Development and maintenance of tendons and ligaments

Tendons and ligaments are closely related dense fibrous connective tissues that play vital roles in musculoskeletal mobility and stability. Tendons connect muscle to bone and facilitate body movement by transmitting tensile forces and storing elastic energy, whereas ligaments join bone to bone, stabilizing joints and guiding movement through the normal range of motion (Benjamin et al., 2008; Connizzo et al., 2013; Murchison et al., 2007). The functions of tendons and ligaments rely on their strong yet flexible structure of collagen fibrils, which are hierarchically organized and bundled by connective tissue sheaths. Though tendon and ligament are similarly composed of collagen 1 (Col1) fibrils, proteoglycans, elastin and glycoproteins, a few differences do exist. Col1 fibrils, which are aligned parallel to one another along the longitudinal axis in tendons, are multidirectional and less densely packed in ligaments to better support tensile load (Amiel et al., 1983; Amis, 2004; Jung et al., 2009; Tozer and Duprez, 2005) (Fig. 1).

Individual tendons and ligaments have unique morphological properties adapted for specific roles. Remarkable variability exists in tendon and ligament shape, which can form cords, strap-like bands, flat ribbons, discs and fan-like structures to accommodate distinct mechanical and anatomical environments (Benjamin et al., 2008; Connizzo et al., 2013). Morphological diversity among tendons and ligaments is further enhanced through their variable association with synovial sheaths, bursae, fibrous retinacula and fat pads that function to lubricate, anchor, support and provide mechanosensory input (Benjamin et al., 2008). In addition, there is morphological variability along tendons, with fibrocartilage forming at bone insertions, sesamoid bones forming within regions that cross over joint surfaces, and myotendinous junctions where tendon unites with muscle (Fig. 1). Genome-wide expression analysis of porcine adult tendon and ligament tissue from distinct regions and subtypes suggests that morphological and functional differences largely correlate with variation in gene expression levels rather than activation of subtype-specific genes (Pearse et al., 2009). The extent to which embryonic factors are sufficient to establish location-specific morphology has not been directly investigated. Undoubtedly, morphological variation among tendon and ligament subtypes will be key to interpreting region-specific differences in development and repair.

Tendons and ligaments arise from distinct embryonic origins depending on their position along the rostro-caudal axis (Fig. 2). Tendons and ligaments in the head arise from cranial neural crest cells (CNCCs), whereas those in the neck and shoulder girdle are derived from lateral plate mesoderm, CNCCs and paraxial mesoderm (Durland et al., 2008; Evans and Noden, 2006; Köntges and Lumsden, 1996; Matsuoka et al., 2005; Nagashima et al., 2016; Valasek et al., 2010; Ziermann et al., 2018). Lineage tracing of CNCCs and mesoderm in the murine neck suggests that the embryonic origin of the tendon matches that of the bone to which it attaches (Heude et al., 2018). In the limb, tendons and ligaments originate from lateral plate mesoderm, whereas those in the trunk are derived from somites (Nassari et al., 2017). The extent to which the developmental origin of tendons and ligaments influences their mechanisms of formation and later properties is unclear.

Fibroblastic-like cells, often called tenocytes and ligamentocytes, are the principal mature cell types within tendons and ligaments, respectively. The fundamental nature of tendon and ligament cell identity and their heterogeneity have yet to be determined. A major challenge in defining these cell types is in their seemingly strong potential to revert to a progenitor-like state or transdifferentiate into osteoblasts or chondrocytes (Howell et al., 2017; Oldfield and Evans, 2003; Tan et al., 2020; Wren et al., 2000). As both tendon and ligament pathologies are commonly associated with aberrant bone formation, there is a need to understand the factors that maintain tendon and ligament cell fate (Magne and Bougault, 2015).

In this Review, we focus primarily on the better-understood mechanism of tendon development, indicating where key findings are also applicable to ligaments. We discuss region-specific developmental mechanisms in the head, trunk and limb that induce and maintain tendon progenitors and regulate tendon cell differentiation and maturation. We also discuss current evidence in the field that points to heterogeneity and regional specificity within tendon cell populations. Finally, we compare tendon progenitors in development to adult stem and progenitor cells that contribute to tendon and tendon attachment repair.

The discovery of scleraxis (Scx), a basic helix-loop-helix transcription factor, as a marker of both progenitors and differentiated cells within tendon and ligament has led to the creation of powerful genetic tools for understanding their development (Cserjesi et al., 1995; Pryce et al., 2007; Schweitzer et al., 2001). The Scx-GFP reporter and Scx-Cre alleles in mice are routinely employed in mechanistic studies that address tendon progenitor patterning, differentiation and maintenance (Blitz et al., 2013; Pryce et al., 2007; Yoshimoto et al., 2017). Similarly, the Scx knockout mouse has underscored its necessity in tendon morphogenesis (Murchison et al., 2007; Roberts et al., 2019). In zebrafish, nonsense mutations in scxa and scxb lead to defects in cranial tendon and ligament morphogenesis, showing a conserved role for Scx in differentiation and maturation of tendons across vertebrates (Kague et al., 2019). Molecular studies indicate that Scx promotes tendon cell differentiation and maturation by transcriptionally activating a suite of tendon cell genes, including the major building blocks of the tendon extracellular matrix (ECM), such as Col1a1 and tenomodulin (Tnmd) (Leéjard et al., 2007; Shukunami et al., 2018).

Scx is expressed in all tendon and ligament progenitors, yet its requirement during development varies depending on the subtype and location of the tendons. Loss of Scx in mice preferentially disrupts formation of long-range, but not short-range, muscle-anchoring of tendons and ligaments in the limb and trunk (Murchison et al., 2007). This confounding aspect of the Scx knockout phenotype has been addressed in a recent study that more closely examined long-range tendon development. Development of both long-range and short-range tendons begins with formation of a tendon anlage that anchors to muscle and bone. In long-range tendons, this anlage subsequently undergoes rapid elongation that is fueled by recruitment of surrounding Scx⁺ progenitors into the tenogenic lineage. Scx, although not required for formation of the initial tendon anlagen, is necessary to recruit tendon progenitors for substantial extension during growth (Huang et al., 2019). Nevertheless, Scx remains the earliest known marker of tendon and ligament progenitors, it does not appear to be necessary for tendon cell fate determination.

A central pathway key to establishment of tendon progenitors is transforming growth factor β (Tgfβ) signaling. Tgfβ2/3 ligands emanating from neighboring cartilage/muscle recruit undifferentiated mesenchymal cells to form tendon progenitors, and then maintain them by activating Scx expression in both mouse and chick (Havis et al., 2014; Pryce et al., 2009) (Fig. 3). Loss of Tgfb2/3 in mice leads to diminished Scx expression, considerable reduction in tendon progenitor numbers, and eventual widespread loss of cranial, limb and trunk tendons (Anthwal et al., 2008; Pryce et al., 2009). In addition, conditional loss of Tgfbr2 in mouse limb mesenchyme (via Prx1-Cre) results in depletion of limb tendons (Pryce et al., 2009). The dramatic reduction of tendon and ligament progenitors in mice with mutations in the Tgfβ signaling pathway is the result of lost cellular identity, either through de-differentiation or acquisition of different cell fates (Pryce et al., 2009).

The distinct embryonic origins of cranial, limb and trunk tendons suggests the possibility that the establishment and differentiation of each population is uniquely regulated (Fig. 2). Although this has been difficult to address given the paucity of information on cranial tendon progenitors relative to those of the limb and trunk, growing evidence supports this idea. For example, fibroblast growth factor (Fgf) signaling from the adjacent myotome promotes the induction of axial tendon progenitors by driving Scx expression in mice and chicks (Brent et al., 2005, 2003; Brent and Tabin, 2004) (Fig. 2). Conversely, induction of cranial and distal limb (autopod) tendon progenitors is muscle independent, with muscle interaction only required later for maintenance of cellular identity (Bonnin et al., 2005; Chen and Galloway, 2014; Edom-Vovard and Duprez, 2004; Grenier et al., 2009; Grifone et al., 2008). In addition, interactions with cartilage are vital to the organization (and in some cases establishment) of tendon and ligament progenitors of the craniofacial and axial skeleton (Chen and Galloway, 2014).

Current research suggests that the relationship between Fgf signaling and Scx expression varies based on species and anatomical region. We have recently found that Fgfr2 promotes establishment of Scx⁺ tendon progenitor cells at tendon-bone insertion sites in the jaw (Roberts et al., 2019) (Fig. 2). Exogenous treatments that activate or block Fgf ligands, or its downstream effector Erk (Mapk1), in limb explants suggest that Fgf signaling promotes Scx expression in chick and inhibits Scx expression in mouse (Edom-Vovard et al., 2002; Havis et al., 2016, 2014). Conversely in zebrafish embryos, inhibition of Fgf signaling leads to loss of all Scx⁺ progenitors contributing to both tendon and ligament (Chen and Galloway, 2014). The reason for this species-specific difference is unclear, but the outcome of Fgf signaling on Scx expression may be related to the intensity and duration of the Fgf-Erk signal (Brent and Tabin, 2004) or, alternatively, differences in experimental approaches between species. It is also not yet clear what the specific role of Fgf signaling is in ligament, although the common embryonic origins and function of tendon and ligament would suggest that findings specific to tendons in various anatomical regions could be extrapolated to nearby ligaments. The limited available literature suggests that Fgf signaling enhances proliferation and differentiation in periodontal ligament cells in vitro by upregulating Scx and Tnmd while suppressing osteogenic differentiation through inhibition of Runx2 expression (An et al., 2015; Hyun et al., 2017). Defining a clearer role for Fgf signaling in tendon and ligament induction will require further analysis of Fgf-related mutant models.

Studies in the limb provide evidence that tendon progenitors of the same embryonic origin can have distinct regulatory programs depending on their location along the dorsal-ventral and proximal-distal axes. Whole-transcriptome expression profiling of limb-derived Scx-GFP⁺ cells during the early stages of development has identified Foxf2 and Six2 as tendon-enriched transcription factors with region-specific expression (Liu et al., 2015). Foxf2 expression is restricted to the dorsal tendons within the distal limb (autopod), whereas Six2 expression is enriched in ventral tendons of both the proximal (zeugopod) and distal (autopod) limb (Fig. 2). However, it is not clear whether this difference in expression is specific to Scx-expressing tendon and ligament populations (Liu et al., 2015). In the proximal limb (zeugopod), differentiation and growth of tendons is dependent on the muscle (Bonnin et al., 2005; Huang et al., 2015; Kardon, 1998; Schweitzer et al., 2001) (Fig. 2). In the autopod, however, the induction, differentiation and maintenance of tendons, progenitors of which are distinctly derived from a Six2 lineage, is instead dependent on the cartilage that prefigures the bone (Huang et al., 2015) (Fig. 2). The conclusion that development of autopod tendons is dependent on skeletal cues is underscored by the finding that autopod tendons are induced in supernumerary digits and fail to form in the absence of cartilage (Huang et al., 2015). Whether cartilage promotes tendon formation by secreting tenogenic factors or providing a source of progenitors with tenogenic potential is unknown; however, candidate signaling pathways for regulation of this process include Tgfβ, Fgf and Wnt (Pryce et al., 2009; Roberts et al., 2019; Yamamoto-Shiraishi and Kuroiwa, 2013).

Once tendon cells are specified, they function primarily in producing the tendon's ECM and establishing the patterning of collagen fibrils and their bundles (Birk et al., 1996) (Fig. 3). The most abundant structural protein in tendon is Col1, which comprises about 95% of the total collagen content. Expression of Col1a1 in tendon cells is regulated by multiple factors, including Scx, which directly binds to E-box sites within the Col1a1 promoter (Leéjard et al., 2007). Other transcription factors that are known to have functional roles in tendon development, such as mohawk (Mkx) and Egr1/2, are primarily required for tendon maturation, including matrix deposition and organization (Ito et al., 2010; Lejard et al., 2011; Liu et al., 2010) (Fig. 3). For example, loss of Mkx decreases Col1a1 expression and impairs collagen fibril formation, reducing tendon mass, but has no discernible effect on ligament development (Ito et al., 2010; Liu et al., 2010). Though expression of Mkx and Egr1/2 has been documented in ligament fibroblasts, it is not yet clear what their roles in ligament cell differentiation and maturation may be (Liu et al., 2010; Tamura et al., 2008).

As tendons mature, their respective fibroblasts arrange in longitudinal rows along the collagen fibrils, flatten and become less dense. Maturing tendon cells develop a complex network of cellular projections, including nanotubes that facilitate intercellular communications via gap junctions (Benjamin et al., 2008; Egerbacher et al., 2020; Kato et al., 2013). In zebrafish, the force of muscle contraction promotes release of Tgfβ, which is required for tendon progenitors to extend microtubule-rich projections (Subramanian et al., 2018). These projections, in turn, promote tendon cell maturation by regulating force-dependent expression of tendon- and ligament-specific ECM components, such as Thrombospondin 4b (Thbs4b) (Jelinsky et al., 2010; Subramanian et al., 2018; Subramanian and Schilling, 2014).

In the mature tendon, tenocytes require continuous feedback to maintain their differentiated state and repress metaplasia into the osteochondral lineage. Feedback mechanisms that maintain tendon cell fate (and have been loosely implicated in ligament fibroblast maintenance) are discussed below, including growth factors, ECM components and mechanical force (Fig. 3). Tendon cell projections are also essential for cell fate maintenance, as their loss leads to the upregulation of scxa in zebrafish, suggesting a reversion to a de-differentiated state (Subramanian et al., 2018).

Tgfβ signaling is the primary growth factor pathway known to be involved in maintenance of tendon cell fate (Fig. 3). Targeted deletion of Tgfbr2 in tendon and ligament cells using Scx-Cre does not appear to affect cell fate during embryonic development. However, tendon and ligament cells in early postnatal conditional knockout mice lose expression of differentiation markers and reactivate genes associated with tendon progenitors at the earliest stages of embryonic tendon induction (Havis et al., 2014; Tan et al., 2020). The role of Tgfβr2 signaling is cell-autonomous in tendon cells, as targeted re-expression of Tgfbr2 in de-differentiated tendon cells is sufficient to reactivate differentiation. Together, this suggests that tendon cell fate is reversible and that tendon cell identity requires extrinsic growth factor signals for continuous maintenance.

Proteoglycans within the tendon ECM control the bioavailability of growth factors, such as Tgfβ/Bmp, and play important roles in maintaining tendon cell fate. The metalloproteinases Adamts7 and Adamts12 are co-expressed in mouse hindlimb tendons starting in the late embryonic period and persist through adulthood to maintain tendon cell fate (Mead et al., 2018). Combined inactivation of Adamts7 and Adamts12 alters collagen fibrillogenesis and leads to progressive heterotopic ossification (HO) in juvenile mouse tendons (Mead et al., 2018). Prior to HO, tendon cells exhibit enhanced Bmp signaling and upregulate cartilage and bone markers. This suggests that Adamts7 and Adamts12 maintain the tendon cell fate by suppressing osteochondral lineage differentiation. Similarly, the small leucine-rich proteoglycans biglycan (Bgn) and fibromodulin (Fmod) promote tendon cell differentiation by repressing Bmp signaling, and their combined loss results in HO within the hindlimb tendons (Ameye et al., 2002; Bi et al., 2007). In a model for tendon HO caused by hyperactive Bmp signaling, ectopic bone is derived from Scx-lineage cells that aberrantly activate bone and cartilage markers (Agarwal et al., 2017). These studies show that tendon ECM preserves the tendon cell fate, in part by providing factors that inhibit bone formation, and strongly supports a role for ECM proteins in the maintenance of tendon cell fate.

Mechanical signals are also necessary to maintain tendon cell fate during development and homeostasis (Arvind and Huang, 2017). Tendon cells respond to physical input in part through activation of the mechanoresponsive transcription factors Egr1/2 and Mkx. Increased load activates Egr1 and Mkx expression, and decreased load reduces their expression (Gaut et al., 2016; Kayama et al., 2016; Maeda et al., 2011). Downstream of mechanical signals, Egr1/2 and Mkx promote tendon cell fate through distinct mechanisms. Egr1 is sufficient to activate gene expression of Scx and Col1a1 during development, and to promote tendon cell fate during postnatal growth (Guerquin et al., 2013; Lejard et al., 2011). In postnatal tendons, Egr1 is enriched at the Tgfb2 promoter and the absence of Egr1 reduces Tgfb2 expression (Guerquin et al., 2013). This suggests that, in response to mechanical signals, Egr1 acts upstream of Tgfβ signaling to maintain tendon cell identity (Havis and Duprez, 2020). Although all of the aforementioned factors have also been found to be expressed in ligament, their roles in ligament cell maintenance remain unclear. However, a wealth of evidence supports the importance of mechanical signaling in controlling ligament cell fate, in part through regulation of Rho/Rock signaling in the ECM microenvironment (Meng et al., 2015; Yamamoto et al., 2018).

In Mkx knockout mice, mechanical stimuli fail to maintain tenogenic gene expression and hindlimb tendons exhibit HO (Kayama et al., 2016; Liu et al., 2019). These ossified sites are derived from aberrant differentiation of Scx-lineage cells into chondrocytes that then undergo endochondral-like bone formation similar to that observed in Adamts7, Adamts12, Bgn and Fmod mutant models. Selective inactivation of Mkx in postnatal Scx⁺ tendon cells recapitulates the HO phenotype (Liu et al., 2019). Although tendon HO is associated with increased Bmp signaling in the mouse models listed above, Hedgehog (Hh) signaling plays a crucial role in tendon ossification in Mkx mutant mice. Interestingly, many cells at the periphery of ossifying tendon nodules express both Scx-GFP and Sox9, similar to developing sesamoid bones and bone eminences (Blitz et al., 2013; Eyal et al., 2015, 2019) (Fig. 4). A chromatin immunoprecipitation assay for Mkx in rat tendon-derived cells identified enrichment of both tendon- and cartilage-related genes, suggesting a dual role for Mkx in transcriptional promotion of tendon cell differentiation, while preventing chondrogenic/osteogenic differentiation (Suzuki et al., 2016). Although these studies did not directly address ligaments, the mechanisms and signals are assumed to be recapitulated, as many human diseases featuring HO affect both types of tissue (Zhang et al., 2020).

Tendons and ligaments attach to bone through a specialized tissue called the enthesis, which functions to dissipate mechanical stress across the hard-soft tissue interface (Fig. 1). There are two major types of entheses: fibrous entheses, whereby tendons and ligaments attach or insert directly to the bone or periosteum, and fibrocartilaginous entheses, which are zonally organized into tendon or ligament, fibrocartilage, mineralized fibrocartilage and bone (Benjamin et al., 2006) (Fig. 4). In fibrocartilaginous entheses, this gradient in cell types, ECM and minerals is essential for the transmission of force during musculoskeletal movement (Lu and Thomopoulos, 2013). The intermediate properties of this tissue arise in development from attachment progenitors (APs), which are established at the tendon-bone interface through a little-understood mechanism requiring Tgfβ signaling (Blitz et al., 2013) (Fig. 4). Tgfbr2 is expressed in APs and its loss in the jaw (condyle and angular process) or limb (deltoid tuberosity) results in loss of these key tendon-bone attachment sites in mice (Anthwal et al., 2008; Blitz et al., 2013; Seo and Serra, 2007). Though the structure and function of tendon and ligament fibrocartilaginous entheses appear to be the same, it is unclear whether developmental mechanisms for enthesis establishment are conserved between these two types. However, it has been shown that Tgfβ signaling plays a crucial role in regulating bone resorption following traumatic injury to both tendon and ligament entheses, pointing to at least some conservation of enthesis development between the two tissue types (Doschak et al., 2005; Tang et al., 2009; Zhen et al., 2013).

Comparative transcriptomic analysis together with mouse genetics has identified that, in the appendicular skeleton, there are global and regional regulators that determine the correct location of APs along the bone's length. Spatial organization of APs throughout the limb requires Gli3, whereas proximal versus distal patterning of APs is regulated by Pbx1/Pbx2 and Hoxa11/Hoxd11, respectively (Eyal et al., 2019) (Fig. 2). Disruption in these global and regional regulators of AP patterning leads to mislocalization of the bone eminence attachment sites of both tendon and ligament. Global and regional regulators working together to fine-tune AP patterning demonstrates modularity in enthesis development.

APs co-express lineage commitment factors for tendon (Scx) and cartilage (Sox9) and have the potential to adopt either cell fate depending on their location along the tendon-bone axis (Fig. 4). During enthesis formation, APs near the tendon form fibroblasts of the tendon terminus, whereas those near the bone form chondrocytes, some of which then ossify into the bone eminence (Sugimoto et al., 2013) (Fig. 4). Loss of Scx or Sox9 disrupts tendon-bone attachment development throughout the skeleton (Blitz et al., 2013; Killian and Thomopoulos, 2016; Roberts et al., 2019; Sugimoto et al., 2013). Genome-wide transcriptomic and chromatin accessibility assays of the limb attachment cells early in their differentiation have shown that APs, as a population, are marked by a hybrid fibroblast-chondrocyte transcriptional profile with active enhancers shared between tenocytes or ligamentocytes and chondrocytes (Kult et al., 2021). This suggests that the graded structure of the enthesis is derived from attachment cells with a hybrid tendon-cartilage or ligament-cartilage cell fate (Fig. 4); however, transcriptional analysis at the single-cell level will be required to rule out the possibility that the graded structure results from intermingling of attachment cells of discrete types.

In recent years, several regulators of AP differentiation have been identified. The Krüppel-like transcription factors Klf2 and Klf4, which bind the proximal regulatory regions of genes expressed in attachment cells, such as Gli1 and Col5a1, are required for tendon AP differentiation in the limb (Kult et al., 2021). In periodontal ligaments, Klf5 has been implicated as a key regulator of progenitor cell proliferation and osteogenic differentiation via activation of both Fgf and Wnt signaling pathways (Gao et al., 2018). Whereas Klf family genes regulate the overall capacity of APs to differentiate, Bmp and Fgf signaling regulate the cell fate choice of APs (Fig. 4). In the limb, Bmp4 derived from the tendon tip induces APs to differentiate into chondrocytes, some of which then undergo endochondral-like ossification to form the bone eminence (Blitz et al., 2009). Conditional inactivation of Bmp4 using Scx-Cre blocks formation of the cartilage anlage prefiguring the bone eminence, despite normal tendon differentiation. Bmp2 and Bmp4 are also necessary for the endochondral ossification of APs into tendon-embedded sesamoid bones (Eyal et al., 2015). Although it is understood that Bmp promotes AP differentiation into chondrocytes, it is not clear what blocks Bmp-induced chondrogenesis in APs that differentiate into tendon cells.

During a similar time frame, APs also express the Bmp family member Gdf5. Gdf5 lineage tracing marks APs that contribute to linear growth of the enthesis and give rise to fibrocartilage (Dyment et al., 2015, 2014). The relationship between this Gdf5 lineage and APs that co-express Scx and Sox9 remains unresolved. However, in the developing joint Sox9⁺ cells give rise to Gdf5⁺ joint interzone progenitors that contribute to various joint tissues, including the menisci, cruciate ligaments, articular cartilage and intra-articular ligaments (Harada et al., 2007; Shwartz et al., 2016). It has been suggested that the onset and duration of Gdf5 signaling may instruct lineage divergence in the joint, and a similar mechanism may also occur in the enthesis. The one pathway currently known to promote tendon cell fate in APs is Fgf signaling (Fig. 4). We recently showed that perichondral Fgfr2 signaling, activated by Fgf2 from the tendon, promotes tendon cell differentiation in cranial APs by regulating Notch signaling (Roberts et al., 2019). Upon CNCC-specific deletion of Fgfr2, cranial APs lose Dll1-Notch2 signaling and undergo biased differentiation into chondrocytes over tenocytes. This indicates that Fgf signaling plays a key role in promoting the differentiation of tendon at the expense of cartilage in the face.

Hh signaling also plays an important role in the cellular complexity and maturation of tendon-bone and ligament-bone attachment cells. Sonic hedgehog (Shh) induces Gli1 expression in a subpopulation of newly differentiated attachment cells, with Gli1 expression maintained postnatally by Indian hedgehog (Ihh). These Gli1-expressing cells give rise to the mineralized fibrocartilage in the mature enthesis (Dyment et al., 2015; Felsenthal et al., 2018; Liu et al., 2013; Schwartz et al., 2015) (Fig. 4). Mouse genetics show that Hedgehog signaling is necessary for fibrocartilage mineralization: loss of the Hedgehog transmembrane effector smoothened using Scx-Cre greatly reduces the number of Gli1⁺ cells and inhibits their subsequent mineralization (Dyment et al., 2015; Liu et al., 2013; Schwartz et al., 2015). In addition, ablation of Gli1⁺ cells early in postnatal development results in loss of mineralized fibrocartilage (Schwartz et al., 2015). The lineage relationship between postnatal Gli1⁺ cells and embryonic APs that precede them differs depending on whether the enthesis remains stationary or migrates during bone growth. In stationary entheses, which are found at bone ends, Gli1⁺ cells are derived from APs; however, in migrating entheses found along the bone shaft, a separate population of Gli1⁺ cells replace APs (Felsenthal et al., 2018). The origin of Gli1⁺ cells in migrating entheses remains unclear, but the periosteum seems a likely source.

The structure and mechanical properties of the myotendinous junction (MTJ) rely on interaction between tendon and muscle via the ECM (Fig. 1). MTJ connections are unique to tendons because ligament, by definition, facilitates the joining of bones exclusively. During MTJ formation in Drosophila, myoblasts migrate to sites of integration and extend actin-based, filopodia-like structures to recognize tendon cells via thrombospondin-mediated cell adhesion (Liu and Geisbrecht, 2012). Interactions between Drosophila tendon and muscle are mediated by cell-cell contact through integrin heterodimers, as well as various ECM proteins that can be classified as force transmitting and structural collagens, scaffolding proteins or crosslinking proteins. Key ECM proteins that make up the MTJ include laminin, fibronectin, thrombospondins and collagens (Frolova et al., 2014; Subramanian and Schilling, 2015). Recent RNAi screens have expanded the predicted list of genes contributing to MTJ formation and highlight Tango1 and Fascin as key regulators of collagen deposition and actin bundling in filopodia, respectively (Camuglia et al., 2018; Tiwari et al., 2015).

It remains unclear to what extent insights made in Drosophila can be applied to vertebrate MTJ formation. In vertebrates, the migratory nature of cells contributing to the MTJ is known to differ based on both species and anatomical region. Avian hindlimb myoblasts elongate and adhere to tenocytes already present at attachment sites, yet in the avian cranium it is the tenocytes that must travel to meet the established muscle (Grenier et al., 2009; Kardon, 1998). In the zebrafish, cranial tenocytes appear to prefigure regions of future muscle attachment, with altered tenocyte migration in cyp26b1 mutants resulting in secondary mislocalization of muscle attachments (McGurk et al., 2017). It is unknown whether differences in MTJ patterning across regions and species are due to differing embryonic origins of tendon and muscle or other, potentially species-specific, factors.

During MTJ development in vertebrates, a careful balance between ECM deposition and remodeling is maintained in response to mechanical signals. This balance is achieved through reciprocal interactions between structural proteins, such as laminin and collagens, and matrix metalloproteinases (Brown, 2000; Matchett et al., 2019; Zhang et al., 2014). As previously discussed, both molecular and mechanical factors originating from the muscle play a vital role in inducing tendon progenitors in the axial skeleton and promoting tendon cell maturation in the limb and craniofacial complex. Mechanical signals also affect tendon cell morphogenesis at the MTJ through a Tgfβ-driven feedback loop that regulates the growth and branching of tenocyte projections (Subramanian et al., 2018). Studies in the zebrafish axial skeleton suggest that interactions of tendon and muscle are mediated by long, microtubule-rich cell projections originating from tenocytes that extend into the intersomitic space during early development of the MTJ (Ma et al., 2018; Subramanian et al., 2018). These projections likely increase the adhesion of tendon and muscle at the MTJ and provide support and structure to the muscle, a hypothesis supported by the phenotypes observed in dystroglycan1 (dag1) mutants and col22a1 knockdown zebrafish, which model muscular dystrophy and feature increased muscle detachment (Charvet et al., 2013; Ma et al., 2018). Loss of function of transmembrane protein 2 (tmem2; also known as cemip2) in zebrafish also results in muscle fiber detachment, coupled with disorganization of key MTJ ECM proteins, such as laminin and fibronectin, indicating its important role in the regulation of muscle cell-ECM interactions that facilitate muscle-tendon attachment (Ryckebüsch et al., 2016). Transcription factors Tbx4 and Tbx5 originating in muscle connective tissue have also been implicated in the patterning of tendon and muscle in the mouse limb through modulation of N-cadherin (cadherin 2) and β-catenin expression (Hasson et al., 2010).

Tendons, ligaments and their entheses are commonly injured and, in adults, heal through the formation of a persistent fibrovascular scar that does not recapitulate the developmental process or mechanical properties of the native tissue (Galatz et al., 2015). Injured tendons and entheses heal through phases of inflammation, proliferation and remodeling (Thomopoulos et al., 2015). During inflammation, the release of growth factors and cytokines initiates vascularization and recruits immune cells and activates resident stem and progenitor cells. In the proliferative stage, resident stem and progenitor cells expand and synthesize a granulation tissue rich in fibroblasts, capillaries and an ECM network of Col3 (Juneja, 2013; Snedeker and Foolen, 2017). The granulation tissue then undergoes remodeling to form scar tissue that has reduced cellularity and vascularity as well as increased Col1 (Juneja, 2013; Snedeker and Foolen, 2017). As reparative strategies for tendon, ligament and enthesis, which are largely limited to surgical repair, fail to recreate the structural and functional characteristics of the native tissue that forms during development, there is a need to identify and harness the regenerative potential of endogenous stem and progenitor cells.

As discussed, tendon and ligament development appears to share key morphological and functional characteristics. In fact, a frequently used surgical intervention to repair torn or injured ligaments is tendon auto- or allografting (Kan et al., 2016; Macaulay et al., 2012). For purposes of simplicity, we focus here on the larger body of current research pertaining to tendon healing with the hope that findings in this field can be extrapolated to improve understanding of ligament healing. Tendon stem and progenitor cells (TSPCs) were first identified through in vitro characterization of cells disassociated from the tendon proper (Bi et al., 2007). These cells were identified as expressing surface markers commonly associated with other stem cells, such as Sca1 (also known as Ly6a) and CD44, as well as tenogenic factors such as Scx. Tissue-isolated TSPCs also exhibit characteristic traits of stem cells, such as clonal serial passaging and multilineage potential in vitro, and give rise to tendon-like tissue upon engraftment. More recently, genetic lineage tracing after tendon injury has begun to characterize potential TSPCs in vivo. Emerging studies reveal that putative in vivo TSPCs that contribute to tendon repair following injury can arise from both the tendon proper and sheath, and that age is a major factor for whether healing is regenerative or fibrotic (Fig. 5).

In neonatal mice, transection injury sustained to the Achilles tendon results in the recruitment of resident Scx-lineage tenocytes, which promote regenerative healing through new tendon formation (Howell et al., 2017). In adult mice, however, the same injury induces fibrotic healing whereby Scx-lineage cells remain quiescent and an extrinsic population of αSMA⁺ cells contribute to a permanent scar (Fig. 5). Similarly, following full thickness transection injury in the mouse patellar and supraspinatus tendons, αSMA-lineage cells are recruited, activate expression of Scx-GFP and become the main contributors to the healing response (Dyment et al., 2014; Yoshida et al., 2016). The αSMA⁺ cells that contribute to adult tendon repair are heterogeneous, arising from a subpopulation of resident tenocytes marked by S100a4, as well as from extrinsic progenitors originating from the tendon sheath (Ackerman et al., 2017, 2019; Best and Loiselle, 2019; Dyment et al., 2014, 2013). Since the αSMA-lineage contributes to multiple cell types within the tendon, only some of which turn on tenogenic genes during repair, αSMA is not a specific marker for TSPCs (Dyment et al., 2014; Harvey et al., 2019) (Fig. 5). Inducible Cre lines that more specifically trace lineages of the tendon sheath support the idea that the sheath may be a source of TSPCs during repair. An inducible Cre line driven by tubulin polymerization-promoting protein family member 3 (Tppp3), a developmental marker of tendon sheath cells, marks a population of Sca1⁺/Pdgfra⁺ cells that are recruited to the site of biopsy punch-injury of the patellar tendon where they activate Scx expression and give rise to tenocytes in a Pdgf-dependent manner (Harvey et al., 2019; Staverosky et al., 2009) (Fig. 5). Interestingly, Tppp3 expression overlaps with that of Bglap (osteocalcin), a marker typical of osteoblasts, in the sheath of some tendons. In these tendons, Bglap-lineage cells from the tendon sheath differentiate and contribute to the population of tenocytes activated during tendon repair (Wang et al., 2017). As Bglap and Tppp3 do not overlap in all tendons, there are likely regional differences in TSPC populations (Fig. 5). Alternatively, heterogeneity in TSPCs might also indicate different states of commitment along the differentiation pathway.

Evidence that TSPC populations differ in their capacity for repair comes from in vitro studies comparing the relative healing potential of putative TSPCs from the tendon proper versus the sheath. Tendon proper-derived TSPCs express higher levels of tenogenic markers and have greater potential to form tendon-like tissue, whereas sheath-derived TSPCs express pericyte-like markers, proliferate faster and have increased potential for myofibroblast differentiation (Cadby et al., 2014; Mienaltowski et al., 2013, 2014). Single-cell RNA-seq analysis is beginning to reveal the heterogeneity of cells within the tendon proper and tendon sheath; however, whether this is an output of diversity in TSPCs is unclear (De Micheli et al., 2020; Yin et al., 2016). Taken together, these studies suggest that TSPCs are heterogeneous depending on their anatomical location, tissue source (tendon proper versus sheath) and degree of lineage commitment.

Similar to tendons and ligaments, entheses heal via the contribution of multiple cell lineages. The putative enthesis stem and progenitor cells (ESPCs), primarily characterized by lineage tracing in repair models, appear to be distinct from TSPCs. For example, αSMA-lineage cells do not contribute to enthesis healing (Dyment et al., 2014; Moser et al., 2018) (Fig. 5). Evidence suggests that the reparative potential of intrinsic enthesis cells is age dependent. In adult mice, Scx-, Sox9- and Gli1-lineage cells, which mark tendon, ligament and fibrocartilage of the enthesis, show limited involvement during mature enthesis healing (Moser et al., 2018; Schwartz et al., 2017). In contrast, in juvenile mice following needle-punch injury, Gli1-lineage cells proliferate and contribute to immature enthesis healing through a regenerative process that more closely mimics normal development (Schwartz et al., 2017) (Fig. 5). This suggests that the adult enthesis healing process could potentially be enhanced by identifying methods to mimic the developmental process.

Although cells intrinsic to the adult enthesis have poor capacity for repair, recent studies in mice suggest that extrinsic cells from the adjacent bone and tendon are a source of ESPCs during enthesis healing. Following enthesis reconstruction surgery, in which tendon grafts are inserted into bone tunnels, Gdf5-lineage cells from the underlying bone proliferate, infiltrate the tendon graft, and give rise to a zonal tendon-bone attachment (Hagiwara et al., 2020) (Fig. 5). Following surgical repair of detached tendons, Axin2-lineage cells from the tendon proper also contribute to healing, although they promote formation of a fibrovascular scar rather than recapitulating the graded structure of the native attachment (Moser et al., 2018).

In this Review, we have discussed a number of mechanisms that regulate tendon and ligament cell fate in the context of both development and repair. Recent advances in our understanding of these mechanisms have opened up a number of interesting questions for future investigation. For example, mechanisms regulating tendon and ligament cell fate in development feature regional differences in the necessity of transcription factors and dependence on inductive signals from muscle versus cartilage. However, it is not yet known how regional and universal developmental programs are integrated to organize the musculoskeletal system, or the extent to which developmental insights learned in one anatomical region can be applied to others. Additionally, tendon cell identity shows evidence of plasticity and reversibility, but it is not yet fully understood how or what transcriptional factors can function to actively induce tendon or ligament cell fate while inhibiting osteoblast and chondrocyte cell fate. There is also a high likelihood that key members of these transcriptional networks have not yet been identified. Additional information is needed to determine the similarities and differences between tendon and ligament cells in a developmental, functional and reparative capacity. For example, can a close study of joint development reveal regulators that are unique to ligament development (Salva and Merrill, 2017; Smeeton et al., 2017)? To what extent are the factors regulating TSPCs conserved between anatomical regions and between organisms with differing regenerative potential? Can TSPCs harvested in one location be relied upon for tendon repair in another location? Answers to these questions will help determine whether repair and tissue engineering approaches need to be designed with regional specificity in mind.

Understanding tendon cell identity using single-cell sequencing technologies that allow for RNA expression profiling and chromatin accessibility will be a vital first step to begin to address many of the questions proposed here about tendon cell fate. For example, single-cell technologies will be able to determine whether the graded connective tissue within tendon-bone attachments arises from intermingling of canonical cell types (tenocyte, osteoblast, chondrocyte), hybrid cell types that feature mixed properties, or a combination of both. Single-cell technologies will allow for a better resolution of the similarities and differences between tendon and ligament cells, which will in turn affect our clinical approaches to injury and repair. A better understanding of the molecular relationship between these cell types would allow for in vivo studies that advance our knowledge of the mechanisms regulating development and repair of ligament and ligament-bone attachments.

Studies in model organisms such as mouse and zebrafish have significantly advanced our understanding of tendon cell differentiation and maintenance, yet we lack a complete picture of how tendon cell fate is regulated. By focusing research on human disorders that feature tendon and ligament defects, such as joint laxity, dislocations and contractures, it may be possible to identify key genes and mechanisms regulating the developmental process. Examples of such disorders include geleophysic dysplasia, which features joint contractures and shortened limbs due to tendon defects. Recent work has identified ADAMTSL2 as the likely affected gene, which had not traditionally been studied as a regulator of tendon and ligament cell identity (Hubmacher et al., 2019; Piccolo et al., 2019). Additionally, disorders such as progressive osseous heteroplasia (POH) (Kaplan et al., 1994) and McCune–Albright syndrome (Chapurlat and Orcel, 2008) feature metaplasia of bone and fibrous connective tissue caused by loss- and gain-of-function mutations in GNAS, respectively. The opposing phenotypes of these disorders (ossification of fibrous tissue in POH and fibrous metaplasia of bone in McCune–Albright syndrome) indicate a key role of GNAS in tendon cell identity and maintenance that is not yet fully understood (Shore et al., 2002; Weinstein et al., 1991). Mechanistic studies of congenital disorders with tendon and ligament defects such as these are likely to reveal new insights into tendon and ligament cell identity and development.