Osteo-chondroprogenitor cells are derived from Sox9 expressing precursors

Abstract

The transcription factor Sox9 is expressed in all chondroprogenitors and has an essential role in chondrogenesis. Sox9 is also expressed in other tissues, including central nervous system, neural crest, intestine, pancreas, testis, and endocardial cushions, and plays a crucial role in cell proliferation and differentiation in several of these tissues. To determine the cell fate of Sox9-expressing cells during mouse embryogenesis, we generated mice in which a Cre recombinase gene preceded by an internal ribosome entry site was inserted into the 3′ untranslated region of the Sox9 gene (Sox9-Cre knock-in). In the developing skeleton, Sox9 was expressed before Runx2, an early osteoblast marker gene. Cell fate mapping by using Sox9-Cre;ROSA26 reporter (R26R) mice revealed that Sox9-expressing limb bud mesenchymal cells gave rise to both chondrocytes and osteoblasts. Furthermore, a mutant in which the Osterix gene was inactivated in Sox9-expressing cells exhibited a lack of endochondral and intramembranous ossification and a lack of mature osteoblasts comparable with Osterix-null mutants. In addition, Sox9-expressing limb bud mesenchymal cells also contributed to tendon and synovium formation. By using Sox9-Cre;R26R mice, we also were able to systematically follow Sox9-expressing cells from embryonic day 8.0 to 17.0. Our results showed that Sox9-expressing cells contributed to the formation of all cell types of the spinal cord, epithelium of the intestine, pancreas, and mesenchyme of the testis. Thus, our results strongly suggest that all osteo-chondroprogenitor cells, as well as progenitors in a variety of tissues, are derived from Sox9-expressing precursors during mouse embryogenesis.

Limb skeleton is formed as a cartilage model that undergoes endochondral bone formation. At the initiation of limb development, undifferentiated mesenchymal cells in the lateral plate mesoderm receive proliferation signals from the apical ectodermal ridge. These cells start to aggregate and form mesenchymal condensations, which are the primordia of the limb skeleton, then differentiate into chondrocytes and generate a cartilage skeleton. Cells surrounding the nascent cartilage form the perichondrium and periosteum, specialized structures consisting of thin layers of mesenchymal cells. The cells surrounding the zone of hypertrophic chondrocytes begin to differentiate into osteoblasts and, together with blood vessels and osteoclasts, invade the mineralized cartilage matrix and replace cartilage by bone.

Specific transcription factors regulate the differentiation pathways of chondrocytes and osteoblasts. Sox9, a transcription factor with a high-mobility group DNA-binding domain, activates chondrocyte-specific marker genes, such as Col2a1, Col11a2, and Aggrecan (1–3). Sox9 is expressed in all chondroprogenitors and chondrocytes except hypertrophic chondrocytes (4, 5). Campomelic dysplasia, a human disease that is caused by heterozygous mutations in the Sox9 gene, is characterized by a general hypoplasia of endochondral bones (6, 7). We have recently shown that inactivation of Sox9 in limb buds by using the Cre recombinase/loxP recombination system before chondrogenic mesenchymal condensations form results in the complete absence of mesenchymal condensations and subsequent cartilage and bone formation (8). Moreover, inactivation of Sox9 during or after mesenchymal condensations results in a very severe chondrodysplasia, which is characterized by an almost complete absence of cartilage in the endochondral skeleton (8). Hence, Sox9 is required at sequential steps in chondrogenesis before and after mesenchymal condensations. These results also suggest that Sox9 may be involved in the determination of both chondrogenic and osteogenic cell lineages in limb development.

Runx2 is a member of the Runt-domain family of transcription factors, which form heterodimers with a single ubiquitous polypeptide called Cbfb (9). In addition to the Runx DNA-binding domain, Runx2 contains an active transactivation domain, rich in glutamine and alanine residues, and activates the Osteocalcin (Oc) and Col1a1 genes (10, 11). Cleidocranial dysplasia, a genetic disease in humans that is characterized by hypoplastic clavicles, large open spaces between the frontal and parietal bones of the skull, and other skeletal dysplasias, is caused by heterozygous mutations in the Runx2 gene (12). Runx2-null mice are characterized by a block in osteoblast differentiation and the absence of both endochondral and intramembranous bone formation, indicating that Runx2 has an essential role in the differentiation of mesenchymal progenitors into osteoblasts (10, 13, 14). Thus, Runx2 is an initial marker of the osteogenic cell lineage.

Recently, we identified a second transcription factor required for osteoblast differentiation, called Osterix (Osx) (15), that is specifically expressed in osteoblasts of all endochondral and membranous bones. Osx contains a DNA-binding domain consisting of three C2H2-type zinc fingers at its C terminus that share a high degree of sequence identity with similar motifs in Spl, Sp3, and Sp4. In addition, Osx also contains a proline- and serine-rich transactivation domain and activates the Oc and Col1a1 genes. In Osx-null mutant mice, no endochondral or intramembranous bone formation occurs (15). The mesenchymal cells in Osx-null mutant mice do not deposit bone matrix, and cells in the periosteum and the condensed mesenchyme of membranous skeletal elements cannot differentiate into osteoblasts, although these cells express normal levels of Runx2. Arrest in osteoblast differentiation in Osx-null mutant mice occurs at a later step than that in Runx2-null mice. Indeed, expression of Osx requires Runx2, indicating that Osx acts downstream of Runx2.

On the basis of our previous mouse genetic studies, we hypothesized that Sox9 is expressed in osteo-chondroprogenitors during endochondral bone formation before chondrogenic and osteogenic cell lineages are distinct (8). To test this hypothesis, we generated mice carrying the Cre recombinase gene inserted into the 3′-UTR of the Sox9 gene and crossed these mice with the Rosa26 reporter (R26R) strain to follow the cell fate of Sox9-expressing cells during limb development. We also crossed this Cre line with mice carrying a floxed Osx allele to inactivate the Osx gene in Sox9-expressing cells. Our results indicated that Sox9 defines common progenitors of chondrogenic and osteogenic cell lineages.

Materials and Methods

Generation of Sox9-Cre Knock-In Mice and Floxed Osx Mice. A Sox9 clone was isolated from a mouse 129/Sv genomic DNA library. The targeting vector spanned a 7.7-kb fragment of the Sox9 gene and an internal ribosome entry site (IRES)-Cre-pA/FRT-flanked PGK-neo-bpA cassette was inserted into a HpaI site within the 3′-UTR in exon3. An MC1-tk-pA herpes simplex virus thymidine kinase expression cassette was added onto the 3′ arm of homology to enrich for homologous recombinants by negative selection with 1-(2′-deoxy-2-fluoro′-β-d-arabinofuranosyl)-5-iodouracil (FIAU). The targeting vector was introduced into 129SvEv AB1 mouse ES cells and G418/FIAU-resistant ES cell clones were initially screened by Southern blot analysis of BamHI-digested genomic DNA by using a 3′ probe external to the region of vector homology. Mouse chimeras were generated by injection of mutant ES cell clones into C57BL/6 host blastocysts, and the chimeras obtained were bred with C57BL/6 mice to generate Sox9-Cre knock-in heterozygous mice.

An Osx clone was also isolated from a mouse129/Sv genomic DNA library. To construct the targeting vector for the floxed Osx allele, an 8.0-kb fragment was subcloned. Exon 2 was flanked by two loxP sites; the first was in intron 1 and the second in the 3′-flanking region. An FRT-flanked PGK-neo-bpA cassette and an IRES-EGFP-pA cassette, which contained a splicing acceptor signal at the 5′ end, were inserted into the BamHI site 3′ to the second loxP site. The target vector was introduced into 129SvEv AB1 ES cells, and G418-resistant ES cell clones were initially screened by Southern blot analysis of EcoRV-digested genomic DNA with a 3′ probe external to the region of homology. Homologous recombination was verified by using a 5′ probe external to the region of homology. Mouse chimeras were generated by C57BL/6 host blastocyst injection of mutant ES cell clones, and chimeras obtained were bred with C57BL/6 mice to generate floxed Osx heterozygous mice. The FRT-flanked PGK-neo-bpA cassette was removed by Flp-mediated recombination with ACTFlpe transgenic mice (16).

In a first cross, Sox9-Cre knock-in mice were mated with mice heterozygous for the floxed Osx allele. The offspring inheriting both the Cre recombinase gene and the floxed allele were then mated with Osx^lacZ/wt heterozygous mice (15) to obtain embryos with the Osx^flox/lacZ and the Sox9-Cre knock-in allele. Routine mouse genotyping was performed by PCR. The following primer pairs were used: floxed Osx allele (5′-CTTGGGAACACTGAAGCTGT-3′ and 5′-CTGTCTTCACCTCAATTCTATT-3′); Osx-lacZ allele (5′-CGGATAAACGGAACTGGAAA-3′ and 5′-TAATCACGACGCGCTGTATC-3′); and Sox9-Cre knock-in allele (5′-TCCAATTTACTGACCGTACACCAA-3′ and 5′-CCTGATCCTGGCAATTTCGGCTA-3′).

Histological Analysis. β-gal staining of embryos was performed as described in ref. 17. For the histological analysis, embryos were fixed with 4% paraformaldehyde and embedded in paraffin. Sections of 7 μm were stained with alcian blue, hematoxylin, and Treosin (Statlab, Lewisville, TX). Immunohistochemical staining of sections was performed by using peroxidase chromogens (Zymed, South San Francisco, CA)/TrueBlue substrate (Kirkegaard & Perry Laboratories, Gaithersburg, MD) with rabbit polyclonal anti-Sox9 antibody (1:100) (1). RNA in situ hybridization analysis was carried out as described in ref. 8. Pictures of hybridization signals were superimposed with blue f luorescence images of cell nuclei stained with Hoechst 33258 dye.

Results

Generation of Sox9-Cre Knock-In Mice. To follow the fate of Sox9-expressing cells throughout embryonic development, we generated mice that express Cre recombinase in a Sox9-specific pattern. The perinatal lethality of heterozygous Sox9 mutant mice precludes the generation of mice in which specific DNAs are inserted into the coding regions of the Sox9 gene (18, 19). Thus, to overcome this problem, we inserted IRES-Cre-pA into the 3′-UTR of the Sox9 gene in exon3 just 5′ to the polyadenylation signal (Fig. 5, which is published as supporting information on the PNAS web site). In contrast to Sox9 heterozygous mutant mice, which died perinatally with cleft palate, respiratory distress, hypoplasia, and bending of many skeletal structures derived from cartilage precursors (18, 19), heterozygous Sox9-Cre knock-in mice were viable and fertile and showed no noticeable phenotypic changes.

Sox9 Expression Precedes Runx2 and Osx Expression During Limb Bud Development. Sox9 was expressed in the chondrogenic cell lineage during limb bud development. Indeed, in embryonic day (E)10.0 Sox9 started to be expressed in limb bud mesenchyme, then it was expressed in the cells in mesenchymal condensations and in chondrocytes (Fig. 1A). In contrast, the expression of Runx2 and Osx, early markers of the osteogenic cell lineage, was first detected in E10.5 and E13.5 in limb bud mesenchyme, respectively, and followed Sox9 expression (Fig. 1 A), indicating that Sox9 expression occurs before the determination of an osteogenic cell lineage in limb buds.

Fig. 1.

Comparison of Sox9, Runx2, and Osx expression during embryonic development with lacZ expression in Sox9-Cre;R26R compound heterozygous embryos. (A) Expression of Sox9, Runx2, and Osx during limb bud development. β-gal activity in limb buds of Sox9 heterozygous (Sox9^lacZ/wt) (19), Runx2 heterozygous (Runx2^lacZ/wt) (14), and Osx heterozygous (Osx^lacZ/wt) (15) embryos from E9.5 to E16.5. The arrows indicate the expression of Sox9 in E10.0. The arrowheads indicate the expression of Runx2 in E10.5 and E11.5. (B) Sections of limb buds in Sox9-Cre;R26R compound heterozygous embryos from E10.5 to E17.0 stained with whole-mount β-gal staining. c, resting and proliferating chondrocytes; h, hypertrophic chondrocytes; p, periosteum; ob, osteoblasts; S, synovium; T, tendon. (C) Sections of limb buds in Sox9 heterozygous (Sox9^lacZ/wt) embryos in E13.5 stained with whole-mount β-gal staining. (D) Sections of limb buds in Prx1-Cre;R26R compound heterozygous embryos in E16.5 with whole-mount β-gal staining. The abbreviations are the same as in B.

Sox9-Expressing Limb Bud Mesenchymal Cells Give Rise to Chondrocytes and Osteoblasts. We hypothesized that Sox9-expressing limb bud mesenchymal cells give rise to both the chondrogenic and osteogenic cell lineages. To confirm this hypothesis, we crossed Sox9-Cre knock-in mice with R26R mice. Expression of Cre recombinase in R26R mice results in activation of the lacZ gene, and, therefore, robust β-gal staining provides an indelible marker of cells that have expressed Sox9 at any time during development (20–22). In Sox9-Cre;R26R E10.5 embryos, β-gal staining was detected in limb bud mesenchyme, and, in E13.5 embryos, all cells in cartilage primordia and perichondrium were β-gal positive (Fig. 1B). Indeed, the β-gal positive region in the limb buds of Sox9-Cre;R26R embryos was markedly wider than in the limb buds of Sox9^lacZ/wt embryos, which represents only a chondrogenic cell population (Fig. 1C). In E17.0 limb buds, all chondrocytes as well as perichondrial, periosteal, and osteoblast cells were β-gal positive (Fig. 1B), indicating that Sox9-expressing limb bud mesenchymal cells give rise to both chondrogenic and osteogenic cell lineages. In addition, β-gal positive cells were detected in tendons and synoviums in E17.0, indicating that Sox9-expressing limb bud mesenchymal cells also give rise to tendon cells and synovial cells in limb bud development.

Inactivation of Conditional Osx Alleles by Sox9-Cre Results in Lack of Bone Formation. Cell fate mapping by using Sox9-Cre knock-in mice indicated that Sox9-expressing limb bud mesenchymal cells give rise to osteoblasts. To confirm this result, we generated mutant mice in which both Osx alleles were inactivated in Sox9-expressing cells. Osx is a transcription factor that is specifically expressed in osteoblasts. We recently reported that Osx is a downstream gene of Runx2 in bone formation and that the Osx-null mutant mice lack bones completely (15). To generate the conditional Osx-null mice, we used mice carrying the floxed Osx allele, in which the DNA segment including exon2 was flanked by loxP sites (Fig. 2A). Both heterozygous and homozygous animals were viable and fertile and showed no noticeable phenotypic change. Using the mating scheme described in Materials and Methods, we produced mutants carrying the Sox9-Cre gene in which one Osx allele was replaced by the lacZ gene (15) and the other was replaced by a floxed Osx allele (Fig. 2B). The conditional Osx-null mutant mice resulting from expression of the Sox9-Cre gene, which were recovered with the expected Mendelian frequency, died in the immediate postnatal period because of the absence of bones. Staining of skeletal preparations of E17.5 conditional null mutant mice embryos with alcian blue and alizarin red indicated virtually no mineralization in any facial and skull bones generated by intramembranous bone formation, although very small parts of the maxilla, mandible, and parietal bones were calcified (Fig. 2C). Other skeletal elements formed by endochondral bone formation, including the ribs, limb skeletons, and vertebrae, were hypoplastic, bent, and often deformed, comparable with the phenotype in the Osx-null mice. Histological analysis showed that the humerus of E17.5 wild-type embryos displayed a typical cellular organization of bone trabeculae, bone collars, and a zone of hypertrophic chondrocytes adjacent to the chondro-osseous junction (Fig. 2D). In contrast, the humerus of E17.5 Osx-null and conditional Osx-null mutant mice had no bone collars and no bone trabeculae. The abnormalities in the humerus were present in all other endochondral skeletal elements (data not shown).

Fig. 2.

Targeting strategy for conditional inactivation of the Osx gene. (A) Structure of the genomic Osx locus, targeting vector, and the homologous recombined allele. Exons are depicted as filled boxes, and intronic sequences are shown as solid lines. The FRT-flanked PGK-neo bpA and the IRES-EGFP-pA cassettes are depicted as open boxes. DNA fragments revealed in Southern analysis are indicated as arrows with the restriction enzymes and the probes. RI, EcoRI; N, NheI; B, BamHI; X, XbaI; RV, EcoRV. (B) Southern blot analysis of genomic DNA. (C–E) Analysis of skeletal phenotypes in Osx ^flox/lacZ;Sox9-Cre knock-in mice. Skeletons in E17.5 embryos stained with alcian blue followed by alizarin red showed no mineralization of bones in Osx-null and the conditional Osx-null mutants (C). Histological analysis of humerus stained by alcian blue, hematoxylin, and Treosin revealed no bone trabeculae or mineralization in Osx-null and the conditional Osx-null mutants in E17.5 (D). Expression of Oc mRNA was not detected in humerus of E17.5 Osx-null and the conditional Osx-null mutants (E).

To better characterize the abnormalities of endochondral bones in the conditional Osx-null mutant mice, we examined the expression of the Oc gene, a late osteoblast marker, by in situ hybridization (Fig. 2E). In E17.5 wild-type embryos, Oc was strongly expressed in bones. In contrast, Oc was not expressed in E17.5 Osx-null and conditional Osx-null embryos. Thus, these results indicate that osteoblast differentiation was arrested in the mutant embryos in which the Osx gene was inactivated in Sox9-expressing limb mesenchymal cells. Hence, Sox9-expressing mesenchymal cells are osteo-chondroprogenitors that differentiate into both chondrocytes and osteoblasts. The absence of alizarin red staining in membranous skeletal elements in Osx^flox/LacZ;Sox9-Cre embryos also indicated that Sox9-expressing cells are the progenitors of osteoblasts in membranous bones.

Paired-Related Homeobox Gene-1 (Prx1)-Expressing Limb Bud Mesenchymal Cells Give Rise to All of the Mesenchymal Cell Lineages in Limb Buds. Prx1 is a homeobox-containing gene that starts to be expressed in all limb bud mesenchymal cells at E9.0 in advance of Sox9 (23, 24). Mice homozygous for a mutant Prx1 allele exhibited the loss or malformation of facial, limb, and vertebral skeletal structures because of a defect in the formation and growth of chondrogenic and osteogenic precursors (25). We previously reported that the conditional Sox9-null mutants harboring a Prx1-Cre transgene showed a complete lack of both cartilage and bone in limbs (8). Therefore, these findings strongly suggest that Prx1-expressing limb bud mesenchymal cells give rise to Sox9-expressing osteo-chondroprogenitors. In contrast to the expression of the Prx1 transgene, Sox9 is only expressed in a subpopulation of limb mesenchymal cells. Indeed, in Prx1-Cre;R26R embryos (23, 24), all of the cells in skeletal elements, the fibroblasts in connective tissues, and the tendon cells and the synovial cells were β-gal positive (Fig. 1D). Hence, Prx1-expressing undifferentiated limb bud mesenchymal cells differentiate into all of the mesenchymal cell populations during limb bud development.

Sox9 Defines Progenitors in Testis, Intestine, Spinal Cord, and Pancreas During Mouse Embryogenesis. We also examined the fates of Sox9-expressing cells in other tissues during embryogenesis. Whole-mount β-gal staining in Sox9-Cre knock-in;R26R embryos was first detected in E8.0 in notochord, node, and endodermal linings (Fig. 3A). From E10.5, β-gal staining was very strong in nonneural ectoderm throughout embryonic development, hampering the observation of β-gal staining of internal organs. For this reason, we examined β-gal staining in sections made from paraffin-embedded whole-mount β-gal stained E17.0 embryos in which the skin had been removed. In the male gonad, Sox9 was expressed in Sertoli cells (Fig. 3B). In contrast, all of the mesenchymal cells including Sertoli cells and Leydig cells in Sox9-Cre;R26R embryos were β-gal positive, but the germ cells were β-gal negative (Fig. 3B). In the intestine, because Sox9 was expressed in stem cells and progenitor cells in the bottom of crypts in villi, all of the epithelial cells were β-gal positive; furthermore, the neurogenic cells derived from the neural crest in which Sox9 was expressed (4, 5, 26) were also β-gal positive (Fig. 3B). In the spinal cord, Sox9 expression was detected throughout the ventricular neuroepithelial layer in which neural progenitor cells are located and also in the cells migrating outward along the dorsoventral axis. Indeed, all of the neural cells, neurons and glia, in the spinal cord of Sox9-Cre;R26R embryos were β-gal positive. In the pancreas, Sox9 was expressed in ductal epithelial cells. In contrast, in Sox9-Cre;R26R embryos, not only the ductal epithelial cells but also all of the endocrine and exocrine cells were β-gal positive (Fig. 3B). These observations strongly support the hypothesis that Sox9 defines progenitors in these different tissues during mouse embryogenesis.

Fig. 3.

Comparison of the pattern of cells expressing Sox9 in E17.0 intestine, testis, spinal cord, and pancreas with the pattern of cells that are derived from Sox9-expressing cells in these embryos. (A) Contribution of Sox9-expressing cells during mouse embryogenesis. β-gal activity in Sox9-Cre;R26R compound heterozygous embryos in E8.0 and E10.5. (B) Comparison between the expression of Sox9, assessed by immunohistochemistry (intestine and testis) of wild-type embryos or by β-gal staining (spinal cord and pancreas) of Sox9 heterozygous (Sox9^lacZ/wt) embryos, and the distribution of cells derived from Sox9-expressing cells in E17.0 Sox9-Cre;R26R embryos. The arrowhead indicates enteric neurons.

Discussion

We generated a mouse strain that expresses Cre recombinase in Sox9-expressing cells during embryogenesis. An IRES-Cre-pA cassette was inserted into the 3′-UTR of the Sox9 gene, resulting in Cre recombinase expression directly under Sox9 cis-regulatory control via production of a bicistronic Sox9-Cre recombinase mRNA. This approach avoided the perinatal lethality of mice that would have resulted from the insertion of Cre in the coding portion of Sox9 and inactivation of the Sox9 gene. Expression of Cre recombinase was assessed by crossing Sox9-Cre knock-in mice to the Cre-dependent R26R strain. The advantage of using this genetic system to examine gene expression patterns is that the R26R strain provides a permanent lineage record through the constitutive β-gal expression in cells that express Cre recombinase even transiently in the precursor cells of different cell lineages (20–22).

During mouse limb development, Sox9 expression was first detected in a subpopulation of limb mesenchyme in E10.0. In contrast, Runx2 expression was detected in the lateral plate mesoderm from E10.5, and Osx expression was seen in limb mesenchyme from E13.5. These observations strongly suggest that the osteogenic lineage appeared after Sox9-expressing cells in limb bud development. Indeed, this hypothesis is supported by β-gal expression in Sox9-Cre knock-in;R26R compound heterozygous embryos. In these embryos, β-gal expression was detected in mesenchymal cells before mesenchymal condensation, whereas after mesenchymal condensations, β-gal expression was found in all of the skeletal components, including not only the chondrocytes but the perichondrium, the periosteum, and osteoblasts. Thus, these observations indicate that Sox9-expressing mesenchymal cells, present before mesenchymal condensations occur, are osteo-chondroprogenitors. Indeed, the inactivation of the Osx gene, which presumably takes place in Sox9-expressing mesenchymal cells before mesenchymal condensations, results in a complete block of osteoblast differentiation. Hence, Sox9 defines osteo-chondroprogenitors in the mesenchyme of developing limb buds.

We speculate that in endochondral skeletal elements, the osteogenic lineage cells segregate from Sox9-expressing osteochondroprogenitors at the stage of mesenchymal condensations. This idea is supported first by histological analysis, which showed that Sox9 was expressed in limb mesenchymal cells before mesenchymal condensations (Fig. 1 A) (4, 5). In contrast, Runx2 expression was detected at the stage of mesenchymal condensations (Fig. 1 A) (27). Second, inactivation of the Sox9 gene in limb buds before mesenchymal condensations prevented mesenchymal condensations and subsequent cartilage and bone formation, although the Sox9-null cells are present in their appropriate locations (8). In contrast, Runx2-null mutants form mesenchymal condensations normally (13, 14). Thus, these observations strongly suggest that the establishment of an osteogenic cell lineage occurs during mesenchymal condensations. We speculate that in endochondral skeletal elements, the process of mesenchymal condensation plays a major role in the establishment of the osteoblast lineage.

During limb bud development in mouse embryos, undifferentiated limb bud mesenchymal cells that were derived from the lateral plate mesoderm differentiate and establish mesenchymal cell lineages, including chondrocytes, osteoblasts, and fibroblasts. Previous studies showed that Prx1 is expressed in the early limb bud mesenchyme in mouse embryos (23) and that Cre recombinase activity is present in all mesenchymal cells, including osteo-chondroprogenitors, in E10.5 forelimb buds of Prx1-Cre transgenic mice (24). In E16.5 Prx1-Cre;R26R compound heterozygous embryos, β-gal activity was detected in all of the skeletal elements, including chondrocytes, perichondrium, periosteum, and osteoblasts, and all fibroblasts in connective tissues, tendon cells, and synovial cells. In these embryos, muscle and skin did not stain for β-gal, indicating that these cell lineages do not derive from Prx1-Cre expressing cells in limb buds (data not shown). In Sox9-Cre;R26R embryos not all tendon cells stained for β-gal. The difference between β-gal staining of tendons in Prx1-Cre;R26R and that of tendons in Sox9-Cre;R26R embryos may be due to the activation of the Prx1 promoter at the tendon insertion points of the skeletal muscles (23). We therefore concluded that Prx1-expressing undifferentiated limb bud mesenchyme gives rise to all mesenchymal cell populations, including Sox9-expressing osteo-chondroprogenitors, which differentiate into Sox9-expressing chondrocytes and Runx2-expressing osteoblasts during limb development (Fig. 4).

Fig. 4.

Model of establishment of mesenchymal cell lineages during limb bud development. Prx1-expressing undifferentiated limb bud mesenchyme gives rise to Sox9-expressing osteo-chondroprogenitors, fibroblasts, tendon cells, and synovial cells. These Sox9-expressing osteo-chondroprogenitors form mesenchymal condensations, in which Runx2-expressing osteogenic cells are separated from Sox9-expressing chondrocytes.

The absence of membranous bones in Osx^flox/LacZ;Sox9-Cre embryos indicates that osteoblasts in membranous bones in the skull derive from Sox9-expressing precursor cells. In previous experiments, we used Wnt1-Cre to inactivate Sox9 in delaminating cranial neural crest cells before frank emigration. In these embryos, endochondral bones did not form, but intramembranous bone formed normally (26). Together, these findings suggest that the segregation of two of the skeletal lineages arising from Sox9-expressing cranial neural crest cells, namely osteoblast progenitors of membranous bones and chondrocyte precursors of endochondral bones, occurs in the cranial neural crest before the expression of Wnt-1 Cre and, thus, before emigration of neural crest cells.

In Sox9-Cre;R26R compound heterozygous mice, β-gal expression was detected first in E8.0 embryos in node and notochord, and then in neural crest and its derivatives, including skin and dorsal root ganglia. Comparison of β-gal expression with Sox9 expression revealed that Sox9-expressing cells gave rise to all mesenchymal cell lineages in the testis. Sox9-expressing cells also gave rise to all cell lineages in the spinal cord, pancreas, and epithelium of the intestines. These observations therefore indicate that Sox9 is expressed in the progenitor cells of all of the cells that are present later in these tissues of mouse embryos. In addition, Sox9 may play crucial roles in cell proliferation and differentiation of these progenitors. Indeed, Sox9-null cells in the testis, spinal cord, and pancreas do not proliferate nor differentiate, as do wild-type Sox9-expressing cells in these tissues (H.A., unpublished data; refs. 28 and 29).