Tendon-bone attachment unit is formed modularly by a distinct pool of Scx- and Sox9-positive progenitors

The musculoskeletal system provides the body with form, stability and mobility. This requires precise and tightly coordinated assembly of tendons, muscles and bones into one functional system. Although the development of each of these components has been extensively studied, the process of musculoskeletal assembly has remained poorly understood.

A fundamental step in the assembly process is the development of an attachment unit between a bone and a tendon. The term ‘tendon-bone attachment unit’ (referred to herein as AU) describes a complex structure that includes the tip of the tendon and the part of the bone into which it is inserted. Tendons are often inserted into specific sites named bone eminences. These are superstructures that grow on the surface of the bone, exhibiting a variety of shapes and sizes depending on bone type and organism (Gray, 1918; Hill, 1964). Bone eminences are vital for the functionality of the musculoskeletal system. They provide stable anchoring points for muscles, which are inserted into the skeleton via tendons, and dissipate the stress exerted on the skeleton by contracting muscles. This effect improves the mechanical resilience of muscle attachment and facilitates movement (Benjamin et al., 2002; Biewener et al., 1996; Thomopoulos et al., 2011; Thomopoulos et al., 2010). Additionally, because of their prominence in the bone landscape, these structures largely contribute to the three-dimensional morphology of bones (Gray, 1918). Obviously, deciphering the process by which bone eminences develop is vital for the understanding of tendon-bone AU formation during the assembly of the musculoskeletal system.

The development of the appendicular skeleton is initiated when a subset of mesenchymal cells, originating in the lateral plate mesoderm, amasses and is specified as chondroprogenitors. Sox9 is the earliest known marker for these progenitor cells and is necessary for their condensation and differentiation (Akiyama, 2008; Akiyama et al., 2002). Interestingly, lineage studies on Sox9-Cre mice suggest that Sox9-expressing cells also serve as progenitors for osteoblasts, tenocytes and synovial cells (Soeda et al., 2010; Akiyama et al., 2005). As development proceeds, chondroprogenitors differentiate into chondrocytes, expressing markers such as type II collagen, and form cartilaginous templates that prefigure the future bones (Karsenty et al., 2009; Lefebvre and Smits, 2005; Provot and Schipani, 2005).

Previously, we showed that bone eminences are formed during the cartilaginous phase of skeletal development, in a process that involves both molecular and mechanical signals. We showed that the basic helix-loop-helix (bHLH) transcription factor scleraxis (SCX) drives the expression of bone morphogenetic protein 4 (Bmp4) at the tendon-skeleton junction to induce formation of cartilaginous bone eminences. Blockage of either Scx or Bmp4 expression in the limb arrested eminence development (Blitz et al., 2009).

Despite this progress, we still lack basic understanding of many aspects of AU development, such as the origin of eminence cells and the molecular mechanisms that control eminence progenitors. In this work, we identify a novel mechanism that regulates AU formation. We show that an external, distinct pool of progenitors that express both Sox9 and Scx form eminences, which are modularly added onto the developing bone. Moreover, we show that TGFβ signaling controls specification of these progenitors, whereas the SCX/BMP4 pathway mediates their differentiation to eminence cells.

The generation of Scx^-/- (Murchison et al., 2007), R26R-lacZ reporter (Soriano, 1999), floxed Tgf-βRII (Chytil et al., 2002), floxed Bmp4 (Liu et al., 2004; Selever et al., 2004), floxed Sox9 (Akiyama et al., 2002), Prx1-Cre (Logan et al., 2002), Col2a1-CreER (Nakamura et al., 2006) and Sox9-CreER (Soeda et al., 2010) mice have been described previously. Scx-Cre transgenic mice were generated by and obtained from R. Schweitzer (Shriners Hospital for Children Research Division, Portland, OR, USA) and R. L. Johnson (The University of Texas MD Anderson Cancer Center, Houston, TX, USA). To create Scx^-/- mice, animals heterozygous for the mutation were crossed; as a control we used heterozygous Scx embryos. To create Scx-Sox9, Prx1-Tgf-βRII and Prx1-Bmp4 mutant mice, floxed Sox9, floxed Tgf-βRII and floxed Bmp4 mice were mated with Scx-Sox9, Prx1-Tgf-βRII and Prx1-Bmp4 mice, respectively. As a control, we used embryos that lack Cre alleles. To create Col2a1-CreER, R26R-lacZ and Sox9-CreER, R26R-lacZ reporter mice, floxed R26R-lacZ mice were mated with Col2a1-CreER and Sox9-CreER mice, respectively.

In all timed pregnancies, plug date was defined as embryonic day (E) 0.5. For harvesting of embryos, timed-pregnant females were sacrificed by cervical dislocation. Tail genomic DNA was used for genotyping by PCR.

Cartilage and bones in whole mouse embryos were visualized after staining with Alcian Blue and Alizarin Red S (Sigma) and clarification of soft tissue with potassium hydroxide (McLeod, 1980).

Adult females were administered 6.5 mg tamoxifen (TM) in corn oil by oral gavage or 6 mg 4-hydroxy-TM in 1:10 (v/v) ethanol:sunflower oil (from a stock of 20 mg/ml) by intraperitoneal injection at the indicated time points.

For immunofluorescence, embryo limbs were fixed overnight in 4% paraformaldehyde (PFA) at 4°C, embedded in paraffin and sectioned at 7 μm. Paraffin sections were collected on Fisherbrand Superfrost Plus slides, de-paraffinized and rehydrated to water. Antigen was retrieved in 10 mM citrate buffer, pH 6, using a microwave. In order to block non-specific binding of immunoglobulin, sections were incubated with 7% goat serum. Following blockage, sections were incubated overnight at 4°C with primary antibody anti-SOX9 (1:200; AB5535, Chemicon). Then, sections were washed in 0.1% Tween 20 in PBS (PBST) and incubated with Cy2-conjugated secondary fluorescent antibodies (1:200; Jackson Laboratories). After staining for SOX9, slides were washed in PBS and fixed in 4% PFA at room temperature for 10 minutes. Then, slides were incubated with protein kinase (P2308, Sigma), washed and post-fixed again in 4% PFA. Next, sections were washed and incubated overnight at 4°C with primary antibody anti-collagen II [1:100; Developmental Studies Hybridoma Bank (DSHB), University of Iowa, USA]. The next day, sections were washed in PBST and incubated with Cy3-conjugated secondary fluorescent antibodies (1:200; Jackson Laboratories). Slides were mounted with Immuno-mount aqueous-based mounting medium (Thermo).

For pulse-chase experiments on Col2a1-CreER, R26R-lacZ heterozygous embryos, whole-mount X-gal staining was carried out. Embryos were fixed for 1 hour in 4% PFA at 4°C, washed three times in rinse buffer (PBS containing 0.01% deoxycholate, 0.02% NP-40, 2 mM MgCl₂ and 5 mM EGTA) at room temperature and stained for 5 hours at 37°C in rinse buffer supplemented with 1 mg/ml X-gal, 5 mM K₃Fe(CN)₆ and 5 mM K₄Fe(CN)₆. For histological examination, stained whole-mount limbs were fixed in 4% PFA overnight, dehydrated, embedded in paraffin and sectioned at 7 μm. Sections were then collected on Fisherbrand Superfrost Plus slides, dehydrated and cleared in xylene. Fast Red (Sigma) was used for counter-staining.

For pulse-chase experiments on Sox9-CreER, R26R-lacZ heterozygous embryos, we used slide X-gal staining. Cryostat sections were cut at a thickness of 10 μm and dried for 1 hour at 37°C. Then, sections were fixed for 10 minutes in 4% PFA, washed three times in rinse buffer containing 0.01% deoxycholate, 0.02% NP-40, 2 mM MgCl₂ and 5 mM EGTA at room temperature. Slides were stained overnight at 37°C in rinse buffer supplemented with 1 mg/ml X-gal, 5 mM K₃Fe(CN)₆ and 5 mM K₄Fe(CN)₆. For histological examination, stained slides were counter-stained with Fast Red solution (Sigma), dehydrated and cleared in xylene. To quantify the prevalence of lacZ-positive chondrocytes, regional density was approximated by counting the number of X-Gal-stained cells in sampled rectangles of equal area in images of the humerus, femur and calcaneus. Manual marking of cells was followed by automatic enumeration using in-house software. At least five sections were analyzed for each of the Sox9-CreER, R26R-lacZ littermate embryos at E15.5. Statistical significance was determined by Student’s t-test.

Section in situ hybridizations were performed as described previously (Murtaugh et al., 1999; Riddle et al., 1993). Double fluorescent in situ hybridizations on paraffin sections were performed using fluorescein- and DIG-labeled probes (see supplementary material Table S1 for probe sequences). After hybridization, slides were washed, quenched and blocked. Probes were detected by incubation with anti-fluorescein-POD and anti-DIG-POD (Roche; 1:300), followed by Cy3- or Cy2-tyramide labeled fluorescent dyes (according to the instructions of the TSA Plus Fluorescent Systems Kit, Perkin Elmer).

Harvested limbs were fixed overnight in 4% PFA in PBS, dehydrated to 100% ethanol and then soaked in 2% iodine solution. Samples were scanned using a microfocus X-ray tomographic system (Micro XCT-400, Xradia), at 40 kV and 200 lA. A thousand projection images at a total integration time of 5 mseconds with a linear magnification of ×4 were taken. The final pixel size was 2.3μm. The volume was reconstructed using a back projection filtered algorithm (XRadia). Following reconstruction, 3D image processing and analysis were carried out using MicroView (MicroView software version 5.2.2, GE Healthcare). Bone and cartilage tissues were manually segmented and segmented structures were rendered as 3D surfaces.

Despite their functional importance (Fig. 1A,A′), little is known about bone eminence development. Examination of the chronological sequence of long bone morphogenesis revealed that different eminences, such as the deltoid tuberosity and great tuberosity of the humerus or the olecranon process of the ulna, could first be observed at E14.5. However, rudimentary cartilaginous elements were visible already at E12.5 (Fig. 1B). This observation prompted us to study the mechanism that regulates the formation of eminences on the surface of developing bones.

We examined two possible scenarios for the genesis of bone eminences (Fig. 1C). The first, referred to as the ‘local model’, suggests that eminences form by outgrowth from the pre-existing primary cartilage template. According to this model, eminence-forming cells are descendants of chondrocytes that form the primary cartilage. An alternative hypothesis, referred to as the ‘modular model’, suggests that the bone eminence is derived from a pool of cells located externally to the primary template.

To determine which of these two models is correct, we performed a pulse-chase cell lineage experiment using the Cre-ER system, in which the expression of Cre recombinase is activated at different stages of development by tamoxifen administration. By crossing mice that express Cre-ER under control of the collagen type II promoter (Col2a1-CreER) with R26R reporter mice (Nakamura et al., 2006; Soriano, 1999), we were able to mark chondrocytes of the primary cartilage and follow their descendants at later stages. Administration of tamoxifen at E10.5 to Col2a1-CreER, R26R-lacZ heterozygous embryos revealed at E17.5 lacZ expression along the main axis of long bones. Interestingly, there was no X-Gal staining in various eminences (Fig. 2A-G), suggesting that eminence cells do not originate from chondrocytes of the primary template.

To exclude the possibility that the lack of staining was due to inefficiency of the Cre-ER system to mark the entire primary template, we administered tamoxifen to Col2a1-CreER, R26R-lacZ heterozygous embryos at E10.5 and followed their descendants at E12.5. Examination showed that the initial induction of the system was effective, as most of the primary cartilaginous template cells were X-Gal-positive and equally distributed (Fig. 2H). To verify that lacZ expression can be induced in eminence cells, we administered tamoxifen to Col2a1-CreER, R26R-lacZ heterozygous embryos at E13.5 and followed their descendants at E15.5. The results showed that the Cre-ER system was active in bone eminence cells, as X-Gal-positive cells were detected in the deltoid tuberosity (Fig. 2I).

These results strongly imply that eminence-forming cells are not descendants of chondrocytes that compose the primary cartilage.

Our findings that bone eminences appear after the primary template is established (Fig. 1B) and that cells that form eminences are not descendants of primary template chondrocytes (Fig. 2A-E) suggest that bone eminences develop from an external pool of cells. To examine this possibility, we analyzed the differentiation state of cells at the presumable locations of three eminences: the deltoid, the great and the lesser tuberosities of the humerus. To that end, we performed immunofluorescence staining with antibodies for SOX9 as a marker for chondrocyte progenitors and for collagen type II alpha 1 (COL2A1) as a marker for differentiated cartilage. At E12.5, although cells in the primary template have already differentiated to chondrocytes and expressed both Sox9 and Col2a1, we identified at presumable eminence sites cells that expressed only Sox9 (Fig. 3A). From E13.5 to E14.5, cells in these domains underwent differentiation and expressed Col2a1 in addition to Sox9 (Fig. 3A′,A′). Examination of presumable locations of other eminences in the forelimb and hindlimb also revealed cells that expressed only Sox9 (Fig. 3B-E). Finding fields of Sox9-positive, Col2a1-negative cells at presumable locations of bone eminences at a stage when primary template cells have already differentiated suggests that eminences develop from external pools of progenitors, thus supporting the modular model (Fig. 1C).

Our finding that bone eminences originate from an external pool of Sox9-positive progenitor cells raised a question about their specification. During limb development, eminence progenitors can be specified as part of a common pool of progenitor cells that will later form both the eminence and the primary cartilaginous element. In this scenario, the common pool of progenitors differentiates to chondrocytes in two steps, first to form the primary cartilage and then to form the eminence (Fig. 4A). Alternatively, specification of a distinct pool of eminence progenitors might be a particular event that is secondary to the specification of primary cartilage progenitors (Fig. 4A′).

In order to distinguish between these two possibilities, we again performed a pulse-chase cell lineage experiment using the Cre-ER system. By crossing mice that express Cre-ER under control of the Sox9 promoter (Sox9-CreER) (Soeda et al., 2010) with R26R reporter mice, we marked Sox9-positive progenitors and followed their descendants. Examination of E15.5 Sox9-CreER, R26R-lacZ heterozygous embryos that had been administered tamoxifen at E9.5-E10.5 revealed lacZ expression by chondrocytes in different skeletal elements; by contrast, cells in various eminences were poorly stained (Fig. 4B-D,F). As a control, tamoxifen administration at E12.5-E13.5 induced lacZ expression by all cells, including in eminences (Fig. 4E). Quantification of X-Gal-positive cells in the eminence field relative to marked cells in the primary structure of cartilage confirmed these observations (Fig. 4G; supplementary material Fig. S1). These results indicate that eminence progenitors are specified separately from and secondary to the pool of cells that gives rise to the primary cartilage (Fig. 4A′). This implies that during limb development, a distinct and previously unknown pool of progenitors is involved in the formation of bone eminences, which are part of the tendon-bone AU.

The specification of eminence-forming cells as a secondary pool of progenitor cells prompted us to examine what distinguishes this population from the progenitor pool of the primary cartilage. Double fluorescence in situ hybridizations on sections of E11.5 forelimbs showed that, unlike cells of the primary cartilage, the Sox9-positive eminence progenitors also expressed Scx (Fig. 5A). As development proceeded, the Scx expression domain in the forming AU became more restricted, concomitantly with tendon patterning (Fig. 5B). By E13.5, Scx and Sox9 expressions were mutually exclusive as the former was expressed only in forming tendons (Fig. 5A-C). To support this observation, we studied the Scx lineage in the limb using the Scx-Cre mouse crossed with the R26R reporter mouse. As expected, tendons and various eminences were positive for X-Gal, supporting the notion that AU progenitors express both Scx and Sox9 (supplementary material Fig. S2). Interestingly, our analysis revealed that some cells in different skeletal elements were X-Gal-positive as well, most prominently in the proximal humeral head and the distal femoral head. This result agrees with a previous report of overlapping Scx and Sox9 expression during initial stages of limb development (Asou et al., 2002).

Next, we sought to evaluate the contribution of the Scx- and Sox9-positive progenitors to bone eminence formation. For that, we targeted Sox9 in eminence progenitors using the Scx-Cre mouse as a deleter (Scx-Sox9). Given the results of the Scx lineage analysis, it was important to determine which cells lost Sox9 expression. Immunofluorescence staining with antibodies for SOX9 on sections from Scx-Sox9 conditional knockout (cKO) embryos revealed loss of Sox9 expression only in eminence progenitors (Fig. 6A), whereas in other skeletal elements it was maintained. This result strongly supports our expression analysis results, indicating that bone eminence progenitors are indeed Sox9 and Scx positive.

Examination of skeletons of Scx-Sox9 mutants revealed loss of bone eminences (Fig. 6B), suggesting that expression of Sox9 by Scx-positive cells is necessary for eminence formation. Finally, because previous lineage studies suggest that some tenocytes originate in Sox9-expressing progenitors (Soeda et al., 2010), we examined in Scx-Sox9 cKO mutants tendon development upon the loss of corresponding bone eminences using tendon markers such as Scx, Tnmd and Col1a1. As shown in Fig. 6C, tendon development was comparable between Scx-Sox9 cKO and control embryos.

Together, these results imply that bone eminences develop from distinct pools of progenitor cells that co-express Sox9 and Scx.

Next, we explored the molecular mechanism that regulates the Sox9- and Scx-positive bone eminence progenitors. The expressions of Scx by eminence progenitors motivated us to examine its role in this mechanism. Analysis of forelimbs from E12.5 Scx^-/- embryos revealed Sox9-positive progenitors at the presumptive eminence site (Fig. 7A). This result indicates that SCX is not involved in the specification of eminence progenitors.

Previously, we showed that BMP4, produced under regulation of SCX, mediates bone eminence formation (Blitz et al., 2009). We therefore examined the involvement of this pathway in the regulation of eminence progenitors. Examination of forelimbs from E12.5 mutants in which Bmp4 was ablated from limb mesenchyme (Prx1-Bmp4) revealed the presence of Sox9-positive eminence progenitors (Fig. 7B), similarly to Scx^-/- mice. However, as development proceeded, these cells maintained Sox9 expression and failed to express Col2a1. That these cells remained in their progenitor state implies that the SCX/BMP4 pathway is necessary for differentiation of eminence progenitors (Fig. 7C,D).

Having found that the SCX/BMP4 pathway regulates differentiation of eminence progenitors, we proceeded to explore their specification. Previously, it was shown that in mice in which Tgfbr2 was ablated in limb mesenchyme (Prx1-Tgf-βRII), prominent bone eminences were absent (Blitz et al., 2009; Seo and Serra, 2007; Spagnoli et al., 2007). This phenotype was attributed to the loss of mechanical load exerted by muscle contraction (Seo and Serra, 2007) or to the loss of tendon formation (Blitz et al., 2009). The observation of Tgfbr2 expression by AU progenitors at E12.5 (supplementary material Fig. S3A) raised the possibility that TGFβ signaling has a direct role in their specification and/or differentiation. To test this hypothesis, we examined the presence of Sox9-positive progenitor cells in Prx1-Tgf-βRII mutants (Baffi et al., 2004; Logan et al., 2002). Our results showed that at E13.5, whereas control embryos exhibited Sox9-positive, Col2a1-negative eminence progenitors, the mutants lacked these cells (Fig. 8A,B). To verify this result we studied the expression of Scx in the limbs of these mutants and, as expected, it was dramatically reduced (supplementary material Fig. S3B). These results indicate that TGFβ signaling is necessary for specification of bone eminence progenitors. Moreover, the formation of a primary cartilaginous template in the absence of TGFβ signaling provides another indication that bone eminences are derived from a separate pool of progenitors, because the specification of one cell population requires TGFβ signaling whereas the other is refractory to it.

To establish further the role of TGFβ signaling in bone eminence formation, we examined Prx1-Tgf-βRII mutants at E18.5 by CT analysis. As can be seen in Fig. 8C-E, various eminences were completely lost in the mutants. These results clearly indicate the extensive effect of TGFβ signaling on eminence formation and, consequently, on attachment unit development.

Several studies have provided a histological and, to some extent, molecular description of the mature tendon-to-bone insertion site (Benjamin et al., 2002; Biewener et al., 1996; Thomopoulos et al., 2011; Thomopoulos et al., 2010). Nevertheless, the fundamental question of how an attachment unit is formed during musculoskeletal assembly has been largely neglected. In order to reveal the mechanisms by which the AU is established, we studied the development of its bony side, the bone eminence.

In this work, we reveal a unique pool of progenitors that form bone eminences. We show that two distinct pools of progenitor cells form the cartilaginous template of the long bone in a modular fashion. Sox9-positive progenitors form the primary, cylindrical structure of the cartilaginous anlage, whereas a previously unknown second pool of Sox9- and Scx-positive progenitors gives rise to bone eminences. Moreover, we show that these two pools are regulated separately, as TGFβ signaling is necessary for specification of bone eminence progenitors, but not of progenitors of the primary cartilage. Subsequently, the differentiation of the Sox9- and Scx-positive eminence progenitors to chondrocytes involves BMP signaling (Fig. 9).

Our finding that eminence-forming cells are not descendants of primary cartilage chondrocytes, but rather originate from a distinct pool of Sox9- and Scx-positive progenitors, is the first indication that bones form in a modular fashion. The early events that lead to the formation of this distinct pool of progenitors and the exact origin of the Sox9- and Scx-positive cells are still not understood.

Although later in development Scx expression is restricted to forming tendons (Schweitzer et al., 2001), previous works reported early and extensive expression of Scx in the limb bud mesenchyme (Asou et al., 2002; Cserjesi et al., 1995). Thus, it is possible that some of the limb mesenchymal cells that express both Sox9 and Scx are the first progenitors of the bone eminence.

Intriguingly, the development of long bones progressively from two pools of progenitor cells bears a resemblance to the developmental modularity of the mammalian heart. During development, the heart is formed by cells from two distinct pools, known as the first and second heart fields. Initially, the heart tube is formed by cells of the first field. Then, cells that originate in the second heart field contribute to its elongation and to the formation of the outflow tract (Kelly et al., 2001; Mjaatvedt et al., 2001; Waldo et al., 2001). This similarity highlights the use of modularity in organogenesis in order to allow morphological and functional complexity.

Our finding that bone development is a modular process provides new perspectives on various aspects of musculoskeletal function and assembly. Functionally, different parts of the same bone are subjected to different mechanical loading conditions. For example, bone eminences are loaded in tension, whereas the joint surfaces experience compression. The finding that bones form modularly by two pools of progenitors may provide a mechanistic explanation for the ability of different anatomical regions of the bone to cope with different mechanical conditions. Developmentally, modularity can facilitate the assembly of the musculoskeletal system by coordinating interactions between the bone and the attaching tendon without interference with the construction of the entire bone. This notion underscores the evolutionary advantage of a modular strategy of bone morphogenesis, which may be easily manipulated by adding, removing or altering modules instead of by reshaping the whole structure.

Our discovery of a second pool of progenitor cells that form bone eminences suggests that a different regulatory mechanism controls specification and differentiation of these progenitors. Indeed, by blocking the expression of Tgfbr2 in limb mesenchyme we demonstrate the central role of TGFβ signaling in regulating eminence progenitors, as this pathway is necessary exclusively for their specification. Previously, we showed that tendon cells are involved in the initiation of bone eminence formation (Blitz et al., 2009). The observation that Sox9- and Scx-positive eminence progenitors are already specified by E11.5 demonstrates that bone eminence development initiates at E11 rather than at E14, as we previously suggested (Blitz et al., 2009). Because TGFβ signaling is necessary for tendon formation (Pryce et al., 2009), one might argue that the absence of eminence progenitors was secondary to the loss of tendons. However, in Prx1-Tgf-βRII mutants, abnormal Scx expression and tendon development were only manifested at E12.5 (Pryce et al., 2009). Therefore, it is unlikely that arrested tendon formation was the reason for the loss of eminence progenitors in the absence of TGFβ signaling.

Finally, it is tempting to speculate that by controlling both tendon and bone eminence formation, TGFβ signaling is a key regulator of tendon-bone attachment unit. Indeed, it has been suggested that TGFβ coordinates cartilage and tendon differentiation during limb development (Lorda-Diez et al., 2009).

The mechanism that underlies the role of TGFβ signaling in specification of eminence progenitors is still unclear, especially in light of the broad expression of Tgfbr2 and TGFβ ligands in the developing limb (Pryce et al., 2009). Nonetheless, our finding may help to resolve the ambiguity surrounding the role of this pathway in skeletogenesis. Previous in vitro studies suggested that TGFβs play a crucial role both in the induction of chondroprogenitors and in their differentiation to chondrocytes (Carrington and Reddi, 1990; Chimal-Monroy and Díaz de León, 1997; Kulyk et al., 1989; Leonard et al., 1991; Merino et al., 1998). However, ablation of TGFβ signaling in limb mesenchyme did not appear to affect these processes (Seo and Serra, 2007; Spagnoli et al., 2007). Our findings may imply that this pathway does regulate chondrogenesis, but its influence is limited to the secondary pools of progenitors that establish bone eminences.

Previously, we showed that the SCX/BMP4 pathway induces eminence formation (Blitz et al., 2009); however, we lacked understanding of the exact contribution of this pathway to the process. Here, we provide the missing information by showing that this pathway regulates differentiation of eminence progenitors. Another aspect of the involvement of SCX/BMP4 signaling is the possibility that its serves as a non-autonomous signal by which tendons regulate bone eminence development.

Our finding that eminence progenitors are Scx positive raises a new hypothesis that Bmp4 expression might be regulated by Scx-positive cells within the AU, thus ruling out a non-autonomous role for tendons. Although our lineage experiment cannot provide a clear verdict in this matter, our expression analysis may offer an important clue. As mentioned, between E11.5 and E12.5 there is an overlap in the expression domains of Sox9 and Scx in the forming AU; yet, by E13.5 clear spatial separation between the domains is visible, as Scx expression is restricted to the tendon. At that exact stage, SCX drives Bmp4 expression in cells at the tip of the tendon (Blitz et al., 2009) and Sox9-positive cells undergo differentiation to chondrocytes. This suggests that in the forming tendon, SCX drives Bmp4 expression to regulate non-autonomously the differentiation of Sox9-positive cells to Sox9- and Col2a1-positive chondrocytes, which form the bone eminence.

Finally, as TGFβ signaling was previously shown to regulate Scx expression (Murchison et al., 2007; Pryce et al., 2009), it is possible that TGFβ and BMP4 signaling together form a hierarchical mechanism that coordinates specification and differentiation of bone eminence progenitors.

In this study, we demonstrate that long bones develop modularly from two separate groups of progenitor cells. The first gives rise to the primary structure of the bone, whereas a previously unknown second pool establishes bone eminences. Expression of both Sox9 and Scx by eminence progenitors provides a mechanistic basis for this modularity. The molecular mechanism that underlies the modular morphogenetic process of bone eminence formation involves TGFβ as well as BMP4 signaling. These findings shed new light on two central developmental processes, namely bone morphogenesis and musculoskeletal assembly

We thank to N. Konstantin for expert editorial assistance; S. Kerief for expert technical support; and Dr R. Schweitzer for his helpful discussion and suggestions. Special thanks to all members of the Zelzer laboratory for their advice and suggestions.