During embryonic development, the formation of the skeleton initiates when mesenchymal cells condense and differentiate into either osteoblasts, generating the intramembranous bones of the skull, or into chondrocytes, building cartilaginous templates of all the future long bones. The cartilaginous mold is gradually replaced by bone in a process called endochondral ossification.1–4 Chondrocytes in the cartilaginous mold, which is also known as fetal growth plate, synthesize collagen type II, are highly proliferative, and while they divide they pile up to form a columnar layer. The most distal cells of the columnar layer stop proliferating, exit the cell cycle, and differentiate into hypertrophic chondrocytes producing collagen type X and mineralizing their surrounding matrix. Terminal differentiation and death of the hypertrophic chondrocytes coincides with invasion by blood vessels and osteoblast precursors and replacement of the cartilage by the bone and marrow cavity (called the primary ossification center or, later, the primary spongiosa).2, 5
The fetal growth plate is a unique mesenchymal tissue in that it is itself avascular, whereas the angiogenic switch of the terminal hypertrophic cartilage is associated with its decay and replacement by bone. As shown previously, the fetal growth plate contains a central hypoxic region, whereas the late hypertrophic chondrocytes at the border with the primary spongiosa are not hypoxic.6–9 This specific spatial distribution of hypoxic and oxygenated areas in the cartilage is consistent with both its avascularity and, conversely, with the extensive vascularization of the surrounding soft tissue and of the primary spongiosa.
In the last few years, studies using genetically modified mice have shed light on how the genetic program controlled by hypoxia modulates endochondral bone development via the essential and nonredundant role of hypoxia-inducible transcription factor 1 (HIF-1) in chondrogenesis in vivo.6, 8, 10, 11 Conditional deletion of HIF-1α in limb bud mesenchyme or in chondrocytes, achieved by the use of the Cre-loxP strategy, caused massive cell death of the inner chondrocyte layer in the developing growth plate.
HIF-1 is a heterodimer of two proteins, HIF-1α and HIF-1β; HIF-1β is constitutively expressed, whereas HIF-1α is the hypoxia-responsive component of the complex.12, 13 Particularly the stability of the HIF-1α protein is hypoxia-sensitive, through oxygen-dependent hydroxylation of specific residues within its amino acid sequence. In nonhypoxic conditions, a family of HIF prolyl 4-hydroxylases is responsible for the hydroxylation of two proline residues (P402 and P564) in the oxygen-dependent degradation domain (ODDD) of HIF-1α.14 The E3 ubiquitin ligase Von Hippel-Lindau (VHL) binds to the hydroxylated HIF-1α, and targets it to the proteasome for degradation.15, 16 In hypoxic conditions (estimated at oxygen dropping below 5%),17 this hydroxylation of HIF-1α does not occur, and HIF-1α is stabilized and able to translocate to the nucleus, heterodimerize with the β subunit, and initiate its transcriptional program along with other cofactors. As a result, a variety of pathways involved in the cellular adaptation to hypoxia are activated, including key regulators of glucose utilization and cell metabolism (stimulating anaerobic glycolysis), angiogenesis, and erythropoiesis.17 Thus, the massive cell death observed centrally in cartilage lacking HIF-1α suggests that HIF-1α is required for turning on pathways that allow chondrocytes to survive and differentiate in a hypoxic environment.
Besides genes related to metabolism, a main downstream target of HIF-1α is vascular endothelial growth factor A (VEGF), a principal regulator of blood vessel formation and hematopoiesis.18 VEGF is a homodimeric glycoprotein of 45 kDA belonging to the dimeric cysteine-knot growth factor superfamily.18, 19 VEGF binds to and activates two tyrosine kinase receptors, VEGFR1 (Flt-1) and VEGFR2 (KDR/Flk-1), which regulate both physiological and pathological angiogenesis.20 The VEGF gene encodes various differentially spliced isoforms, the three main ones in mice being VEGF120, VEGF164, and VEGF188. In contrast to VEGF188, the soluble isoform VEGF120 does not bind the extracellular matrix component heparan sulfate or the coreceptor neuropilin; VEGF164 displays combined properties, ie, it is both soluble and able to bind heparan sulphate and neuropilin.18 In the embryo, very strict regulation of the levels of VEGF signaling is essential for normal angiogenesis; for instance, heterozygous deletion of the VEGF gene results in early embryonic lethality because of defective vascular development.21, 22
Studies by us and others have demonstrated that in the fetal growth plate VEGF is expressed not only in late hypertrophic chondrocytes, where it is critical for blood vessel invasion and replacement of cartilage by bone,23–28 but also, even if at a considerably lower level, in the center of the proliferative and upper hypertrophic layers, ie, in the hypoxic zones of the growth plate where also HIF-1α is present.7, 27, 29 Notably, both the universal knockout of VEGF164 and VEGF120 and the conditional knockout of all three VEGF isoforms in chondrocytes cause a cell death phenotype in the center of the fetal growth plate that closely mimics what is observed in the HIF-1α–deficient growth plates, though it is overall milder and presents a little later in development.7, 27 Moreover, it was shown that hypoxia increases VEGF accumulation in chondrocytes in an HIF-1α–dependent manner in vitro.30–32
Although several of these previous findings suggest the possibility that the expression of VEGF in the hypoxic domain of the growth plate is HIF-1α–dependent, this has not unambiguously been demonstrated. In fetal growth plates in which HIF-1α is constitutively stabilized as a result of conditional inactivation of VHL, the expression of VEGF is upregulated and its pattern perturbed.29 Surprisingly, however, VEGF expression appeared also to be upregulated in the HIF-1α–deficient growth plate, at least in the remaining viable chondrocytes when cell death was marked in the center.6 This may be explained by the severe hypoxia in HIF-1α–deficient growth plates6 because hypoxia per se is a strong inducer of VEGF expression by a variety of transcriptional, posttranscriptional, and translational mechanisms including, but not limited to, actions posed by HIF-1α.33–35
In recognition of these complexities, we here took a genetic approach to analyze the HIF-1α–VEGF network in the developing growth plate by assessing whether transgenic expression of VEGF164 could rescue the survival of chondrocytes lacking HIF-1α.
Materials and Methods
Animals and experimental design
To generate the in vivo models used in this study, the following previously published mouse lines were crossed according to the schemes outlined in Results: Collagen type II (Col2)-Cre transgenic (Tg),36 VEGF floxed,37 HIF floxed,6 and Rosa26-VEGF164 Tg mice.38 Mice were bred under conventional conditions and mated overnight to obtain timed pregnancies, with the morning of the vaginal plug appearance being defined as embryonic day (E)0.5. Embryos were harvested at E14.5, E16.5, and E18.5 by cesarean section of timed-pregnant female mice that received hypoxyprobe (60 mg/kg body weight pimonidazole; Natural Pharmacia International, Inc., Belmont, MA, USA) 2 hours before being sacrificed by CO2 intoxication. Genotyping was done on DNA extracted from tail or liver biopsies by PCR using the primers published in the articles referenced above. Animal breeding and experimental sacrifices were performed at Ghent University and KULeuven, with the approval by the respective ethical committees.
Skeletal preparation and histological methods
Skeletons were stained whole-mount with alizarin red and alcian blue according to the standard protocol as before.28, 38 For histological analysis, dissected limbs from E14.5, E16.5, and E18.5 embryos were fixed overnight at 4°C in freshly prepared 2% paraformaldehyde (PFA) in PBS. After subsequent PBS rinses and stepwise dehydration to 70% ethanol, the specimens were processed for paraffin embedding. Sections were cut at 4 to 5 µm, stained with hematoxylin and eosin (H&E) or safraninO, or subjected to one of the following specific stainings. Immunohistochemistry to detect the endothelial cell antigen CD31 (PECAM-1) was done as described before.28 Detection of hypoxyprobe binding was performed using the Hypoxyprobe-1 Plus Kit (Natural Pharmacia International, Inc.) with the fluorescein isothiocyanate (FITC)-conjugated antibody diluted 1:125 in blocking buffer and employing a peroxidase-conjugated anti-FITC secondary antibody (Chemicon, Billerica, MA, USA) at 1:50. In situ hybridization using complementary 35S-labeled riboprobes was performed as described previously.10 For TdT-mediated dUTP-biotin nick end labeling (TUNEL) assay, sections were permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate, followed by the In Situ Cell Death Detection Kit (Roche, Mannheim, Germany) procedure. Each staining was qualitatively analyzed in a large number of sections retrieved at various locations throughout the specimens, employing as far as possible an initial blind semi-scoring analysis and using a minimum of three embryos per genotype. The vascularization of the fetal growth plate cartilage was quantified on CD31-stained sections from E16.5 control (n = 7), HIF-1α cKO (n = 5), and VEGF164 Tg (n = 2) hind limbs, analyzing tibias and femurs (three sections per bone) independently. Histomorphometric analysis consisted of measuring the percentage of the cartilage surface that was immediately flanked by blood vessels, discriminating the periarticular cartilage surface, the cartilage-perichondrium interface enclosing the columnar chondrocyte region, and the metaphyseal hypertrophic chondrocyte border or chondro-osseous junction, as illustrated in detail in Supplemental Fig. S4.
For gene expression analysis, E14.5, E16.5, and E18.5 embryonic tibias and femurs were dissected free of surrounding soft tissues and snap-frozen immediately in liquid nitrogen. Samples were subjected to RNA extraction, cDNA synthesis, and real-time quantitative (q)RT-PCR for VEGF and the normalization housekeeping gene HPRT as described.28 Note that the primers and probe used in this study were designed to detect total VEGF, with inclusion of the various isoforms. PGK1 mRNA expression levels were determined by qRT-PCR using the commercial assay #Mm00435617_m1 (Applied Biosystems, Foster City, CA, USA). Tibias isolated from E16.5 and E18.5 pups were subjected to standard procedures of protein extraction, followed by determination of the VEGF levels by the Quantikine ELISA kit (R&D Systems, Minneapolis, MN, USA) and of the total protein levels in the bone samples (Bio-Rad DC kit, Bio-Rad Laboratories, Hercules, CA, USA). The VEGF levels were expressed relative to the protein content in each of the samples. The mRNA and protein expression data were analyzed for statistical significant differences from the control values by two-sided t test (*p < 0.05 and **p < 0.01). The data are represented as bars designating averages ± standard error of the mean (SEM).
Conditional transgenic expression of VEGF164 effectively prevents cell death induced by VEGF deficiency in cartilage
To address the question whether the chondrocyte survival effect of HIF-1α was mediated via its downstream target VEGF, we assessed whether the cell death phenotype induced by HIF-1α deficiency could be prevented by forced expression of VEGF in cartilage. We therefore used mice carrying a conditional transgene (Tg) expressing the major isoform VEGF164 from the genomic Rosa26 locus after Cre-mediated recombination (Fig. 1A).38 First, we tested the efficacy of the transgene-driven expression of VEGF164 by assessing whether it was able to overcome the cell death phenotype of mice lacking (all isoforms of) VEGF in cartilage. Collagen 2 (Col2)-Cre(Tg/–);VEGF(lox/ + ) mice were crossed to VEGF(lox/lox);Rosa-VEGF164(Tg/–) mice to generate embryos conditionally lacking VEGF (conditional knockout [cKO]), either or not in combination with transgenic VEGF164 expression (VEGF cKO + VEGF164 Tg) as schematically outlined in Fig. 1B. Molecular analysis by qRT-PCR and ELISA in whole isolated E18.5 tibias and femurs, respectively, indicated that the VEGF164 transgene significantly increased the mRNA and protein levels of (total) VEGF above normal, both in VEGF164 Tg and VEGF cKO + VEGF164 Tg mice (Fig. 1C). As shown before,38 Col2-Cre-directed overexpression of VEGF164 did not disrupt cartilage development (also see Supplemental Fig. S1), but it dramatically induced ossification resulting in malformations of the limbs and rib cage (Fig. 1D, Alcian blue and Alizarin red staining for cartilage and bone, respectively). Compared with control mice, VEGF cKO mice had 30% to 40% lower VEGF mRNA and protein levels in their long bones on the whole (Fig. 1C). Their skeletal preparations showed a reduced overall skeletal size and typical cartilage abnormalities, including lengthening of the bone collar–surrounded cartilage (indicated in magnified views of distal radius/ulna, Fig. 1D) and the presence of central areas in the growth cartilage with notably pale Alcian blue staining (arrows in Fig. 1D, bottom panel). The skeletal preparations suggested that these manifestations were rescued by transgenic expression of VEGF164 (Fig. 1D, right panels). Histology indeed revealed that the central death area in the growth cartilage of VEGF cKO mice was completely rescued by the Rosa26-VEGF164 transgene. Transgenic overexpression of VEGF164 by itself did not affect chondrocyte proliferation, differentiation, or apoptosis (Maes and colleagues38 and Supplemental Fig. S1), but was effective in preserving the mRNA expression of the cartilage collagens type II and X in the central chondrocytes of VEGF cKO + VEGF164 Tg mice and fully prevented the ectopic accumulation of TUNEL-positive chondrocytes (Fig. 2). Of note, although transgenic VEGF164 expression corrected the cartilage phenotype of embryos with conditional VEGF inactivation, the excessive amount of VEGF caused the ossified bone center to develop prematurely and abnormally as described in mice carrying the transgene alone (Maes and colleagues38 and Supplemental Fig. S2).
These data indicate that Col2-Cre–mediated conditional overexpression of VEGF164 completely prevented the chondrocyte cell death induced by VEGF deficiency, confirming the efficacy of this transgenic approach.
Reduced extent and delayed appearance of cell death in cartilage lacking HIF-1α by conditional transgenic expression of VEGF164
Next, we similarly crossed the Rosa26-VEGF164 transgene along with the Col2-Cre transgene into a mouse line carrying floxed alleles of HIF-1α (Fig. 3A). Control, HIF-1α cKO, VEGF164 Tg, and HIF-1α cKO + VEGF164 Tg embryos were harvested at E16.5 to assess a potential rescue effect of the VEGF164 transgene on the chondrocyte death phenotype in HIF-1α embryos.
Molecular analysis of whole tibias and femurs confirmed that the presence of the VEGF164 transgene caused a significant upregulation of the VEGF mRNA and protein levels compared with control levels (Fig. 3B). The endogenous VEGF mRNA and protein levels were similar or even slightly increased in HIF-1α cKO bones compared with controls (Fig. 3B), likely explained by the phenotypic consequences of HIF-1α inactivation (see further). Skeletal preparations revealed that the overexpression of VEGF164 in HIF-1α cKO + VEGF164 Tg embryos caused the typical ossification defects resulting from VEGF overexpression, but was unable to rescue the reduced overall body growth of HIF-1α cKO mice (Fig. 3C, top panels). Moreover, whole mount HIF-1α skeletons showed evident signs of cartilage disruption, which was also seen in HIF-1α cKO + VEGF164 Tg embryos, albeit to a lesser extent (Fig. 3C, bottom panels, arrows).
Histological analysis at E16.5 confirmed that HIF-1α cKO mice display abnormal morphological appearance of the chondrocytes in a far-stretching area central in the cartilage, lack of collagen type II mRNA expression in this region, distortion of the collagen type X-expressing hypertrophic chondrocyte zone, and massive ectopic TUNEL staining (Fig. 4A and Schipani and colleagues6). Similar features were detected in HIF-1α cKO + VEGF164 Tg embryos at E16.5 but consistently to a reduced extent compared with HIF-1α cKO littermates: location-matched cell death lesions in the growth cartilages of the forelimbs and hind limbs were smaller upon overexpression of VEGF164 in addition to HIF-1α deficiency, with fewer numbers of chondrocytes undergoing cell death (Fig. 4A). Transgenic expression of VEGF164 remained to only partially rescue tissue integrity in growth cartilage lacking HIF-1α later in development, as judged by the cartilage damage observed at E18.5 that was only slightly improved compared with HIF-1α cKO mice (Fig. S3).
Analysis of embryos retrieved at E14.5 furthermore revealed that HIF-1α cKO mice already showed pronounced disruption of the cartilage at this early stage (Fig. 4B, left). In contrast, the integrity of the cartilage of HIF-1α cKO + VEGF164 Tg embryos was largely preserved at E14.5 (Fig. 4B, right).
Together, these data indicate that the onset of the chondrocyte death phenotype induced by HIF-1α deficiency was delayed and its severity reduced by introducing transgenic VEGF164 expression in the cartilage. Forced expression of VEGF164 hence only partially rescued cell survival in growth cartilage lacking HIF-1α.
VEGF164 overexpression increases vascularization in the surrounding soft tissue but does not fully prevent excessive hypoxia in cartilage lacking HIF-1α
In search of an explanation of the partial rescue phenotype, we next stained consecutive histological sections of E16.5 hind limbs with CD31 (PECAM-1) and hypoxyprobe antibodies to analyze the vascularization and presence of hypoxia, respectively (Fig. 5). CD31-stained blood vessels were observed surrounding the avascular growth cartilage of the proximal tibia (Fig. 5A) and distal femur (Supplemental Fig. S4A) of control bones and in the primary ossification center. Hypoxyprobe staining was limited to chondrocytes of the lower proliferating and prehypertrophic zones, particularly those located in the center of the growth cartilage (Fig. 5A, left panels). Transgenic VEGF164 overexpression per se drastically increased the density of blood vessels around the cartilage, and hypoxia in the growth cartilage was extremely scarce (Fig. 5A, second set of panels from the left, and Supplemental Fig. S4B). In contrast, hypoxyprobe staining was very strong and widespread in the growth cartilage of HIF-1α cKO mice, excluding only those chondrocytes located at the periphery closest to the cartilage surface (Fig. 5A, third set of images). The number of blood vessels appeared to be decreased in the soft tissue surrounding the joint space and flanking the growth cartilage of HIF-1α cKO long bones (Fig. 5A), as was confirmed by quantifying the CD31-positive vessel density at the cartilage surfaces of the E16.5 tibias and femurs (see Supplemental Fig. S4 for data and detailed outline of the histomorphometric methodology). Yet remarkably, the induction of massive vascularization around the cartilage by the additional expression of the VEGF164 transgene in mutants also lacking HIF-1α only modestly reduced the intensity and pattern of hypoxyprobe staining compared with the HIF-1α cKO single mutant (Fig. 5A, right panels).
To circumvent the influence of the severe cartilage damage seen in the E16.5 knee, we next examined the earlier stages of the endochondral bone development process by analyzing E16.5 metatarsals (outlined in detail in Fig. 5B) and E14.5 hind limb bones (similar results, data not shown). The bone templates at these stages characteristically consist of fully cartilaginous anlagen, preceding the initial invasion by blood vessels to generate the primary ossification center. As seen in Fig. 5B (left panels, control), the mid-diaphyseal hypertrophic cartilage region of the metatarsals inherently experienced a phase of hypoxia. Vascularization was at this stage still restricted to the interdigital space, blood vessels being separated from the cartilage surface by the avascular fibrous perichondrium (Fig. 5B, left panels and arrowhead). HIF-1α cKO metatarsals did not show evidence of tissue damage at E16.5, and tissue vascularization in the HIF-1α cKO autopod appeared normal (Fig. 5B, third set of panels from the left), as also reported previously.6, 8, 10 However, HIF-1α cKO cartilaginous templates already displayed excessive hypoxia, as judged from the hypoxyprobe staining (Fig. 5B) that was in all cases more intense as well as more extensively spread throughout the anlagen, particularly along the central axis, than in control samples. Transgenic expression of VEGF164 induced premature intercalation of blood vessels into the perichondrial tissue layers (Fig. 5B, arrows). This blood vessel accumulation immediately adjacent to the cartilage was associated with a reduction of the physiological hypoxia in VEGF164 Tg embryos (Fig. 5B, second from left) but did not fully rescue the HIF-1α cKO + VEGF164 Tg cartilages from aberrant hypoxia (Fig. 5B, right panels). These data indicate a function of HIF-1α in regulating the oxygen consumption in cartilage, preceding and independent of VEGF and angiogenesis. The role of HIF-1α in inducing vascularization becomes visible at a later time point, as indicated by the delayed vascular invasion of the cartilaginous templates in E18.5 metatarsals that could be rescued by transgenic VEGF164 (Supplemental Fig. S5).
The functionality of the blood vessels induced by the VEGF164 transgene and the existence of VEGF-independent functions of HIF-1α was further supported by the fact that the transgene was sufficient to prevent excessive hypoxia in the growth cartilage of mice deficient in endogenous VEGF expression. Indeed, VEGF cKO embryos at E18.5 displayed markedly reduced density of CD31-stained blood vessels around the growth cartilage (Fig. 5C). Concomitantly, hypoxyprobe staining revealed drastically increased hypoxia throughout the VEGF cKO cartilage. Interestingly, double mutants with the VEGF cKO + VEGF164 Tg genotype showed the expected enhanced blood vessel density, as well as a complete reversal of the hypoxyprobe staining to the restricted pattern and mild intensity seen in control limbs (Fig. 5C). These data suggest that the complete rescue of the viability of chondrocytes lacking endogenous VEGF by transgenic VEGF164 overexpression was likely the result of restoration of the oxygen supply by the enhanced blood vessel growth around the developing growth plate, preventing excessive hypoxia and cell death.
Altogether, these data indicate that excessive hypoxia in the cartilage of HIF-1α cKO mice appears to precede vascularization defects and is most likely caused by an excessive, nonadapted level of oxygen consumption by chondrocytes lacking HIF-1α that cannot be fully compensated by increasing the oxygen supply from the surrounding blood vessels as in the case of VEGF164 overexpression.
In general, cells respond to hypoxia by increasing the activity of the oxygen-sparing glycolytic component of metabolism, mediated by HIF-1α–dependent upregulation of glycolytic enzymes such as phosphoglycerate-kinase 1 (PGK1).39, 40 Consistent with our previous findings,6 PGK1 mRNA expression was observed in the growth plates of control mice but severely impaired by lack of HIF-1α. The levels of PGK1 expression were reduced irrespective of the absence or presence of the Rosa26-VEGF164 transgene, as analyzed by in situ hybridization (ISH) (Fig. 6A) and qRT-PCR (Fig. 6B), suggesting that HIF-1α also exerts functions independent of VEGF and angiogenesis, namely regulating a switch to an anaerobic metabolism in avascular cartilage.
In this study, we tested the hypothesis that upregulation of VEGF is one of the main modalities HIF-1α exploits to promote survival of hypoxic chondrocytes in fetal development.
HIF-1α is a major orchestrator of the cellular responses to low oxygen or hypoxia, with more than 100 putative HIF-1α target genes identified, most of them involved in energy metabolism and angiogenesis.41, 42 HIF-1α–mediated adaptation to hypoxia is important not only in relation to pathological conditions but also in normal fetal development and cell differentiation.43, 44 In endochondral bone development, the avascular cartilage molds or fetal growth plates become inherently hypoxic as they enlarge. Here, HIF-1α unambiguously functions as an essential “survival factor” for hypoxic chondrocytes, as shown previously.6 The downstream mediators of this HIF-1α survival function have, however, not been defined. VEGF, a classical downstream target of hypoxia and HIF-1α, represents a plausible candidate. Yet, whereas the hypoxic upregulation of VEGF in chondrocytes in vitro was shown to be HIF-1α–dependent,30–32 the regulation of VEGF by HIF-1α in vivo remains elusive because VEGF expression is very sensitive to the phenotypic manifestations of HIF-1α mutants, particularly the severe hypoxia (Schipani and colleagues6 and current study, Fig. 5). Indeed the increased VEGF expression observed in HIF-1α null growth plates and bones, at later stages, could be secondary to the increased hypoxia, whereas the reduction in perichondrial blood vessel formation and the delayed initial vascular invasion of HIF-1α–deficient cartilaginous templates may be possibly caused by a primary dysfunction in VEGF upregulation. Of interest, HIF-1α heterozygote E16.5 femur samples (Col2-Cre; HIF-1αf/+) showed a significant downregulation of VEGF mRNA, to 58% of control values (p < 0.02), arguing in favor of a regulation of VEGF expression by HIF-1α in vivo (data not shown). Our findings do demonstrate that expression of VEGF in chondrocytes is both necessary and powerful in inducing angiogenesis around the avascular growth cartilage. Yet, transgenic VEGF164 expression could only mildly and temporarily protect, but not rescue, HIF-1α–deficient chondrocytes from aberrant hypoxia and central cell death, despite massive induction of angiogenesis. Hence, the genetic approach employed in our study clearly revealed that HIF-1α–mediated regulation of VEGF is not the sole mechanism by which HIF-1α regulates chondrocyte viability and differentiation in vivo. The massive cell death phenotype in cartilage lacking HIF-1α turns out in fact to be largely attributable to VEGF-independent cell-autonomous mechanisms by which HIF-1α prevents overconsumption of oxygen in this challenged avascular tissue.
On the other hand, Col2-Cre–directed conditional overexpression of VEGF164 did partially rescue the survival of chondrocytes lacking HIF-1α, suggesting that additional aspects of HIF-1α functioning in the developing growth cartilage may be mediated via its regulation of VEGF. We have previously reported that chondrocytes do not show detectable expression of either of the main signaling receptors VEGFR1 and VEGFR2, which raises the intriguing question of how VEGF functions as a survival factor in the developing growth plate.7, 27 Although the contribution of a direct, cell-autonomous (autocrine and/or intracrine) role of VEGF in cell survival cannot be excluded, our current data foremost reinforce its angiogenic function in ensuring adequate oxygen availability within the avascular cartilage, as highlighted before.7 Indeed, the number of blood vessels was evidently decreased in the soft tissue surrounding VEGF-deficient growth plates; consistently, chondrocytes lacking VEGF were severely hypoxic. Moreover, transgenic expression of VEGF164 was able to vastly increase the blood vessel growth in the soft tissues around the cartilage, thereby fully restoring the oxygenation of the VEGF-deficient chondrocytes and completely preventing their death.
The fact that overexpression of only the VEGF164 isoform was able to compensate for the lack of endogenous expression of all the VEGF isoforms in the growth plate is consistent with the previous finding that mice expressing exclusively VEGF164 (by inactivation of VEGF120 and VEGF188) do not present any signs of chondrocyte cell death.7 VEGF164 and VEGF120 are the most abundant isoforms in normoxic as well as hypoxic chondrocytes in vitro and in growth cartilage in vivo, whereas VEGF188 is expressed to a much lower degree.7, 31 The relative weight of the various isoforms in the total amount of VEGF expressed is highly tissue-specific and considered to reflect functional significance, thus suggesting particular importance of VEGF120 and VEGF164 in cartilage. Interestingly, spatially localized loss of chondrocyte viability was also absent in mice that universally express exclusively VEGF120,7, 26, 28 whereas it was massive in mice expressing specifically the VEGF188 isoform (at normal total VEGF levels).7 Whereas VEGF164 is able to bind the coreceptor neuropilin, so far the only VEGF receptor found to be expressed in chondrocytes, VEGF120 does not. On the other hand, VEGF120 and VEGF164 do share with each other the characteristic of being able to diffuse away from the site of secretion, eg, a hypoxic region, to attract endothelial cells and stimulate angiogenesis at a distance; conversely, VEGF188 is not soluble and remains largely bound to the cells and matrix of the VEGF-secreting area. Thus, our data support and extend the previously suggested model of isoform-specific functioning of VEGF in chondrocyte biology,7 highlighting the importance of paracrine-acting diffusible VEGF released by chondrocytes to stimulate expansion of the vascular network in the surrounding perichondrium and ensure a viable level of oxygenation in the growing cartilage (see schematic model provided in Fig. 7). These conclusions are consistent with a recent report showing that VEGF produced by the condensed limb bud mesenchyme controls the vasculature in the uncondensed mesenchyme.45
Despite the functionality of the transgenic VEGF164-induced vascularization in vastly increasing the level of oxygenation in cartilage of control mice and of mice lacking endogenous VEGF, increased angiogenesis was associated with only a modest effect on reversing the excessive hypoxia in HIF-1α–deficient chondrocytes. These data suggest that chondrocytes lacking HIF-1α become oxygen-deprived particularly by a nonadapted, hence excessive level of oxygen consumption; the partial rescue inflicted by VEGF164 overexpression is likely owing to a partial and/or temporal compensational effect, ie, counteracting the buildup of a lethal level of hypoxia by enhancing the oxygen supply via the increased surrounding blood vessels.
Although cell nonautonomous induction of vascularization may thus mediate the reduced severity and delayed manifestation of the cell death phenotype in mice conditionally lacking HIF-1α and overexpressing VEGF164, the remaining chondrocyte death likely reflects VEGF-independent cell-autonomous mechanisms reliant on HIF-1α (Fig. 7). Beside its effects on angiogenesis, HIF-1α indeed uses a variety of mechanisms to promote cell survival in hypoxic conditions. Some of them involve regulation of glucose metabolism. Louis Pasteur was the first to notice that oxygen-deprived cells exhibit increased conversion of glucose to lactate, the so-called “Pasteur effect.” Activation of the Pasteur effect in mammalian hypoxic cells is HIF-1α–dependent because HIF-1α upregulates glucose transporters such as Glut1, glycolytic enzymes that support an increase in ATP production by anaerobic glycolysis, and the enzyme lactic dehydrogenase (LDH) that converts pyruvate to lactate.46, 47 It has also been reported that HIF-1α inhibits mitochondrial oxidative phosphorylation by negatively modulating the entry of pyruvate into the tricarboxylic acid (TCA) cycle.48, 49 By inhibiting the entry of pyruvate into the mitochondria, HIF-1α attenuates mitochondrial respiration and oxygen consumption. In addition, HIF-1α is able to induce mitochondrial autophagy and to negatively regulate mitochondrial biogenesis,50, 51 possibly adding to an outcome of reduced oxygen consumption.
Consistent with this model and with the hypoxic status of the fetal growth plate, we have previously shown that PGK, a key enzyme of anaerobic glycolysis, is strikingly stronger expressed in developing cartilage than in the surrounding tissues.6 Lack of HIF-1α nearly completely abolished PGK mRNA expression in chondrocytes, despite and/or adding to the increased hypoxia in these growth plates (current study and Schipani and colleagues6). Hence, lack of HIF-1α could inhibit chondrocytes to switch to oxygen-sparing glycolytic metabolism because of impaired transcriptional activation of key anaerobic energy pathway components such as PGK. Thus, metabolic changes regulated by HIF-1α may on the one hand affect the oxygen consumption and hence the level of hypoxia, and/or affect the ability of chondrocytes to survive in a hypoxic environment (Fig. 7). Additional studies will be necessary to further test this working hypothesis.
Taken together, we have demonstrated that VEGF is not sufficient to fully prevent the cell death of HIF-1α–deficient growth plates. Our findings strongly suggest that growth plate chondrocytes have an important role in regulating the number of blood vessels in the surrounding soft tissues through secretion of VEGF, and that this oxygen-supplying function is likely to be the prime modality exploited by VEGF to support chondrocyte survival during development. Yet, VEGF is not the sole critical downstream effector of the HIF-1α survival function. In the current study, we unequivocally demonstrate that the severity of hypoxia in HIF-1α–deficient growth plates is only modestly corrected by increasing the oxygen supply through increased vascularization. This finding suggests that a vital homeostatic function of HIF-1α in developing cartilage is to protect from excessive hypoxia by limiting the oxygen consumption. VEGF and HIF-1α thus are critical regulators of the balance between the availability and the handling of oxygen in developing growth cartilage, thereby preserving chondrocyte viability (Fig. 7).
All authors state that they have no conflicts of interest.
We thank Dr Henry M Kronenberg for helpful discussions, and Kim Atkin (Histology Core, Endocrine Unit, MGH), Karen Moermans, Ingrid Stockmans, and Sophie Torrekens (Legendo, KULeuven) for excellent technical help. We thank Napoleone Ferrara and Andras Nagy for the VEGF floxed and Rosa26-VEGF164 Tg mice used in this study. This work was supported by National Institutes of Health RO1 AR048191-06 (to ES) and Fund for Scientific Research (FWO)-Flanders G.0569.07 (to GC and JH) and G.0500.08 and G.0982.11 (to GC).
Authors' roles: ES designed the study concept; CM and EA performed experiments and analyzed data; KH, RK, and RVL gave technical support; AJG, JH, and GC provided reagents and mice; CM, GC, and ES interpreted the data and discussed the results; CM and ES wrote the paper. CM, GC, and ES take responsibility for the integrity of the data analysis.