scRNA-Seq captures muscle-resident SCs across healthy and denervated muscles. SC regenerative responses are activated by axonal injury, but the extent and persistence of the injury can lead to a spectrum of cellular processes within the regenerative timeframe. Utilizing a sciatic nerve crush injury model in SC reporter mice (S100GFP-tg), we induced complete muscle denervation, characterized by a total loss of force upon nerve stimulation at 7 days post injury (7 dpi) (Figure 1A). In uninjured 2-month-old Sod1–/– mice, the force generated during nerve stimulation was 33% lower than that of Sod1WT/WT controls, while the force generated in response to direct muscle stimulation did not differ between genotypes (Figure 1B). This finding is indicative of the presence in Sod1–/– mice of denervated muscle fibers that were highly contractile but nonresponsive to nerve stimulation. Consistent with published reports of elevated markers of muscle fiber denervation at 2 months of age in Sod1–/– mice (10), the neurotransmission impairments we observed in Sod1–/– SC reporter mice (S100GFP-tg Sod1–/–) were specifically noted at 2 months of age. The denervation event observed in the muscles 2-month-old Sod1–/– mice appears to be transient in nature, as evidenced by the lack of differences in force elicited by nerve and direct muscle stimulation at either 1 or 3 months (Figure 1C). In contrast, there was no functional evidence of denervation in Sod1WT/WT controls at any time point (Figure 1C). This temporal pattern of innervation loss and subsequent reinnervation in Sod1–/– mice correlates with direct assessments of NMJ denervation observed histologically, where the percentage of denervated NMJs was significantly elevated at age 2 months (~23%) and returned to baseline by age 3 months (Figure 1D). Morphological assessments of NMJs in both denervation models and healthy innervated controls confirmed the presence of tSCs across all conditions (Figure 1E). Despite distinct triggering events and varying degrees of denervation between models, tSCs remained a constant feature at NMJs, suggesting underlying transcriptional commonalities among the groups that warranted further investigation.

Characterization of muscle denervation in WT and Sod1–/– mice and muscle-resident Schwann cell subtypes across different denervation states via scRNA-Seq. (A and B) Maximum tetanic contraction forces (N/cm2) generated by nerve stimulation and direct muscle stimulation in WT mice 7 days post sciatic nerve crush injury (7 dpi) (A) and in S100GFP-tg Sod1–/– mice (B), compared to age-matched WT controls (n = 4 per genotype). (C) Comparison of maximum isometric force ratios elicited by nerve versus direct muscle stimulation across 1–3 months in S100GFP-tg Sod1–/– (n = 3–5) and control mice (n = 3–5). (D) Percentage of denervated NMJs in gastrocnemius muscles of S100GFP-tg and S100GFP-tg Sod1–/– mice aged 1–3 months. **P < 0.01; ***P < 0.001; ****P < 0.0001 by 2-way ANOVA with Tukey test for multiple testing correction. (E) Representative staining of NMJs showing Schwann cells (S100B; green), nerve terminals (NF/SV2; cyan), acetylcholine receptors (AChR; red), and nuclei (DAPI; blue) in 2-month-old S100GFP-tg control mice, S100GFP-tg mice 7 dpi and Sod1–/– mice without nerve injury. Scale bars: 25 μm. (F) Experimental workflow: Bilateral gastrocnemius (GTN) and tibialis anterior (TA) muscles were harvested from 2-month-old S100GFP-tg, S100GFP-tg Sod1–/–, and S100GFP-tg mice 7 dpi, followed by FACS to isolate GFP+ and GFP– cells, then processed for scRNA-Seq using the 10X Chromium platform. (G) FACS plots showing gating strategies for GFP+PI– single cells, with FMO (PI only) controls on the left. (H) UMAP plot visualizing 15 distinct cell clusters. (I) UMAP plots displaying transcript levels for myelin Schwann cell (mSC) markers (Mbp, Mpz), general Schwann cell markers (Sox10), Schwann cell repair phenotype (Ngfr), and terminal Schwann cells (tSC) (Cspg4, Kcnj10).

While the classification of SCs has largely relied on histological examination and scRNA-Seq from nerve extracts (11), transcriptomic data on muscle-resident SCs, especially tSCs, remains scant (12–15). A particular challenge faced by these studies is the rarity of tSCs limiting the cell numbers previously analyzed to as few as 100 individual cells. To overcome this limitation and gain a thorough understanding of the transcriptomic changes in muscle-resident SCs, as well as other cells in the muscle microenvironment during NMJ remodeling, we used scRNA-Seq on pooled tibialis anterior (TA) and gastrocnemius (GTN) muscles from S100GFP-tg mice, S100GFP-tg Sod1–/– mice, and S100GFP-tg mice at 7 dpi, all at 2 months of age (Figure 1F). After digesting the muscles, we sorted GFP+PI– and GFP–PI– single cells using fluorescence-activated cell sorting (FACS) (Figure 1G) and then analyzed the cells by droplet-based scRNA-Seq. Our unbiased clustering approach using uniform manifold approximation and projection (UMAP) classified 54,273 cells (S100GFP-tg [n = 5,330]; S100GFP-tg Sod1–/– [n = 9,577]; S100GFP-tg 7 dpi [n = 39,366]) into 12 non-SC clusters and 2 SC clusters, including mSC and tSC clusters (Figure 1H). Both SC clusters (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.195917DS1) displayed classic SC markers such as S100b and Sox10 (Figure 1I and Supplemental Figure 1A) and no SC clusters were identified in GFP– cells. Notably, clusters of mSCs were marked by expression of myelin-associated genes like Mbp and Mpz (Figure 1I), while the tSC cluster presented expression of recently reported non-mSC markers, Cspg4 (Ng2) and Kcnj10 (Kir4.1) (Figure 1I) (12, 14). Additionally, the tSC cluster prominently expressed Ngfr (p75NTR), a classical SC repair marker. S100GFP-tg and S100GFP-tg Sod1–/– samples displayed a comparable total number of SCs (1,649 and 2,101, respectively); however, relatively few SCs (n = 369) were isolated from S100GFP-tg 7-dpi mice, despite having nearly 4-fold more total cells than the other groups. These muscles (7 dpi), however, were highly enriched for mesenchymal cells (+30%), monocytes (+10), and a Spp1+Cora1a+ cell cluster we termed as phagocytic (Supplemental Figure 1B). Interestingly, we found many GFP+ non-SC clusters by FACS that expressed Gfp mRNA but had little to no expression of S100b (Supplemental Figure 2). These cells were abundant in S100GFP-tg 7-dpi mice, and likely suggestive that SC dedifferentiation was underway and prevalent at this injury time point, but this remains to be experimentally validated.

Differential SC dynamics during NMJ remodeling. Reclustering of SC transcriptomes revealed 5 subtypes, including 3 mSCs (mSC-A, mSC-B, mSC-C) and 2 distinct tSC subgroups, termed tSC-A and tSC-B (Figure 2, A–C). Notably, there was a significant increase in the tSC-B subgroup under both denervation conditions, representing approximately 20% of the SCs in 2-month-old S100GFP-tg Sod1–/– mice and approximately 50% in the S100GFP-tg 7-dpi mice but only 4% in the S100GFP-tg controls. Also common to both denervation models was a dramatic reduction in the proportion of mSC-A, representing 63% of the SCs in S100GFP-tg, but only 31% in S100GFP-tg Sod1–/–, and not detected in S100GFP-tg 7-dpi mice, as was also true for tSC-A. Additionally, the mSC-B and mSC-C clusters were more prevalent in the denervation models than in controls.

Dynamics of muscle-resident Schwann cell subtypes in healthy and remodeling neuromuscular junctions. (A) UMAP visualization of reclustered muscle-resident Schwann cell clusters, identifying 2 terminal Schwann cell subclusters (tSC-A and tSC-B) and 3 myelin Schwann cell subclusters (mSC-A, mSC-B, mSC-C). (B) Proportions (%) of each Schwann cell subcluster, with corresponding UMAP plots (C) across conditions in WT, Sod1–/–, and 7-dpi mice. (D) Heatmap illustrating gene ontology (GO) pathway analysis results, with enriched biological processes for each Schwann cell cluster presented as z scores of normalized –log(P value) for each GO term. (E) Diffusion map showing Schwann cell subclusters with 3 trajectory lineages from slingshot analysis superimposed. (F) UMAP feature plots indicating expression patterns of key genes (Mpz, Gap43, Cola1a), highlighting their distribution within lineages 1, 2, and 3, respectively. (G) Variation in cell density over pseudotime for each lineage in WT, Sod1–/–, and 7-dpi mice, reflecting differential engagement in nerve repair processes.

Pathway analysis highlighted that the tSC-A cluster was associated with synapse organization and structural pathways (Figure 2D and Supplemental Figure 3A), while tSC-B showed enrichment for biological processes tied to migration, cell polarization, and morphogenesis (Figure 2D and Supplemental Figure 3B). The mSC-A cluster primarily engaged classical myelin production pathways, including axon ensheathment, myelination, and cholesterol metabolism (Figure 2D and Supplemental Figure 3C). Interestingly, our data suggest that mSC-B is characterized by gene expression programs defined by mesenchymal differentiation (Figure 2D and Supplemental Figure 3D), consistent with findings from studies of peripheral nerve injury in which SCs are reported to express epithelial-mesenchymal transition (EMT) markers and respond to transforming growth factor β (TGF-β) signaling, a well-known inducer of EMT genes (16, 17). Finally, mSC-C was linked with extracellular matrix (ECM) organization and collagen production (Figure 2D and Supplemental Figure 3E).

To better understand the dynamic response of SC subpopulations during NMJ remodeling, we performed Slingshot trajectory analysis (18) on a diffusion map–reduced space of muscle-resident SCs (Figure 2E). This analysis revealed 3 primary lineages. Lineage 1, which originates in the mSC-A cluster, demonstrated enrichment for myelin-associated genes such as Mpz (Supplemental Figure 3C), underscoring its involvement in myelination and axonal support (Figure 2, E and F). The other 2 lineages arise from the mSC-B cluster and coursed through nonmyelinating tSC subpopulations. Lineage 2 was characterized by an enrichment of synapse-related genes, such as Gap43 (Figure 2E and Supplemental Figure 3A), highlighting its role in synaptic maintenance and regeneration. Meanwhile, Lineage 3 extended toward the mSC-C cluster, which showed enrichment for ECM and mesenchymal differentiation genes, including Col1a1, suggesting its involvement in structural remodeling and repair (Figure 2, E and F, and Supplemental Figure 3E).

Our analysis across different experimental groups revealed a complete absence of myelination-associated Lineage 1 SCs in WT (S100GFP-tg) mice at 7 dpi. This finding aligns with established knowledge that acute nerve injury not only suppresses myelination-specific pathways but also leads to the reduction of SCs contributing to these pathways, while concurrently activating myelin clearance, a typical immediate response to nerve damage (19) (Figure 2G). Conversely, control noninjured S100GFP-tg mice progressed more rapidly along the myelination trajectory, indicating robust myelination activity from the onset of the predicted trajectory. Sod1–/– mice–derived muscle-resident SCs, representing a model of spontaneous noninjurious denervation of a fraction of the fibers within the muscle, displayed a moderated progression in this lineage. Lineage 2, associated with synaptic maintenance and regeneration, revealed increasing cell density with pseudotime, beginning with the S100GFP-tg controls and progressively increasing in S100GFP-tg Sod1–/– and 7-dpi SCs. The pattern of increasing cell density in Lineage 2 for both S100GFP-tg Sod1–/– and 7-dpi SCs is consistent with increased demands for synaptic remodeling in response to denervation through the remodeling trajectory. Lastly, Lineage 3, linked to ECM remodeling and mesenchymal differentiation, showed pronounced enrichment in the 7-dpi SCs, indicative of a shift toward repair functions following acute denervation. Both S100GFP-tg and Sod1–/– SCs exhibited phases of activation within this lineage, indicating ongoing remodeling activities during neuromuscular adaptation and repair (Figure 2G).

The remodeling of NMJs is associated with increased numbers of tSCs and larger synaptic areas. Given the substantial capture of muscle-resident SCs in 2-month-old Sod1–/– mice compared with the 7-dpi mice, and the recognition of this period as a crucial phase for SC-mediated repair against denervation, we undertook a focused examination of cellular and morphological changes during the denervation-reinnervation cycle in Sod1–/– mice. We performed detailed imaging analyses on fixed fiber bundles from S100GFP-tg Sod1–/– and S100GFP-tg control mice at 1, 2, and 3 months of age (Figure 3, A and B). Our comprehensive analysis defined NMJ structure, accounting for 16 distinct features related to pre- and postsynaptic structures, including tSC number and morphology, allowing us to track the dynamics of NMJ remodeling and SC activity over a transient repair period. Detailed descriptions of each morphological feature appear in Supplemental Figure 4. In muscles from 2-month-old mice, our findings revealed overall smaller values for S100GFP-tg Sod1–/– mice compared with fully innervated controls in nerve terminal perimeter and overlap of the nerve terminal with AChRs by 33% and 60%, respectively (Figure 3, C and D). Meanwhile, the area of AChRs was observed to be 30% larger in S100GFP-tg Sod1–/– mice compared with controls, and the total tSC area and number of tSCs per NMJ were 60% and 3-fold greater, respectively, in the S100GFP-tg Sod1–/– group (Figure 3E). As a result of the large increase in tSC number, the terminal area per tSC was 38% smaller in S100GFP-tg Sod1–/– mice. Given that both central and peripheral glia respond to neuronal injury by increasing cell number (2, 20), increased tSC number and area at the NMJ are likely acting to protect or promote NMJ integrity during the early neuromuscular events linked to the Sod1–/– NMJ remodeling phenotype.

The remodeling of neuromuscular junctions (NMJs) is associated with greater tSC numbers, larger synaptic areas, and enhanced proliferation in S100GFP-tg Sod1–/– mice. (A) Schematic of NMJ analysis, consisting of collecting NMJs images from muscle fiber bundles followed by the generation of feature masks and their measurements. Additional details on the generation of the masks are provided in Supplemental Figures 3 and 4 and Methods. (B) NMJs stained for S100B (Schwann cells; green), NF/SV2 (nerve; cyan), AChRs (α-bungarotoxin, BTX; red), and nuclei (DAPI; blue) in 2-month-old S100GFP-tg control and S100GFP-tg Sod1–/– mice. (C–E) Quantification of nerve terminal area, nerve terminal perimeter, AChR area, percentage overlap between AChR area and nerve terminal area, and tSC number and tSC area. Muscle fiber imaging from S100GFP-tg (F) and S100GFP-tg Sod1–/– (G) mice, immunostained for S100B (green), Ki67 (magenta), AChR (red), and nuclei (blue). White arrows point to extrasynaptic nuclei positive for Ki67 but lacking endogenously expressed GFP and or S100B immunostaining, while the yellow arrow highlights a perisynaptic Ki67+GFP+ nucleus. (H) Enlarged view of 2 NMJs from the highlighted region in G, detailing the S100B, Ki67, and BTX stains. (I) The same NMJs from H, but focused on BTX and Ki67, revealing multiple Ki67+ nuclei in close proximity to the endplate. (J) Quantification of extrasynaptic and perisynaptic nuclei either singly labeled for Ki67 or double labeled for Ki67 and GFP. Values for all features across all NMJs analyzed are provided in Supplemental Figure 4. Open circles indicate average for each individual mouse of no fewer than 20 NMJs analyzed per muscle and bars represent means across animals ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by 2-tailed unpaired t test (C–E, vs. S100GFP-tg Sod1–/–) and (J). Scale bars: 25 μm (B) and 50 μm (F and G). In C–E, n = 4–5/group; J, n = 3/group.

Both the degree of overlap and synaptic area serve as indicators of the physical congruence of the pre- and postsynaptic components (21). These markers are robust histological indicators of nerve-to-muscle connections. In S100GFP-tg Sod1–/– mice, the overlap between the nerve terminal and AChRs was markedly decreased at 2 months compared with controls; however, by 3 months, this overlap aligned with that of the S100GFP-tg controls (Supplemental Figure 5). Reestablishment of control levels of overlap could be explained by presynaptic responses, compensatory postsynaptic changes, or a combination. The present analysis revealed an expansion in synaptic area between 2 and 3 months in S100GFP-tg Sod1–/– mice driven largely by an enlargement in nerve terminal area at 3 months (Supplemental Figure 4). Collectively, the restoration at 3 months in the degree of overlap coupled with the reduction in denervated NMJs (characterized by NMJs with less than 10% overlap) (Figure 1D) strongly support reinnervation.

Based on our findings that NMJs in 2-month-old Sod1–/– mice were characterized by higher numbers of tSCs than observed in the other groups, we aimed to verify proliferation of S100GFP+ cells at the NMJs in these mice. To directly test this, we stained muscle bundles with an anti-Ki67 antibody (Figure 3, F–I) and found numerous Ki67+ nuclei in Sod1–/– mice compared with age-matched WT mice. Moreover, the majority of Ki67+ nuclei originated from cells that were also S100GFP+. Ki67+ cells were observed both in perisynaptic (overlapping AChR staining) and extrasynaptic locations (within 50 μm of the nearest NMJ) (Figure 3J). These observations corroborate our earlier finding of a 3-fold increase in tSCs in Sod1–/– mice at 2 months of age when compared with WT controls (Figure 3E).

Our functional and morphological data strongly support the existence of a key regenerative window centered around 2 months of age in Sod1–/– mice that is ideal for studying the cellular and molecular changes in muscle SCs during NMJ remodeling and reinnervation. We also present compelling evidence that this remodeling process includes, at least in part, proliferation of SCs located in the muscle at the NMJ and a promotion of increased synaptic area through nerve terminal growth. Thus, we next aimed to define key intercellular signals that regulate SC dynamics during the remodeling of the NMJ.

Intercellular communication network analysis reveals an SPP1 signaling dynamic between mSCs and tSCs. Based on the unique location of tSCs at the NMJ, signaling to these cells is key to understanding the dynamics of this specialized environment. To explore how tSCs interact with other cellular components within the niche during NMJ remodeling, we conducted an intercellular communication network analysis employing CellChat (22). Utilizing our scRNA-Seq data from tSCs, mSCs, mesenchymal progenitors, macrophages, and smooth muscle cells we identified shared secreted signaling pathways that target tSCs in all experimental groups, including a combined denervation group encompassing both S100GFP-tg Sod1–/– and S100GFP-tg 7-dpi models (Figure 4A). TGF-β and SPP1 signaling emerged as the pathways with the highest normalized communication probabilities in the denervation group, implying important roles in NMJ remodeling in both spontaneous and injurious denervation conditions.

Intercellular communication suggests an SPP1 signaling dynamic between mSCs and tSCs. (A) Heatmap showing the common significantly enriched secretion signaling pathways across WT, Sod1–/–, 7 dpi, and Denervation (combined Sod1–/– + WT 7 dpi) targeting tSCs. Coloring of the heatmap is based on the normalized (z score) communication probabilities. (B) Circle plots displaying the SPP1 signaling network across cell clusters for S100GFP-tg, S100GFP-tg Sod1–/–, and S100GFP-tg (7 dpi) mice. The thickness of connecting lines represents the communication likelihood between paired cell clusters, with arrowheads demarcating communication directionality. (C) Representative immunofluorescence images of NMJs stained for SPP1 (red), GFP (green), AChRs (BTX, white), and nuclei (DAPI, blue). Scale bars: 20 μm.

Our finding of marked induction of SPP1 signaling in our denervation groups is compelling in light of a recent study finding that SPP1 promotes SC proliferation and survival, and its expression is notably upregulated in mSCs following human peripheral nerve injury (23). Exploring SPP1 signaling in our dataset, we identified a predominant expression of Spp1 in mSCs (Figure 4B). While SPP1 appears to communicate with several cell types in S100GFP-tg mice, a greater number of cell type receivers was inferred in Sod1–/– mice (5 vs. 7) (Figure 4B), which included the addition of proliferating and phagocytic (Spp1+) cells. Data from S100GFP-tg 7-dpi mice also revealed inferred SPP1 signaling originated from phagocytic (Spp1+), mSCs, and proliferating cell clusters, while the phagocytic (Spp1+) cell cluster in Sod1–/– mice was not predicted to communicate via SPP1 with other cells, likely owing to the limited number of these cells identified in these mice. The SPP1 signaling was primarily predicted to act through Cd44, and various integrin dimer combinations with Itgav, Itgb1, Itgb3, and Itgb5 (Supplemental Figure 6A).

To validate our CellChat findings, we performed targeted RT-qPCR to evaluate the expression levels of key genes within the SPP1 pathway in GTN muscles isolated from 2-month-old S100GFP-tg and S100GFP-tg Sod1–/– mice (Supplemental Figure 6B). We showed upregulation of Tgfb1, Tgfbr2, and Spp1 in Sod1–/– mice. While Cd44 levels remained unchanged, we detected a borderline significant elevation in Cd44v6 (P = 0.052), the specific receptor variant of CD44 known to bind SPP1. We next sought to confirm our scRNA-Seq and bioinformatic findings by pinpointing the protein localization of SPP1 in muscle fibers. Histological analysis using an anti-SPP1 antibody on fixed muscle fiber bundles revealed pronounced localization of SPP1 associated with GFP+ SCs near the NMJ, with intensified staining observed in Sod1–/– mice compared with controls (Figure 4C). In contrast to muscles from control and 2-month-old Sod1–/– mice, in S100GFP-tg 7-dpi muscles, SPP1 protein localization was observed in both GFP+ cells and GFP– cells. It seems likely that the GFP– cells represent the Spp1+ cell clusters in our scRNA-Seq dataset. Immunostaining also confirmed the localization of CD44 protein at the NMJ and within S100B+ SCs (Supplemental Figure 6C). Taken together, our findings illuminate SPP1 cell signaling as a pathway involving mSCs, phagocytic (SPP1+) cells, and tSCs, which could play a role in enhancing the tSC proliferation and survival important for successful NMJ remodeling and reinnervation.

SPP1 gene expression is markedly increased in muscles following nerve injury. To explore whether an acute recoverable nerve injury alters the gene expression dynamics of SPP1 signaling in skeletal muscle, we performed sciatic nerve crush procedures on 2-month-old mice (Figure 5A) and assessed transcript levels of denervation response pathways and SPP1 signaling in naive noninjured controls, and 7, 14, and 28 dpi. We observed a striking elevation in Spp1 expression and its receptors Cd44 and Itgav peaking at 7 dpi and subsequently reverting to baseline levels by 14 dpi (Figure 5A). This temporal trend closely parallels the known post-nerve-injury gene expression pattern of Chrna1 (9), suggesting an inverse relationship with muscle innervation. In addition, Tgfb1 and its receptor Tgfbr2, proposed to be upstream of SPP1 signaling, were also elevated at 7 dpi, with Tgfbr2 expression being reduced at 28 dpi compared with uninjured controls. Ngfr and Gdnf, which are associated with SC-mediated nerve regeneration, were also highly expressed at 7 dpi, while the cell proliferation gene Ccnd1 was elevated at 7 dpi and persisted at high levels out to 14 dpi. The pronounced Spp1 expression during the initial nerve regeneration phase and its subsequent normalization consistent with known muscle reinnervation milestones implicates the potential involvement of SPP1 in muscle reinnervation.

SPP1 signaling promotes tSC proliferation and muscle fiber reinnervation after nerve injury. (A) Following sciatic nerve crush injuries, gastrocnemius (GTN) muscles from C57BL/6J mice were collected at 0 (control), 7, 14, and 28 days postinjury (dpi). mRNA expression levels of denervation markers (Chrna1), components of SPP1 signaling (Spp1, Cd44, Itgav, Tgfb1, Tgfbr2), genes linked to SC-mediated nerve regeneration (Ngfr, Gdnf), and cell proliferation (Ccnd1) at each time point are presented. (B) Peroneal nerve injuries were induced, and tibialis anterior (TA) muscles were intramuscularly injected with either SPP1-nAb or IgG at time of injury and every 2 days thereafter. (C) Data are shown for force (mN) evoked by direct muscle stimulation, with nerve stimulation, and ratio of force elicited by nerve and direct muscle stimulation at 7 dpi for IgG- (gray) and SPP1-nAb–treated (blue) groups. (D) Representative immunofluorescence images of NMJs at 7 dpi stained for S100B (green), nuclei (DAPI, blue), AChRs (BTX, red), and NF/SV2 (cyan). (E) Quantification of NMJ nerve terminal area and perimeter, synaptic area, percentage of denervated (>10% overlap) NMJs, tSC number and area. Open circles indicate values for individual mice and bars represent the mean across animals ± SEM. Scale bars: 25 μm. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001 by 2-tailed unpaired t test (IgG vs. SPP1-nAb) or 1-way ANOVA with Dunnett’s test; comparison against the single control uninjured group (7, 14, 28 dpi vs. control) for multiple comparisons. In A, n = 4/group; C, n = 8/group; E, n = 4–5/group.

Inhibition of muscle SPP1 after acute nerve injury results in reduced muscle reinnervation and fewer tSCs. To determine the role of SPP1 signaling in muscle reinnervation and its potential mediation of tSC responses following nerve injury, we performed nerve crush injuries on control S100GFP-tg mice and administered intramuscular injections of either an SPP1-neutralizing antibody (SPP1-nAb) or a species-matched control IgG (Figure 5B). To expedite the onset of muscle reinnervation and thereby reduce the number of muscle injections, nerve crushes were performed near the nerve entry point to the TA muscle, with injections administered at the time of injury and every 2 days thereafter. At 7 dpi, recovery of functional neurotransmission was evaluated by comparing nerve- and muscle-evoked contractile responses. Direct muscle stimulation revealed no difference in maximal force production between IgG- and SPP1-nAb–treated groups; however, nerve-evoked muscle forces were 34% lower in the SPP1-nAb group (Figure 5C). Likewise, the ratio of nerve- to muscle-evoked force was 29% lower in muscles treated with SPP1-nAb compared with those receiving control IgG, suggesting that neutralization of intramuscular SPP1 impaired or delayed functional reinnervation.

We next assessed tSC morphology and NMJ structure in fixed muscle fiber bundles at 7 dpi (Figure 5E). Muscles treated with SPP1-nAb exhibited fewer tSCs per NMJ and reduced tSC area compared with IgG controls. Additionally, the SPP1-nAb group displayed marked decreases in nerve terminal area, perimeter, and total synaptic area. The proportion of denervated muscle fibers was substantially higher in SPP1-nAb–treated mice relative to IgG controls (71% vs. 34%), and NMJ synaptic area was correspondingly reduced. Together, these findings indicate that SPP1 signaling promotes effective muscle reinnervation and supports tSC expansion at the NMJ following nerve injury.

Single-cell profiling reveals that SPP1 neutralization stalls tSCs mid-trajectory and blunts late repair programs. To define how Spp1 shapes the tSC response in vivo during denervation and reinnervation, we performed scRNA-Seq on TA muscles 7 days following peroneal nerve crush with administration of either IgG or SPP1-nAb. The atlas contained expected immune and stromal populations and a discrete tSC cluster (Figure 6, A and B). Density overlays on the common UMAP showed that, consistent with our scRNA-Seq analyses following nerve crush injury in untreated muscles (Figure 1H), Spp1+ phagocytic/myeloid cells persisted in both IgG- and SPP1-nAb–treated muscles. Interestingly, this cell population was enriched in SPP1-nAb–treated muscles compared with IgG controls (Figure 6, C and D), consistent with compensatory recruitment/retention of Spp1-expressing cells when extracellular SPP1 is neutralized. Compared with the uninjected dataset in Figure 1, both IgG and SPP1-nAb cohorts showed a clearer representation of conventional dendritic cell type 1 (cDC1), conventional DC type 2 (cDC2), and mature regulatory DCs (mregDCs) (Figure 6, A and B). This likely reflects contextual differences between assays, including sciatic versus peroneal injury and the repeated intramuscular injections that can recruit or mature DCs.

scRNA-Seq suggests that SPP1 neutralization stalls tSC state transitions after nerve injury. (A) UMAP of whole-muscle scRNA-Seq at 7 dpi (IgG vs. SPP1-nAb combined), annotated by major cell types, including terminal Schwann cells (tSCs) and an Spp1+ phagocytic population. (B) Dot plot of canonical markers across annotated populations. (C and D) Density maps of IgG-treated (C) and SPP1-nAb–treated (D) cells projected onto the same manifold; dashed lines highlight the Spp1+ cells. (E) Reclustering of the tSC subset identifies 4 transcriptional states (1 to 4). (F) Pseudotime ordering of tSCs with inferred trajectory (black arrow), indicating progression toward a late state and return toward a homeostatic node. (G) Pseudotime distributions by treatment reveal a significant left shift in SPP1-nAb–treated cells (2-tailed unpaired Student’s t test; P < 0.001), consistent with a stall in state progression. (H) GO pathway heatmap (z scored within term) across tSC states. Early Cluster 1: axon/neuron guidance and PI3K/AKT; Cluster 3: ECM organization/adhesion and TGF-β; Cluster 4: glial maturation/ensheathment; Cluster 2: neural-crest/motility. Purple ticks indicate pathways unique to a single state. (I) Aggregated, scaled expression heatmap (genes × clusters) for state-defining markers with callout labels.

Focusing specifically on the tSCs, reclustering of tSCs yielded 4 transcriptional states (Figure 6E). Pseudotime analysis of the tSC clusters inferred a progression from an early injury/guidance state through an ECM/adhesion remodeling phase to a late glial maturation state, followed by a return toward a homeostatic node (Figure 6F). While tSCs from IgG-treated muscles traversed this loop, reaching the late maturation state and then turning back, cells from muscles in which SPP1 was neutralized accumulated earlier along the path, with a significant left-shift in pseudotime density (P < 0.001; Figure 6G), indicating stalled progression.

State-specific enrichment and aggregated expression (Figure 6, H and I) support a stepwise tSC program from Cluster 1 → 2 → 3 → 4. Cluster 1 showed neuron/axon guidance and PI3K/AKT signatures with transcripts including Reln, Met, Sema5a, Fgfr2, and Hmga2, consistent with tSCs directing reinnervation at NMJs and with growth/AKT modules linked to axon-glia interactions (24–29). Cluster 2 featured a motility/trophic crest–like program (Kitl, Cdh2, Gdnf, Prkg1, and Gfra3), in line with studies showing that trophic cues and cadherin signaling promote SC migration and neurite outgrowth during regeneration (30–32). Cluster 3 was enriched for ECM organization, adhesion, and TGF-β processes, marked by Postn, Itga5, Serpine1, Tgfb2, and Chl1; prior work shows that ECM-integrin pathways enable SC adhesion/migration after injury, periostin is induced in SCs and enhances remodeling, CHL1 guides regenerating motor axons, and TGF-β orchestrates injury-evoked SC states (17, 33–35). Cluster 4 expressed late glial/ensheathment genes (Mpz, Pmp22, Mbp, Ncmap, and Ptprz1), matching classic mSC markers (36–38).

To test the hypothesis that SPP1 neutralization affects migration and proliferation programs in tSCs, we performed a targeted pathway analysis on the IgG versus SPP1-nAb single-cell dataset. Within each tSC cluster, we computed differentially expressed genes, ran Gene Ontology (GO) Biological Process enrichment, and then selected terms annotated to migration or motility and to proliferation or cell cycle. We summarized enrichment as raw −log10(P)-adjusted values in cluster-by-term heatmaps (Supplemental Figure 7, A and B). Migration and motility terms were most prominent in Cluster 3, with additional neural-crest and glial migration terms in Cluster 2, consistent with sustained guidance and remodeling programs under SPP1 neutralization. In contrast, proliferation and cell cycle terms were concentrated in Cluster 4 and were minimal or absent in Cluster 2, and these proliferation signals were reduced in SPP1-nAb relative to IgG. Together with the left shift in pseudotime, this pattern indicates that lowering SPP1 activity in the muscle niche maintains tSCs in migration-rich states in Clusters 2–3, while blunting the late proliferative features that characterize Cluster 4 at the same postinjury time point.

Overall, these data suggest that SPP1 neutralization arrests tSC progression before the late glial program. Whereas IgG cells proceed from guidance to motility/trophic to ECM/adhesion and ultimately ensheathment, SPP1-nAb–treated cells accumulate in earlier states and exhibit diminished ECM/TGF-β and maturation modules, providing a mechanistic explanation for the reduced reinnervation observed in Figure 5.