conducted immunohistochemical experiments, and reviewed and edited the manuscript. M.F.P. provided technical expertise, maintained mouse colonies, and reviewed and edited the manuscript. S.A.B. and W.R. developed the web resource and performed bioinformatic analysis. R.A.S. conceptualized and designed the study, supervised the research, contributed to the data interpretation, and provided critical revisions to the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by NIH grants R01-NS050674 and R01-CA205255 (RAS), T32-AG000222 (ES), the Rita Allen Foundation, and the Burroughs Wellcome Fund (WR).

Data availability

All raw and processed TRAP-seq and RIP-seq data from this study were deposited to Gene Expression Omnibus at GEO:GSE289550 and GEO:GSE289551 and are publicly available: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE289550. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE289551. TRAP-seq data are also available for browsing and analysis via the Pain-seq multi-omic data resource at http://painseq.shinyapps.io/CompartmentTRAP/ (or trap.painseq.com). The TRAP-seq, RIP-seq, and integrative cross dataset analyses performed in this study are available in the Supplementary Data files. All visual data analyzed in this study can be found in the Source Data file. Requests for further information and resources should be directed to the lead contact, Rosalind Segal (Rosalind_segal@dfci.harvard.edu). Source data are provided in this paper.

Code availability

Novel code used to generate the web resource has been deposited to GitHub (http://github.com/Renthal-Lab/CompartmentTRAPShiny): https://doi.org/10.5281/zenodo.21176607.

Competing interests

The authors declare the following competing interests:W.R. declares research support from Eli Lilly and company. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77192-x.