HIF1α regulation of Sox9 is necessary to maintain differentiation of hypoxic prechondrogenic cells during early skeletogenesis

The formation of the limb skeleton is initiated when pluripotent mesenchymal cells, derived from the lateral plate mesoderm, commit to the chondrogenic lineage and aggregate to form condensations, which, through a series of differentiation steps, form cartilaginous template of the future skeleton.

During the initial stages of this process, the limb vasculature undergoes a remodeling process that renders the condensing mesenchyme avascularized(Feinberg et al., 1986; Hallmann et al., 1987). As the condensations increase in size, cells differentiate into chondrocytes, forming a cartilaginous template of the future bones. The cartilaginous elements of the autopod develop last, as each digit originates from a single condensation known as the digital ray (Oster,1988). As the digital ray increases in size, it undergoes segmentation, giving rise to the carpal, tarsal and the phalangeal elements. The ensuing formation of joints between the separating segments begins with the appearance of a higher cell density domain, called the interzone, at the site of the future joint. Cells in this region lose typical chondrocyte characteristics, as they reduce the expression of collagen type II (also known as procollagen, type II alpha 1 - Mouse Genome Informatics) and instead express markers such as Wnt9a, Gdf5, Bmp2 and noggin(Hartmann and Tabin, 2001; Seemann et al., 2005). Next,the joint cavitates within the interzone, separating the two skeletal elements(Archer et al., 2003; Mitrovic, 1977; Pacifici et al., 2005).

As development proceeds, the avascularized cartilaginous template is eroded and replaced by vascularized bone in a process termed endochondral ossification (Karsenty and Wagner,2002; Kronenberg,2003; Olsen et al.,2000).

Mesenchymal condensation is the initial step in cartilage formation, and the transcription factor SOX9 is an essential regulator of this process(Bi et al., 1999). Inactivation of Sox9 in limb mesenchymal and neural crest cells results in complete absence of mesenchymal condensation and subsequent failure in cartilage formation (Akiyama et al.,2002; Mori-Akiyama et al.,2003). Furthermore, SOX9 is needed during the sequential steps that follow mesenchymal condensation. Inactivation of Sox9 after the condensation step results in chondrodysplasia with severe reduction in cartilage-specific extracellular matrix protein and attenuation in chondrocyte proliferation (Akiyama et al.,2002).

Two other members of the Sox family, namely L-SOX5 and SOX6, are necessary to maintain the chondrocyte differentiation process. Whereas targeting the expression of either Sox5 or Sox6 resulted in limited skeletal abnormalities, mutant embryos that lacked both genes showed severe aberrations in cartilage formation (Smits et al., 2001). The precise mechanism that regulates the expression of Sox9, Sox5 and Sox6 is unknown; nevertheless, normal expression of Sox9 observed in Sox5- and Sox6-null mice and the loss of Sox5 and Sox6 expression in Sox9-deficient mesenchymal cells position Sox9 upstream from its two family members (Akiyama et al.,2002; Smits et al.,2001).

The regression of blood vessels from sites where mesenchyme condense is likely to induce a localized reduction in oxygen tension at those vessel-free domains, thus forming hypoxic niches. Numerous studies on a variety of cell types have reported that hypoxia has an inhibitory effect on cell differentiation. In view of that, the expected consequence of hypoxic niche formation at the condensation sites would be differentiation arrest. The implication of mesenchymal differentiation into chondrocytes is the existence of a unique mechanism that enables this process to take place under hypoxic conditions.

The transcription factor complex hypoxia-inducible factor 1 (HIF1) is a key mediator of adaptive responses to changes in cellular oxygen level(Semenza, 1998). HIF1 is a heterodimer that consists of HIF1α, the oxygen sensitive subunit, and the constitutively expressed HIF1β (also referred to as ARNT). Under normoxia, HIF1α is hydroxylated by prolyl hydroxylases that act as oxygen sensors (Semenza,2004). The hydroxylation of proline residues is followed by rapid proteasomal degradation (Jaakkola et al.,2001). Conversely, when under hypoxic conditions HIF1α is stabilized, as a result of reduced proteasome-mediated degradation. It then binds to HIF1β and enhances the transcription of genes that are involved in glucose metabolism, angiogenesis, and cell survival(Schofield and Ratcliffe,2004; Semenza,2003).

A previous study identified HIF1α as a critical factor in chondrocyte survival (Schipani et al.,2001). In that study, Hif1α expression was abolished in collagen type II-expressing chondrocytes. Finding that at initial stages of chondrogenesis cells of the forming condensations are hypoxic led us to hypothesize that HIF1α has an additional and yet unidentified role in earlier stages of skeletogenesis.

This study describes a novel role for HIF1α as a regulator of Sox9 expression in hypoxic prechondrogenic condensations. Hif1α deletion in mouse limb mesenchyme led to differentiation arrest of prechondrogenic condensation and resulted in severe skeletal malformations. Moreover, the dramatic reduction in Sox9 expression in the prechondrogenic condensation, accompanied by misexpression of Gdf5 and noggin in Hif1α-depleted limb,provides a molecular mechanism to explain the joint abnormalities observed in Hif1α- depleted limbs. Micromass cultures experiments further supported the role of HIF1α in chondrogenesis: under hypoxic conditions Sox9 expression increased in control cells; this increment was lost in Hif1α- depleted cells. Furthermore, under normoxic conditions Hif1α overexpression induced an increase in Sox9 expression.

Chromatin immunoprecipitation (ChIP) assay provided evidence for direct interaction of HIF1α with the Sox9 promoter, thus supporting direct regulation of HIF1α on Sox9 expression. Our findings establish HIF1α as a key component in the mechanism that regulates chondrogenesis by regulating Sox9 expression in the hypoxic prechondrogenic condensations.

The generation of floxed-Hif1α(Ryan et al., 2000) and Prx1 (also known as Prrx1 - Mouse Genome Informatics)-Cre mice (Logan et al., 2002) have been described previously. In all timed pregnancies the day of the vaginal plug appearance was defined as E0.5. For harvesting of embryos, timed-pregnant female mice were sacrificed by CO₂ intoxication. The gravid uterus was dissected out and suspended in a bath of cold PBS, and the embryos were harvested after amnionectomy and removal of the placenta. Tail genomic DNA was used for genotyping.

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).

Three-dimensional high-resolution images were obtained from the left limb of Prx1-Hif1α and control embryos using microcomputed tomography (GE Healthcare, London, Ontario, Canada). Scans were taken at 8μm isotropic resolution. Images were reconstructed and thresholded to distinguish bone voxels with MicroView software version 5.2.2 (GE Healthcare). One threshold was chosen for all specimens.

For histology and section in situ hybridization, embryos were fixed overnight in 4% PFA-PBS, dehydrated to 100% ethanol, embedded in paraffin and sectioned at 7 μm. Section and whole-mount in situ hybridizations were performed as described previously(Murtaugh et al., 1999; Riddle et al., 1993). All probes are available on request. Hematoxylin and Eosin (H&E) staining was performed following standard protocols.

For immunofluorescence, embryos were embedded in OCT (Tissue-Tek) and 7μm cryostat sections were made. Cryosections were fixed for 20 minutes in 4% PFA-PBS, permeabilized with 0.1% Triton X-100 and incubated with anti-CD31(BD PharMingen), monoclonal anti-HIF1α (Novus Biologicals, Littleton,CO, USA), anti-collagen type II (Developmental Studies Hybridoma Bank, The University of Iowa, IA, USA). Secondary antibodies were purchased from Jackson Laboratories. All experiments were performed with at least three different wild-type (WT) and knockout (KO) limbs from different litters.

Animals were injected with 60 mg/kg hypoxyprobe-1 (Chemicon) and sacrificed 30 minutes after injection. Paraffin sections (7 μm) were stained with FITC-conjugated Hypoxyprobe-1 Mab-1 according to the manufacturer's protocols.

Female mice were injected with 100 mg/kg BrdU (Sigma) and sacrificed 2 hours later. Embryo limbs were collected, fixed with 4% PFA-PBS, embedded in paraffin and 7 μm sections were made. Further processing was performed with a BrdU staining kit (Zymed). To quantify the rate of cell proliferation,serial images of the same digits were collected and BrdU-positive cells (red)and negative cells (gray) in the phalangeal region were counted in four control and four Prx1-Hif1α limbs from two different litters. Statistical significance was determined by Student's t-test.

For micromass cultures, limbs of E11.0-E11.5 floxed-Hif1αembryos were collected, digested with 0.1% collagenase IV, 0.1% trypsin(Sigma) and 2% FCS for 15 minutes. The cell suspension was placed in DMEM-F12,10% FCS. Cells were plated as 10 μl droplets at 2×10⁷cells/ml. Cells were allowed to attach for 75 minutes and were then overlaid with 300 μl of DMEM-F12, 10% FCS containing 6.5×10⁷ viral particles/μl of Adeno-Gfp, Adeno-Cre, Ad-βgal (Gene Transfer Vector Core, University of Iowa) or Adeno-Sox9 (kindly provided by Dr H. Akiyama,Kyoto University, Japan). Medium was changed daily. Cells were cultured either with 20% oxygen (normoxia) or 1% oxygen (hypoxia) balanced with N₂in a 3-Gas incubator (Heraeus) in a humidified atmosphere. Cells were moved to hypoxia 24 hours after the initial plating, and after 96 hours the cultures were either stained with Alcian Blue (pH 1) to visualize chondrogenic nodule formation or harvested to extract RNA. For lentivirus production, cDNA encoding stabilized human HIF1α was digested from PEF-HIF1αP564A/N803A plasmid (kindly provided by Dr M. Whitelaw, University of Adelaid,South Australia) and subcloned into lentiviral transfer vector (kindly provided by Dr Inder M. Verma, Salk Institute, California). Lentivirus production and purification was carried out according to the method of Tiscornia et al. (Tiscornia et al.,2006).

For immunohistochemical staining of micromass cultures, cells were fixed for 15 minutes at room temperature with 4% paraformaldehyde in PBS and then washed twice with PBS. Endogenous peroxidase activity was inactivated by incubating the cells for 30 minutes in 1% H₂O₂ in PBS. Cells were subsequently washed three times with PBS, blocked for 30 minutes with PBS, 10% FCS and 0.1% Triton X-100, and incubated with the primary antibodies against collagen type II (II-II6B3 supernatant, 1:30) from the Developmental Hybridoma Bank (Iowa). The signal was detected using a biotinylated anti-mouse secondary antibody (dilution 1:250; Vector Laboratories) in combination with the ABC Kit (Vector Laboratories) and DAB(Vector Laboratories) as a substrate.

For quantitative RT-PCR analysis, 1 μg total RNA was used to produce first-strand cDNA. Reverse transcription was performed with SuperScriptII(Invitrogen) according to the manufacturer's protocol. Quantitative PCR was performed using SYBR green (Roche). Values were calculated using the second derivative method and normalized to 18S rRNA expression. All primers are available on request.

For western blot analysis, protein was extracted from micromass cultures. Protein concentration was determined using the BCA assay (Pierce). SOX9(1:1000; Santa Cruz Biotechnology) and α-tubulin (1:1000; Sigma)antibodies were used, followed by the appropriate HRP-conjugated secondary antibodies (1:10,000; Jackson ImmunoResearch) and luminol detection.

Micromass lysates were prepared as follows: 20 drops, each 10 μl at 2×10⁷ cells/ml were plated and either cultured under 1%oxygen (hypoxia) or 20% oxygen (normoxia) for 12 hours. Cells were cross-linked in vivo with 1.5% formaldehyde for 10 minutes in the incubator chamber. The cells were washed once with PBS and incubated with 0.25%trypsin-EDTA for 20 minutes. Cells were washed with 2.5 ml of cold PBS and homogenized in 1 ml of buffer I [10 mM Hepes (pH 6.5), 10 mM EDTA, 0.5 mM EGTA, 0.25% Triton X-100]; lysates were washed once with buffer II [10 mM Hepes (pH 6.5), 1 mM EDTA, 0.5 mM EGTA, 200 mM NaCl]. Cell extracts were prepared for ChIP as described previously(Ainbinder et al., 2004). For immunoprecipitation, 5 μl of either polyclonal anti-HIF1α (Abcam) or control IgG were added to 0.5 ml of the soluble chromatin (corresponding to 5×10⁵ cells) and the mixture was incubated overnight at 4°C. Purified DNA from immunoprecipitates, as well as of the input material, was analyzed by real-time PCR using the Roche Sybr green quantification method. Results were normalized and presented as percentage of input DNA. The primers sequences used for amplification of potential HIF1 binding sites in Sox9 and Pgk1 (phosphoglycerate kinase 1)promoters are the following [relative to transcription start site (+1)]: Sox9 amplicon A forward (-1949) GCCTTTGTGCCAGAATACGTGA, reverse(-1695) ACCCTGTAGCCTGTTTACGAGT; amplicon B forward (-917)TGTGACTCAGTCAGGAGGCAAGAA, reverse (-723) TGAAAACCAAAGCCGAGCACCA; amplicon C forward (-495) CATTGCTGTAAACGCCAGCGAA, reverse (-312)GTTTTGGAACGGTCTCCGTGTGAA; amplicon D forward (-102) TCAGCGACTTGCCAACACTGAT,reverse (+46) CCCACAGAAGTTTCCAGGCAGTT; Pgk1 forward (-302)CCTCGCACACATTCCACATCCA, reverse TCAGCGACTTGCCAACACTGAT.

pGL3-basic vector containing 2.8 kb of mouse Sox9 proximal promoter was kindly provided by Dr C. Hartmann (Research Institute for Molecular Pathology, Vienna, Austria). Mutant constructs encoding a 4-nt substitution (CGTG to AAAA) were prepared in the context of the full-length Sox9 promoter by PCR using the following primers: forward 5′-ATAGGTACCACGGAGACAGCATCGAAAAGTGGGGGTGGGGGGTTGTGGAGGGTCCTAGTCTAGACACGCTCGAAAACACGCGCACACACACAC-3′,reverse 5′-TCTCTCGAGCGACTTCCAGCTCAGGGTCTCTA-3′.

PCR was performed under the following conditions: 2 minutes at 94°C;0.3 minutes at 94°C; 0.5 minutes at 55°C; 0.5 minutes at 72°C for 33 cycles and 5 minutes at 72°C. The PCR product was then digested with ACC65I and XhoI and ligated into pGL3-basic promoter. A 2.3 kb ACC65I-ACC65I was digested from the WT mouse Sox9 promoter and ligated into ACC65I to construct a 2.8 kb Sox9 promoter with mutated HRE.

At 70% confluence, 293A cells in a 100-mm dish were transfected with 1μg of either WT or mutated pGL3 Sox9 promoter and 3 μg of PEF-HIF1α P564A/N803A. At 30 hours after transfection, cells were fixed in normal culture medium with formaldehyde at a final concentration of 1% for 10 minutes at 37°C. Plasmid IP was perform as described previously(Ainbinder et al., 2004).

Mesenchymal cells differentiate into chondrocytes in an avascular environment (Fig. 1A)(Feinberg et al., 1986; Hallmann et al., 1987). In order to examine whether the absence of vasculature at the condensation sites forces the condensed mesenchymal cells to differentiate under hypoxic conditions we used hypoxyprobe, a molecular marker that detects hypoxic cells(Arteel et al., 1998). Analysis of E11.5-E12.5 forelimbs identified a signal in the differentiating chondrocytes in the humerus, radius and ulna and in the condensing mesenchyme of the autopod (Fig. 1B-D). By E13.5 the signal was reduced, more prominently in the autopod, and it was elevated in the interzone of the forming joints(Fig. 1E). These results suggest that mesenchyme differentiation into chondrocytes and joint formation occur under low oxygen conditions.

The observation that differentiating mesenchymal cells, chondrocytes and cells of the forming joints are hypoxic led us to examine the expression of Hif1α and Pgk1, a bona fide Hif1αtarget gene, in these cells. Immunofluorescence analysis detected Hif1α expression in differentiating chondrocytes in the E13.5 autopod (Fig. 1F). Pgk1 expression followed the pattern we observed using the hypoxyprobe, as it was detected in proximal elements of the limb and in the forming joints (Fig. 1G,I). Interestingly, Pgk1 expression was observed in cells located at the center of the digits, whereas in the periphery the expression was reduced dramatically. Pgk1 expression was lost in the autopod and the zeugopod upon inactivation of Hif1α in the limbs, indicating that the expression of Pgk1 in the skeletal elements is HIF1αdependent (Fig. 1H,J). These results demonstrate that differentiating prechondrogenic cells in the limb are hypoxic and express Hif1α.

Finding that mesenchymal cell differentiation into chondrocytes and joint formation did take place under oxygen deprivation led us to study whether HIF1α is involved in the mechanism that supports differentiation under these conditions.

Using the Prx1 promoter to drive the expression of Crerecombinase (Logan et al.,2002), we analyzed mice with a conditional deletion of Hif1α in limb bud mesenchyme. Embryos homozygous for floxed-Hif1α and heterozygous for Prx1-Crealleles (Prx1-Hif1α) were compared with embryos heterozygous for floxed-Hif1α and Prx1-Cre alleles(control). Skeletal preparations of E18.5 Prx1-Hif1α embryos demonstrated a significant retardation in skeleton development relative to control: long bones were shorter, severely deformed and less mineralized, with joint fusion in elbow, knee and phalangeal joints(Fig. 2A-C). Examination of the autopod revealed severe defects in carpal and tarsal and digit formation. Cartilage formation was mostly identified in the periphery of the forming digits (Fig. 2C).

Histological examination of the Prx1-Hif1α limb confirmed the previously described role of HIF1α in chondrocyte survival(Schipani et al., 2001). Cell death in the proximal part of the skeleton was initiated at the joint region by E12.5 and was clearly visible at E13.5(Fig. 2D and see Fig. S1 in the supplementary material). By E18.5 cell death at the joint region was extensive in most of the cartilaginous elements of the limb with the exception of the autopod (data not shown). In the autopod we observed only minor cell death,starting at E15.5 (see Fig. S1 in the supplementary material). At that stage we observed cells at the center of the digits that failed to differentiate, as they did not stain with Alcian Blue (see Fig. S2 in the supplementary material).

Blocking the expression of Hif1α in limb mesenchyme led to severe abnormalities in all skeletal elements(Fig. 2). However, the extensive chondrocyte cell death in the proximal skeletal elements prevented sufficient analysis of the direct roles of HIF1α in chondrogenesis. Our observation that cartilage formation in the autopod of the Prx1-Hif1α mouse occurred without noticeable cell death enabled us to study the possible roles of HIF1α in skeleton development regardless of its role in chondrocyte survival.

Further histological examination of the Prx1-Hif1α autopod revealed loss of phalanges in some of the digits(Fig. 2F). Some phalangeal joints were missing or only partially cavitated, whereas some of the joints that were cavitated lacked articular cartilage(Fig. 2E and see Fig. S1 in the supplementary material).

To evaluate the developmental abnormalities in the Prx1-Hif1α ossified skeleton we examined skeletons of post-natal day 21 mice by micro-CT. As can be seen in Fig. 2, the Prx1-Hif1α autopod digit five (D5) is missing one phalanx(Fig. 2G,H). In D3, the joint between the distal and the intermediate phalanges is partially fused(Fig. 2I). In addition, the metacarpophalangeal joints are severely deformed(Fig. 2G). In D1, the distal and the intermediate phalanges are fused and the joint is deformed(Fig. 2G). The sesamoid bones,which are located adjacent to the metacarpophalangeal joint, are fully or partially fused with the contiguous elements in the Prx1-Hif1αskeleton (Fig. 2G-I).

The severe abnormalities of Prx1-Hif1α skeleton strongly suggest that HIF1α plays a key role in the mechanism that regulates cartilage and joint formation.

Histological examination of E13.5 Prx1-Hif1α sections of the autopod revealed cells that appeared as undifferentiated mesenchyme in the center of the forming digits. Furthermore, while the interzone, which marks the forming joint, had emerged in the control digits, it failed to appear in the Prx1-Hif1α digits (Fig. 3A,B).

To study the possibility that prechondrogenic cells in the Prx1-Hif1α forming digits cease to differentiate, we examined the expression of Sox9 in E11.5-E13.5 Prx1-Hif1αautopods by section in situ hybridization. To date, Sox9 is the earliest known marker for condensing mesenchyme. Sox9 expression pattern in E11.5 Prx1-Hif1α limbs was comparable with the control limbs (Fig. 3C); by contrast, Sox9 expression at E12.5 was noticeably altered: expression was observed only in the periphery outlining the forming phalanges, whereas at the center, where we observed undifferentiated mesenchymal cells, Sox9 expression was dramatically reduced(Fig. 3D). By E13.5 the effect we had observed at E12.5 became much more noticeable(Fig. 3E). Concomitantly with the loss of Sox9 expression, the expression of additional markers for chondrocyte differentiation, Sox6, collagen II and aggrecan, were lost as well (Fig. 3F-H). Interestingly, unlike in the control, when Sox9 and collagen II expression was reduced at the sites where joints were forming, the Prx1-Hif1α sections lacked the Sox9 and collagen II segmentation profile. Sox6 expression at E13.5 seemed to increase in the forming joints of the control, but was missing in the Prx1-Hif1α sections, where digits also lacked the interzone(Fig. 3F).

The loss of Sox9, followed by the loss of Sox6, collagen II and aggrecan expression in the Prx1-Hif1α autopod strongly suggest that under hypoxic conditions HIF1α is necessary to maintain the differentiation program of prechondrogenic cells to chondrocytes.

Previous experiments where Sox9 expression was abolished in chondrocytes resulted in reduced chondrocyte proliferation(Akiyama et al., 2002). The reduction in both Sox9 expression and the size of Prx1-Hif1α skeletal elements prompted us to examine whether the loss of HIF1α affected cell proliferation by assessing the incorporation of BrdU into the cells of the forming digits at E14.5. Whereas cell proliferation in the regions outside the forming digits was comparable in Prx1-Hif1α and control autopods, we observed a 3.5-fold reduction in the percentage of cell proliferation in Prx1-Hif1αdigits, including the regions where joints were forming(Fig. 4B).

These results show that the abnormal differentiation of Prx1-Hif1α prechondrogenic cells to chondrocytes is associated with reduced cell proliferation.

Our observations of abnormal joint formation in the Prx1-Hif1α limb (Figs 2, 3 and 4) led us to explore the involvement of HIF1α during early events of joint formation by examining the expression of Gdf5, a marker for joint formation(Storm and Kingsley, 1999). In E12.5 control autopods, Gdf5 expression was mainly observed at joint formation sites and in the interdigital zone surrounding the forming condensation (Fig. 5A). By E13.5-E15.5, Gdf5 expression was reduced outside the forming digits,and it was mostly observed in the developing joints(Fig. 5B-D,G). Gdf5expression in E12.5 Prx1-Hif1α limb was missing from the domains in which joints should have been developed, and was observed instead on the distal side of the developing digits; in the interdigital zone Gdf5 expression was higher than in the control(Fig. 5A). By E13.5, the differences became more obvious: Gdf5 expression in the interdigital zone was still prominent. In the Prx1-Hif1α digits, Gdf5 expression could be observed on the distal side of the metacarpals, but instead of outlining the forming joint the expressing cells were located in the center of the digits(Fig. 5B,C). By E14.5-15.5, in some of the digits we detected indications of aberrant interzone formation; Gdf5 expression domains were broader relative to the control, with indistinct borders (Fig. 5D,G). However, unlike in the control, we could not detect at that stage the expression of interzone markers such as Wnt9a or Bmp2 in the Prx1-Hif1α joints (Fig. 5F-I).

Interestingly, the expression pattern of noggin, a GDF5 antagonist that is known to regulate cartilage and joint formation(Brunet et al., 1998), was also altered in Prx1-Hif1α limb. In E14.5 control autopod, noggin expression could be observed in prehypertrophic chondrocytes, epiphyseal chondrocytes and in the forming joints(Fig. 5E). In the Prx1-Hif1α limb, noggin expression was lost in the center of the forming digits but was instead present in the cells that outlined the digits (Fig. 5E). These results raise the possibility that joint abnormalities in Prx1-Hif1αlimb are a consequence of interference with the GDF5-noggin signaling.

Vegf (vascular endothelial growth factor, also known as Vegfa - Mouse Genome Informatics) is a well-documented transcriptional target of HIF1α(Forsythe et al., 1996; Liu et al., 1995). To examine the possibility that HIF1α regulates Vegf expression in the limb and, as a consequence, regulates limb vasculature, we analyzed Vegf expression in E12.5 Prx1-Hif1α and control limbs by quantitative RT-PCR analysis. Vegf expression in the in Prx1-Hif1α limb was reduced by 30% relative to the control(Fig. 6C). To evaluate whether the reduction of Vegf in the Prx1-Hif1α limb caused vasculature abnormalities we examined vasculature development and patterning in the Prx1-Hif1α limb using sections double-stained with antibodies for PECAM (CD31) and collagen II to identify endothelial cells and chondrocytes, respectively. The vasculature in the Prx1-Hif1αautopod was comparable with that of the control(Fig. 6A,B); however, we observed a substantial decrease in collagen II expression that was well correlated with the reduction we detected in the collagen II mRNA level(Fig. 6B).

These results suggest that the regulation of Vegf and the vasculature in the limb is only partially regulated by HIF1α.

The reduction in the expression of Sox9, Sox6 and collagen II in the Prx1-Hif1α autopod strongly implies that HIF1α is required to maintain the differentiation of prechondrogenic cells to chondrocytes.

To unambiguously demonstrate that HIF1α is cell-autonomously required in mesenchymal precursors for their differentiation into chondrocytes we used high-density micromass culture as an in vitro model(DeLise et al., 2000). Micromass cultures derived from limb buds of floxed-Hif1α embryos were infected by either adeno-Cre virus (AdCre) to delete HIF1α, or adenovirus expressing GFP (AdGfp) as a control. To assess the efficiency of HIF1α deletion by AdCre we measured the expression of Hif1α and Pgk1, a defined HIF1α target gene, using real-time PCR. Micromass cultures infected by AdCre or AdGfp were cultured under hypoxic or normoxic conditions. Under normoxic or hypoxic conditions the expression level of Hif1α in AdCre-infected cells was 25% and 12%, respectively, relative to the control(Fig. 7A). Pgk1expression level in control cells under hypoxic conditions increased more than twofold compared with normoxic levels, whereas in Hif1α-depleted cells this elevation was lost, as the level of expression was similar to normoxic values, suggesting an efficient blockage of HIF1α activity (Fig. 7B).

Next, we studied the ability of Hif1α-depleted mesenchymal precursors to form cartilage nodules. Micromass cultures infected by AdCre or AdGfp were cultured under hypoxic or normoxic conditions and stained with Alcian Blue or tested by immunohistochemistry using collagen II antibody. As can be seen in Fig. 7C, in Hif1α-depleted cells under normoxia there was a mild reduction in nodule formation relative to control cells, whereas under hypoxic conditions, nodule formation by Hif1α-depleted cells was dramatically reduced. Next we examined the expression of collagen II and aggrecan, markers for chondrocyte differentiation, by quantitative RT-PCR. Interestingly, their expression showed a similar pattern to that of Pgk1. In control cells under hypoxic conditions their expression was elevated 1.6- and 2.3-fold, respectively, compared with normoxic levels,whereas in hypoxic Hif1α-depleted cells this increment was lost(Fig. 7D,E).

The reduction in Sox9 expression in vivo along with the expression pattern of Sox9 target genes collagen II and aggrecan in vitro, led us to investigate whether HIF1 regulates Sox9 expression. First we studied the ability of Hif1α overexpression to increase Sox9 expression under normoxic conditions. Sox9 expression in micromass cultures infected with either lentivirus (Lv)-HIF1α or Lv-Gfp (control) was examined by quantitative RT-PCR. In cells infected with Lv-HIF1α the expression of Sox9 was elevated twofold, similar to Pgk1, a bona fide HIF1α target gene(Fig. 8A).

To further establish the regulation of Sox9 by HIF1α, the expression of Sox9 in Hif1α-depleted cells was examined by quantitative RT-PCR and western blot analysis. Under normoxia, in Hif1α-depleted cells there was a 29% reduction in the expression of Sox9 relative to the control(Fig. 8B). Under hypoxic conditions there was a twofold increment in Sox9 expression in control cells, whereas in Hif1α-depleted cells the elevation was lost as the level of expression was similar to normoxic values(Fig. 8B). Concomitantly with the lack of Sox9 mRNA induction in Hif1α-depleted cells, under hypoxic conditions SOX9 protein level was dramatically reduced(Fig. 8C). Next we examined the expression of Sox5 and Sox6; as can be seen in Fig. 8D,E, their expression profile followed Sox9 expression. Under normoxia, in Hif1α-depleted cells there was a mild reduction in the expression of Sox5 and Sox6 relative to the control. Under hypoxic conditions there was a 2.3-fold increment in Sox5 expression and 1.6-fold increment in Sox6 expression in control cells, whereas in the Hif1α-depleted cells the level of expression was similar to values measured under normoxic conditions.

In order to examine whether HIF1α directly regulated Sox9expression we searched the mouse Sox9 promoter for HIF1 consensus binding sites (also referred to as hypoxia response elements or HRE)(Wenger et al., 2005). We identified four putative binding sites within 3.0 kb upstream to the transcription initiation site (Fig. 8F). In order to examine whether HIF1α binds to one or more of the putative sequences, chromatin immunoprecipitation (ChIP) was performed using lysate from micromass cells cultured under either hypoxia or normoxia. The lysate was incubated with either an anti-HIF1α antibody or antiβ-galactosidase antibody (as control). As a positive control we demonstrated binding of HIF1α to Pgk1, in chromatin from hypoxic cells (see Fig. S2 in the supplementary material). Our analysis revealed that HIF1α bound to an HRE sequence located 398 bp upstream of the transcription initiation site (HRE 398). The other three HRE binding sites did not reveal any significant binding(Fig. 8G,H). Interestingly,HIF1α bound to HRE 398 sequence under normoxia as well, although with lower affinity. This result might explain the reduction in Sox9 mRNA level that we observed in cells cultured under normoxia(Fig. 8B).

To further demonstrate the specificity of the HIF1α consensus binding site that we identified in the Sox9 promoter, we substituted four nucleotides in the core consensus sequence of HRE 398 and examined HIF1αbinding to the mutated Sox9 promoter using a plasmid IP experiment. HIF1α binding was evaluated in 293A cells that were co-transfected with an HIF1α-expressing plasmid and with plasmids that contained either the WT Sox9 promoter or a Sox9 promoter with mutations in HRE 398. We detected HIF1α binding to the HRE 398 region in cells that were transfected with the control Sox9 promoter; the binding was lost once the HRE 398 site was mutated (Fig. 8I).

These results indicate that HIF1α directly regulates Sox9expression. In addition, they show that HIF1α-dependent regulation of Sox9 is necessary to maintain Sox5 and Sox6expression under hypoxic conditions.

In this study we describe how an early step in skeletogenesis, namely mesenchymal cell differentiation, which occurs under oxygen deprivation, is regulated by HIF1α. Hif1α deletion in limb mesenchyme led to dramatic reduction in Sox9 expression followed by differentiation arrest that resulted in severe skeletal malformations. Using micromass cultures as an in vitro model for chondrogenesis we found that HIF1αdirectly regulated Sox9 expression, thus providing a molecular mechanism for the abnormalities observed in Prx1-Hif1α limbs. These findings establish HIF1α as a key component in the mechanism that regulates embryonic chondrogenesis.

Until the establishment of a connection with maternal blood supply at E8.5,the murine embryo experiences low oxygen tension within the hypoxic range. At later stages of development, organ growth that precedes vascular development leads to hypoxic micro-environments(Maltepe and Simon, 1998; Mitchell and Yochim, 1968; Rodesch et al., 1992). During evolution, several organs have adapted to hypoxic developmental conditions and integrated hypoxia into their intrinsic genetic program as an external regulatory signal. Organs such as the neural tube(Hogan et al., 2004), placenta(Cowden Dahl et al., 2005; Ambati et al., 2006) and skeleton develop in the absence of embedded vasculature. The hypoxic niches that are formed directly affect the developmental process of each specific organ. The most profound effect of hypoxia during organogenesis is the regulation of differentiation and proliferation of progenitor cells. Hypoxia may either promote or inhibit differentiation in a cell-type-specific manner. For example, whereas hypoxia prevents the differentiation of hES cells(Ezashi et al., 2005) and inhibits myogenesis (Gustafsson et al.,2005), osteogenesis (D'Ippolito et al., 2006; Salim et al.,2004) and adipogenesis (Yun et al., 2002), it promotes the differentiation of mesencephalic precursors (Studer et al.,2000) and enhances hemangioblast specification(Ramirez-Bergeron et al.,2004).

In order to sense oxygen tension and convert the information into a cellular response, cells have developed a molecular signaling pathway in which HIF1 is an essential component (Semenza,2004). Evidence for the significance of this pathway in embryonic development came from genetic studies. Null mutations in Hif1subunits led to early embryonic lethality due to placental failure, neural tube and vascular defects (Maltepe et al.,1997; Semenza et al.,1999). More recent studies have provided molecular insight into the role of HIF1 in the developmental response to hypoxia. Under hypoxia, HIF1 upregulates the expression of Vegf, Flk-1 (also known as Kdr- Mouse Genome Informatics) and erythropoietin, as well as other genes involved in vascular development (Maltepe and Simon, 1998). This molecular response is essential for the proper differentiation and maintenance of the cardiovascular system. HIF1α inhibits the differentiation of myogenic and neural precursor cell lines by enhancing Notch signaling(Gustafsson et al., 2005) and prevents adipocyte differentiation by inhibiting PPARγ2 expression(Yun et al., 2002).

During initial stages of skeletogenesis the prechondrogenic condensations are avascularized and, as shown in our work, hypoxic. Our study provides direct evidence for the key role of HIF1α in the mechanism that has been developed by prechondrogenic cells to support their differentiation into chondrocytes and joint-forming cells. More specifically, under hypoxic conditions HIF1α is necessary to regulate the expression of the key chondrogenic regulator Sox9, in order to maintain chondrogenesis.

However, one interesting question that still remains to be resolved is the evolutionary explanation for the selection of a genetic program that dictates and requires that the chondrogenic process should take place under hypoxic conditions. Although we have no definite answer, we favor the possibility that the driving force behind this selection is to enhance the robustness of the genetic program that regulates chondrogenesis. Limb mesenchyme can differentiate to various lineages including chondrocytes, osteoblasts and tendon-forming cells. It is possible that whereas hypoxia inhibits differentiation of limb mesenchymal cells as a whole(D'Ippolito et al., 2006; Salim et al., 2004), the chondrogenic lineage escapes this inhibition because of the regulation of Sox9 by Hif1α.

The emergence of the interzone is the first histological indication of joint formation (Mitrovic,1977). Molecularly, the expression of the chondrogenic markers Sox9, collagen II and aggrecan decreases in interzone cells, whereas the expression of Gdf5, noggin, Wnt9A and Bmp2 is elevated. In our study, both histological and molecular examinations of Prx1-Hif1α autopods showed abnormal interzone formation(Fig. 5).

Abnormal joint formation in the Prx1-Hif1α limb may result from the reduction in Sox9 and the consequent failure of prechondrogenic cells to differentiate to chondrocytes. This suggests that the ability of condensed mesenchymal cells to adopt interzone cell fate depends on proper differentiation of the flanking cells into chondrocytes. Studies where SOX9 was inactivated in prechondrocytes support this possibility: severe reduction in cartilage formation was followed by fusion of the carpal elements and low expression levels of noggin, Gdf5 and Wnt9a(Akiyama et al., 2002).

An alternative explanation may lie in our observation of abnormal expression of Gdf5 and noggin(Fig. 5A-E). It has been shown that alterations in either the expression or activity of these two genes resulted in multiple joint defects (Kjaer et al., 2006; Lehmann et al.,2003; Seemann et al.,2005). Gdf5 misexpression by implantation of beads into the interdigital region of E12.5 embryos resulted in interference in metacarpophalangeal joint development, with reduction in the expression of joint markers and increase in the expression of chondrogenic markers(Storm and Kingsley,1999).

Noggin haploinsufficiency was recently reported to lead to carpal and tarsal joint fusions (Tylzanowski et al.,2006). Human genetic studies further support this hypothesis:point mutations that altered the activity of GDF5 and its antagonist noggin resulted in brachydactyly and symphalangism(Gong et al., 1999; Marcelino et al., 2001; Seemann et al., 2005). The expression of Gdf5 in Prx1-Hif1α limbs in our study rules out the possibility that HIF1α is necessary for its expression(Fig. 5A-D). Nevertheless, the alterations we observed in Gdf5 and noggin expression patterns in Prx1-Hif1α limbs may indicate that in the absence of HIF1α the fine balance between noggin and GDF5 is disturbed, causing aberrations in joint formation.

Mesenchymal condensation is the initial step in cartilage formation, and SOX9 is an essential regulator of this process(Akiyama et al., 2002). Our finding that Prx1-Hif1α limb mesenchymal cells did condense and initially expressed Sox9 (Fig. 3C) suggests that HIF1α is not necessary for chondrocyte cell fate determination. However, later in development Sox9expression is further required to regulate the differentiation of prechondrogenic cells into chondrocytes. Our histological observation of cells that appeared as undifferentiated mesenchyme and lacked Sox9expression in the Prx1-Hif1α autopod implies that HIF1αis necessary to sustain the chondrogenic program by maintaining Sox9expression in these cells (Fig. 3D,E)

Micromass culture experiments further supported the role of HIF1α in regulating the transition of prechondrogenic cells to chondrocytes by regulating Sox9 expression. Under hypoxia, cartilage nodule formation by Hif1α-deleted mesenchymal cells was dramatically reduced relative to the control (Fig. 7C). Quantitative real-time PCR revealed an HIF1α-dependent induction of Sox9 expression under hypoxic conditions(Fig. 8B). Moreover, forced expression of HIF1α in these cells resulted in a twofold increase in Sox9 mRNA level. These results suggest that HIF1α is necessary to maintain the Sox9 mRNA level under hypoxic conditions(Fig. 8B,C).

Following the same pattern, the expression of Sox5, Sox6(Fig. 8C,D), collagen II and aggrecan (Fig. 7D,E) were elevated in control mesenchymal cells under hypoxic conditions in a HIF1α-dependent manner, as this elevation failed to occur in Hif1α-deleted cells.

A previous study demonstrated that hypoxia could increase the activity of the Sox9 proximal promoter in ST2 cell line. The increment was lost when the HIF1 consensus binding site in the promoter was mutated(Robins et al., 2005). Our analysis revealed four putative HIF1 consensus binding sites in a genomic region spanning 3.0 kb upstream to transcription start site. ChIP and plasmid IP analyses provided evidence for direct interaction of HIF1α with one out of the four sites identified (Fig. 8H,I). Interestingly, Robins et al. identified the same element as a potential HIF1 binding site, thus providing additional and independent support for the direct regulation of Sox9 expression by HIF1α.

With the exception of the three members of the SOX transcription factors family, namely: SOX9, SOX5 and SOX6, very little is known about the transcriptional machinery that regulates the various differentiation steps leading to the formation of a functional chondrocyte. Finding both in vitro and in vivo that the differentiation of mesenchymal cells to chondrocytes required Hif1α expression suggests that HIF1α is an essential component in the transcriptional mechanism that regulates the transition of prechondrogenic cells to chondrocytes.

We are grateful to Dr C. Tabin for the Prx1-Cre mice, to Dr C. Hartmann for the mouse Sox9 proximal promoter and RNA probes, to Dr V. Lefebvre for RNA probes, to Dr Inder M. Verma for lentiviral transfer vector, to Dr A. Chen for helpful suggestions in constructing and producing lentiviruses, to Dr H. Akiyama for Ad-Sox9 virus constructs and to Dr M. Whitelaw for HIF1α expression plasmids. We thank Ms S. Kerief for expert technical support, Mr N. Konstantin for expert editorial assistance and members of the Zelzer laboratory for advice and suggestions. This work was supported by Israel Science Foundation grant 499/05, Minerva grant M941, The Leo and Julia Forchheimer Center for Molecular Genetics, The Stanley Chais New Scientist Fund and The Women's Health Research Center. E.Z. is the incumbent of the Martha S. Sagon Career Development Chair.