Endogenous CCN family member WISP1 inhibits trauma-induced heterotopic ossification.

Heterotopic ossification (HO) is defined as abnormal differentiation of local stromal cells of mesenchymal origin, resulting in pathologic cartilage and bone matrix deposition. Cyr61, CTGF, Nov (CCN) family members are matricellular proteins that have diverse regulatory functions on cell proliferation and differentiation, including the regulation of chondrogenesis. However, little is known regarding CCN family member expression or function in HO. Here, a combination of bulk and single-cell RNA sequencing defined the dynamic temporospatial pattern of CCN family member induction within a mouse model of trauma-induced HO. Among CCN family proteins, Wisp1 (also known as Ccn4 ) was most upregulated during the evolution of HO, and Wisp1 expression corresponded with chondrogenic gene profile. Immunohistochemistry confirmed WISP1 expression across traumatic and genetic HO mouse models as well as in human HO samples. Transgenic Wisp1 LacZ/LacZ knockin animals showed an increase in endochondral ossification in HO after trauma. Finally, the transcriptome of Wisp1 -null tenocytes revealed enrichment in signaling pathways, such as the STAT3 and PCP signaling pathways, that may explain increased HO in the context of Wisp1 deficiency. In sum, CCN family members, and in particular Wisp1 , are spatiotemporally associated with and negatively regulate trauma-induced HO formation. In our in vivo results, Wisp1 -null mice showed a conspicuous increase in cartilage after injury, while bone formation was increased to a lesser degree. Further study must be performed to understand the cell- and context-dependent effects of Wisp1 in other models of endochondral ossification. analysis steps were performed using Seurat 2.3.0. Cells with few-er than 500 genes per cell, with more than 60,000 UMIs, or expressing a fraction of mitochondrial UMIs higher than 0.1 were filtered for quality control. The number of UMIs (UMI fraction related to mitochondrial genes >0.1) and the number of genes, mitochondrial genes, and cell cycle genes (S and G2M scores by CellCycleScoring Seurat function) were regressed before further analysis. The downstream analysis was performed with Seurat 2.3.0. The time point sets were merged with canonical correlation analysis based on the intersection of the variable genes for time point set. Unsupervised clustering (Lou-vain algorithm) was used to identify cell populations.


Introduction
Heterotopic ossification (HO) is the abnormal collection of bone and cartilage matrix in soft tissues, incited by traumatic insult (1). In its severe manifestations, HO can be disabling, owing to joint contracture and chronic pain (2). Although HO can form in diverse anatomic sites, including tendon, ligament, and muscle, it consistently derives from tissue-resident connective tissue mesenchymal progenitor cells (1,(3)(4)(5)(6). Here, predominantly local stromal/fibroblastic cells of mesenchymal origin within connective tissue (3,4,(6)(7)(8) undergo abnormal osteochondral differentiation producing pathologic extracellular matrix of bone and cartilage (1,9). Matricellular proteins are one link between the extracellular matrix and cell fate determination decisions. The Cyr61, CTGF, Nov (CCN) family consists of 6 structurally related matricellular proteins with cysteine-rich regulatory proteins (10) that bind to a plethora of extracellular ligands, in turn regulating cell proliferation and differentiation. Some CCN family members regulate cartilage and bone formation, including, for example, CTGF (also known as CCN2) (11)(12)(13) and WISP1 (also known as CCN4) (14,15). In particular, WISP1 is expressed in perichondral and preosteoblastic cells during mammalian skeletal development, and this finding is recapitulated in mouse fracture repair (16). Although WISP1 exerts clear positive regulation of osteogenic differentiation (15,17), its role in cartilage differentiation is more context dependent. For example, WISP1 overexpression antagonizes BMP2-or growth differentiation factor 5-induced (GDF5-induced) chondrogenesis (16), while, in other observations, WISP1 enhances TGF-β3-induced chondrogenesis (18). In addition, WISP1 appears to play an aggravating role in cartilage degradation in osteoarthritis, potentially through an increase in expression of matrix metalloproteinases (19,20). Despite their roles in modulating osteochondral cell proliferation and differentiation, virtually nothing is known regarding CCN family members, including WISP1, in the aberrant cell fate determination seen in HO.
Heterotopic ossification (HO) is defined as abnormal differentiation of local stromal cells of mesenchymal origin, resulting in pathologic cartilage and bone matrix deposition. Cyr61, CTGF, Nov (CCN) family members are matricellular proteins that have diverse regulatory functions on cell proliferation and differentiation, including the regulation of chondrogenesis. However, little is known regarding CCN family member expression or function in HO. Here, a combination of bulk and single-cell RNA sequencing defined the dynamic temporospatial pattern of CCN family member induction within a mouse model of trauma-induced HO. Among CCN family proteins, Wisp1 (also known as Ccn4) was most upregulated during the evolution of HO, and Wisp1 expression corresponded with chondrogenic gene profile. Immunohistochemistry confirmed WISP1 expression across traumatic and genetic HO mouse models as well as in human HO samples. Transgenic Wisp1 LacZ/LacZ knockin animals showed an increase in endochondral ossification in HO after trauma. Finally, the transcriptome of Wisp1-null tenocytes revealed enrichment in signaling pathways, such as the STAT3 and PCP signaling pathways, that may explain increased HO in the context of Wisp1 deficiency. In sum, CCN family members, and in particular Wisp1, are spatiotemporally associated with and negatively regulate trauma-induced HO formation.
In the present study, we hypothesized that CCN family members would be dynamically regulated during the evolution of HO and would modulate disease progression. We first surveyed CCN family member expression within a mouse model of extremity trauma-induced HO. Bulk and single-cell RNA (scRNA) sequencing demonstrated an enrichment in most CCN family members during HO formation, and, in particular, Wisp1 was most upregulated. scRNA sequencing, in combination with immunohistochemistry in mouse and human tissues, demonstrated that mesenchymal progenitor cells and their osteochondral progeny were the major source of WISP1 during HO formation. Transgenic Wisp1 LacZ/LacZ -knockin mice demonstrated more exuberant endochondral ossification during HO, and downstream pathway analysis identified putative signaling changes associated with increased endochondral ossification within Wisp1-deficient cells.

Results
Wisp1 is enriched in traumatic HO in vivo. CCN family members were first examined in a model of trauma-induced HO. Complete Achilles tenotomy, coupled with cutaneous dorsal burn injury, results in a well-characterized incidence of HO at the tenotomy site, replicating human disease (21)(22)(23)(24)(25). First, bulk RNA sequencing was performed at the tenotomy site and compared with the contralateral tendon site of the same animal (serving as an internal control) at 3 weeks after injury ( Figure 1). A 3-week time point was chosen for analysis as most representative of the cartilaginous phase of HO in this model (24). Five of six CCN family members showed an increase in the HO site, as compared with the contralateral uninjured limb ( Figure 1A). Of these CCN family members, Wisp1 showed the highest increase (2.8694 fold change [FC], +6 SD). Wisp2 (also known as Ccn5) was also increased (1.8642 FC, +3 SD), while Ctgf (0.5615 FC, +0.5 SD), Cyr61 (0.2931 FC, +0.5 SD), and Nov (0.5231 FC, +0.5 SD) were marginally increased. Wisp3 (also known as Ccn6) showed a modest reduction (-0.2347 FC, -0.5 SD).
scRNA sequencing identifies Wisp1 enrichment in chondrocyte-like mesenchymal cells within HO. Expression of CCN family members, including Wisp1, was next examined in detail across the dynamic time course of HO formation by scRNA sequencing (Figure 2 and Supplemental Figures 1 and 2; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.135432DS1). Nine cell clusters were defined (Figure 2A), which broadly included 3 mesodermal populations (mesenchymal, pericyte/vascular smooth muscle, and neuromuscular junction cells), 2 endothelial populations (endothelial and endothelial lymphatic cells), and 4 inflammatory cell populations (macrophage, B cell, T cell, and neutrophil). A list of markers used to define each cluster is shown in Supplemental Table 1. Single-cell sequencing at the site of HO formation was performed and analyzed before injury (day 0) and at the early time points of 3, 7, and 21 days after injury in relation to uninjured control (Figure 2, B and C). Overall, Wisp1 was principally expressed in the mesenchymal cell cluster, typified by expression of platelet-derived growth factor receptor α (Pdgfra), paired related homeobox 1 (Prrx1), and leptin receptor (Lepr) (Figure 2, B and C). Within mesenchymal cells, Wisp1 showed mean normalized expression of 0.46, and it was expressed in 52% of cells. In contrast, in nonmesenchymal cells (8 other cell clusters combined), the average normalized expression of Wisp1 was 0.02, and it was expressed in 3% of cells. Changes in Wisp1 expression over the time course of HO development was next assessed ( Figure  2, D and E). As shown by t-distributed stochastic neighbor embedding (t-SNE) plots ( Figure 2D) or violin plots ( Figure 2E), Wisp1 gene expression was inconspicuous at baseline and overtime increased in the mesenchymal cluster, peaking at day 7, and remaining elevated at day 21 after injury when cartilage is visible histologically.
For comparison, other CCN family members were also assessed by scRNA sequencing (Supplemental Figure 1). With the exception of Wisp3, all CCN family members were detected within the mesenchymal cluster of the HO site, ranging from 35% (Nov) to 72% (Cyr61) of the mesenchymal cells. Unlike Wisp1, other CCN family members, such as Ctgf and Cyr61, were identified across multiple cell clusters (Supplemental Figure 1).
Next, Wisp1 expression among 9 cells clusters (markers used to identify different clusters are shown in Supplemental Table 1) in HO was compared with cartilage-associated markers, including Sox9 ( Figure  2, F-I) and Acan and Col2a1 (Supplemental Figure 2). Sox9, the master chondrogenic transcription factor, showed an expression pattern that closely mirrored that of Wisp1 (  Figure 2). These findings were replicated when the mesenchymal cluster was isolated and subclustering was performed (Supplemental Figure 3). Four subclusters were identified within the mesenchymal cluster (Supplemental Figure  3A). Overlap between Wisp1 expression and cartilage-associated genes was again observed. Thus, among CCN family members, Wisp1 demonstrated a prominent increase in expression after trauma, particularly in mesenchymal cells with a chondrocyte-like gene profile.
WISP1 immunolocalization in mouse models of HO. To validate these transcriptional analyses, WISP1 expression was next studied using immunohistochemistry in mouse models of HO ( Figure 3). First, WISP1 immunohistochemical staining was performed at sequential time points after trauma, which highlighted the fibroproliferative phase (3 days, Figure 3, A and B), cartilaginous phase (3 weeks, Figure 3, C and D), and osseous phase of HO (9 weeks, Figure 3, E and F). WISP1 immunoreactivity was seen within fibroblastic mesenchymal cells of the injury site at the 3-day time point and was weak to moderate in intensity (asterisk, Figure 3B). In the cartilaginous phase, stronger and more widespread WISP1 immunoreactivity was present within chondrocytes ( Figure 3D). In the osseous phase, bone-lining cells and some osteocytes showed some WISP1 immunoreactivity, which was weak to intermediate in intensity (Figure 3F). Multinucleated osteoclasts also showed WISP1 immunostaining. A similar pattern was observed using Wisp1-LacZ reporter, in which X-gal staining was predominantly seen in bone lining of the HO site at 9 weeks after injury (Supplemental Figure 4).
Finally, WISP1 protein expression was examined within an fibrodysplasia ossificans progressiva-like (FOPlike) mouse model (Figure 3, G and H). In this model, the mutant Acvr1 allele (Q207D mutation) is driven by expression of Nfatc1-Cre (26)(27)(28)(29). Resulting pups carrying both transgenes (Nfatc1-Cre; ca-ACVR1 fl/WT ) developed widespread HO by postnatal days 4-5. On H&E sections, the histologic appearance was similar to that of the osseous phase of trauma-induced HO. Likewise, WISP1 immunoreactivity was similar to traumatic HO and predominantly found in bone-lining cells. In sum, WISP1 was robustly expressed across multiple cells types, giving rise to HO, including chondrocytes and bone-lining osteoblasts.
WISP1 immunolocalization among human nongenetic HO. Next, to translate our findings to human HO, the surgical pathology case files of 12 human HO specimens were examined for WISP1 expression. Bulk RNA sequencing of Achilles tendon tenotomy site. Tissue was collected from the soft tissue around the injury site from the posterior compartment between the muscular origin and calcaneal insertion of Achilles tendon. Corresponding soft tissue, including intact Achilles tendon, was also harvested from the uninjured contralateral hind limb. In each image, data are presented in volcano plots and logarithmic fold change is shown on the x axis (log 2 ). The y axis depicts P value (log 10 ). Data were derived from n = 3 pooled samples per group, with expression examined at 3 weeks after injury. (A) CCN family members, including Ctgf, Cyr61, Nov, Wisp1, Wisp2, and Wisp3. (B) Selective cartilage-associated genes. (C) Selective bone-associated genes. All analyses were performed on the same data set. Light red and dark red dots indicate genes with a >2 SD increase and a >3 SD increase in comparison with uninjured control, respectively. Light blue and dark blue dots indicate genes with a >2 SD decrease and a >3 SD decrease in comparison with uninjured control, respectively.
All clinical cases were examples of sporadic (nongenetic) HO. Patient age, sex, and location of HO are summarized in Supplemental Table 2. An example of plain film radiographic appearance of 1 case is shown, demonstrating flocculent radiodensity within the soft tissue ( Figure 4A). WISP1 immunoreactivity was noted among bone-lining cells as well as osteocytes within bone matrix ( Figure 4, B-E). Cartilaginous areas of HO also showed strong WISP1 immunoreactivity ( Figure 4, F-I). Spatial overlap with transcription factors associated with bone and cartilage differentiation were next assessed ( Figure 4, J-M). WISP1immunostained sections again highlighted predominantly bone-lining cells ( Figure 4J, brown), which corresponded with spatial distributions of both RUNX2 ( Figure 4K, red) and SOX9 ( Figure 4L, red). In all cases, immunostaining without primary antibody served as a negative control ( Figure 4M). Thus, and similar to mouse HO models, WISP1 protein expression was primarily found in bone-and cartilage-associated cell types in human HO.
Wisp1 -/mice demonstrate increased trauma-induced HO. We next returned to the mouse trauma-induced HO model in the context of Wisp1 global knockin animals (30), in which exons 1-5 of the allele are replaced with a LacZ-neo cassette. Wisp1 +/+ and Wisp1 -/-(Wisp1 LacZ/LacZ ) littermates were subjected to burn/tenotomy, followed by examination of the resulting HO 3 and 9 weeks later ( Figure 5). Safranin O/fast green staining on samples 3 weeks after tenotomy was used to assess cartilage ( Figure 5, A and B). Histomorphometric analysis of the mean area of cartilage using histologic cross sections of the injury site showed an increase in cartilage formation among Wisp1 -/animals ( Figure 5, C and D). Masson's trichrome staining on samples 9 weeks after tenotomy was used to highlight bone matrix, which was likewise more conspicuous among Wisp1 -/animals ( Figure 5, E and F). μCT was used to analyze new bone formation in HO at 9 weeks after tenotomy, again demonstrating a significant (P < 0.05) increase in bone formation in HO among Wisp1 -/animals, which was notable at the proximal and distal tenotomy sites ( Figure 5, G and H). μCT quantification of bone volume at the HO injury site demonstrated a significant increase in bone volume among Wisp1 -/animals ( Figure 5I).
Immunohistochemical staining among injury sites from Wisp1 +/+ and Wisp1 -/animals was next performed. Foci of stronger expression of the cartilage-associated antigens COL2 ( Changes in the transcriptome of Wisp1 -/soft tissue mesenchymal progenitor cells. Having observed an increase in HO within Wisp1 -/animals, we sought to better explain this with isolated in vitro analysis of soft tissuederived stromal cells. For this purpose, adipose tissue derived-stromal cells (ASCs) were first chosen, as ASCs are an easily cultured cell population that can serve as a representative extraskeletal multipotent stromal cell. Wisp1 +/+ and Wisp1 -/adipose stromal cells were examined for global changes in gene expression via Clariom D microarray among cells in confluent monolayer culture at equivalent passage. Evaluation of CCN family members was first performed, which confirmed deletion of Wisp1 among Wisp1 -/mice (blue box, Figure 6A). Interestingly, the CCN family member Nov was significantly increased among Wisp1 -/cells (0.7162 FC, >+3 SD) (red box, Figure 6A). No significant change in other CCN family members was seen. Representative genes associated with chondrogenic differentiation were next assessed among Wisp1 -/cells ( Figure 6B). Enrichment in several cartilage-associated genes was observed among Wisp1 -/cells, including the transcription factor Sox9 (0.3709 FC, +3 SD) and the Hedgehog signaling-responsive gene Patched1 (Ptch1) (0.4054 FC, +3 SD) (red boxes, Figure 6B). Other cartilage-associated genes were relatively similar among undifferentiated Wisp1 -/versus Wisp +/+ cells, such as markers of chondrocyte differentiation, including Col2a) (0.1811 FC, +2 SD), Acan (0.1156 FC, +1 SD), alkaline phosphatase (Alpl) (0.085 FC, +1 SD), and Col10a1 (0.501 FC, +0.5 SD).
Conversely, downregulated signaling pathways among Wisp1 -/cells included several pathways that negatively regulate chondrogenesis or promote cartilage matrix degradation, such as AMP-activated protein  kinase (AMPK) signaling (Z score = -0.447) (36) and pigment epithelium-derived factor (PEDF) signaling (Z score = -0.447) (Supplemental Table 5) (37). Pathways with no change between Wisp1 +/+ and Wisp1 -/cells are shown in Supplemental Table 6. In sum, isolated Wisp1 -/cells have a gene expression profile associated with an increase in osteochondral markers and changes in several pathways that would predispose to a chondrogenic cell fate.
Osteochondrogenic differentiation is skewed among Wisp1 -/mesenchymal progenitor cells. As mentioned, WISP1 has context-dependent effects on chondrogenic differentiation (18). To further investigate, isolated Wisp1 +/+ and Wisp1 -/-ASCs were exposed to chondrogenic or osteogenic culture conditions (Figure 7 and Supplemental Figure 5). Wisp1 -/cells showed an increase in chondrogenic differentiation in micromass culture (Figure 7). This included an increase in chondrogenic gene expression by qPCR, including Sox9 ( Figure 7A), Comp ( Figure 7B), and Aggrecan ( Figure 7C). Similar findings were observed using sections of micromasses after 21 days in culture, as assessed by Alcian blue (Figure 7, D and E) and Safranin O red staining (Figure 7, F and G).

Discussion
Here, we observed that CCN family members, including Wisp1, demonstrated dynamic induction during the process of trauma-induced HO. These findings were observed across mouse and human tissues. Further, transgenic Wisp1-null mice demonstrated a tendency to form more endochondral bone formation after soft tissue trauma. Several potential downstream signaling pathways altered among Wisp1 -/adipose stromal cells were identified that would predispose to a chondrogenic phenotype, including STAT3 (32), PKA (31), and p38 MAPK signaling (34).
Our findings are just one example of how the matricellular protein WISP1 alters cellular fate and differentiation decisions. Past studies have implicated WISP1 in the balance of osteoblast versus adipocyte formation (38,39) and, more broadly, epithelial cells adopting a mesenchymal cell fate (epithelial-mesenchymal transition) (40)(41)(42)(43). As CCN family members, including WISP1, have numerous extracellular binding partners it is understandable that they may have context-dependent effects on cellular differentiation. In terms of chondrogenesis, WISP1 overexpression antagonizes BMP2-or GDF5-induced chondrogenesis (16), while, in other observations, WISP1 enhances TGF-β3-induced chondrogenesis (18). Further, deficiency of Wisp1 appears to be protective against articular cartilage catabolism in mouse models. For example, previous data has shown a significant decrease in cartilage degeneration in collagenase-induced arthritis using another Wisp1-knockout mouse (19). Data in osteoarthritis have also shown a decrease in cartilage degeneration and an increase in cartilage matrix formation in Wisp1-knockout mice, and similar findings in WISP1 stimulated osteoarthritic human chondrocytes (19,20). In our models, Wisp1 deletion led to a significant increase in aberrant endochondral bone formation after trauma, and isolation of Wisp1-null adipose stromal cells demonstrated signaling pathway changes consistent with a chondrogenic cell fate. Understanding that the effects of WISP1 signaling on cartilage differentiation have been somewhat nonuniform across studies, it is important to recognize that the tissue source, species for cell derivation, and method of Wisp1 inhibition (acute knockdown versus global genetic knockout) all may have contributory roles in any discrepant findings.
was upregulated, while no change in canonical TGF-β signaling was observed among Wisp1-null cells. The STAT3 signaling pathway has shown to positively regulate chondrogenic differentiation in human mesenchymal stromal cells (45). Likewise, and in agreement with findings in other cell lines (46), noncanonical BMP signaling was activated via P38 MAPK signaling among Wisp1-null cells. The differential expression of selected target genes of the chondrogenic differentiation-related signaling pathway were confirmed among primary tenocytes. For instance, Rela has been reported to be involved in chondrogenic differentiation and cell survival and prevent chondrocyte apoptosis (47). The confluence of these pathway changes would be expected to include a predisposition to cartilage differentiation, as we observed in vivo after soft tissue injury.
The chondrogenic and osteogenic induced ASC differentiation also confirmed the observation in mouse HO. Our chondrogenic differentiation results showed a significant increase in chondrogenic associated genes in Wisp1-null cells. On the other hand, osteogenic induced culture showed a decrease in osteogenic differentiation in Wisp1-null ASCs. This suggested that in a context-dependent manner Wisp1 can negatively regulate chondrogenic differentiation but positively regulate osteogenic differentiation.  Tables 4-6 for complete pathway analysis. (D) Selected target genes of the planar cell polarity (PCP) signaling pathway and (E) STAT3 signaling pathway among Wisp1 -/and Wisp1 +/+ primary tenocytes. *P < 0.05, **P < 0.01 in comparison with Wisp1 +/+ cells. Differences were calculated between groups by Student's t test. Data represent mean ± SD and triplicate experimental replicates in biological duplicate.
In our in vivo results, Wisp1-null mice showed a conspicuous increase in cartilage after injury, while bone formation was increased to a lesser degree. Further study must be performed to understand the cell-and context-dependent effects of Wisp1 in other models of endochondral ossification.
Important limitations exist in the study. Our in vitro findings were notable for a prominent and possibly compensatory increase in the CCN family member Nov among Wisp1-null cells. In previous studies with genetic deletion of CCN family member proteins, other CCN family members can play a compensatory role (48). For example, overexpression of Ccn3 in murine embryonic skin fibroblastderived NIH3T3 cells results in a significant decrease in expression of Ccn2 and Ccn4, suggesting the compensatory nature of these CCN family member proteins in organ fibrosis (49,50). As a second potential limitation, global knockout of Wisp1 not only affects mesenchymal cells, but likely has secondary effects on inflammatory cell types, such as macrophages and T lymphocytes (51). In future studies, use of tissue-specific Wisp1-knockout animals may better elucidate mesenchymal cell intrinsic functions of WISP1 in the regulation of HO. However, based on our scRNA-sequencing data, Wisp1 is predominantly expressed in adipose stromal cells and only expressed at low levels in other cell subpopulations. This suggests that a tissue-specific knockout for mesenchymal cells would likely have the same phenotype as the global knockout used here.
In conclusion, among CCN family members, Wisp1 expression is strongly associated with bone and cartilage across mouse and human HO. Studies in transgenic Wisp1-knockout animals suggest a role for Wisp1 in the negative regulation of trauma-induced HO and implicate pathways associated with enhanced cartilaginous differentiation in this process. To date, there are no clinically approved treatments for trauma-induced HO that use specific molecular targets (1). Improved understanding of CCN family signaling in HO may open the door to more targeted therapies for this common and potentially severe condition.

Methods
Bulk RNA sequencing. Bulk RNA sequencing of RNA was performed using RNA from the Achilles tendon tenotomy site. Tissue was collected from the soft tissue around the injury site. Tissue was dissected between the muscular origin and calcaneal insertion of Achilles tendon. Corresponding soft tissue, including intact Achilles tendon, was also harvested from the uninjured contralateral hind limb. n = 3 mouse tissues were isolated at 3 weeks after injury. Tissues were immediately stabilized after isolation using RNAlater (Ambion). Total cellular RNA was extracted from the specimen using the TRIZOL reagent kit (Ambion) and was examined per condition with the Affymetrix Gene 2.1 ST.
Mouse ASC and tenocyte isolation, differentiation, and analyses. For ASC collection, Wisp1 +/+ and Wisp1 -/mice were sacrificed and the inguinal fat pad tissues were harvested. Tissues were sequentially washed in PBS. The tissues were placed on a clean Petri dish lid and finely minced. Digestion of fat was achieved by 1 mg/mL collagenase type II. For tenocyte collection, Wisp1 +/+ and Wisp1 -/mice were sacrificed and the Achilles tendon was harvested and digested with 3 mg/mL collagenase type I and 4 mg/ mL dispase type II (52). Both tissues were incubated in a 37°C shaker (160 g) for 1 hour. Equal amounts of growth media were added (DMEM, 10% FBS, 1% penicillin/streptomycin), and cells were passed through a 40-μm strainer. The filtered cells were centrifuged at 450 g for 10 minutes, and the cell pellet was resuspended in 10 mL growth media. Cells were seeded into a 10-cm dish. Media were changed every 3 days unless otherwise noted.
For chondrogenic differentiation, mouse ASCs at passages 3-5 were used. Cells were cultured in high-density micromass as modified protocol (53) or 1 × 10 4 cells/well in 24-well plates for monolayer culture. The MesenCult-ACF Chondrogenic Differentiation Kit (StemCell Technology, catalog 05455) was used. Medium was changed every 3 days, and micromasses were analyzed for up to 14 days. Micromasses were fixed in 4% PFA, placed in 30% sucrose and embedded with optimal cutting temperature compound. Serial sections of 10 μm were made, hydrated, and subsequently stained with Alcian blue (MilliporeSigma) or Safranin O (MilliporeSigma) using previously published methods (54).
Osteogenic differentiation medium consisted of 10 mM β-glycerophosphate and 50 μM ascorbic acid in DMEM + 15% FBS. Alkaline phosphatase staining was performed using the Leukocyte Alkaline Phosphatase Kit (MilliporeSigma). Briefly, cells were seeded in 24-well plates at a density of 1 × 10 4 cells/well. Cells were cultured under osteogenic differentiation conditions for 14 days before staining. Cells were then washed with PBS and fixed with formalin for 10 minutes at room temperature. Following fixation, cells were stained using the Leukocyte Alkaline Phosphatase Kit (MilliporeSigma) according to the manufacturer's protocol. Cells were incubated in alkaline phosphatase stain for 15 minutes at 37°C and then washed with PBS. Relative staining was quantified using Adobe Photoshop CC2015.
For the detection of mineralization, cells were seeded in growth medium in 24-well plates at a density of 1 × 10 4 cells/well. Twenty-four hours after seeding cells, basal medium was replaced with ODM in triplicates per each treatment for 12 days. Cells were washed with PBS and fixed with 4% paraformaldehyde. Following fixation, cells were stained with 2% alizarin red (MilliporeSigma) at room temperature for 15 minutes and then washed with deionized water and allowed to dry.
RNA isolation and quantitative real-time PCR. Gene expression was assayed by quantitative RT-PCR. Primer sequences are shown in Supplemental Table 7. Briefly, total RNA was extracted using Trizol (Invitrogen). 1 μg total RNA from each sample was subjected to first-strand cDNA synthesis using the Super-Script III Reverse-Transcriptase Kit (Life Technologies) to a final volume of 20 μL. The reverse transcription reaction was performed at 65°C for 5 minutes, followed by 50°C for 50 minutes and 85°C for 5 minutes.
The quantitative real-time PCR (qRT-PCR) reaction was performed using SYBR green RT-PCR master mix (Thermo Fisher Scientific) and an ABI PRISM 7300 qRT-PCR system instrument (Applied Biosystems). qRT-PCR was performed using 384-well optical plates at 95°C for 10 minutes, followed by 40 cycles at 95°C for 15 seconds and at 60°C for 60 seconds. The relative quantification of gene expression was performed using a comparative CT method according to the manufacturer's protocol and was normalized to the expression levels of Gapdh in each sample.
RNASeq -10x single-cell genomics. For scRNA-sequencing analysis, the tenotomy site was harvested by microdissection (days 3, 7, and 21 after injury) and compared with baseline uninjured tissue at the same site. n = 3-4 mice per time point were used. Baseline (t0) and tissue samples harvested on days 3, 7, and 21 were digested for 45 minutes in 0.3% type 1 collagenase and 0.4% Dispase II (Gibco Invitrogen) in RPMI medium at 37°C under constant agitation at 160 g. Digestions were subsequently quenched with 10% FBS RPMI and filtered through 40-μm sterile strainers. Cells were then washed in PBS with 0.04% BSA, counted, and resuspended at a concentration of approximately 1000 cells/μL. Cell viability was assessed with trypan blue exclusion on a Countess II (Thermo Fisher Scientific) automated counter, and only samples with >85% viability were processed for further sequencing.
Single-cell 3′ library generation was performed on the 10× Genomics Chromium Controller following the manufacturer's protocol for the v2 reagent kit (10× Genomics). Cell suspensions were loaded onto a Chromium Single-Cell A chip along with reverse-transcription (RT) master mix and single-cell 3′ gel beads, aiming for 2000-6000 cells per channel. The cDNA was amplified, and SPRIselect (Beckman Coulter) beads were cleaned and quantified and then enzymatically fragmented and size selected using SPRIselect beads to optimize the cDNA amplicon size before library construction. An additional round of double-sided SPRI bead cleanup was performed after end repair and A-tailing. Another single-sided cleanup was done after adapter ligation. Indexes were added during PCR amplification and a final double-sided SPRI cleanup was performed. Libraries were sequenced on a HiSeq 4000 (Illumina) using a HiSeq 4000 PE Cluster Kit (PN PE-410-1001) with HiSeq 4000 SBS Kit (100 cycles, PN FC-410-1002) reagents, loaded at 200 pM following Illumina's denaturing and dilution recommendations. Cell Ranger Single Cell Software Suite 1.3 was used to perform sample demultiplexing, barcode processing, and single-cell gene counting (alignment, barcoding, and unique molecular identifiers [UMI] count) at the University of Michigan Biomedical Core Facilities DNA Sequencing Core.
The sequencing data were first preprocessed using the 10× Genomics software Cell Ranger and aligned to mm10 genome. Downstream analysis steps were performed using Seurat 2.3.0. Cells with fewer than 500 genes per cell, with more than 60,000 UMIs, or expressing a fraction of mitochondrial UMIs higher than 0.1 were filtered for quality control. The number of UMIs (UMI fraction related to mitochondrial genes >0.1) and the number of genes, mitochondrial genes, and cell cycle genes (S and G2M scores by CellCycleScoring Seurat function) were regressed before further analysis. The downstream analysis was performed with Seurat 2.3.0. The time point sets were merged with canonical correlation analysis based on the intersection of the variable genes for time point set. Unsupervised clustering (Louvain algorithm) was used to identify cell populations.
Histological and immunohistochemical analyses. Five-mm-thick paraffin sections of mouse and human tissues were stained with H&E. Additional sections were analyzed by indirect immunohistochemistry, with antibodies summarized in Supplemental Table 3. Briefly, unstained sections were deparaffinized in xylene and a series of graded ethanol solutions and rehydrated using PBS. The slides were treated with Trypsin Enzymatic Antigen Retrieval Solution (ab970, Abcam) for 10 minutes at 37°C and then incubated with 3% hydrogen peroxide (20 minutes at room temperature) and 5% goat serum (1 hour at room temperature). Next, slides were probed with primary antibodies at a dilution of 1:200 overnight. Subsequently, for slides probed with rabbit host primary antibodies, the sections were washed and incubated with biotinylated secondary antibody (BA-1000, Vector Laboratories) and VECTASTAIN Elite ABC HRP reagent (PK-6100, Vector Laboratories) for 1 hour and 30 minutes at room temperature. Color development was achieved by treatment with the chromogen DAB (SK-4100, Vector Laboratories) and was carried out for 5-10 minutes. The slides were then blocked with BLOXALL Blocking Solution (SP-6000, Vector laboratories) and 2.5% normal horse serum (MP-5402, Vector Laboratories) for 10 and 20 minutes at room temperature. For slides probed with mouse host primary antibodies, the sections were washed and incubated with IMPRESS-AP reagent (MP-5402, Vector Laboratories) for 30 minutes at room temperature. Color development was achieved by treatment with the Alkaline Phosphatase Substrate (SK-5100, Vector Laboratories) and was carried out for 5-10 minutes. Finally, the sections were rinsed in distilled water, counterstained with hematoxylin (H-3401, Vector Laboratories), and mounted on glass slides before microscopic evaluation.
Positive immunoreactivity was detected following ABC complex (PK-6100, Vectastain Elite ABC Kit, Vector Laboratories Inc.) incubation and development with AEC chromogen (K346911-2, Dako). Negative controls for each antibody consisted of incubation with secondary antibody in the absence of primary antibody. Sections were counterstained in Modified Mayer's Hematoxylin (Thermo Fisher Scientific) for 30 seconds and placed under running water for 5 minutes. Slides were mounted using aqueous mounting medium (Dako). Photomicrographs were acquired using Olympus BX51 (UPLanFL).
Identification of human nongenetic HO samples. Twelve cases of nongenetic HO were identified in our surgical pathology archives (Johns Hopkins University and University of California, Los Angeles). All material was coded so as to protect the confidentiality of personal health information. All cases were reviewed by 2 independent pathologists verifying the diagnostic accuracy. Patient demographics are provided in Supplemental Table 2.
RNA Clariom D microarray analyses. RNA from Wisp1 +/+ and Wisp1 -/adipose stromal cells was isolated as per prior methods (39). RNA isolation was performed as previously described (55). RNA transcript levels were quantified by microarray analyses on the Clariom D Array GeneChip (Affymetrix; Thermo Fisher Scientific). Raw expression was RNA normalized and underwent analysis of variance using the PARTEC GS 7.0 platform. Microarray analysis was conducted in the Johns Hopkins Transcriptomics and Deep Sequencing Core Facility. Gene expression values for Wisp1 -/to Wisp1 +/+ animals are shown as log 2 FC.
Data availability. Expression data that support the findings of this study have been deposited in Gene Expression Omnibus (GEO) (accession GSE 126060, GSE 127956, and GSE 126118).
Statistics. Statistical analyses were performed in IBM SPSS Statistics 16.0. Quantitative data are expressed at mean ± SD. Two-tailed Student's t test was used for 2-group comparisons, and 1-way ANOVA test was used for comparisons of 3 or more groups, followed by Tukey's post hoc test.
Study approval. All animal procedures were carried out in accordance with the guidelines provided in the Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and were approved by IACUC of the University of Michigan (PRO0007390). All animals were housed in IACUC-supervised facilities at 18°C-22°C on a 12-hour light/dark cycle, with ad libitum access to food and water. Cases were obtained under approval from the University of California, Los Angeles, IRB with a waiver of informed consent.

Author contributions
GCYH collected and assembled data, provided data analysis and interpretation, and wrote the manuscript. S. Marini collected and assembled data. CAM collected and assembled data. JX collected and assembled data. YW collected and assembled data. SN collected and assembled data. CH collected and assembled data. DS collected and assembled data. S. Miller collected data and wrote the manuscript. CP collected and assembled data. EM provided study materials. KML provided study materials. SN, CH, and DS collected and assembled data. BL provided study materials and financial support and approved the final manuscript. AWJ provided study materials, conception, and design; wrote the manuscript; provided financial support; and approved the final manuscript.