Scx⁺/Sox9⁺ progenitors contribute to the establishment of the junction between cartilage and tendon/ligament

In vertebrates, coordinated body movement is ensured by a close functional and physical association of bones, muscles, tendons and ligaments. Tendons connect muscles to the skeletal components and function as force transmitters, whereas ligaments bind bones together to stabilize joints (Benjamin and Ralphs, 2000; Rumian et al., 2007). Cells in tendons and ligaments are categorized as special types of fibroblasts known as tenocytes and ligamentocytes (Benjamin and Ralphs, 2000). Unlike randomly distributed fibroblasts in loose connective tissues, tenocytes and ligamentocytes in dense connective tissues are highly organized and align in rows between parallel thick fibers, mainly consisting of type I collagen, that provide the major resistance to tensile forces (Amiel et al., 1984; Canty et al., 2004). By inserting dense regular type I collagen fibers into muscle from the myotendinous junction and into bone from the osteo-tendinous/ligamentous junction, which is termed the enthesis, tendons and ligaments integrate each musculoskeletal component into a single functional unit (Benjamin and Ralphs, 1998; Benjamin and Ralphs, 2000). To achieve this integration, progenitors for these cells need to be coordinately distributed at both sides of the junction, and then execute each differentiation program there. However, it is still unclear how the coordinated connection of the musculoskeletal components is established by tendon/ligament progenitors during development.

Progenitors for tendons, ligaments, cartilage and bone arise from the sclerotome, the lateral plate mesoderm and the neural crest (Akiyama et al., 2005; Christ et al., 2004; Mori-Akiyama et al., 2003; Smith et al., 2005), whereas myogenic progenitors are derived from the myotome (Brent and Tabin, 2002). During the early stages of musculoskeletal development, these progenitor populations migrate and settle in the prospective region to give rise to cartilage, muscle, tendon and ligament primordia (Kardon, 1998). Each primordium for the musculoskeletal component initially develops as an individual unit, but subsequently they integrate with each other by an unknown mechanism.

Sox9, an SRY-related transcription factor that contains a high-mobility-group box DNA-binding domain, is an important regulator of cartilage formation. In Sox9-deficient chimeric embryos generated by the injection of Sox9^−/− embryonic stem cells into Sox9^+/+ blastocysts, Sox9^−/− cells are eliminated from cartilaginous primordia and are instead incorporated into the surrounding connective tissues (Bi et al., 1999). Conditional inactivation studies of Sox9 using Prx1Cre or Col2a1Cre mice have revealed that Sox9 is required for multiple steps of chondrogenic differentiation before and after cartilaginous condensation (Akiyama et al., 2002). In the tendon and ligament cell lineage, scleraxis (Scx), a basic helix-loop-helix transcription factor, is persistently expressed throughout differentiation (Pryce et al., 2007; Schweitzer et al., 2001). In Scx^−/− mice, the intermuscular and force-transmitting tendons in the limbs and the tail tendons become hypoplastic, although the short appendicular anchoring tendons and ligaments are not significantly affected (Murchison et al., 2007). Such differential dependence on Scx expression suggests that tendons consist of distinct cell populations that have thus far not been defined.

At the early stages of musculoskeletal development, both Sox9 and Scx are detected in the subpopulation of tendon/ligament progenitors and chondroprogenitors (Akiyama et al., 2005; Brent et al., 2005; Sugimoto et al., 2013). Sox9 is upregulated during chondrogenesis (Zhao et al., 1997), whereas its expression is downregulated in association with the formation of the cruciate ligaments of the knee joint, the Achilles tendon and patella tendon (Soeda et al., 2010). Scx expression in the cartilaginous primordia is transient during chondrogenesis (Cserjesi et al., 1995; Sugimoto et al., 2013). Lineage analysis crossing ScxCre transgenic mice with reporter mice revealed that Scx⁺ chondroprogenitors differentiate into chondrocytes near the chondro-tendinous/ligamentous junction (CTJ/CLJ) during mouse development (Sugimoto et al., 2013). These lines of evidence suggest that the expression of Scx and Sox9 is coordinately regulated in the cell population bridging between cartilage and tendon/ligament. However, very little is known about the cellular origin or molecular mechanism that regulates the formation of the junction between cartilage and tendon/ligament.

Through detailed Sox9 lineage tracing in Scx⁺ cells, we found that the Scx⁺ cell population can be divided into two distinct populations with or without their Sox9 expression history, i.e. Scx⁺/Sox9⁻ and Scx⁺/Sox9⁺ progenitors. Tenocytes are derived from both Scx⁺/Sox9⁻ and Scx⁺/Sox9⁺ progenitors, whereas ligamentocytes arise from Scx⁺/Sox9⁺ progenitors. Chondrocytes around the CTJ/CLJ are descendants of Scx⁺/Sox9⁺ progenitors. The closer the tendon is to the cartilaginous primordium, the more tenocytes arise from Scx⁺/Sox9⁺ progenitors. Using loss-of-function approaches, we demonstrate that Scx⁺/Sox9⁺ progenitors functionally contribute to the establishment of the junction between hyaline cartilage and tendon/ligament.

Mice were purchased from Japan SLC (Shizuoka, Japan) or from Shimizu Laboratory Supplies Co. (Kyoto, Japan). ROSA26R (R26R) (Soriano, 1999) or Rosa-CAG-LSL-tdTomato (Ai14) (Madisen et al., 2010) strains obtained from The Jackson Laboratory were crossed to generate the Sox9^Cre/+;R26R and Sox9^Cre/+;Ai14 mice for Sox9 lineage tracing. Ai14 mice harbor a targeted mutation of the Gt(ROSA)26Sor locus with a loxP-flanked STOP cassette preventing transcription of a CAG promoter-driven red fluorescent protein variant, tdTomato. Generation of ScxGFP and ScxCre transgenic strains is reported elsewhere (Sugimoto et al., 2013). To generate Sox9 conditional knockout mice, Sox9-flox (Kist et al., 2002) and ScxCre transgenic strains were crossed. All animal experimental procedures were approved by the Animal Care Committee of the Institute for Frontier Medical Sciences, Kyoto University and conformed to institutional guidelines for the study of vertebrates.

Antisense RNA probes were transcribed from linearized plasmids with a digoxigenin (DIG) RNA labeling kit (Roche) as previously described (Takimoto et al., 2009). For RNA probes, cDNAs for Scx and Myog were amplified by RT-PCR based on sequence information in GenBank (Scx, BC062161; Myog, BC068019). Mouse Sox9 cDNA was described previously (Wagner et al., 1994). For frozen section in situ hybridization, mouse embryos were treated in 20% sucrose without any fixation and then embedded in Tissue-Tek OCT compound (Sakura Finetek) and sectioned at 8 μm. Frozen sections were postfixed with 4% paraformaldehyde in phosphate-buffered saline (PFA/PBS) for 10 minutes at room temperature and then carbethoxylated in PBS containing 0.1% diethylpyrocarbonate twice. Sections were treated in 5× SSC, and hybridization was performed at 58°C with DIG-labeled antisense RNA probes. Immunological detection of DIG-labeled RNA probes was with an anti-DIG antibody conjugated with alkaline phosphatase (anti-DIG-AP Fab fragment; Roche) and BM Purple (Roche).

Embryos were fixed with 4% PFA/PBS at 4°C for 3 hours, immersed in a series of sucrose solutions (12%, 15% and 18% sucrose in PBS), frozen, and cryosectioned at 8 μm. For Sox9^Cre/+;R26R mice, specimens were treated with 20% sucrose at 4°C for 3 hours without prefixation, cryosectioned at 10 μm, and then fixed with ice-cold acetone. After washing with PBS, the slides were incubated with 2% skimmed milk in PBS for 20 minutes and incubated overnight at 4°C with primary antibodies diluted with 2% skimmed milk in PBS. After washing, the sections were incubated with goat anti-rat and anti-rabbit secondary antibodies conjugated to Alexa Fluor 488 or with goat anti-rabbit and anti-mouse secondary antibodies conjugated to Alexa Fluor 594 (Molecular Probes), and washed again in PBS. The primary antibodies used were anti-GFP (diluted 1:1000; Nakarai), anti-Sox9 (1:600; Chemicon), anti-tenomodulin (Tnmd) (1:1000) (Oshima et al., 2004; Shukunami et al., 2008), anti-chondromodulin 1 (Chm1) (1:1000; Cosmo Bio) and anti-type I collagen (1:500; Rockland). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). The images were captured under a Leica DMRXA microscope equipped with a Leica DC500 camera.

For Toluidine Blue staining, deparaffinized and/or hydrated sections were stained with a 0.05% Toluidine Blue solution (pH 4) for 2-5 minutes as described (Takimoto et al., 2012). For X-gal staining, embryos were treated with 20% sucrose in PBS at 4°C, and embedded in Tissue-Tek OCT compound. Frozen sections were prepared at 14 μm. Before staining, the sections were treated with fixation solution (0.2% glutaraldehyde, 5 mM EGTA, 2 mM MgCl₂) at 4°C for 5 minutes. After washing (phosphate buffer containing 2 mM MgCl₂, 0.01% sodium deoxycholate, 0.02% Nonidet P40), the sections were incubated with X-gal staining solution (5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 1 mg/ml X-gal) at 37°C overnight.

After fixation with 4% PFA/PBS, mouse embryos were dehydrated with ethanol. Skin and soft tissues were removed and the embryos were then stained with 0.015% Alcian Blue 8GX (Sigma). After clearing with 2% KOH, the embryos were stained with 0.05% Alizarin Red S (Wako) in 1% KOH and then cleared with 1% KOH.

Scx is expressed in the tendogenic/ligamentogenic regions, as well as in the chondrogenic regions (Cserjesi et al., 1995; Sugimoto et al., 2013). In situ hybridization analysis revealed that Scx⁺/Sox9⁺ chondrogenic cells are predominantly distributed in and around the primordial enthesis between cartilage and tendon/ligament (supplementary material Fig. S1). To compare the expression domains of Sox9 in Scx⁺ cells in more detail, we performed double immunostaining using antibodies against Sox9 and GFP in transgenic ScxGFP embryos that express enhanced green fluorescent protein (EGFP) under the control of the promoter and enhancer of mouse Scx (Fig. 1) (Sugimoto et al., 2013).

During axial musculoskeletal development, the paraxial mesoderm separates into the somites that eventually give rise to the vertebrae, ribs, tendons, ligaments, the dermis of the dorsal skin and muscles (Christ et al., 2004; Christ et al., 2000). In the thoracic somite at E10.5, Sox9 was expressed in the entire sclerotome, notochord and neural tube (Fig. 1A), whereas the dorsolateral sclerotome containing tendon progenitors was composed of Scx⁺ cells (Fig. 1B). The dorsolateral sclerotome was positive for Sox9, but small numbers of Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ cells were observed in the dermomyotome (Fig. 1C). At E11.5, Scx⁺/Sox9⁺ cells were observed in the vertebral and rib primordia (supplementary material Fig. S2A,B), and Scx⁻/Sox9⁺ cells were surrounded by Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ cells (supplementary material Fig. S2C). At E13.5, Sox9 was detected in the cartilaginous primordia of the vertebral body, neural arch and ribs (Fig. 1D). By contrast, Scx was exclusively expressed in the vertebral and costal tendon primordia (Fig. 1E). As typically seen in the costal region, Sox9 and Scx exhibited non-overlapping expression patterns at this stage (Fig. 1F).

Appendicular and abdominal muscles are derived from the hypaxial myotome, whereas lateral plate mesoderm gives rise to the skeletal elements, tendons and ligaments of limbs (Brent and Tabin, 2002). At E10.5, overlapping expression of Sox9 and Scx was observed in the limb bud mesenchyme, except for the distal region (Fig. 1G-I). In the forelimb at E11.5, the primordia of the radius, ulna, carpal and metacarpal bone express Sox9. Scx⁺/Sox9⁺ or Scx⁺/Sox9⁻ cells then rearranged into the dorsal and ventral superficial regions surrounding the Sox9⁺ region (supplementary material Fig. S2D-F). Scx⁺/Sox9⁺ cells were observed at the most proximal region of the limb (supplementary material Fig. S2F). At E13.5, the appendicular cartilaginous elements were positive for just Sox9 (Fig. 1J), but the collateral ligaments and the interzone of the metacarpophalangeal joint were double-positive for Sox9 and Scx (Fig. 1K). At E14.5, Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ cells were present in the cartilage of the nasal septum and the fibrous cells of the turbinate primordia, respectively (Fig. 1L). In the vertebral column, the outer layer of the intervertebral discs was Scx⁺ (Fig. 1M). Scx⁺/Sox9⁺ chondrogenic cells were found in the entheseal region of the Achilles tendon and the patella (Fig. 1N,O).

Based on these data, we conclude that the Scx⁺ cell population can be subdivided into two distinct subpopulations of Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ cells, and that Sox9 expression later disappears from tendons and ligaments as differentiation proceeds.

For lineage tracing of Sox9⁺ cells in the Scx⁺ domains during tendon and ligament formation, we crossed Sox9^Cre/+ mice (Akiyama et al., 2005) with the reporter line Rosa-CAG-LSL-tdTomato (Ai14) (Madisen et al., 2010) to generate Sox9^Cre/+;Ai14 mice, which were then crossed with ScxGFP mice to obtain Sox9^Cre/+;Ai14;ScxGFP embryos (Fig. 2A-E; Fig. 3A-G).

In Sox9^Cre/+;Ai14;ScxGFP embryos at E14.5, cells of the Sox9⁺ lineage were found in the tendons near the vertebral column and ribs, joints between the ribs and vertebrae, and in the developing lung (Fig. 2A). The outer fibrous region of the vertebrae and the surrounding membranous regions consisted of Scx⁺ cells that retained their Sox9 expression history (Fig. 2B). In the tendinous diaphragm near the heart, most cells were negative for Sox9 and positive for Scx (Fig. 2C). Abdominal tendons were positive for just Scx (Fig. 2D). In the tail, the insertion sites of the tendons into the vertebrae were derived from Scx⁺/Sox9⁺ progenitors, whereas tendons located further away from vertebrae were almost exclusively Sox9⁻ and Scx⁺ (Fig. 2E).

We then analyzed the contribution of Sox9⁺ progenitors in the lumbar vertebrae and their associated tendons/ligaments of Sox9^Cre/+;Ai14 neonates at the level of the vertebral body, the articular process or the spinous process (Fig. 2F-H; supplementary material Table S1). At P0, tendons and ligaments in the vicinity of vertebrae were derived from the Sox9⁺ progenitor population (Fig. 2F-H). The lateral region of the thoracolumbar fascia enclosing the erector spinae muscles and tendons anchoring the latissimus dorsi muscle were negative for Sox9 at P0 (Fig. 2H, T4, T3). Thus, Scx⁺/Sox9⁺ progenitors contribute to the formation of ligaments and tendons in the vicinity of ribs and vertebrae, whereas abdominal tendons are derived from the Scx⁺/Sox9⁻ cell lineage.

In the distal part of the hindlimb of Sox9^Cre/+;Ai14;ScxGFP embryos at E14.5, the ligaments arose from Scx⁺/Sox9⁺ progenitors, but both Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ progenitors contributed to tendon formation (Fig. 3A). Scx⁺/Sox9⁺ progenitors contributed to the formation of collateral ligaments (Fig. 3B, L3) and the entheseal side of tendons (Fig. 3C,D), whereas other parts of tendons mainly arose from Scx⁺/Sox9⁻ progenitors and the proportion of these cells varied between the individual tendons; for example, the extensor digitorum longus tendon was derived from Scx⁺/Sox9⁻ progenitors except for the prospective enthesis (Fig. 3B, T8), whereas Achilles tendons arose from both the Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ cell lineages (Fig. 3C,D, T9).

In the knee joint at E14.5, the primordia for cruciate and patella ligaments were visible as an Scx⁺ region within the prospective joint cavity (Fig. 3E). All of the articular components and cartilage were positive for Sox9 (Fig. 3F). The developing cruciate ligaments and capsular ligaments including the patella ligament and tibial collateral ligament were Sox9⁺ (Fig. 3G). In the cruciate and the patella ligament at P0, Sox9 protein was no longer detectable in these tendons or ligaments, except for cartilage (Fig. 3H,I). Thus, all appendicular ligamentocytes arise from Scx⁺/Sox9⁺ progenitors, whereas appendicular tenocytes are derived from both the Scx⁺/Sox9⁺ and the Scx⁺/Sox9⁻ cell lineages.

Thus, the Scx⁺/Sox9⁺ progenitor pool is a unique multipotent cell population that gives rise to Scx⁻/Sox9⁺ chondrocytes and Scx⁺/Sox9⁻ tenocytes/ligamentocytes.

To investigate how Sox9⁺ progenitors contribute to limb tendon formation, we analyzed the distribution of cells of the Sox9⁺ lineage in Tnmd⁺ mature tendons and Chm1⁺ mature cartilage in the forearm of Sox9^Cre/+;R26R mice at P0 (Fig. 4A-L; supplementary material Table S1). Chm1 and Tnmd are markers of mature chondrocytes and tenocytes/ligamentocytes, respectively (Oshima et al., 2004; Shukunami et al., 2008; Shukunami et al., 2006).

Within the proximal parts of the ulna and radius, the sheet-like anchoring tendons consisted of tenocytes derived from Sox9⁺ progenitors (Fig. 4A,G). By contrast, at the medial side of the ulna and radius, Sox9⁺ progeny were absent from the proximal region of the cord-like force-transmitting tendons that were inserted into the individual muscles (Fig. 4B,H). However, tenocytes derived from Sox9⁺ progenitors were found in the bundled force-transmitting tendons in the dorsal region of the distal ulna and radius and of the carpal levels (Fig. 4C,D,I,J). The forearm at the wrist level can be subdivided into several extensor tendon compartments with thick fascia. Within the same compartment, each tendon was derived from both Sox9⁺ and Sox9⁻ progenitors, and the ratio of Sox9⁺ to Sox9⁻ progenitor-derived tenocytes was similar (Fig. 4I,J). In carpal tendons, more of the tenocytes were derived from Sox9⁺ progenitors (Fig. 4D,J). Tendons containing many Sox9⁺ progenitor-derived tenocytes were inserted into the proximal edges of the metacarpals or carpals, whereas tendons containing fewer or no Sox9⁺ progenitor-derived tenocytes were inserted into the middle or distal phalanges, in the more distal region of the autopod (Fig. 4D,J).

Around the metacarpal level, bundled tendons separate into individual tendons that insert into the end point of each digit (Fig. 4E,F). More tenocytes of the Sox9⁺ cell lineage were observed in the tendons at the palmar side, including tendons of the flexor digitorium profundus (T26), flexor digitorium sublimis (T27) and interosseous (T30) (Fig. 4K,L), whereas most dorsal tendons (T18, T19) were Sox9⁻ (Fig. 4E,K). In the collateral ligaments (L9) of the metacarpophalangeal joint, all ligamentocytes were strongly positive for Sox9 (Fig. 4F,L). Although the force-transmitting tendons were derived from both Sox9⁺ and Sox9⁻ progenitors, the anchoring tendons near the elbow wholly arose from Sox9⁺ progenitors (Fig. 4M).

Taken together, although the proportion of tenocytes that retain their Sox9⁺ expression history varies between the individual force-transmitting tendons, in general the number of these Sox9⁺ tenocytes decreases with increasing distance from the skeletal element.

We then focused our analysis on the transitional zone between cartilage and tendon/ligament in order to reveal the contribution of Sox9⁺ progenitors to the entheses. Entheses are classified into two groups: fibrous and fibrocartilaginous (Benjamin and Ralphs, 2001). Collagen fibers in the fibrous entheses are inserted into bone via the periosteum, which gives a firmer hold to tendons and ligaments. Fibrocartilaginous entheses have four zones during the transition from tendon/ligament to bone, consisting of tendon/ligament, fibrocartilage, mineralized fibrocartilage, and bone. Fibrous entheses are mainly present in short ligaments or tendons. Since periosteum has been reported to be derived from Sox9⁺ progenitors (Akiyama et al., 2005), we examined the prospective fibrocartilaginous entheses of quadriceps femoris tendon, cruciate ligaments and the Achilles tendon in Sox9^Cre/+;R26R mice (Fig. 5).

Type I collagen (Col1) and Chm1 were localized to tendons/ligaments including the prospective entheseal region and hyaline cartilage, respectively (Fig. 5A,D,G), whereas Tnmd was expressed in tendons and ligaments except for the region just adjacent to hyaline cartilage (Fig. 5B,E,H). These Col1⁺/Tnmd⁻ cells were positive for X-gal staining (Fig. 5C,F,I). Hence, near the joint region, tenocytes, ligamentocytes and chondrocytes were derived from Sox9⁺ progenitors, but the prospective entheseal region abutting hyaline cartilage was negative for both Tnmd and Chm1, suggesting the presence of a distinct population in the prospective fibrocartilaginous enthesis bridging between hyaline cartilage and tendon/ligament.

We have generated two transgenic mouse lines that express Cre recombinase in the Scx⁺ domains at high (ScxCre-H) or low (ScxCre-L) levels (Sugimoto et al., 2013). Owing to the expression gradient and transient expression of Scx around the entheseal cartilage (supplementary material Fig. S1), more chondrocytes in ScxCre-H are Scx⁺ than in ScxCre-L (Sugimoto et al., 2013). To investigate the functional role of Sox9 in Scx⁺/Sox9⁺ cells by a loss-of-function approach, we crossed these lines with Sox9-flox mice (Kist et al., 2002) to inactivate Sox9 in Scx⁺ cells. Both ScxCre-L;Sox9^flox/+ and ScxCre-H;Sox9^flox/+mice were viable and fertile, but ScxCre-L;Sox9^flox/flox and ScxCre-H;Sox9^flox/flox mice died after birth. In ScxCre-H;Sox9^flox/flox mice, severe skeletal hypoplasia was observed beyond the prospective entheseal cartilage, thus causing the secondary defects observed in tenocytes derived from Scx⁺/Sox9⁻ cells (not shown). Hence, we analyzed the ScxCre-L;Sox9^flox/flox mice with skeletal defects around the entheseal cartilage in more detail.

In ScxCre-L;Sox9^flox/flox neonates, the sternum and ribcage except for the proximal region were missing (Fig. 6A-D). In the vertebral column of ScxCre-L;Sox9^flox/flox neonates, the vertebral bodies, the intervertebral discs, the articular processes of the neural arch and the transverse processes were hypoplastic (Fig. 6E,F). Severe hypoplasia in the ribcage is expected from the expression of Scx during the early stages of costal cartilage formation (Fig. 6M-O). The appendages of ScxCre-L;Sox9^flox/flox mice were hypoplastic and shorter than those of controls and the joint cavity was smaller (Fig. 6A,B,G-L). In the forelimb of ScxCre-L;Sox9^flox/flox mice, hypoplasia of carpal bones at the ulnar side, elbow joint, cartilage around the shoulder joint and deltoid tuberosity of the humerus was evident and curvature of the wrist was observed (Fig. 6G,H). Interestingly, abnormal mineralization occurred in the olecranon (Fig. 6G,H). In the hindlimb of ScxCre-L;Sox9^flox/flox mice the tarsal bones, cartilage around the hip and the knee joint and tibial tuberosity were defective (Fig. 6I,J) and the patella was missing (Fig. 6K,L). These results suggest that skeletal dysplasia occurs in the Scx⁺ cartilaginous region that is closely associated with tendons and ligaments.

Double immunostaining of Tnmd and Chm1 revealed defective formation of the junction between cartilage and tendon/ligament in ScxCre-L;Sox9^flox/flox at E18.5. Transverse processes of the lumber vertebrae (Fig. 7C) and the lateral region of sacral vertebrae (Fig. 7A) provide the attachment sites for axial tendons, but these sites were missing in ScxCre-L;Sox9^flox/flox embryos (Fig. 7B,D). In control mice, Sox9⁺ cells were scattered in the outer annulus fibrosus near the inner annulus fibrosus (Fig. 7E), whereas these cells were missing in ScxCre-L;Sox9^flox/flox embryos (Fig. 7F). In ScxCre-L;Sox9^flox/flox intervertebral discs, the formation of the inner annulus fibrosus, which shows metachromatic staining with Toluidine Blue, was defective, whereas the outer annulus fibrosus became wider (Fig. 7H) compared with that of control mice (Fig. 7G). Thus, in ScxCre-L;Sox9^flox/flox embryos, the prospective entheses were either missing or hypoplastic in the axial skeleton.

In the forelimb, hypoplastic tendon formation in association with defective cartilage formation at the ulnar side was observed in ScxCre-L;Sox9^flox/flox embryos at E16.5 (not shown). In the knee joint, the patella and the frontal region of the femoral condyle were missing (Fig. 7I,J). In the heel, the attachment site for the Achilles tendon was defective (Fig. 7K-N). Interestingly, cells just adjacent to the tendon attachment site, which are Sox9⁺ in control mice, were missing in ScxCre-L;Sox9^flox/flox embryos (Fig. 7O,P).

In this study, we have demonstrated for the first time that the Scx⁺ cell population can be subdivided into two distinct populations with regard to their Sox9 expression history: Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ progenitors. The Scx⁺/Sox9⁺ progenitor pool is a unique multipotent cell population that gives rise to Scx⁻/Sox9⁺ chondrocytes and Scx⁺/Sox9⁻ tenocytes/ligamentocytes (Fig. 8A). The closer the tendon and cartilage are to the prospective enthesis, the more tenocytes and chondrocytes originate from Scx⁺/Sox9⁺ progenitors (Fig. 8B). Further analyses of ScxCre-L;Sox9^flox/flox mice revealed that the Scx⁺/Sox9⁺ cell population functionally contributes to the establishment of the junction between cartilage and tendon/ligament.

Tenocytes are descendants of Scx⁺/Sox9⁺ and Scx⁺/Sox9⁻ progenitors (Fig. 8A). In general, the number of tenocytes that retain their Sox9 expression history decreases with increasing distance from the skeletal element. Ligamentocytes and annulus fibrosus cells in the intervertebral discs are derived from Scx⁺/Sox9⁺ progenitors, whereas chondrocytes are derived from Scx⁺/Sox9⁺ and Scx⁻/Sox9⁺ progenitors (Fig. 8A). The closer the cartilage is to the prospective entheses, the more chondrocytes arise from Scx⁺/Sox9⁺ progenitors. Thus, the Scx⁺/Sox9⁺ progenitor population is predominantly distributed across the enthesis to form the CTJ/CLJ during development (Fig. 8B).

In contrast to axial tendon formation, very little is known about axial ligament formation. We show that the Scx⁺ axial ligaments are derived from the Sox9⁺ cell lineage and thus conclude that these Scx⁺ ligamentocytes originate from the Sox9⁺ sclerotome. However, the timing of Scx expression in Sox9⁺ ligament progenitors needs to be investigated further in order to clarify whether the axial ligament progenitors are derived from the Scx⁺/Sox9⁺ dorsolateral domain of the sclerotome or are recruited from the Scx⁻ sclerotome to express Scx at later stages of development.

The appendicular tendons include a considerable number of tenocytes derived from Scx⁺/Sox9⁻ progenitors, particularly in the distal part of the limbs. Unlike the sclerotome, which consists of Sox9⁺ progenitors, both Sox9⁺ and Sox9⁻ progenitors are present in the lateral plate mesoderm of the E10.5 limb bud. The Scx⁺/Sox9⁻ population in the lateral plate mesoderm might represent prospective distal tendon progenitors, although we cannot exclude the possibility that another, as yet unknown, population of tendon progenitors is recruited from the surrounding tissue to become Scx⁺ tenocytes later in development.

In ScxCre-L;Sox9^flox/flox mice, the attachment sites of the tendons/ligaments to the cartilaginous primordia and the annulus fibrosus of the intervertebral discs are impaired. The most notable phenotype of ScxCre-L;Sox9^flox/flox embryos is the absence of the ribcage. Chondrogenic cells in the developing costal cartilage have the ability to differentiate into entheseal chondrocytes, as evidenced by the expression of Scx in the entire rib cartilaginous primordium. This is compatible with the histological feature that costal chondrocytes are located very close to the tendinous attachment site of the surrounding intercostal muscle to each rib cartilage. Likewise, the cartilaginous bone primordium of the patella embedded in the tendon is missing. Thus, our loss-of-function analysis of Sox9 in the Scx⁺ domain reveals the functional significance of the Scx⁺/Sox9⁺ progenitor population in the establishment of the CTJ/CLJ, especially on the cartilaginous side.

In Sox9^flox/flox;Prx1Cre mice, inactivation of Sox9 in limb bud mesenchyme causes the complete absence of cartilage and bone (Akiyama et al., 2002). Severe chondrodysplasia also occurs in Sox9^flox/flox;Col2a1Cre mice upon inactivation of Sox9 in precartilaginous condensing cells and chondrocytes (Akiyama et al., 2002). Based on these findings, functional roles of Sox9 in chondrogenesis could be discussed at three key stages: the chondroprogenitor stage, cartilaginous condensation stage and chondrocyte stage. Similarly, we consider tendo/ligamentogenesis in three distinct stages: the tendon/ligament progenitor stage, the tendon/ligament primordium formation stage, and tenocyte/ligamentocyte stage. In ScxCre-L;Sox9^flox/flox embryos, we observed hypoplasia of the entheses of tendons/ligaments, the annulus fibrosus of the intervertebral discs, and of cartilages arising from Scx⁺/Sox9⁺ chondroprogenitors. The longer that Sox9 expression continues, the more severe the defects within the Scx⁺/Sox9⁺ domain of ScxCre-L;Sox9^flox/flox embryos become. Unlike chondrogenic cells, which continuously express Sox9, Sox9 was downregulated in the migrating tendon/ligament progenitors before their arrival at the presumptive tendon/ligament-forming site. Considering the timing of Sox9 downregulation in the tendon/ligament cell lineages, it is unlikely that the last two stages during tendo/ligamentogenesis critically depend on the function of Sox9. Loss of Sox9 in Scx⁺/Sox9⁺ progenitors is likely to be a principal cause of the hypoplastic CTJ/CLJ in ScxCre-L;Sox9^flox/flox mice.

Intervertebral discs and joints connect adjacent vertebrae. Each intervertebral disc is composed of an external annulus fibrosus surrounding an internal nucleus pulposus. Cells in the annulus fibrosus of intervertebral discs can be traced back to somitocoele cells that are included in the central core of the somite, distinct from the progenitor population for the vertebral body (Mittapalli et al., 2005). The annulus fibrosus consists of the inner annulus with chondrocytic cells and the outer annulus with tenocytic cells. We have shown here that both types of cells arise from Scx⁺/Sox9⁺ progenitors. In ScxCre-L;Sox9^flox/flox mice, the inner annulus fibrosus is defective but expansion of the outer annulus fibrosus takes place on the ventral side. Thus, it is suggested that Sox9 maintains the proper balance between the inner and the outer cell numbers by regulating the survival and differentiation of cells in the inner annulus fibrosus during intervertebral disc formation.

During postnatal growth, entheseal fibrocartilage develops in response to compressive loads (Benjamin and Ralphs, 1998). Fibrocartilage is an important connective structure between tendon and hyaline cartilage, but its cellular origin remains uncertain. We show that Chm1 and Tnmd are expressed in hyaline cartilage and tendon/ligament, respectively, whereas the transitional region just adjacent to hyaline cartilage or tendon/ligament is negative for Chm1 and Tnmd, consistent with our previous observation in rabbits (Yukata et al., 2010). Our lineage analysis further revealed that cells in this Tnmd⁻/Chm1⁻ zone are positive for Sox9 and Scx. Therefore, it is likely that cells in this transitional zone give rise to fibrochondrocytes during postnatal development. Further studies to reveal the cellular origin of fibrochondrocytes are now underway.

We thank Mr T. Matsushita and Ms K. Kogishi for histological studies and Ms H. Sugiyama for valuable secretarial help.