Functional tendon regeneration is driven by regulatory T cells and IL-33 signaling

Abstract

Tendon injuries heal by scar, leading to poor function. To date, the role of immune cells remains underexplored. Using a neonatal mouse model of functional tendon healing compared to adult scar–mediated healing, we identified a regenerative immune profile that is associated with type 1 inflammation followed by rapid polarization to type 2, driven by macrophages and regulatory T cells (T_reg cells). Single-cell and bulk RNA sequencing also revealed neonatal T_reg cells with an immunomodulatory signature distinct from adult. Neonatal T_reg cell ablation resulted in a dysregulated immune response, failed tenocyte recruitment, and impaired regeneration. Adoptive transfer further confirmed the unique capacity of neonatal T_reg cells to rescue functional regeneration. We showed that neonatal T_reg cells mitigate interleukin-33 (IL-33) to enable tenocyte recruitment and structural restoration, and that adult IL-33 deletion improves functional healing. Collectively, these findings demonstrate that T_reg cells and IL-33 immune dysfunction are critical components of failed tendon healing and identify potential targets to drive tendon regeneration.

Neonatal T_reg cells enable tendon regeneration by generating a favorable immune environment.

INTRODUCTION

Tendons are dense connective tissues that transmit muscle forces to the skeleton to enable movement. This load-bearing function is enabled by a highly aligned, organized collagenous matrix that is synthesized and maintained by resident cells termed tenocytes. Despite the high prevalence of tendon injuries, current treatments (including physical rehabilitation and surgical repair) result in sustained functional deficits even at 12 months post-treatment (1). Although the limitations of functional adult tendon healing are well established, pediatric human tendon injuries heal with improved restoration in function and minimal complications (2). Since the mechanisms and targets to regenerate tendons remain unknown due to the paucity of tendon regeneration models, we applied the clinical observation that tendon regeneration occurs in pediatric patients to the neonatal mouse and established a model of functional tendon healing using the Achilles tendon (3–5). Outcomes associated with effective neonatal tendon healing included full restoration of gait function and mechanical properties, improved collagen structure, and distinctive cellular processes such as the recruitment and differentiation of resident tenocytes (5, 6). Although intrinsic differences in neonatal versus adult tenocytes may be one determinant limiting adult tendon healing, we were intrigued by several studies suggesting that a dysregulated adult immune response to tendon injury is also a critical factor (7–9). To date, neonatal immunity has been extensively studied in the context of infection and autoimmunity; however, the relationship between the unique neonatal immune environment and tissue regeneration is largely unknown. Generally, immunity during the neonatal window is highly tolerant, relatively quiescent, and biased toward a type 2 immune response in response to pathogens (10–12). Whether this neonatal environment confers a regenerative advantage has not been studied, outside limited contexts.

Although the immune response to wounding is driven by many immune cell types, most wound healing studies (including tendon) focus predominantly on macrophages, which are early responders and mediate debris clearance, inflammation, and granulation (13, 14). While there are many distinct macrophage subtypes, observations in wound healing have demonstrated a role for inflammatory (Ly6C^hi) macrophages in initiating the healing response and anti-inflammatory (Ly6C^lo) macrophages in resolving inflammation and promoting tissue repair (15, 16). In regenerative tissues (e.g., neonatal heart, neonatal tendon, adult muscle, and adult bone) and regenerative organisms (e.g., salamander and zebrafish), depletion of macrophages results in failed regeneration (17–21), while in adult tissues macrophages are indispensable for scar-mediated fibrotic healing (22, 23). Polarization of macrophages toward a Ly6C^lo anti-inflammatory profile appears to improve tissue healing (17, 24, 25); however, how this polarization is controlled has not been fully defined.

Compared to macrophages, the role of other immune cells in wound healing (including tendon) is much less well established. A few studies across different tissues suggest that T cell subtypes of the adaptive immune system contribute to regenerative or scar-mediated healing and may also regulate macrophage activation (12, 26–30). For tendon, in vitro studies suggest that T cells contribute to tendon disease by promoting a chronic cycle of inflammation (31, 32). However, the role of specific T cell subtypes in tendon healing and their role in mediating inflammation in the context of regeneration is largely unknown. The signaling pathways that regulate immune polarization in tendon healing have also not been fully elucidated. To date, most tendon studies focus on well-established type I [such as interleukin-1β (IL-1β), tumor necrosis factor–α (TNFα), and IL-6] and type II cytokines (IL-4 and IL-10) (32, 33). While loss-of-function studies are rarely carried out, deletion of IL-4 and IL-6 showed no difference in tendon function after injury at 6 and 12 weeks (34).

To address these questions, we applied multiple transcriptomic approaches to establish a unique neonatal immune response associated with effective tendon healing. We further show through loss-of-function and adoptive transfer studies that regulatory T cells (T_reg cells) derived from neonatal mice are necessary for tendon regeneration by creating an immune environment conducive to resident tenocyte recruitment. We also identify IL-33 signaling as a regulator of macrophage polarization and tenocyte recruitment and establish a potential role for neonatal T_reg cells in resolving IL-33 signaling to enable tenocyte recruitment and structural tendon regeneration. Finally, we show that IL-33 deletion improves adult functional tendon healing through attenuation of inflammatory macrophage polarization.

RESULTS

Neonatal tendon regeneration is defined by rapid polarization from type 1 to type 2 immune response

To determine whether the neonatal immune response to tendon injuries is distinct from the adult response, we profiled immune-related gene expression 3 days post–Achilles tendon injury (3 DPI) in neonatal [postnatal day 5 (P5)] and adult mice (4 to 6 months), using the NanoString PanCancer Immune Panel (770 immune targets). While we initially expected a dampened immune response in neonates compared to adults, neonates showed increased activation of immune genes (127 genes; Fig. 1A). Gene ontology (GO) analysis of differentially expressed genes (DEGs) identified type 1 immune signatures, including production of inflammatory cytokines IL-1β and TNF, and angiogenesis in injured adult tendon. In contrast, type 2 immune response and tissue remodeling signatures were identified in injured neonatal tendon (Fig. 1A). Type 2 immune genes included receptors for type 2 cytokines (Il4ra, Il10ra, I13ra1, and Il1rl1), which were uniquely up-regulated in injured neonatal tendon compared to adult (Fig. 1A and fig. S1). Since transcriptional changes can precede cellular response, cytokine protein arrays were performed to confirm neonatal and adult immune signature. In neonatal injured tendon, cytokine protein arrays showed increased abundance of both type 1 (IL-1β; fig. S2) and type 2 polarizing cytokines, IL-4 and CCL5, at 3 DPI compared to adult (Fig. 1B). Enzyme-linked immunosorbent assay (ELISA) analysis for the inflammatory cytokine IL-1β further showed sustained IL-1β protein in the adult injured tendon compared to the neonatal injured tendon at 14 DPI, which was at baseline control levels (Fig. 1C). These data suggest that neonatal tendons mount a robust type 1 inflammatory response at 0 to 3 DPI before rapidly transitioning to an anti-inflammatory type 2 response by 3 to 14 DPI.

Fig. 1. Rapid polarization toward an anti-inflammatory type 2 immune response in neonates after tendon injury.

(A) Left: Volcano plot comparing gene expression from NanoString analysis of neonatal versus adult injured tendon isolated at 3 DPI (n = 3 mice, P values adjusted for multiple comparisons). Right: (Top) GO analysis on DEGs significantly up-regulated (P_adj < 0.05) from neonatal injured and adult injured tendon. (Bottom) Expression levels of representative type II immune genes. One-way ANOVA, Tukey’s post hoc. (B) Significantly different representative type II cytokines quantified from cytokine proteome profiling blots of neonatal and adult injured tendon at 3 DPI (n = 3 mice, one-way ANOVA, Tukey’s post hoc). (C) ELISA quantification of IL-1β at 14 DPI. (D) Flow cytometry profiling of CD11b⁺ and Ly6C macrophages in neonatal and adult tendon before and after injury (n = 3 to 7, two-way ANOVA, Sidak’s post hoc). (E) qPCR analysis of inflammatory genes expressed in Ly6C^lo and Ly6C^hi macrophages isolated by FACS from injured neonatal tendon at 3 and 14 DPI (n = 3 to 4 mice, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Following adult tendon injury, there is robust recruitment of innate and adaptive immune cells (35). To define these cell types during neonatal tendon regeneration, we performed flow cytometry and profiled macrophage and T cell subpopulations (fig. S3). Analysis of the pan-macrophage marker CD11b showed no difference in total macrophage recruitment between neonates and adults at any time point (Fig. 1D). In both neonates and adults, there was a rapid increase in macrophage accumulation in the injured tendon peaking at 3 DPI (Fig. 1D). When we distinguished anti-inflammatory and proinflammatory macrophages using the marker Ly6C (gated on CD11b⁺, Gr1⁻), we observed polarization from proinflammatory Ly6C^hi macrophages in neonates at 3 DPI (which were greater in number than adults at the same stage) toward anti-inflammatory Ly6C^lo macrophages by 14 DPI. By contrast, adults polarized toward proinflammatory Ly6C^hi macrophages by 14 DPI (Fig. 1D). Neonatal tendons were enriched for Ly6C^hi macrophages at baseline (and in sham tendons at all time points), while adult tendons were enriched for Ly6C^lo macrophages (Fig. 1D and fig. S4). Macrophage polarization was specific to the injured tendon as differences in macrophages polarization were not observed between neonatal and adult spleen or sham-operated tendons (fig. S4), indicating that signals specific to the injured tendon direct polarization. Analysis of the inflammatory cytokines Il1β and Tnfα in sorted macrophages showed elevated expression at 3 DPI in Ly6C^hi macrophages (Fig. 1E), confirming the proinflammatory phenotype of the Ly6C^hi population consistent with literature (36). While Tnfα expression levels in Ly6C^hi cells reverted to Ly6C^lo levels at 14 DPI, Il1β remained markedly elevated in Ly6C^hi cells (Fig. 1E).

Transcriptional profiling by single and bulk RNA-seq reveals unique T_reg cell populations and anti-inflammatory signatures enriched in neonatal T_reg cells compared to adult T_reg cells

Although T cells play critical roles in the healing of tissues including muscle, lung, and nerve, their role in tendon healing is largely unexplored (28, 37). Therefore, to define T cell populations after injury, we performed single-cell RNA sequencing (scRNA-seq). Using positive selection, we isolated CD3⁺ T cells from injured adult and neonatal tendon at 14 DPI by fluorescence-activated cell sorting (FACS) and successfully sequenced 6741 adult and 1635 neonatal CD3⁺ T cells (Fig. 2A and fig. S5). This difference in CD3⁺ T cells sequenced may be due to the difference in tendon size between neonates and adults or due to higher baseline presence of these cells in adult tendons. Unsupervised clustering of the combined neonatal and adult tendon T cell datasets resulted in six clusters. Clusters were then annotated using an unsupervised reference-based approach and verified based on expression of cluster markers (Fig. 2B). As expected, most cells expressed Cd3, although there was a small macrophage population (cluster 5, genes: Csf1r and Cd68) that did not express Cd3 or other T cell markers, suggesting minor contamination during the sorting process. Within Cd3⁺ cells, we observed two populations of Cd4⁻/Cd8⁻ γδ T cells (clusters 0 and 3, genes: Trdc, Tcrg, and Trdc1), natural killer (NK)–like T cells (cluster 1, genes: Nkg7, Cd8b1, Klrd1, Klrc1, and Klrk1), Cd8⁺ effector T cells (cluster 4, genes: Cd8b1 and Cd8a), and Cd4⁺/Foxp3⁺ T_reg cells (cluster 2, genes: Foxp3, Il2ra, Ikzf2, and Klrg1). Additional genes defining each cluster are shown in fig. S6. The abundance of relative T cell populations varied between neonatal and adult tendons, particularly γδ T cells, which were enriched in the neonatal group (neo: 60%, adult: 37%) and NK-like T cells, which were enriched in the adult (neo: 14%, adult: 31%).

Fig. 2. Single-cell RNA and bulk RNA sequencing reveals a distinct role for neonatal T_reg cells in immune polarization.

(A) Single-cell RNA sequencing of FACS-sorted CD3⁺ T cells after neonatal and adult tendon injury at 14 DPI. (B) Violin plots of gene expression markers used to distinguish individual clusters. (C) Flow cytometry quantification of CD8⁺, CD4⁺, and T_reg cells (n = 3 to 5, two-way ANOVA, Sidak’s post hoc). (D) Pearson’s correlations of Ly6C^lo macrophages and T_reg cells in tendons after neonatal and adult tendon injury (n = 3 to 4 mice). (E) Subclustering of the Foxp3⁺ T_reg cell cluster and dot plot analysis of T_reg cell subcluster 1. For all quantifications, **P < 0.01 and ****P < 0.0001. (F) PCA comparing bulk RNA sequencing of T_reg cells isolated from neonatal spleen (NS), adult spleen (AS), neonatal tendon (NT), or adult tendon (AT) at 14 DPI (n = 3 mice). (G) Comparison plots of log₁₀-fold gene expression. Numbers indicate the number of up-regulated DEGs. Red lines indicate twofold difference in expression. (H) Volcano plot comparing gene expression of NT versus AT after subtracting a T_reg cell activation gene signature dataset (38). (I) Venn diagram of gene signatures identified by up-regulated DEGs. (J) GSEA for genes involved in “T_reg cell IL-4 Stimulation” are highly correlated in T_reg cells isolated from injured neonatal tendon (NT). NES, normalized enrichment score; FDR, false discovery rate. (K) Hierarchical clustering and GO analysis of interleukin DEGs between NT and AT, corresponding to Human Genome Organization (HUGO) gene group “Interleukins.”

Flow cytometry identified slightly higher baseline levels of both cytotoxic CD8⁺ and helper CD4⁺ T cells as well as minor differences in their recruitment after injury between neonates and adults; however, there were no differences by 14 DPI (Fig. 2C). There was gradual accumulation of CD4⁺, FOXP3⁺ T_reg cells in injured neonatal tendon with increased numbers by 14 DPI relative to the total population (Fig. 2C). Analysis of transverse cryosections from Foxp3^GFP mice at 14 DPI further confirmed localization of Foxp3^GFP+ T_reg cells at the tendon injury site (fig. S7). Since a small number of tendon-resident T_reg cells were observed at baseline, we next determined whether neonatal tendon T_reg cells were peripherally recruited or expanded from tendon-resident T_reg cells after injury. Inhibition of lymphocyte egress by FTY720 (daily injections for 2 weeks following P5 injury) resulted in reduced neonatal T_reg cell numbers at 14 DPI, indicating peripheral recruitment as the primary mode of T_reg cell accumulation in injured tendon (fig. S7). Tendon T_reg cell accumulation strongly correlated with anti-inflammatory Ly6C^lo macrophage polarization after neonatal injury, but not adult (Fig. 2D), suggesting a potential role for neonatal T_reg cells in regulating macrophage polarization.

Since CD4⁺, Foxp3⁺ T_reg cells were present in both adult and neonatal tendon following injury, we performed subclustering analysis of the T_reg cell population in our scRNA-seq dataset to further define T_reg cell populations that may be differentially recruited between neonate and adult. Unsupervised k-means clustering identified three clusters (Fig. 2E). Of these three, T_reg cell subcluster 1 had the highest expression of genes associated with tissue repair (Areg and Tgfb1), nonlymphoid homing (Ccr5, Ccrl2, and Ccr7), and type 2 inflammatory activation (Il1rl1, Il4ra, Gata3, and Stat3). Subcluster cluster 1 T_reg cells were largely composed of neonatal T_reg cells, while adult T_reg cells made up the majority of subcluster 0 (Fig. 2E). GO analysis of subcluster 1 T_reg cells revealed biologic processes related to cell cycle and regulation of NK cell activation (fig. S8). Cluster 0, which was predominantly composed of adult T_reg cells, was enriched for few significant GO terms, which may indicate decreased cellular response (fig. S8). Intermediate subcluster 2 T_reg cells demonstrated enrichment for GO terms associated with cell cycle activation, with no GO terms associated with regulation of inflammation. Despite subcluster 1 T_reg cells being predominantly composed of neonatal T_reg cells, substratification of T_reg cells by age revealed increased expression of subcluster 1 marker genes (Il1rl1, Stat3, and Ccr5) in adult subcluster 1 T_reg cells as well (Fig. 2E and fig. S8).

To better define the transcriptional differences in neonatal and adult T_reg cells and identify potential regenerative programs, we isolated Foxp3^GFP+ T_reg cells from injured adult and neonatal tendons at 14 DPI and performed bulk RNA-seq. Splenic T_reg cells isolated from the same mice served as age-matched controls. Principal components analysis (PCA) showed the neonatal tendon T_reg cell (NT) transcriptome was distinct compared to T_reg cells isolated from neonatal spleen (NS), adult spleen (AS), and adult tendon (AT) (Fig. 2F). Differential gene expression analysis identified 1929 significantly up-regulated genes (P_adj < 0.05) between NT and NS, and 513 significantly up-regulated genes between AT and AS, indicating that neonatal T_reg cells are transcriptionally more dissimilar to splenic T_reg cells in comparison to adult T_reg cells (Fig. 2G, fig. S9, and data S1). To determine whether differences in NT and AT signatures were due to differences in T_reg cell activation, we compared DEGs following subtraction of the canonical T_reg cell activation gene expression signature identified by Hill et al. (38). Differences in NT and AT T_reg cell signatures were not solely attributable to differences associated with an activated T_reg cell signature, as 985 genes remained differentially expressed between NT and AT (Fig. 2H) (38, 39). Therefore, we performed comparative analysis of gene expression signatures unique to NT and AT (relative to respective spleen samples to minimize generic age-specific differences), which revealed an NT-specific transcriptional signature of 1560 enriched genes (Fig. 2I and data S2). GO analysis of the NT gene signature identified GO terms associated with an anti-inflammatory myeloid response, consistent with anti-inflammatory macrophage polarization specific to neonatal tendon injury (Figs. 1D and 2I). We next performed gene set enrichment analysis (GSEA) on NT and AT DEGs to query immune signatures that distinguish adult and neonatal T_reg cells. GSEA revealed enrichment for genes associated with “T_reg cell IL-4 conversion” (40) in NT versus AT, indicating that neonatal tendon T_reg cells resemble a stimulated type 2 state (Fig. 2J). Hierarchical clustering and GO analysis of interleukin gene signatures (P_adj < 0.05) identified terms associated with type 2 and type 1 immune profiles in NT and AT, respectively (Fig. 2K). Collectively, these results show that neonatal tendon T_reg cells are transcriptionally distinct from adult T_reg cells and may be uniquely capable of polarizing the immune environment toward a type 2 response after injury.

Neonatal T_reg cells are required for anti-inflammatory macrophage polarization

Since our results suggested that neonatal T_reg cells may be uniquely anti-inflammatory, we next tested the requirement for neonatal T_reg cells in macrophage polarization. Neonatal T_reg cells were therefore selectively ablated using the Foxp3^DTR mutant, in which T_reg cells express the human diphtheria receptor (DTR) (41). T_reg cell depletion was carried out by delivery of diphtheria toxin (DT) for the first 6 days post-injury (every other day; Fig. 3A) (26, 41). This depletion regimen was selected based on previous literature showing that extended T_reg cell depletion results in autoimmunity and lethality (26, 41). To confirm effective depletion at 14 DPI with this regimen, we used flow cytometry for the Foxp3^GFP reporter in Foxp3^DTR mutants and showed nearly >90% depletion in the spleen and ~40% depletion in tendon, indicating that T_reg cell recruitment remained suppressed past 6 DPI (fig. S3). While total CD11b⁺ macrophage numbers were unchanged with T_reg cell depletion, neonatal macrophages were polarized toward inflammatory Ly6C^hi macrophages in the injured tendon, with no systemic effects observed in the spleen (Fig. 3B). Collectively, these data show that there is a tendon-specific requirement for T_reg cells in type 2 macrophage polarization during neonatal tendon regeneration.

Fig. 3. Neonatal T_reg cells are required for functional neonatal tendon healing.

(A) Schematic depicting T_reg cell depletion strategy. (B) Flow cytometry analysis and quantification of macrophages at 14 DPI after tendon injury in Foxp3^DTR mice treated with PBS or DT (n = 4 mice, two-tailed Student’s t test). (C) Hindlimb images following Achilles tendon injury or sham surgery in PBS- or DT-treated Foxp3^DTR neonates. Black arrowheads indicate site of tendon injury or sham operation. (D) Left: Gait schematic and waveform tracings from representative PBS- or DT-treated Foxp3^DTR neonatal mice before and after tendon injury. Quantification of gait parameter brake stride (n = 5 to 7 mice, two-tailed Student’s t test). (E) Representative force versus displacement tracings and stiffness quantification from uniaxial mechanical tensile testing of Achilles tendons from PBS- and DT-treated Foxp3^DTR neonates at 28 DPI (n = 5 to 6 mice, two-tailed Student’s t test). (F) Fluorescence microscopy and quantification of CHP binding in transverse Achilles tendon sections at 14 DPI from PBS- or DT-treated Foxp3^DTR neonatal mice (n = 3 to 4 mice, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, and ****P < 0.0001. For all scale bars, 100 μm. Panel (A) was created using icons from Biorender.com.

Neonatal T_reg cells are required for structural and functional tendon regeneration

Having established a role for neonatal T_reg cells in macrophage polarization following tendon injury, we next tested the requirement for T_reg cells in neonatal tendon regeneration. Following Achilles tendon transection, impaired healing in T_reg-ablated neonates was apparent as early as 7 DPI. While normal wound closure was observed following tenotomy in phosphate-buffered saline (PBS)–treated limbs and sham operation in DT-treated limbs, wounds exhibited delayed skin healing with large scabs apparent at 7 DPI compared to PBS littermate controls (Fig. 3C). Functional testing using gait analysis and direct tensile testing of injured tendon showed impaired functional recovery of brake stride and tensile stiffness (but not max force) in T_reg-ablated neonates at 28 DPI (Fig. 3, D and E, and fig. S10). To assess structural recovery, we performed collagen hybridizing peptide (CHP) staining to quantify the presence of denatured or damaged collagen (42). Increased CHP staining was observed in injured tendons of T_reg-ablated neonates at 14 DPI, indicating persistently damaged collagen structure (Fig. 3F). Together, these data revealed a necessary requirement for neonatal T_reg cells in structural and functional neonatal tendon regeneration.

To identify the cellular basis for impaired regeneration, we next determined the recruitment of Scleraxis-lineage (ScxLIN) tenocytes since our previous studies implicated these cells in functional tendon regeneration (5, 6). To determine whether T_reg cells are required for tenocyte recruitment, we incorporated the inducible Scx^CreERT2 allele into the Foxp3^DTR background, in combination with the Rosa26-LSL-tdTomato Ai14 Cre reporter (RosaT) and the ScxGFP tendon reporter. ScxLIN tenocytes were labeled by tamoxifen administration 3 days before injury (P2 and P3), and T_reg cell depletion initiated with DT administration on the day of injury (P5) for 6 days (Fig. 4A). Under uninjured conditions, recombination is observed in ~95% of the cells residing within the tendon fascicles and cells of the epitenon are not labeled (5, 6). At 14 DPI, the presence of ScxLIN⁺, ScxGFP⁺ tenocytes in the neo-tendon region was greatly reduced with DT treatment compared to PBS control (Fig. 4B). There was no significant difference in the recruitment of ScxLIN⁻, ScxGFP⁺ tenocytes, suggesting that T_reg cells are required for the recruitment of intrinsically derived but not extrinsically derived tenocytes. Analysis of α-smooth muscle actin–positive (αSMA⁺) myofibroblasts further revealed persistent αSMA⁺ myofibroblast cells in T_reg-ablated mice at 14 DPI, consistent with a nonregenerative, scar-forming phenotype (Fig. 4C). Finally, since we previously showed that transforming growth factor–β (TGFβ)–SMAD2/3 signaling in tenocytes is required for functional neonatal tendon healing (42, 43), we quantified phospho-SMAD2/3 immunostaining at 14 DPI and found decreased staining with neonatal T_reg cell depletion, suggesting impaired TGFβ signaling (Fig. 4D).

Fig. 4. Neonatal T_reg cells are required for Scx-lineage tenocyte recruitment.

(A) Schematic depicting tamoxifen labeling and T_reg cell depletion strategy. (B) Lineage tracing and quantification of ScxLIN tenocytes visualized by whole-mount fluorescence microscopy and transverse Achilles tendon sections from PBS- or DT-treated Foxp3^DTR; Scx^CreERT2; RosaT; ScxGFP neonatal mice at 14 DPI. Tamoxifen was administered at P2 and P3 (n = 3 mice, two-tailed Student’s t test). (C) Immunofluorescence images and quantification of αSMA⁺ and (D) pSMAD2/3⁺ and ScxGFP⁺ cells in transverse Achilles tendon sections from PBS- or DT-treated Foxp3^DTR neonatal mice at 14 DPI (n = 3 mice, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, and ****P < 0.0001. For all scale bars, 100 μm. Panel (A) was created using icons from BioRender.com.

Adoptive transfer of neonatal but not adult T_reg cells rescues functional tendon regeneration

Since transcriptional data suggested that neonatal T_reg cells may be uniquely regenerative compared to adult T_reg cells, we next determined whether neonatal T_reg cells were sufficient for functional healing. Since we previously found that T_reg cells active in tendon injury were peripherally recruited (fig. S7), neonatal and adult Foxp3^GFP+ T_reg cells were isolated from spleens after tendon injury at 3 DPI (fig. S11) and transferred into Rag2^−/− neonatal injured mice at 5 DPI (Fig. 5A). Rag2^−/− mice are deficient in T and B cells but retain innate immune cells including macrophages (43, 44). Analysis of transverse cryosections and flow cytometry from injured Rag2^−/− hindlimbs showed detectable Foxp3^GFP+ T_reg cells in the injured tendon at 14 DPI with T_reg cell transfer at 5 DPI, indicating successful T_reg cell engraftment (Fig. 5B and fig. S12). The prevalence of adoptive transferred T_reg cells detected at the site of tendon injury was comparable to autologous T_reg cells recruited after neonatal tendon injury at this time point (~1% of live cells). Delayed transfer at 5 DPI was motivated by initial studies showing that immediate transfer of neonatal T_reg cells at 0 DPI (time of injury) results in a chronic wound that never healed (fig. S13). These data suggest that suppression of initial inflammation is detrimental to healing. We therefore used delayed transfer at 5 DPI since macrophage polarization reverses between 3 and 7 DPI.

Fig. 5. Neonatal T_reg cells are uniquely regenerative and promote tendon healing.

(A) Experimental design schematic showing adoptive transfer of neonatal and adult mouse T_reg cells into neonatal immunodeficient Rag2^−/− mice. (B) Detection of adoptively transferred neonatal and adult Foxp3^GFP T_reg cells (red arrows) in injured Achilles tendon sections at 14 DPI. Scale bar, 100 μm. (C) (Left) Flow cytometry analysis and quantification of Ly6C macrophages in injured tendons at 14 DPI. (D) Gait analysis and tensile stiffness quantification of injured tendons at 28 DPI (n = 3 to 7 mice, one-way ANOVA, Tukey’s post hoc). (E) Experimental design schematic showing adoptive transfer of neonatal and adult T_reg cells into adult Rag2^−/− mice. Gait analysis and stiffness quantification of adult WT and Rag2^−/− mice at 28 DPI (n = 3 to 5 mice, one-way ANOVA, Tukey’s post hoc). (F) CHP staining for collagen damage at 14 DPI (n = 4 mice, one-way ANOVA, Tukey’s post hoc). For all quantifications, *P < 0.05, **P < 0.01, and ****P < 0.0001. C, control Rag2^−/− without transfer; NT, neonatal T_reg cell transfer; AT, adult T_reg cell transfer. Panels (A) and (E) were created with icons from BioRender.com.

In Rag2^−/− mice without T_reg cell adoptive transfer, we observed impaired immune polarization, resulting in sustained presence of proinflammatory Ly6C^hi macrophages at 14 DPI [~80% in Rag2^−/− versus ~15% in wild type (WT); Figs. 1C and 5C]. Impaired macrophage polarization resulted in poor functional recovery of Rag2^−/− neonates compared to WT neonates in terms of both gait and tendon mechanical properties at 28 DPI (Fig. 5D). Adoptive transfer of neonatal T_reg cells into neonatal Rag2^−/− hosts at 5 DPI effectively shifted macrophage polarization toward Ly6C^lo phenotype (~30%; Fig. 5C). While transfer of adult T_reg cells also improved macrophage polarization relative to control Rag2^−/− mice, Ly6C^hi cells were still elevated compared to neonatal T_reg cell transfer (Fig. 5C). Remarkably, adoptive transfer of neonatal but not adult T_reg cells was sufficient to rescue gait and improve tendon tensile stiffness (Fig. 5D). No significant differences were observed between uninjured contralateral hindlimbs (fig. S14).

We next determined whether neonatal T_reg cells could promote tendon healing in an adult host environment. In contrast to neonates, there was no difference in functional recovery between WT and Rag2^−/− adult mice, suggesting a minimal role for adult adaptive immune cells in adult tendon healing (Fig. 5E). Adoptive transfer of 3 DPI neonatal splenic T_reg cells into adult Rag2^−/− mice at 5 DPI significantly improved gait as well as direct tendon tensile stiffness at 28 DPI, while transfer of adult T_reg cells had no effect (Fig. 5E). No differences in gait and tendon tensile stiffness were observed between uninjured tendons (fig. S14). This improvement in functional healing was reflected in structural properties; analysis of collagen damage by CHP staining showed reduced damage with neonatal T_reg cell transfer, which was not observed with adult T_reg cell transfer (Fig. 5F). Together, these exciting results suggest that neonatal T_reg cells are uniquely capable of promoting functional tendon repair following tendon injury in both neonatal and adult contexts.

Neonatal T_reg cells resolve inflammatory IL-33 signaling to enable neonatal tenocyte recruitment

To determine the molecular basis for T_reg-mediated regeneration, we focused next on IL-33, a cytokine that was previously implicated in the recruitment and expansion of adult T_reg cells for muscle regeneration and in tendinopathy (45, 46). We were further motivated by our observation that Il1rl1, the gene encoding the IL-33 receptor, ST2, was up-regulated in neonatal tendon samples in the NanoString screen and was enriched in the T_reg cell subcluster 1 population in the scRNA-seq dataset (fig. S1). To determine the relevance of IL-33 signaling for regenerative tendon healing, we first carried out immunostaining for IL-33 and found elevated levels at 3 DPI in both neonatal and adult injured tendons (Fig. 6, A and B). Contrary to our expectations, IL-33 levels remained high in adults at 14 DPI compared to neonates, suggesting that low adult T_reg cell recruitment was not due to poor IL-33 production. IL-33 immunostaining also increased in DT-treated, T_reg-depleted Foxp3^DTR neonatal tendons at 14 DPI (Fig. 6C), indicating that neonatal T_reg cells function to resolve or mitigate IL-33 signaling to facilitate functional tendon regeneration. This was further supported by the reduction in collagen damage at 14 DPI following transfer of neonatal T_reg cells into neonatal Rag2^−/− hosts, with reduced IL-33⁺ cells and improved functional healing (fig. S15 and Fig. 5).

Fig. 6. Neonatal T_reg cells mitigate inflammatory IL-33 signaling associated with poor tendon healing.

(A) Schematic showing cryosection levels through the injured Achilles tendon gap at 3 DPI [level 1 (L1)] and through the neotendon at 14 DPI [level 2 (L2)]. (B) Transverse cryosection images and quantification of injured neonatal and adult WT Achilles tendons immunostained for IL-33 (n = 3 mice, one-way ANOVA with Tukey’s post hoc tests). (C) Transverse cryosection images and quantification of injured neonatal T_reg-depleted Achilles tendon immunostained for IL-33 (n = 3 mice, two-tailed Student’s t test). (D) Schematic showing neonatal T_reg cell depletion with and without Il33 deletion (Il33^−/−). (E) Structural analysis of collagen damage by CHP immunostaining at 14 DPI (n = 3 to 5 mice, two-tailed Student’s t test). (F) Flow cytometry analysis at 14 DPI for macrophage polarization (n = 4 to 5, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, and ***P < 0.001.

Since IL-33 appeared to be associated with poor tendon healing, we next tested whether neonatal T_reg cells may mitigate the detrimental effects of IL-33 signaling to facilitate functional tendon regeneration. Il33^−/− mice were generated by crossing Il33^f/f mice with CMV^Cre mice. We then incorporated the Foxp3^DTR allele to generate Il33^−/−; Foxp3^DTR mutants and depleted T_reg cells after neonatal tendon injury (Fig. 6D). At 14 DPI, we found that Il33 deletion rescued the structural collagen damage effects observed following tendon injury with T_reg cell depletion in neonatal mice. To determine whether polarization of macrophages toward an inflammatory Ly6C^hi state following T_reg cell depletion depends on IL-33, we also assessed macrophage polarization in T_reg-ablated Il33^−/−; Foxp3^DTR mutants. Inflammatory macrophage polarization was significantly reduced, implicating T_reg–IL-33 signaling in regulating macrophage polarity following injury (Fig. 6, E and F).

To test the hypothesis that excessive IL-33 signaling is detrimental for functional tendon regeneration, we administered recombinant IL-33 protein to neonates from 5 to 14 DPI to maintain high levels of IL-33 signaling during a phase when it is normally resolved (Fig. 7A). Analysis of collagen organization at 28 DPI using polarized light imaging of Picrosirius Red–stained tendon longitudinal sections showed significantly reduced alignment and increased disorganization in IL-33–treated neonates (Fig. 7B). Although we expected that the adverse effects of IL-33 treatment were due to enhanced polarization of inflammation macrophages, flow cytometry analysis at 14 DPI surprisingly showed reduced polarization of inflammatory Ly6C^hi macrophages with IL-33 treatment, contrary to our hypothesis (Fig. 7C). Although previous research in muscle showed that IL-33 induces expansion of regenerative T_reg cell (45, 47), we did not observe any difference in overall Foxp3⁺ T_reg cell numbers by flow cytometry. These results suggest that exogenous IL-33 administration induced a skewed type 2 response in macrophages that was detrimental for tendon regeneration.

Fig. 7. IL-33 treatment results in impaired neonatal tendon healing in vivo.

(A) Schematic showing neonatal IL-33 injections from 5 to 14 DPI. (B) Structural analysis of collagen organization in Picrosirius Red–stained tendons with PBS and IL-33 treatment (n = 3 mice, two-way ANOVA with Tukey’s post hoc test). (C) Flow cytometry analysis at 14 DPI for macrophage polarization and T_reg cell recruitment (n = 5, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, and ***P < 0.001.

Since macrophage polarization appeared independent of T_reg cell proliferation in response to exogenous IL-33, we assessed the presence of ST2⁺ macrophages after neonatal tendon injury by flow cytometry at 14 DPI and confirmed that a large population of macrophages expressed ST2 (Fig. 8A). We also observed increased ST2⁺ CD4⁺ T helper cells as well as ST2⁺ ScxGFP⁺ tenocytes with injury. To determine whether tenocytes could directly respond to IL-33, we used an in vitro scratch assay for neonatal tenocyte migration and found that IL-33 treatment suppressed migration to similar or greater levels as classic proinflammatory cytokines such as IL-1β and TNFα, respectively (Fig. 8B). Inhibition of the downstream signaling transducer of IL-33, MyD88, rescued the migration defects of IL-33 treatment (Fig. 8C). Since TGFβ signaling is key to tendon regeneration and cellular recruitment, we next tested whether excess IL-33 inhibited TGFβ in tenocytes (47). Immunostaining and Western blot for phospho-SMAD2/3 showed that IL-33 inhibited TGFβ signaling in tenocytes (Fig. 8D and fig. S16), suggesting a direct mechanism by which IL-33 may inhibit tenocyte migration to prevent functional neonatal tendon regeneration. To test this in vivo, we treated Scx^CreERT2; RosaT neonates with IL-33 (Fig. 8E) and found reduced recruitment of ScxLIN cells at 14 DPI with IL-33 treatment as well as reduced immunostaining of pSMAD2/3 (Fig. 8F). Finally, to determine the relevance of IL-33 signaling in the context of adult tendon healing, we injured adult Il33^−/− mice (Fig. 9A). Analysis of functional healing at 28 DPI showed improved tensile stiffness properties compared to WT injured adults, although not to baseline uninjured levels (Fig. 9B). Analysis of collagen alignment at 56 DPI also showed improved tendon collagen structure in Il33^−/− mice (Fig. 9C). Improved healing was likely partially mediated by a reduction in inflammatory macrophage polarization at 14 DPI (Fig. 9D). Overall, these data indicate that IL-33 signaling has multiple roles in TGFβ-mediated tenocyte recruitment and macrophage polarization, with potential as a therapeutic target to improve adult tendon healing.

Fig. 8. IL-33 inhibits tenocyte migration via MyD88 and TGFβ/Smad signaling in vitro and in vivo.

(A) Flow cytometry quantification of ST2⁺ macrophages (CD11b⁺), T helper cells (CD4⁺), and tenocytes (ScxGFP⁺) isolated from injured neonatal tendons at 14 DPI. (B) Scratch assay quantification of tenocyte migration with cytokine treatment at 12 hours (n = 4 independent samples, one-way ANOVA with Tukey’s post hoc test). (C) Scratch assay quantification of tenocyte migration with MyD88 inhibition in the presence of IL-33. (D) Quantification of tenocytes treated with IL-33 for 2 hours and immunostained for phosphorylated SMAD2/3 (n = 160 to 200 cells from four independent samples, two-tailed Student’s t test) (n = 4 to 5 independent samples, one-way ANOVA with Tukey’s post hoc test). (E) Scx^CreERT2; RosaT neonates were treated with recombinant IL-33. Immunostaining quantification of ScxLIN cells with IL-33 treatment at 14 DPI (n = 4 mice, two-tailed Student’s t test). (F) Nuclear immunostaining for pSMAD2/3 with IL-33 treatment at 14 DPI (n = 4 mice, two-tailed Student’s t test). For all quantifications, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 9. Adult tendon healing is improved with IL-33 deletion.

(A) Schematic showing adult tendon injury of Il33^−/− mice. (B) Stiffness was improved at 28 DPI, although not to uninjured levels (dotted line) (n = 8 to 10 mice, two-tailed Student’s t test). (C) Collagen structure was improved in adult Il33^-/- tendons at 56 DPI (n = 8 mice, two-tailed Student’s t-test). (D) Il33^−/− adult tendons showed suppression of inflammatory Ly6C^hi macrophages at 14 DPI (n = 4 mice, two-tailed Student’s t test). For all quantifications, *P < 0.05 and ****P < 0.0001.

DISCUSSION

We showed that neonatal tendon healing is characterized by acute inflammation that is rapidly resolved, while adult tendon healing is characterized by chronic inflammation. Using loss-of-function and adoptive transfer studies, we found that neonatal tendon T_reg cells exhibit a unique capacity to polarize tendon macrophages to resolve inflammation and promote functional healing via tenocyte recruitment (Fig. 10). While we initially posited that neonates would exhibit a dampened immune response to injury, we observed a robust initial inflammatory response. Adoptive T_reg cell transfer experiments suggested that this initial type 1 response is equally important for proper regeneration since neonatal T_reg cell transfer at the time of injury resulted in an exacerbated wound that never healed. This is supported by our previous studies showing that ablation of macrophages before injury also significantly impaired functional neonatal regeneration, likely by interfering with the initial inflammatory response (48). The rapid transition toward a type 2 response and Ly6C^lo macrophages is critical, however, since poor healing was consistently observed in cases of sustained type 1 polarization (such as in adult tendon healing, neonatal Rag2^−/−, or neonatal T_reg cell depletion). Similarly, a skewed type 2 response with hyperpolarization of macrophages toward Ly6C^lo from sustained administration of IL-33 in excess resulted in impaired neonatal tendon healing. Thus, a proper balance in type 1 and 2 immune polarization must be achieved. We found that tendon injury in neonatal Rag2^−/− mice (without adoptive transfers) resulted in impaired functional regeneration compared to WT with a sustained increase in IL-33. In adult mice, healing was not altered in Rag2^−/− adults, suggesting a unique role for neonatal adaptive immune cells in modulating IL-33 signaling during tendon repair.

Fig. 10. Conceptual schematic highlighting role of neonatal T_reg cells and IL-33 signaling in tendon healing.

(A) Neonatal T_reg cells mitigate the effects of IL-33 signaling and polarize Ly6C^lo anti-inflammatory macrophages to enable tenocyte recruitment and functional tendon regeneration. Excessive IL-33 signaling results in excessive type II polarization, which also inhibits tenocyte recruitment, leading to poor tendon healing. (B) Neonatal T_reg cells delivered into the adult environment polarize macrophages toward a Ly6C^lo anti-inflammatory profile to improve functional tendon healing, which is normally inhibited by IL-33 signaling.

Since T_reg cells have long been recognized as potent regulators of inflammation, other studies have investigated whether T_reg cells promote tissue regeneration (49). In zebrafish, T_reg cells are required for organ-specific regenerative programs in the spinal cord, heart, and retina (50). However, in nonregenerative mammalian tissues, the evidence is less clear. Among regeneration-competent adult tissues (e.g., muscle, lung, and skin), release of alarmins following injury stimulates secretion of the epidermal growth factor receptor (EGFR) ligand amphiregulin by T_reg cells that is critical for tissue repair (36, 50, 51). In contrast, adult T_reg cells are not sufficient to promote regeneration in nonregenerative tissues (e.g., adult heart and brain) despite recruitment and activation (37, 51, 52). While impaired regeneration may be due in part to limited stem/progenitor cell pools, recent studies suggest that inflammatory dysregulation also contributes to nonregenerative healing (53, 54). Compared to adult T_reg cells, neonatal T_reg cells are known to play a critical role in the maintenance of self-tolerance and have a distinct transcriptional signature from adult T_reg cells (55). This may be tissue specific, however; while neonatal cardiac T_reg cells are required for neonatal cardiac regeneration, transcriptional profiling of regenerative and nonregenerative cardiac T_reg cells showed surprising similarity (56). Transfer of adult T_reg cells capably rescued cardiac regeneration in the neonatal immunodeficient background (56). In contrast, we not only observed decreased abundance of adult tendon T_reg cells 14 days post-injury but also identified marked differences in transcriptional signatures, and adult transfer was not sufficient for rescue. Since we also found that most tendon T_reg cells after injury are peripherally recruited, these differences suggest that the local injury environment may somehow mediate T_reg cell response and phenotype.

Although our studies showed that neonatal but not adult T_reg cells rescued tendon regeneration and fully restored gait, biomechanical properties were improved but not completely restored, despite restoration of macrophage polarization. Since adoptive transfer studies were carried out in the Rag2^−/− background, this may implicate other T or B cells in tendon regeneration. Transfer of adult T_reg cells into the neonatal background also improved macrophage polarization to some extent, but no effect on functional healing was observed. These data suggest that there may be an anti-inflammatory threshold that must be achieved or that neonatal T_reg cells may have additional functions independent of immune polarization that regulate regeneration. In adult muscle, T_reg cells are a source of amphiregulin and exogenous application of amphiregulin stimulates satellite cell activation (56, 57). With T_reg cell depletion, we found that neonatal tenocyte recruitment was substantially impaired, resulting in poor structural and functional recovery. Notably, ScxGFP⁺ cells from a non-ScxLIN population were not affected by T_reg cell depletion; although the source(s) of these non-ScxLIN cells is unclear, multiple studies suggest that the epitenon harbors a population of stem cells with tenogenic potential (57, 58). If ScxGFP⁺/ScxLIN⁻ cells are epitenon derived, our data suggest that distinctive signaling mechanisms may regulate their recruitment and differentiation compared to the ScxLIN⁺ fascicle-residing tenocytes.

Although chronic IL-33 signaling in the neonate (via exogenous delivery of IL-33) resulted in impaired structural and functional tendon healing, we were surprised to observe excessive polarization of Ly6C^lo type 2 macrophages, contrary to our initial expectations that IL-33 would induce type 1 macrophage polarization. Since neonatal T_reg cell ablation resulted in Ly6C^hi type 1 polarization, this may suggest that high levels of IL-33 in the presence of T_reg cells enhanced the anti-inflammatory functions of neonatal T_reg cells while also directly suppressing neonatal tenocyte recruitment. There may also be a dosing effect, where levels of “chronic IL-33” in the absence of neonatal T_reg cells are normally lower than what were exogenously delivered, and there is a threshold at which an inflammatory versus anti-inflammatory macrophage response is normally induced. This is supported by the fact that Il33 deletion was able to rescue the detrimental effect of neonatal T_reg cell depletion. While we focused largely on the effect of IL-33 on macrophages, T_reg cells, and tenocytes here, IL-33 has the potential to signal to many cell populations and some of these results may be indirectly due to signaling in these other populations. The sources of IL-33 after injury also remain to be fully elucidated. In other systems, IL-33 can be expressed by fibroblasts, endothelial cells, epithelial cells, and a number of immune cells, including inflammatory macrophages (59–61). Whether these cell types are also implicated in the context of tendon healing will be determined in future studies.

We consistently observed that elevated IL-33 was associated with poor tendon healing. This is supported by evidence where tendon healing is improved with abrogation of IL-33 signaling, but contrary to previous studies in adult muscle that showed that IL-33 signaling is required for T_reg cell recruitment and muscle regeneration (5, 59). For tendon, IL-33 inhibited tenocyte migration with no effect on T_reg cell numbers, indicating critical cell-specific differences. We showed that neonatal T_reg cells mitigate the effects of IL-33 signaling to enable tenocyte migration in vitro and in vivo. We also showed that Il33 deletion can improve functional adult tendon healing, suggesting that IL-33 may be a therapeutic target for future studies. One limitation of these studies was that the full Il33^−/− knockout was used; it may be that there is a temporal aspect of IL-33 signaling, where early signaling may be beneficial, while chronic signaling is detrimental. Future studies will determine whether timed deletion of Il33 to bypass the initial response can further improve adult tendon healing. Future studies will also determine whether improved adult tendon healing with neonatal T_reg cell transfer is due to adult tenocyte recruitment or the activities of other cell types. It remains unclear whether neonatal T_reg cells may directly interact with resident tenocytes independent of their role in macrophage polarization. A few recent studies showed that fibroblast populations can serve as antigen-presenting cells under inflammatory stimuli. This intriguing possibility will be addressed in future studies. Finally, we carried out our functional assessments using hindlimb gait analysis and mechanical testing, but focused primarily on tensile stiffness as our mechanical readout. The focus on stiffness represents a limited view of the mechanical properties of the tendon, which does not capture other parameters associated with tendon function such as toughness, stress relaxation, or dynamic properties.

MATERIALS AND METHODS

Mice

Scx^GFP mice were used to identify tenocytes as previously described (5, 62). Foxp3^DTR/EGFP (stock no. 016958) and Rag2^−/− (stock no. 008449) mice were obtained from The Jackson Laboratory. All animal studies were conducted in accordance with Institutional Animal Care and Use Committee (IACUC) at Icahn School of Medicine at Mount Sinai (ISMMS) and Columbia University (IACUC-2014-0031 and AC-AABN0552), and all mice were housed in sterile barrier facilities operated by the Center for Comparative Medicine and Surgery at ISMMS or the Institute of Comparative Medicine at Columbia University. All studies were performed with at least three different mice, with specific sample sizes listed in figure legends. All experiments were performed in both sexes (by random allocation). Potential confounders (i.e., cage location and order of treatment) were not controlled. All sample treatments and conditions were anonymized during analyses and performed in a blinded fashion.

Scx^CreERT2; R26^LSL-tdTomato (Scx^CreERT2; RosaT) mice were crossed with Foxp3^DTR/EGFP to generate Scx^CreERT2; Foxp3^DTR/EGFP mice in a C57BL/6 background. To analyze IL-33 effects in vivo, we performed IL-33 injections on Scx^CreERT2; RosaT; ScxGFP mice. Lineage tracing was carried out by tamoxifen gavage at P2 and P3 before injury at P5 in neonatal mice. Tamoxifen (100 mg/ml) dissolved in ethanol was dissolved 1:10 in corn oil to achieve tamoxifen (10 mg/ml) in corn oil. Twenty-five microliters of tamoxifen (10 mg/ml) in corn oil was administered by pharyngeal gavage at P2 and P3, before injury at P5.

CMV^Cre mice (stock no. 006054) were crossed with Il33^f/f-eGFP mice (stock no. 030619) to obtain a global Il33^−/− knockout. These mice were then crossed with the Foxp3^DTR/EGFP mentioned above, and subsequent generations were intercrossed to obtain homozygous Il33^−/−; Foxp3^DTR/EGFP mice for T_reg cell depletion studies.

Tendon injury

Neonatal (P5) and adult (4 to 5 months old) mice were anesthetized by hypothermia or isoflurane inhalation, respectively. Achilles tendon injury was achieved by full transection without repair. In mouse hindlimbs, a small incision was made followed by complete tendon transection at the midsubstance. After injury, the skin was closed using Prolene sutures and animals were returned to full cage activity. Buprenorphine was administered intraperitoneally after injury for pain management. Sham injuries were performed in contralateral hindlimbs by skin incision to expose the Achilles tendon, followed by closure with Prolene sutures, without tendon transection.

FACS isolation and analysis of macrophages and T_reg cells

Tendon

Tendon digestion and single-cell suspension preparation were performed before FACS isolation and analysis. At the prespecified endpoint, tendons were dissected out and incubated in collagenase solution for 4 hours at 37°C on a rocker. Collagenase solution was prepared with collagenase I (5 mg/ml) (catalog no. LS004196, Worthington Biochemical) and collagenase IV (1 mg/ml) (catalog no. LS004188, Worthington Biochemical) in serum-free medium. Following tendon digestion, tendons were triturated, spin washed (500g for 5 min at 4°C), and resuspended in FACS buffer [2% fetal bovine serum (FBS) in 2 mM ethylenediaminetetraacetic acid (EDTA) in PBS supplemented with penicillin and streptomycin]. Resuspended cells were passed through a 70-μm sterile filter to force cell clumps into a single-cell suspension. Single-cell suspensions were then kept on ice and used for FACS.

Spleen

Spleens were harvested and macerated in FACS buffer using a sterile syringe plunger. Spleen suspensions were then passed through a 70-μm sterile filter to force cell clumps into a single-cell suspension. Splenic single-cell suspensions were kept on ice and used for FACS.

Macrophage polarization

Cells were stained against CD45 [BioLegend, catalog no. 10311, phycoerythrin (PE)/Cy7, 1:100], CD11b (BioLegend, catalog no. 101222, AF700, 1:100), Gr1 (BioLegend, catalog no. 108407, PE, 108407), and Ly6C [BioLegend, catalog no. 128008, fluorescein isothiocyanate (FITC), 1:100] for 30 min at 4°C in the dark. Gating on CD11b⁺, Gr1⁻ cells was to select for myeloid mononuclear cells (excluding granulocytes including eosinophils, basophils, and neutrophils) (37). DAPI (4′,6-diamidino-2-phenylindole) was use for live/dead cell identification.

CD4, CD8, and T_reg cell staining

Cells were stained against CD8 [BioLegend, catalog no. 100731, peridinin chlorophyll protein (PerCP), 1:100], CD4 (BioLegend, catalog no. 116005, PE, 1:100), and CD25 (BioLegend, catalog no. 102041, BUV510, 1:100). In mice without Foxp3^DTR/EGFP, cells were stained against FOXP3 (BioLegend, catalog no. 126407, A647, 1:100) using the FOXP3 Transcription Factor Staining Buffer Kit (eBioscience, catalog no. 00-5523-00). Otherwise, FOXP3 was identified using enhanced green fluorescent protein (EGFP). DAPI was added for live/dead cell identification.

Cells were analyzed and sorted using either a BD LSR II, BD Fortessa, or BD FACSAria III Cell Sorter at the ISMMS Flow Core Facility or the NovoCyte Quanteon and Sony MA900 Cell Sorter at the Columbia Stem Cell Initiative Flow Cytometry Core. Cell sorting was conducted using purity precision mode. FACS profiles were analyzed using FCS Express software.

Peripheral lymphocyte sequestration

Peripheral lymphocyte recruitment was restricted with treatment of FTY720 as per Dempsey (60). Briefly, FTY720 (Sigma-Aldrich, catalog no. SML0700) was reconstituted to 1 mg/ml and injected intraperitoneally at 25 mg/kg body weight every day until the prespecified endpoint following tendon injury.

T_reg cell depletion

T_reg cell ablation was carried out in Foxp3^DTR/EGFP mice, where administration of DT (Sigma-Aldrich, catalog no. 322326) results in apoptosis of Foxp3⁺ T_reg cells (63). DT was administered intraperitoneally at 6 ng/g body weight every other day for 6 days following tendon injury, starting on the day of injury.

T_reg cell adoptive transfer

For mouse T_reg cell transfer, spleens were harvested 3 DPI and processed to prepare a single-cell suspension. Untouched mouse CD4⁺ T cells were enriched using magnetic bead isolation as per the manufacturer’s protocol (Thermo Fisher Scientific, catalog no. 11416D). T_reg cells were then isolated via FACS as per above. If insufficient T_reg cells were collected from a single spleen, T_reg cell populations were pooled from several spleens to achieve sufficient cell quantity. Following isolation, mouse T_reg cells were injected at 5 days post-injury via intravenous tail injection at 10³ cells/3 g body weight in 150 or 500 μl of PBS for neonates and adults, respectively. Murine T_reg cells were transferred into immunodeficient Rag2^−/− mice. T_reg cell engraftment in injured tendons after adoptive transfer was assessed at 14 DPI. Briefly, injured tendons were digested and processed as described above, and stained with Sytox Blue (Thermo Fisher Scientific, catalog no. S11348), CD4 (BioLegend, catalog no. 116013), and GFP (BioLegend, catalog no. 338003). The anti-GFP antibody was added for straightforward detection of a linear FOXP3 GFP⁺ and anti-GFP⁺ population in the injured tendon to confirm engraftment of adoptively transferred T_reg cells into the Rag2^−/− adult mice that do not express GFP.

In vitro IL-33 experiments

Scratch assay studies were performed by culturing P7 tendon cells in 24-well plates until 90% confluence was reached, followed by serum starvation for 24 hours and mitomycin C treatment (5 μg/ml) for 2 hours. The medium was replaced with PBS, and scratch was created in the center of the well; subsequently, cytokine treatment was added in Dulbecco’s modified Eagle’s medium (DMEM) (TGFβ1, 10 ng/ml; TNFα, 10 ng/ml; IL-33, 20 ng/ml; IL-1β, 10 ng/ml) and migration was analyzed over the course of 8 hours. For co-inhibition analysis, a Myd88 inhibitor (Novus Biologicals, NBP2-29328-1mg) was used at 100 μM. For pSMAD2/3 immunostaining, tendon cells were grown at 90% confluency on eight-well chambered slides. After fixation with 4% paraformaldehyde (PFA) for 15 min at room temperature, immunostaining was completed as described in the immunofluorescence section below. For Western blotting, tendon cells were obtained from pooled tendon samples (tenocytes from three to four tendons per well) grown to 90% confluence on six-well plates. Cells were collected with 150 μl of Pierce radioimmunoprecipitation assay buffer (Thermo Fisher Scientific, catalog no. 8990) and scraping, and lysed with vortexing every 10 min for 30 min while on ice. Supernatants were collected after centrifuging 12,400 rpm for 15 min at 4°C. Proteins were quantified with a BCA kit (Thermo Fisher Scientific, catalog no. 23225) for SDS–polyacrylamide gel electrophoresis separation and transfer to nitrocellulose membranes (Bio-Rad, catalog no. 1704158). Primary [Cell Signaling Technology (CST), no. 3108S, 5174S] and secondary (CST, 7074S) antibody incubation was followed by WestFemto (Thermo Fisher Scientific, catalog no. 34096) substrate treatment for chemiluminescence visualization.

Recombinant IL-33 injections

After P5 complete Achilles transections without repair, we performed 10 ng/g weight intra-peritoneal injections with carrier-free recombinant mouse IL-33 (R&D Systems, 3626-ML-010/CF) or PBS injections. We completed injections for 10 consecutive days from 5 to 14 DPI.

RNA isolation, reverse transcription, and qRT-PCR

For RNA isolation from bulk tendon, tendons were dissected out and snap-frozen in liquid nitrogen (three tendons from three mice were combined per sample). Frozen tendons were then pulverized in a Geno/Grinder (SPEX Sample Prep) at 1500 rpm for 30 s at room temperature. Following pulverization, RNA isolation was carried out using TRIzol/chloroform extraction. For FACS-sorted cells, RNAs were extracted on pelleted cells using TRIzol/chloroform. After TRIzol/chloroform RNA isolation, all RNAs were quantified using NanoDrop2000. Reverse transcription was performed using SuperScript VILO (Thermo Fisher Scientific, catalog no. 11754050), and quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed using SYBR Green PCR Master Mix (Thermo Fisher Scientific, catalog no. 4309155). Mouse primer sequences used are the following: TNFα (forward: 5′-TACTGAACTTCGGGG-TGATTGGTCC-3′, reverse: 5′-CAGCCTTGTCCCTTGAAGAGA-ACC-3′), IL-1β (forward: 5′-AGTTGACGGACCCCAAAAGAT-3′, reverse: 5′-GTTGATGTGCTGCTGCGAGA-3′), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (forward: 5′-TGATGACAT-CAAGAAGGTGGTGAAG-3′, reverse: 5′-TCCTTGGAGGCCATGTAGGCCAT-3′). RNA samples were collected from three to five independent mice and ran in triplicate. The qPCR results reported on the y axis were obtained with the 2^−ΔΔCT method. Briefly, the median reference gene (GAPDH) expression was subtracted from each sample to obtain ΔC_T. Then, GAPDH-normalized triplicates were averaged, and reference sample (Ly6C^low) average expression was subtracted from ΔC_T to calculate ΔΔC_T. After, expression values normalized to both the reference gene and the reference sample were obtained by extrapolating 2^−ΔΔCT.

NanoString gene expression and analysis

RNA was isolated from bulk tendons from neonatal and adult mice as described above. Gene expression analysis was performed using the NanoString nCounter platform (NanoString Technologies). Gene counts were normalized using the NanoString nSolver software using background thresholding (threshold count value: 10). Counts were normalized to positive control and housekeeping genes. Normalized gene counts were exported, and gene expression was compared by adjusting for multiple comparisons using a Bonferroni correction. Significantly up-regulated genes (P_adj < 0.05) were identified, and GO analysis was performed using g:Profiler (64).

Cytokine proteomic profiling

Protein was isolated from dissected out bulk tendons at prespecified time points following injury. Each sample represents three tendons combined from three separate mice. Following dissection, tendons were snap-frozen in liquid nitrogen and pulverized using a Geno/Grinder (SPEX Sample Prep) at 1500 rpm for 30 s at room temperature. Pulverized tendons were resuspended in ice-cold tissue protein extraction reagent (Thermo Fisher Scientific, catalog no. 78510) supplemented with 1× HALT protease and phosphatase inhibitor cocktail and 1× EDT (Thermo Fisher Scientific, catalog no. 78446) for 5 min on ice. After tissue digestion, samples were spun at 10g for 5 min at 4°C, and supernatant was stored at −80°C until use. Bradford protein assays were performed, and 150 μg of protein was used from each sample for cytokine profiling. Cytokine proteomic profiling was performed as per the manufacturer’s instructions (R&D Systems, catalog no. ARY006). Densitometry analysis was performed using ImageJ to quantify protein abundance (65). Additional protein samples were used for IL-1β ELISA quantification, which was carried out according to the manufacturer’s directions (R&D Systems).

Single-cell library preparation and mRNA sequencing (scRNA-seq) of CD3⁺ T cells from injured adult and neonatal tendon with 10X Genomics

Twelve injured neonatal Achilles tendons and 4 injured adult tendons were harvested 14 DPI and digested as described elsewhere. CD3⁺ cells were isolated by FACS, after immunostaining with the following antibodies: PE CY7-CD45, PerCP eFluor 710-CD3, and Sytox Blue (viability). Less than 10,000 live T cells were obtained by gaiting on CD45⁺ CD3⁺ Sytox Blue⁻ cells. After, we resuspended these in 50 μl of buffer, which consisted of 10% FBS in PBS. These live, single-cell suspensions were then provided to the Columbia Sulzberger Genomics Center (New York, NY) for single-cell sequencing. Briefly, cells underwent routine quality control check by the core to ensure that enough viable cells were present, and then were input into a 10X Genomics Chromium device, and single-cell RNA libraries were prepared using the Chromium Single Cell 3′ v3 Reagent Kit. Sample libraries were pooled together at equimolar ratios and sequenced on an Illumina NovaSeq 6000 using 26–base pair (bp) paired-end reads (R1 barcodes). Approximately 1635 neonatal and 6741 adult cells were sequenced, with an average of 510,160 reads per cell for the neonatal sample and 49,230 reads per cell for the adult one.

Single-cell RNA-seq data analysis

All subsequent data processing was conducted using the cellranger suite (pipeline version 6.1.2, 10X Genomics). The obtained FASTQ reads were aligned to the reference genome using cellranger. Multi-mapping reads were ignored during quantification. Gene-cell matrices were extracted and further analyzed in R using Seurat (version 4.3.0).

To exclude low-quality cells for subsequent analysis, the following quality control filters were applied: Cells with >10% of transcripts mapping to the mitochondrial genome, <200 transcripts per cell (low quality), and >5000 transcripts per cell (likely doublets/multiplets) were excluded. To account for batch effect differences, we used the Seurat alignment method to integrate datasets from separate runs. Data were normalized using the “LogNormalize” method and scaled. PCA was subsequently performed on the integrated batch-corrected dataset, using the first 30 dimensions for unsupervised clusters and projection of the data in two-dimensional space for visualization. Unsupervised clustering was performed using a nearest neighbor approach (K = 10). Cell cluster identities were identified using a reference-based unsupervised approach SingleR, thereby avoiding initial labeling bias. Cluster-specific cell markers were identified using a one-versus-all differential gene expression approach using DESeq2 method with a log fold change of 0.25 with a Bonferroni adjusted P value of <0.05, which are detected in at least 10% of cells in each specific cluster. Final cluster identities were established after manual assessment of cluster markers with appraisal of the immune and tendon literature. UMAPs were generated using the same number of dimensions as with initial clustering. Other gene expression plots were generated using Seurat base functions and ggplot2 v3.4.1.

Differences in gene expression between clusters (cell types or subtypes) were statistically quantified through differential gene expression analyses. The MAST test was run in Seurat to identify DEGs between clusters of interest. A log fold change of 2 or 1.5 and a Bonferroni-adjusted P value of 0.05 were applied to determine DEGs.

Bulk RNA sequencing and analysis

RNA was isolated using the Arcturus PicoPure RNA Isolation Kit (Thermo Fisher Scientific, catalog no. KIT0204). RNA concentrations were measured with a NanoDrop spectrophotometer (Thermo Fisher Scientific), and quality was assessed with an Agilent TapeStation with RNA integrity number (RIN) > 8 for all samples. RNA amplification, library preparation, and sequencing were performed by Genewiz. Samples were sequenced on the Illumina HiSeq using a 2 × 150 bp sequencing setting at Genewiz. Sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36 (66). The trimmed reads were mapped to the Mus musculus GRCm38 reference genome (ENSEMBLE) using STAR aligner v.2.5.2b (67). Unique gene counts were calculated using featureCounts from Subread package v.1.5.2 (68). After quantification of gene hit counts, differential gene expression analysis was performed using DESeq2 (69). DEGs were identified with Benjamini-Hochberg correction for multiple comparisons with a significance of adjusted P value (P_adj) < 0.05.

PCA and hierarchical clustering were performed on all DEGs using sci-kit learn v.0.24.1. T_reg cell gene signatures were defined by identifying DEGs with P_adj < 0.05 and log₂ fold change >1 compared to respective populations (e.g., NT versus NS, AT versus AS, and NS versus AS). Intersectionality analysis was then performed to identify shared and unique DEGs that comprise distinct T_reg cell signatures. GO analysis was performed on DEGs using g:Profiler (64). RNA-seq data are deposited in Gene Expression Omnibus GSE173770.

Histology and immunofluorescence

Immunostaining

For immunofluorescence histology, samples were fixed in 4% PFA overnight at 4°C and then were decalcified in 0.5 M EDT, replaced every 3 days, until bones were pliable. Limbs were embedded in optimal cutting temperature medium (OCT) and frozen and stored at −80°C until use. Alternating transverse cryosections were collected at 12-μm thickness, along the length of the tendon to capture the tendon bony insertion to muscle origin. Immunostaining against GFP (Life Technologies, catalog no. A11122, 1:200), αSMA (Sigma-Aldrich, catalog no. A5228, 1:100), or IL-33 (R&D Systems AF3626-SP, 10 μg/ml) was performed by incubating overnight at 4°C, following cellular permeabilization with 0.1% Triton X-100 for 5 min at room temperature and blocking for 1 hour. Secondary antibody staining was performed with either Cy3 or Cy5 (Jackson ImmunoResearch), and slides were counterstained with DAPI to visualize nuclei. pSMAD2/3 immunostaining of cryosections was performed with tyramide signal amplification (TSA) Cy3 kit (PerkinElmer, NEL704A001KT), following the manufacturer’s immunohistochemistry instructions. Briefly, sections were fixed, washed, treated with hydrogen peroxide, permeabilized, and stained with pSMAD2/3 antibody (1:1000, CST, catalog no. 8828S) overnight at 4°C, followed by four wash steps before and after anti-rabbit horseradish peroxidase (1:1000) treatment for final Cy3 signal amplification with the kit’s reagents. For quantification, immunofluorescence microscopy (Zeiss Apotome) was performed on serial transverse sections along the tendon sample. Quantifications were performed across transverse sections spanning the injured portion of the tendon using ImageJ and then averaged to achieve an individual result per tendon.

Collagen structure damage

Denatured or damaged collagen was labeled using a fluorescently conjugated CHP as per the manufacturer’s instructions (3Helix, catalog no. RED60 or RED600).

Polarized light imaging of collagen structure

For Picrosirius Red staining, limbs were fixed in zinc formalin fixative (z-fix) (Sigma-Aldrich, catalog no. Z2902) overnight at 4°C. Following fixation, tendons were dissected out, dehydrated, and embedded in methacrylate monomer. Longitudinal sections (6 μm) were collected and stained with Picrosirius Red as per the manufacturer’s instructions (Abcam, catalog no. ab150681). Picosirius red birefringence was imaged using a polarized light filter to visualize collagen alignment and integrity. Images were taken with specimen angling that yielded the highest birefringence. For each biological variable, birefringent fiber orientation at the injury site was quantified from three different sections, from which three images were obtained, and alignment statistics were averaged. Alignment parameters were obtained using the FiberFit software, previously described elsewhere (70).

Whole-mount fluorescence imaging

Limbs were first fixed in 4% PFA and incubated at 4°C overnight. For whole-mount fluorescence imaging, skin was dissected away and tendons were visualized using a Leica M165FC stereomicroscope with filters for fluorescence.

Functional gait analysis

To assess gait, right hindlimb Achilles tendons were injured in mice, with the uninjured left contralateral limb serving as a control (68). DigiGait Imaging System (Mouse Specifics Inc. Quincy MA) was used to analyze mouse gait following injury. Without pretraining, neonatal and adult mice were gaited at 10 and 50 cm/s, respectively. Mice were gaited for 3 to 5 s on a transparent treadmill, with paw contact and position captured using a high-speed camera. Mice were gaited in both directions, and the average of both results was used for analysis. Footage was analyzed using the DigiGait Analysis Software (DigiGait v12.4). To normalize for age or sex differences, gait parameters (brake) were measured and normalized to stride length to calculate %brake stride.

Biomechanical tendon testing

For biomechanical testing, limbs were harvested and stored at −25°C until use. On the day of biomechanical testing, limbs were thawed, and Achilles tendons were dissected out carefully with the bony calcaneal tuberosity. Mechanical tensile testing of mouse Achilles tendons was performed using custom grips to clamp the calcaneal tuberosity and Achilles tendon origin (71). The tendons were then immersed in a PBS bath at room temperature and preloaded to 0.05 N for ~1 min followed by a ramp to failure at 1% strain/second based on initial gauge length. Force and displacement were recorded using an Instron 8872 Universal Testing System (Instron). To quantify tendon midsubstance stiffness, the force-displacement curve was plotted on Microsoft Excel for Mac (version 16.92), and the slope of the linear region was calculated (∆y/∆x).

Statistics

All quantitative data are presented as means ± SD, with each point representing independent biological replicates. For comparison of two groups, two-tailed unpaired Student’s t tests were performed. Analysis of variance (ANOVA) with Tukey’s honestly significant difference (HSD) correction for multiple hypothesis testing was performed for comparison of >2 groups. All statistical analysis was performed using GraphPad Prism v9.0.0. Sample sizes were determined based on our previous data, and no data were excluded.

Supplementary Materials

The PDF file includes:

Figs. S1 to S16

Legends for data S1 and S2

Other Supplementary Material for this manuscript includes the following:

Data S1 and S2