Our previous studies showed that disruption of TGFβ signaling in mouse limb mesenchyme resulted in the complete failure of tendon formation (Pryce et al., 2009). To examine later roles of TGFβ signaling in mouse tendon development, the floxed Tgfbr2 gene was targeted conditionally with ScxCre (Tgfbr2^f/-;ScxCre; called hereafter Tgfbr2;ScxCre mutant) to bypass the early role of TGFβ signaling in tendon development. ScxCre activity in tenocytes is not uniform during embryogenesis (Figure 1—figure supplement 1A) and complete targeting of tenocytes is achieved only in early postnatal stages. Indeed, immunostaining for TGFβ type II receptor revealed that by P0 mutant tendons displayed a nearly complete loss of receptor expression (Figure 1—figure supplement 1C). Consequently, Tgfbr2;ScxCre mutant embryos developed a complete network of tendons by E14.5, indicating they have bypassed the early requirement for TGFβ signaling in tendon development (Figure 1A).

Mutant tendon development was not perturbed through embryogenesis and mutant pups appeared normal at birth (Figure 1C). However, by day 3 after birth (P3), mutant pups showed physical abnormalities that manifested in abducted paws, splayed limbs (Figure 1C, black arrowhead) and severe movement limitations. Examination of forelimb tendons of P7 mutant pups using the tendon reporter ScxGFP revealed severe tendon disruptions. A few lateral limb tendons, for example the extensor carpi radialis longus tendon underwent fragmentation and disintegrated (Figure 1B, yellow arrowhead and Figure 1—figure supplement 3), whereas the majority of other tendons, notably the extensor digitorium communis tendons, retained structural integrity with a substantial loss of ScxGFP signal (Figure 1B, white arrowhead). Substantial loss of ScxGFP was also detected in all tendons and related tissues, including hindlimb and tail tendons, ligaments and the annulus fibrosus of the intervertebral disc (Figure 1D,E). Loss of ScxGFP signal was gradual starting around P2-P3, that is before the manifestation of physical abnormalities (Figure 1B and Figure 1—figure supplement 2). All mutant tendon cells lost ScxGFP at P7. We therefore performed most analyses of the mutant phenotype in this fully-phenotypic stage. The progressive nature of the phenotype also manifested in exacerbated movement limitations as mutant pups became older. This phenotypic progression was observed in most mutant pups but intriguingly, in rare cases (~2%) the mutant pups showed physical abnormalities and severe tendon phenotypes already at birth. Regardless, all mutants died at or before P14 likely due to ScxCre activity in developing cardiac valves (Levay et al., 2008), leading to enlarged heart as evidenced by gross examination and histological analysis (data not shown).

A closer examination of the mutant embryos identified the first indication of a tendon phenotype already at E16.5. The flexor carpi radialis tendons of mutant embryos were consistently torn by E16.5 (Figure 1—figure supplement 4). Interestingly, this phenotype was highly reproducible while the patterning and development of other tendons in mutant embryos was not perturbed through embryogenesis. Moreover, expression of the prototypic tenocyte markers Scx, tenomodulin and collagen I (Figure 2A–D) and the development of the collagen matrix were not disrupted in any tendon of mutant embryos (Figure 2E,F), including the flexor carpi radialis tendon before it snapped. A direct cause for the specific tear of the flexor carpi radialis tendon in mutant embryos was not identified to date.

Tendons are rich in collagen fibers that provide structural integrity to the tendons and transmit the forces generated by muscle contraction (Kannus, 2000). Since young mutant pups exhibited movement difficulties, we first examined possible structural effects in the collagen matrix. The ultrastructure of mutant tendons that remained intact was therefore analyzed by transmission electron microscopy (TEM). Surprisingly, despite the functional defects and loss of ScxGFP signal starting around P3, collagen fibers in mutant tendons appeared organized and indistinguishable from those of wild-type (WT) littermates at this stage (Figure 3A,B). Apparent collagen degradation was observed only in older mutant pups (≥P7) (Figure 3C–G), suggesting the disruption to the matrix of these tendons may be a secondary consequence of the cellular changes in these mutants and/or of their movement difficulties. Furthermore, epitenon, a monolayer of cells that engulf and define the boundary of the tendon (Kannus, 2000) (Figure 3F, black arrowhead), was gradually disrupted and in some cases was almost undetectable in older mutant pups (Figure 3G, white arrowhead), suggesting that loss of the tendon boundary is an additional feature of the phenotype in these mutants.

As mentioned earlier, the ScxGFP signal in mutant tendons appeared patchy contrary to the smooth appearance of WT tendons (Figure 1B), suggesting a disruption at the cellular level. To examine this phenotype at the cellular level, we analyzed cross-sections through the extensor communis tendons of P7 WT and mutant pups. In P7 WT pups, all tendon cells were positive for ScxGFP, Tnmd and Col1a1 (Figure 4A,C). Conversely, most cells in mutant tendons lost the ScxGFP signal and tendon marker gene expression (Figure 4B, white arrowhead and Figure 4C). Interestingly, some cells in mutant tendons retained ScxGFP signal and appeared rounded and enlarged from P3 onwards (Figure 4B, yellow arrowhead). Some of these cells exhibited weak or no expression of the Ai14 Rosa26-tdTomato (RosaT) Cre reporter (Madisen et al., 2010), suggesting a recent activation of the Scx enhancer in these cells and therefore that they are newly recruited tendon cells. Analysis of this aspect of the mutant phenotype will be published in a separate manuscript (Tan et al. in preparation).

The fact that most cells in the mutant tendons do not express tendon markers is surprising, since the cells in these tendons were functional tenocytes at embryonic stages as evidenced by tendon marker gene expression and by the development of a functional collagen matrix (Figure 2). We next sought to determine if the ScxGFP-negative cells were indeed tendon cells that lost tendon gene expression or if the mutant tendons were simply repopulated by non-tenogenic cells. Using TUNEL assay, we did not detect cell death in mutant tendons and the rate of tenocyte proliferation as examined by EdU assay was also not altered in these tendons during different developmental stages ranging from E14.5 to P10 (Figure 2—figure supplement 1A,B), suggesting the cell population of mutant tendons was not altered. To directly determine if the cells in mutant tendons were tenocytes whose cell fate was altered, we took advantage of the fate mapping feature of the RosaT Cre reporter system (Madisen et al., 2010). When the reporter is activated by ScxCre, expression of the RosaT reporter is restricted to the Scx-expressing cells and their progeny. We found that all ScxGFP-negative cells within mutant tendons were positive for the RosaT Cre reporter (Figure 4D, white arrowhead). Notably, some non-tendon cells are also positive for the RosaT Cre reporter at this stage. However, given that there is no apparent elimination of the existing tenocytes, even if some of these cells were recruited into the mutant tendons that would not explain the absence of the original tenocytes in mutant tendons. This result thus indicates that the ScxGFP-negative cells in the mutant tendons were derived from tenocytes, and highlighted an unexpected reversibility for the tendon cell fate where it was possible for committed and functional tenocytes to lose their differentiation status.

Next, we wanted to ask if these results reflected that maintenance of the tendon cell fate was dependent on continuous activation of TGFβ signaling. Since the cellular phenotype manifestated mainly in post-natal stage, we targeted Tgfbr2 in all cells using the ubiquitous tamoxifen-inducible Rosa^CreERT2 driver (Hameyer et al., 2007). Tgfbr2;Rosa^CreERT2 pups were fed with tamoxifen at P1,P2 and P5, P6 (1.25 mg per pup for each time point) and harvested at P7-P14. Efficient recombination of the Tgfbr2 gene was confirmed by immunostaining of the receptor (Figure 1—figure supplement 1D,E). Interestingly, the cell fate of targeted cells was not disrupted in these mutants as evidenced by retention of tendon marker expression (Figure 1—figure supplement 1F). This result suggests that a mere loss of TGFβ signaling is not sufficient to cause tenocyte dedifferentiation, and additional factors associated with the loss of Tgfbr2 in the spatial and temporal features determined by ScxCre activity may also play a critical role in this process.

Loss of cell differentiation marker can be the outcome of a few cellular processes, including most notably cell death, change of cell fate (transdifferentiation) or reversion to a less differentiated state (dedifferentiation) (Cai et al., 2007; Talchai et al., 2012; Tata et al., 2013). As aforementioned, we found no apparent cell death in mutant tendons (Figure 2—figure supplement 1A). Using histological staining for the prototypic markers of osteocytes, adipocytes and chondrocytes, we found that loss of tendon gene expression in the cells of mutant tendons was also not due to transdifferentiation (Figure 2—figure supplement 1C), suggesting that the changes in mutant tendons may reflect a process of cellular dedifferentiation.

One hallmark of cellular dedifferentiation is the loss of differentiation markers, which is what we observed in mutant tendon cells. When cells dedifferentiate they also assume stemness features for example colony-forming potential, and in most cases these cells also acquire expression of stem/progenitor cell markers (Sun et al., 2011; Tata et al., 2013; Nusse et al., 2018). To date, very little is known about the specific gene expression and cellular behavior of embryonic tendon progenitors. The only defined feature of these cells is the expression of the Scx tendon progenitor marker (Schweitzer et al., 2001), which was evidently lost in the mutant tendon cells. We therefore next directed our attention to similarities with stem/progenitor cells isolated from tendons (tendon-derived stem/progenitor cells) (Bi et al., 2007; Rui et al., 2010; Murchison et al., 2007; Mienaltowski et al., 2013) and with stem/progenitor cell markers reported in other studies (Blitz et al., 2013; Dyment et al., 2013; Tan et al., 2013; Runesson et al., 2015; Yin et al., 2016).

To test the colony-forming capacity of the mutant tendon cells, P7 mutant tendons were dissociated and FACS-sorted to collect ScxGFP-negative and RosaT-positive cells, which were then seeded at one cell per well in 96-well plates. As shown in Figure 5A, about 1–2% of cells (ScxGFP-positive and RosaT-positive) isolated from tendons of P7 WT and heterozygous controls formed colonies in culture, similar to the frequency of colony forming cells reported in other studies (Bi et al., 2007; Rui et al., 2010). On the other hand, we found a significant eightfold increase (p<0.01) in the frequency of colony-forming cells in mutant tendons (Figure 5A).

We next screened the mutant tendons for expression of stem/progenitor cell markers and found that the Tgfbr2;ScxCre mutant tendon cells gradually acquired expression of stem cell antigen-1 (Sca-1) and CD44 in postnatal stages (Figure 5B). Notably, expression of Sca-1 was undetectable and CD44 was detected only in very few WT tendon cells, but surprisingly robust expression of these markers was detected in the epitenon (Figure 5C, white arrowheads), a possible source of progenitor cells (Mendias et al., 2012; Dyment et al., 2013; Mienaltowski et al., 2013; Harvey et al., 2019). The similarity of marker expression between the mutant tenocytes and epitenon cells therefore reinforces the notion that the mutant tenocytes acquired progenitor features.

Dedifferentiation is frequently associated with reversion to an earlier progenitor cell fate (Cai et al., 2007). We therefore next examined the expression of these markers during embryonic tendon development. At E12.5, when tendon progenitors are first detected (Pryce et al., 2009), expression of Sca-1 and CD44 could not be detected in ScxGFP-positive tendon progenitors (data not shown). At E14.5, at the onset of tendon cell differentiation, we found low or no expression of both markers in the differentiating tendon cells. Robust positive staining for both markers was however detected in the cells that surround the tendon at this stage, likely the precursors of the epitenon/paratenon (Figure 5—figure supplement 1). Similar expression patterns were also found in mutant embryos (data not shown). These findings suggest that Sca-1 and CD44 are not markers for tendon progenitor in vivo, and possibly simply reflect a generic stemness state of the dedifferentiated mutant tendon cells.

Taken together, our findings show that mutant tendon cells acquired some generic stem/progenitor properties while losing their cell fate. It should be noted however that although these dedifferentiated tendon cells demonstrate some stem/progenitor properties, absence of TGFβ signaling in these cells might prevent them from acquiring the full spectrum of stemness or plasticity.

We next performed single-cell RNA sequencing analysis (scRNASeq) to establish a comprehensive profile of the cellular state and molecular changes in mutant tenocytes. A targeted retention of 2300–2600 cells from P7 WT and mutant tendon was obtained, and the transcriptomes were analyzed using the 10X Genomics platform. Using unsupervised hierarchical clustering analysis, we identified two major clusters corresponding to WT tenocytes and mutant (dedifferentiated) tendon cells in the respective samples (Figure 6A, Supplementary file 1A,B). The WT tenocyte cluster was defined by the expression of tendon markers including Scx, Fmod, Col11a1, Col1a1 and Tnmd. The mutant (dedifferentiated) tendon cell cluster is enriched for Ly6a (encoding Sca-1) and expresses undetectable level of tendon markers. Expression of close to 1000 genes (mean UMI count ≥0.5, adjusted p-value<0.05) was identified in each of these clusters.

Pairwise comparison of the gene sets between the P7 WT tenocyte and mutant cell clusters was next performed to determine changes in gene expression associated with tenocyte dedifferentiation. In total, expression of 186 genes was significantly different between the two cell populations (≥2 fold change and adjusted p-value<0.05), in which expression of 89 genes was upregulated and 97 genes was downregulated in the mutant tendon cells (Supplementary file 2). Almost 30% of the downregulated genes (29 genes) were identified in transcriptome analyses as tendon distinctive genes [(Havis et al., 2014) and our unpublished data]. Notably, the genes Scx, Fmod, Tnmd, Pdgfrl, Col1a1, Col1a2, Col11a1 and Col11a2 were among the top 25 down-regulated genes in Tgfbr2;ScxCre mutant tendon cells (Table 1), further confirming the loss-of-cell fate phenotype in these cells. On the other hand, expression of the Ly6a gene (encoding Sca-1) was greatly enriched in P7 mutant cells, corroborating the IHC findings presented above (Table 2 and Figure 5B). Moreover, we also found a significant increase in the expression of the Cd34 gene, another common marker for diverse progenitor cells. This observation was further confirmed at protein level, where positive immunostaining for CD34 was detected in mutant cells but not in normal tendon cells (Figure 6B). Interestingly, the genes upregulated in the mutant cells included several genes (Dpt, Anxa1, Cd34, Cd44, Mgp and Mfap5) whose expression was previously reported to be enriched during embryonic tendon development (Havis et al., 2014). These findings thus do not only lend support to our notion that the mutant cells lost their differentiation state, but also suggest the possibility of induction of some developmental programs in these cells, a general feature in cellular dedifferentiation (Tata et al., 2013; Stocum, 2017; Nusse et al., 2018).

To gain insights into biological functions activated in the P7 mutant cells, differentially expressed genes (DEGs) in these cells (Supplementary file 2) were further analyzed via GO enrichment tools clusterProfiler (Yu et al., 2012) and PANTHER Classification System (http://pantherdb.org/). Intriguingly, GO enrichment analysis revealed that one of the prominent biological changes observed in P7 mutant cells was upregulation of gene sets associated with wound healing (Figure 6C and Supplementary file 3A). These genes include protease inhibitors (Serpine2, Serping1), inflammatory mediator Anxa1 and extracellular matrix (Col3a1 and Col5a1). This finding suggests a possible role for tendon cells in the responses to pathological conditions, in line with findings reported by others (Dakin et al., 2015; Stolk et al., 2017; Schoenenberger et al., 2018). On the other hand, many biological processes downregulated in P7 mutant cells involved collagen synthesis and organization (Figure 6D and Supplementary file 3B). Since tendon biology is not annotated in most databases, changes in the collagen matrix, the most prominent structural component in tendons is the best indicator for the disruption of the tendon cell fate. Disruption of the collagen matrix in tendons was also detected in older mutant pups by ultrastructural analysis using TEM (Figure 3E,G).

Using PANTHER, we also investigated which protein classes were significantly altered in P7 mutant cells relative to WT tenocytes. Genes found to be most downregulated in mutant cells encode for receptors, signaling molecules, membrane traffic proteins and ECM (Table 3A). On the other hand, the upregulated genes in the mutant cells encode most prominently for proteins involved in nucleic acid binding, enzyme modulators, cytoskeletal protein, signaling molecules and transcription factors (Table 3B). Notably, expression of the activating protein 1 (AP-1) transcriptional complex, associated with numerous cellular processes including cell fate regulation (Hess et al., 2004), was significantly induced in mutant cells. Expression of both AP-1 components, that is the Fos and Junb genes was induced more than twofold, and the Jun gene was induced only slightly less than twofold. Moreover, the Id3 gene encoding for a general bHLH transcription factor inhibitor was also induced. Due to its broad selection of targets, Id3 was also implicated in numerous cellular processes including the regulation of cellular differentiation (Norton, 2000). A possible role for these transcriptional activities in tenocyte dedifferentiation will be addressed in future studies.

We next conducted PANTHER Pathway Analysis using different values of the filter parameter (mean UMI count and fold change) for enriching DEGs in P7 mutant cells. In general, we found that pathways that stood out as relevant for this study included integrin signaling, insulin/IGF, Wnt and inflammation mediated by chemokine and cytokine signaling pathways (Table 4). Insulin/IGF and Wnt signaling are often implicated in cell proliferation and cell fate specification (Stewart and Rotwein, 1996; Sadagurski et al., 2006; Goessling et al., 2009; Salazar et al., 2016). It is interesting to note that their activation has also been associated with cellular dedifferentiation in skin, gut and neuron (Weber et al., 2003; Zhang et al., 2012; Perekatt et al., 2018). Further investigation is required to determine the specific roles of these signaling pathways in tenocyte dedifferentiation.

Lastly we wanted to ask if tenocyte dedifferentiation in these mutants reflected a cell autonomous requirement for TGFβ signaling in tenocytes, or if it was the result of global changes that occurred in mutant tendons. To address this question, we wanted to reactivate TGFβ signaling in isolated mutant tendon cells that will therefore still be exposed to the mutant tendon environment and determine the effects on tenocyte dedifferentiation. We previously found that transuterine injection of AAV viruses into embryonic limbs resulted in sporadic infection of limb tendons [(Huang et al., 2013) and unpublished data]. We therefore decided to address this question by injection of a Cre-activatable virus encoding an epitope tagged version of the receptor, AAV1-FLEX-Tgfbr2-V5 (Figure 7A). Injection of this virus into embryonic mutant limbs would result in expression of Tgfbr2-V5 only in infected tendon cells due to the tendon-restricted activity of ScxCre in mutant embryos.

AAV1-FLEX-Tgfbr2-V5 was injected into mutant limbs at two stages during embryonic tendon development: (a) E12.5 at the onset of ScxCre activity, ensuring that Tgfbr2-V5 expression will be activated in infected cells concurrent with the loss of the endogenous Tgfbr2, resulting in isolated Tgfbr2-expressing cells surrounded by mutant cells. (b) E16.5, before the onset of tenocyte dedifferentiation in mutant embryos. Infected limbs were harvested at P5-P7, and the effects of Tgfbr2 expression on mutant tendon cells was evaluated by analyzing cells with positive V5 signal. Interestingly, targeted re-expression of Tgfbr2-V5 in individual mutant tendon cell at both developmental stages was able to prevent the loss of tendon markers as observed in postnatal pups (Figure 7B–D), suggesting a cell autonomous role for TGFβ signaling in maintenance of the tendon cell fate.

Recognizing that cell autonomous activity of Tgfbr2-V5 was sufficient to prevent dedifferentiation of mutant tenocytes, we next wanted to test if reactivation of TGFβ signaling in a dedifferentiating tenocyte could also reverse the process and rescue a tenocyte from dedifferentiation. Activity of ScxCre may be lost in the dedifferentiating tenocytes due to the loss of Scx expression and therefore of Scx enhancer driven expression of Cre in tendons of Tgfbr2;ScxCre mice. We therefore used in this case a virus encoding constitutive expression of Tgfbr2 in which the virus was tagged with a FLAG Tag (AAV1-Tgfbr2-FLAG). The virus was injected locally into P1 mutant limbs and the limbs were harvested at P7. We found again that all infected mutant tendon cells expressed the tendon markers ScxGFP and tenomodulin (Figure 7D and Figure 7—figure supplement 1A), suggesting that reactivation of TGFβ signaling was indeed sufficient to rescue the dedifferentiated tenocytes. Taken together, these findings demonstrate that TGFβ signaling is sufficient to prevent and to rescue the loss-of-tendon cell fate in a cell autonomous manner.

The constitutive expression of Tgfbr2-FLAG driven by the AAV1-Tgfbr2-FLAG virus ensured that the neonatal infection with this virus resulted in Tgfbr2-FLAG expression both within and outside of tendons. Notably, induction of tendon gene expression following activation of Tgfbr2-FLAG expression was detected only in dedifferentiated tenocytes and not in cells located outside of tendons (Figure 7—figure supplement 1B). It was previously shown that TGFβ signaling is a potent inducer of ScxGFP and other tendon markers (Pryce et al., 2009; Maeda et al., 2011; Sakabe et al., 2018). This result however, reflects the fact that induction of tendon markers by TGFβ signaling is context-dependent and further indicates that the tenocytes in mutant pups have dedifferentiated to a state that retained tenogenic potential and the capacity to respond to TGFβ signaling.

Taken together these results highlight a surprising cell autonomous role for TGFβ signaling in maintenance of the tendon cell fate. In Tgfbr2;ScxCre mutants tenocyte differentiation and function are normal during embryonic development but the tenocytes dedifferentiate in early postnatal stages. Tenocyte dedifferentiation is directly dependent on the loss of TGFβ signaling since retention or reactivation of the TGFβ receptor in isolated cells prevents or reverses the process of dedifferentiation. TGFβ signaling is thus essential for maintenance of the tendon cell fate.

All mouse works were performed in accordance to the guidelines issued by the Animal Care and Use Committee at Oregon Health and Science University (OHSU). Floxed TGFβ type II receptor (Tgfbr2^f/f) mice (Chytil et al., 2002) were crossed with mice carrying the tendon deletor Scleraxis-Cre recombinase (ScxCre) (Blitz et al., 2013) to disrupt TGFβ signaling in tenocytes (called hereafter Tgfbr2;ScxCre mutant). All mice in this study also carried a transgenic tendon reporter ScxGFP (Pryce et al., 2007), and a Cre reporter Ai14 Rosa26-tdTomato (RosaT) (Madisen et al., 2010) for the lineage tracing of Scx-expressing cells. For embryo harvest, timed mating was set up in the afternoon, and identification of a mucosal plug on the next morning was considered 0.5 days of gestation (E0.5). Embryonic day 14.5 to postnatal day 13 (E14.5-P13) limb tendons were used for analysis. Mouse genotype was determined by PCR analysis of DNA extracted from tail snip using a lysis reagent (Viagen Biotech, Cat 102 T) and proteinase K digestion (55°C, overnight).

Skinned mouse forelimbs were fixed intact for several days in 1.5% glutaraldehyde/1.5% formaldehyde, rinsed, then decalcified in 0.2 M EDTA with 50 mM TRIS in a microwave (Ted Pella, Inc) operated at 97.5 watts for fifteen 99 min cycles. Samples were fixed again in 1.5% glutaraldehyde/1.5% formaldehyde with 0.05% tannic acid overnight, then rinsed and post-fixed overnight in 1% OsO₄. Samples were dehydrated and extensively infiltrated in Spurr’s epoxy and polymerized at 70°C (Keene and Tufa, 2018). Ultrathin sections of tendons of interest were cut at 80 nm, contrasted with uranyl acetate and lead citrate, and imaged using a FEI G20 TEM operated at 120 kV with montages collected using a AMT XR-41 2 × 2K camera. The acquired images were stitched using ImageJ software (https://imagej.nih.gov/ij/) (Preibisch et al., 2009). Three pups per time point were harvested for TEM analysis.

Dissected forelimbs were fixed with 4% paraformaldehyde in PBS, decalcified in 5 mM EDTA (1–2 weeks at 4°C) and incubated with a 5–30% sucrose/PBS gradient. The tissues were then embedded in OCT (Tissue-Tek, Inc), sectioned at 10 µm or 12 µm using a Microm HM550 cryostat (Thermo Scientific, Waltham, MA) and mounted on Superfrost plus slides (Fisher). In situ hybridization was performed as previously described (Murchison et al., 2007).

For immunofluorescence staining, sections were air-dried, rinsed thrice with PBS and blocked with 2% BSA and 2% normal goat serum in PBS for 1 hr at RT. The sections were then incubated overnight at 4°C with specific primary antibody as listed in Key Resources Table. This was followed by incubation with the matching Cy3- or Cy5/AlexaFluor647-conjugated secondary antibody (Jackson ImmunoResearch; diluted at 1:400 to 1:800; see Key Resources Table) in PBS containing 2% normal goat serum for 1 hr at RT. DAPI (4′,6-diamidino-2-phenylindole, dihydrochloride; Thermo Fisher Scientific) was used to counterstain cell nuclei. Immunolabelled sections were mounted in Fluorogel (Electron Microscopy Sciences, PA; Cat 17985–10) and visualized using a Zeiss ApoTome microscope. A washing step with PBS containing 0.1% Triton-X 100 was performed after the change of antibodies. Controls included omission of primary antibodies.

For examination of cell death and proliferation, TUNEL and EdU assays were performed using Click-iT EdU (Life Technologies) and In Situ Cell Death Detection (Roche) kits, respectively, following manufacturer’s instructions. For all studies, sections from two to four pups were examined to ensure reproducibility of results.

Mice at P7 were used for tendon progenitor cell isolation using a protocol modified from that in Mienaltowski et al. (2013). Briefly, both forelimbs and hindlimbs were harvested from euthanized mice, skinned and exposed to 0.5% collagenase type I (Gibco, Cat 17100–017) and 0.25% trypsin (Gibco, Cat 27250–018) in PBS for 15 min at 37°C with gentle shaking. The surfaces of tendons were then scraped carefully with a pair of forceps to remove epitenon/paratenon cells. The middle portion of tendons was then harvested, cut into small pieces and tendon cells were released by digestion for 30 min at 37°C with gentle shaking in a solution of 0.3% collagenase type I, 0.8% collagenase type II (Cat 17101–015), 0.5% trypsin and 0.4% dispase II (Cat 17105–041) (all from Gibco). The released cells were strained with a 70 μm cell strainer (BD Falcon, Cat 352350) and collected by centrifugation for 5 min at 300 g. The cells were then resuspended in PBS with 1% BSA, and fluorescence-activated cell sorting (FACS) was used to separate the cells for colony-forming assay.

CFU assay was used to examine the self-renewal potential of cells (Bi et al., 2007). The enzymatically-released WT and heterozygous tenocytes as well as dedifferentiated mutant tendon cells (i.e. ScxGFP-negative and RosaT-positive cells) were sorted by FACS and plated at one cell per well in a 96-well plate using a BD Influx cell sorter (BD Bioscience, USA). About 10–12 days into the culture, the colonies were fixed in 4% paraformaldehyde (10 min, RT), stained with 0.5% crystal violet for 30 min, and rinsed twice with water. Percentage of colony-forming unit was calculated as: Number of wells with colonies ÷ 96 × 100. Each data point represents the mean of duplicate plates from 3 to 5 separate experiments. Each experiment represents limb tendons collected from 2 to 4 pups.

FLAG or V5 epitope tag sequences were added at the C-terminus of the murine TGFbR2 Consensus Coding Sequence (CCDS23601). The Tgfbr2-FLAG (Tgfbr2-FLAG) and reverse-compliment Tgfbr2-V5 (FLEX-Tgfbr2-V5) insert sequences were synthesized and subcloned by GenScript into an AAV1 vector. The FLEX backbone vector (Atasoy et al., 2008) was purchased from AddGene and modified. Vectors were then packaged into AAV1 capsid, purified, and titered by the OHSU Molecular Virology Support Core. AAV1 insert expression was under the control of a chicken beta-actin (CBA) promoter and an SV40 polyadenylation sequence. All experimental procedures were evaluated and approved by the institutional Animal Care and Ethics Committee.

Re-expression of Tgfbr2 in embryos was done by delivery of AAV1-FLEX-Tgfbr2-V5, a Cre-dependent expression cassette, specifically to Tgfbr2;ScxCre mutant tendon cells. Transuterine microinjection of the viral vector into embryos was performed according to a published protocol (Jiang et al., 2013). Briefly, a laparotomy was performed on anesthetized pregnant females to expose the uterus. The left wrist field of the forelimb bud of each embryo was injected with ~2 µl of concentrated viral inoculum (3.8 × 10¹³ vg/ml) using a borosilicate glass capillary pipette (25–30 µm outer diameter and 20 degree bevel). The abdominal and skin incisions were closed with resorbable sutures. The dams were recovered overnight with supplementary heating and then returned to main colony housing.

For postnatal constitutive re-expression of Tgfbr2,~10 µl of AAV1-Tgfbr2-FLAG inoculum (4.1 × 10¹² vg/ml) was injected subcutaneously into the left forelimb of P1 pups using an 8 mm x 31G BD Ultra-Fine insulin syringe and needle (Becton Dickinson and Company, NJ). For both experiments, forelimbs from P5 to P7 mutant pups (n = 3 pups for each stage) were harvested, fixed, cryosectioned and examined for expression of tendon differentiation markers in infected tendon cells.

Tendons were collected and pooled from both forelimbs and hindlimbs as described above from two pups with the omission of tissue-scraping step. The enzymatically released cells were centrifuged, resuspended in α-MEM with 5% FBS and submitted to the OHSU Massively Parallel Sequencing Shared Resource (MPSSR) Core facility. scRNASeq analysis was then performed using the 10x Genomics Chromium Single Cell 3’ Reagent Kits and run on a Chromium Controller followed by sequencing using the Illumina NextSeq 500 Sequencing System (Mid Output), as per the manufacturer's instructions (10x Genomics Inc, CA; Illumina Inc, CA).

Sequencing data processing and downstream analysis were performed using Cell Ranger version 2.0 (10x Genomics, CA) (Zheng et al., 2017) with the default settings. Briefly, sequencing reads were aligned to the mm10 genome and demultiplexed and filtered using total UMI count per cell to generate the gene barcode matrix. Principle component analysis was performed and the first ten principle components were used for the t-distributed stochastic neighbor embedding (tSNE) dimensional reduction and clustering analysis. Cells were clustered using K-means clustering. For each cluster, genes with an average UMI count ≥0.5, fold change ≥1.5 and p-value<0.05 were identified as signature genes for each cluster. Gene Ontology (GO) enrichment analysis (clusterProfiler) (Yu et al., 2012) and the PANTHER Classification System (http://pantherdb.org/) were used to elucidate the biological process and signaling pathway associated with individual gene. Enriched canonical pathways were defined as significant if adjusted p-values were <0.05.

Unless stated otherwise, all graphs are presented as mean ± standard deviation (SD). Student’s t-tests were performed to determine the statistical significance of differences between groups (n ≥ 3). A value of p<0.05 is regarded as statistically significant.