HIF1α is a central regulator of collagen hydroxylation and secretion under hypoxia during bone development

Hypoxic conditions occur normally in various tissues during embryonic development, as well as postnatally (Dunwoodie, 2009; Maltepe and Simon, 1998; Mitchell and Yochim, 1968; Rodesch et al., 1992; Simon and Keith, 2008). To maintain homeostasis in this oxygen-poor environment, mechanisms that allow cells to adjust their metabolism and function had to evolve. Extensive studies performed over the past two decades have identified the complex mechanisms that cells use to adapt their metabolism to changing oxygen levels (Semenza, 2009; Semenza, 2012; Weidemann and Johnson, 2008). However, we know little about the mechanisms that underlie the ability of cells to perform their specific roles, such as synthesis and secretion of extracellular matrix (ECM) proteins, under hypoxia.

Low oxygen levels are found in the growth plates, located at the two extremities of developing bones (Schipani et al., 2001). The growth plate regulates bone ossification and elongation by maintaining balance between chondrocyte proliferation and differentiation (Mackie et al., 2008; Provot and Schipani, 2005). Chondrocytes are professional secretory cells that produce large amounts of ECM, which includes collagen type II (COL2A1), the most prevalent collagen in cartilage, as well as non-collagenous components, such as the proteoglycans aggrecan and perlecan (Kiani et al., 2002; Kvist et al., 2008). The ECM that surrounds the cells is essential for the mechanical properties and integrity of cartilage, as manifested by various pathological conditions associated with abnormal biosynthesis, secretion or turnover of ECM components (Bateman et al., 2009; Kiani et al., 2002; Roughley, 2006; Schwartz and Domowicz, 2002).

One of the crucial steps in the biosynthesis of ECM proteins is their folding in the endoplasmic reticulum (ER). The integrity of this process usually depends on post-translational modifications (PTMs), which are alterations to the composition or structure of proteins. For example, a key step in collagen production is the hydroxylation of proline residues (Berg and Prockop, 1973; Cooper and Prockop, 1968; Juva et al., 1966; Lamandé and Bateman, 1999). This reaction is catalyzed by a group of enzymes known as collagen prolyl 4-hydroxylases (cP4Hs) (Myllyharju, 2003; Myllyharju and Kivirikko, 2004). Aberrant hydroxylation of collagen results in reduced secretion and its accumulation in the ER (Uitto et al., 1975). Interestingly, and perhaps paradoxically, although collagen secretion requires oxygen-dependent proline hydroxylation, the growth plate that hosts secreting chondrocytes is avascular – hence, hypoxic (Myllyharju, 2003; Schipani et al., 2001). The growth plate can therefore serve as an excellent model system with which to study mechanisms that ensure collagen synthesis and secretion under hypoxia.

The basic helix-loop-helix (bHLH)-PAS transcription factor hypoxia-inducible factor 1 (HIF1) is a key regulator of cellular and physiological adaptation to hypoxia. It is a heterodimer consisting of HIF1α, the oxygen-sensitive subunit, and the constitutively expressed HIF1β, also known as ARNT. Under hypoxic conditions, HIF1 activates, for example, the transcription of glucose transporters and glycolytic enzymes, which are important for transition to anaerobic metabolism. HIF1 can also increase oxygen delivery and thereby facilitate adaptation to hypoxia by inducing the expression of genes such as erythropoietin, which regulates erythrocyte production, and vascular endothelial growth factor (Vegf), which regulates angiogenesis (Semenza, 2009; Semenza, 2012; Weidemann and Johnson, 2008).

The importance of HIF1α in chondrocyte biology was demonstrated at the onset of chondrogenesis, as well as after the growth plate is established (Amarilio et al., 2007; Provot et al., 2007; Schipani et al., 2001). In all these studies, loss of Hif1a resulted in massive cell death in the hypoxic regions of cartilaginous elements. These studies firmly establish that HIF1α is necessary for chondrocyte survival under hypoxia. However, the cell death phenotype has hampered further investigation into the involvement of HIF1α in the function of growth plate chondrocytes during development.

Although the deciphering of mechanisms that regulate ECM composition and secretion is fundamental for the understanding of embryogenesis, as well as of human pathologies such as cancer, little is known about these mechanisms. In this study, we identify a novel role for HIF1α in the regulation of cartilage ECM secretion during bone development. We show that Hif1a loss of function in growth plate chondrocytes results in ER stress and arrested secretion of ECM proteins, including COL2A1. We then demonstrate that HIF1α regulates collagen hydroxylation and subsequent secretion by maintaining the fine balance between the hydroxylation enzyme and its substrate oxygen. This molecular mechanism explains how chondrocytes maintain their functionality as professional secretory cells under the challenging conditions of the hypoxic growth plate.

The generation of floxed-Hif1a (Ryan et al., 2000), Col2-CreER^T (Nakamura et al., 2006) and Rosa26R-lacZ reporter mice (Soriano, 1999) has been described previously. In all timed pregnancies, plug date 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 ice-cold PBS and the embryos were harvested after amnionectomy and removal of the placenta. Tail genomic DNA was used for genotyping. Tamoxifen (Sigma) was dissolved in corn oil (Sigma) at a concentration of 20 mg/ml. The solution was administered to time-pregnant female mice by intraperitoneal injection of 300 μl at embryonic day (E) 15.5.

For in situ hybridization, aggrecan immunofluorescence, TUNEL assay and Hematoxylin and Eosin staining, embryos were fixed overnight in 4% PFA-PBS, decalcified in a solution containing equal parts of 0.5 M EDTA (pH 7.4) and 4% PFA in PBS overnight, dehydrated to 100% ethanol, embedded in paraffin and sectioned at 7 μm. Hematoxylin and Eosin staining was performed following standard protocols. TUNEL assay was performed using In Situ Cell Death Detection Kit (Roche) according to the manufacturer's protocol. Section in situ hybridization was performed as described previously (Murtaugh et al., 1999; Riddle et al., 1993). All probes are available by request.

For immunofluorescence of aggrecan, sections were rehydrated to PBS and then digested with chondroitin ABC lyase [0.1 U/ml in 0.01% BSA, 50 mM Tris/HCl (pH 8.0), 60 mM NaAc; Sigma] for 40 minutes at 37°C. Sections were then washed twice for 5 minutes with PBS, permeabilized with 0.5% Triton X-100, 1% BSA in PBS for 15 minutes and blocked with 5% goat serum, 1% BSA in PBT for 1 hour. Following blockage, sections were incubated overnight at 4°C with primary antibody rabbit anti-aggrecan (Millipore AB1031, 1:100). Then, sections were washed three times for 5 minutes with PBS and incubated with biotin donkey anti rabbit secondary antibody (Jackson ImmunoResearch, 1:100) for 1 hour at room temperature. After washing with PBS (three times for 5 minutes), sections were incubated with streptavidin Cy3 (Jackson ImmunoResearch, 1:100) for 1 hour at room temperature and washed again with PBS (three times for 5 minutes). The sections were subsequently stained with DAPI, washed with PBS three times for 5 minutes and mounted with Immu-Mount (Thermo Scientific).

For immunofluorescence of COL2A1 and perlecan, embryos were embedded in OCT (Tissue-Tek) and 10 μm cryostat sections were made. Alternatively, primary chondrocytes were fixed in 1% PFA-PBS for 5 minutes. The rest of the procedure was performed as described previously (Smits et al., 2010), with some adjustments. Digestion with chondroitin ABC lyase was limited to 40 minutes and biotinylated secondary antibodies (1:100) and streptavidin Cy3 (1:100) were used. Primary antibodies used were mouse anti-COL2A1 (Developmental Studies Hybridoma Bank, the University of Iowa, II-II6B3, 1:100) and rat anti-perlecan/HSPG2 (Lifespan Biosciences LS-C15756, 1:100). Secondary antibodies used were biotin donkey anti-rat (Jackson ImmunoResearch) and biotin goat anti-mouse (Jackson ImmunoResearch).

For chondrocyte primary culture, hindlimbs of E17.5 floxed-Hif1a embryos were dissected and soft tissue, skin and particularly muscles were removed. The growth plates were dissected and placed in DMEM 4500 mg/l glucose (Invitrogen) with 1% pen-strep solution (Biological Industries). Growth plates were digested in trypsin containing 0.25% EDTA (Biological Industries) for 30 minutes at 37°C and in 1 mg/ml collagenase type V (Sigma) in DMEM 4500 mg/l glucose with 1% pen-strep solution for 2 hours. Chondrocytes were plated at a density of 125×10³ cells/ml and grown in monolayer cultures in high glucose DMEM supplemented with 10% fetal bovine serum (FBS, Biological Industries) and 1% pen-strep solution. Cells were infected with 300 viral particles/cell of Ad5CMVeGFP or Ad5CMVCre-eGFP virus (Gene Transfer Vector Core, University of Iowa). After viral infection, cells were cultured either in 20% oxygen (normoxia) or in 1% oxygen (hypoxia) balanced with N₂ in a 3-Gas incubator (Heraeus) in a humidified atmosphere for 48 hours. During all experiments, except for exposure to hypoxia or normoxia, medium was changed daily. As a positive control, chondrocytes (primary cells) were treated with the ER stress inducer Brefeldin-A (Sigma) at a final concentration of 0.4 μg/ml for 24 hours.

Tibiae growth plates were prepared for electron microscopy analysis as described previously (Hunziker et al., 1983), with some adjustments. The growth plates were shaken in a primary fixative solution containing 2% glutaraldehyde, 0.05 M sodium cacodylate buffer (pH 7.4) with 0.7% ruthenium hexammine trichloride (RHT) (Polysciences) for 3 hours at room temperature, then washed three times for 10 minutes with 0.1 M sodium cacodylate buffer (pH 7.4). Post fixation was carried out with 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (pH 7.4) with 0.7% RHT for 2 hours at room temperature, and samples were washed three times for 5 minutes with the same buffer. For the control, primary fixation and post fixation were carried out without RHT. Following fixation, samples were dehydrated in ethanol (70% twice for 15 minutes, 96% once for 15 minutes, 100% twice for 25 minutes) and washed with propylene oxide twice for 15 minutes at room temperature. The tissue was then subjected to propylene oxide with increasing concentration of epon (30%, 50% and 75%, each overnight at room temperature) and then 100% epon for 5 hours before polymerization. Ultrathin sections (80 nm) were examined with a transmission electron microscope operated at 120 kV.

Total RNA was purified from long bone growth plates of E17.5 embryos or from tissue culture cells using the RNeasy Kit (Qiagen). Reverse transcription was performed with high capacity reverse transcription kit (Applied Biosystems) according to the manufacturer's protocol. Analysis of Xbp1 splicing was performed as described previously (Ho et al., 2007).

qRT-PCR was performed using Fast SYBR Green master mix (Applied Biosystems) on the StepOnePlus machine (Applied Biosystems). Values were calculated using the StepOne software version 2.2, according to the relative standard curve method. Data was normalized to TATA-box binding protein (Tbp) in all cases, except in chondrocyte primary culture experiments, where it was also normalized to 18S rRNA. Primer sequences can be found in supplementary material Table S2. Statistical analysis was carried out using Prism 5 software and statistical significance was determined by Student's t-test as P<0.05.

For western blot analysis, protein was extracted from long bone growth plates of E17.5 embryos or from primary chondrocytes. Protein concentration was determined by Bio-Rad protein assay (Bio-Rad). GADD 153 (R-20; Santa Cruz Biotechnology, 1:500) and α-tubulin (Sigma, 1:10,000) antibodies were used, followed by the appropriate HRP-conjugated secondary antibodies (1:10,000; Jackson ImmunoResearch) and HRP detection using the EZ-ECL kit (Biological Industries).

E17.5 embryos were sacrificed and growth plates from the long bones were isolated and subjected to collagenase type V digestion (Sigma), diluted 1 mg/ml in PBS for 20 minutes at 37°C. Samples were then washed with PBS three times and hydrolyzed, using constant-boiling vapor HCl under vacuum at 110°C for 30 hours. Alternatively, primary chondrocytes were subjected to trypsin containing 0.25% EDTA digestion for 6 minutes at 37°C and were separated from medium by centrifugation. Cell samples were also hydrolyzed for 30 hours. Amino acid analysis was performed on Waters Alliance 2695 instrument equipped with Waters 474 fluorescence detector (Waters, Milford, MA, USA), using Waters AccqTag amino acids analysis kit (Waters). The ratio between areas of hydroxyproline and proline was calculated for each tissue sample. Statistical analysis was carried out using Prism 5 software and statistical significance was determined by Student's t-test as P<0.05.

X-gal staining was performed as described previously (Eshkar-Oren et al., 2009).

Hypoxia detection and BrdU assay were performed as described previously (Amarilio et al., 2007).

Transcription factor binding site prediction was performed with MatInspector (Cartharius et al., 2005), in the Genomatix Genome Analyzer package (Genomatix Software GmbH, Germany) with the search set specifically to find HIF binding sites (V$HIFF). Conservation of binding sites was analyzed using the UCSC genome browser (genome version mm9) 30-way Conservation track (Fujita et al., 2011).

Growth plates from E17.5 wild-type embryos were digested in trypsin containing 0.25% EDTA for 15 minutes at 37°C and in 1 mg/ml collagenase type V in DMEM 4500 mg/l glucose with 1% Pen-Strep solution for approximately 1 hour. During collagenase digestion, cells were collected every 10 minutes for crosslinking, suspended with high glucose DMEM supplemented with 10% FBS and 1% Pen-Strep solution containing 1% formaldehyde and kept shaking on ice. After all cells were collected (3×10⁶ cells), they were left for additional 10 minutes shaking at room temperature. ChIP was the performed as described previously (Ainbinder et al., 2002), with some adjustments. Chromatin collected from the cells was sonicated (10 cycles of 30 seconds each) by Bioruptor UCD-200 sonicator (Diagenode). For immunoprecipitation, rabbit polyclonal anti-sera to HIF1α (PM14, kindly provided by Prof. Christopher W. Pugh, University of Oxford, UK) and whole rabbit serum (IgG control, kindly provided by Prof. Groner Yoram, Weizmann Institute, Israel) were used. HIF1α binding was analyzed using PCR; primer sequences can be found in supplementary material Table S2.

The centrality of HIF1α in cell adaptation to hypoxia makes it a strong candidate to partake in mechanisms that allow professional secretory cells, such as chondrocytes, to maintain functionality under these conditions. However, previous attempts to explore the role of HIF1α in the growth plate of developing bones by a loss-of-function approach resulted in massive chondrocyte cell death, which prevented further analysis (Schipani et al., 2001). To overcome this obstacle, we used Cre recombinase fused to an estrogen receptor under regulation of collagen type II promoter, allowing for temporal control of recombinase activity in chondrocytes (Hif1a^f/fCol2-CreER^T) (Nakamura et al., 2006).

Using TUNEL assay, we assessed chondrocyte cell death in the Hif1a^f/fCol2-CreER^T growth plate. At E18.5, 72 hours after tamoxifen (TM) administration, cell death was prominent in the center of the growth plate (Fig. 1A-B′). This result was similar to the cell death phenotype previously reported (Schipani et al., 2001). Importantly, at E17.5, 48 hours after TM administration, although Cre activity and Hif1a ablation were verified (Fig. 1C,D), chondrocytes were viable, as was evident by the expression of growth plate markers and TUNEL analysis (supplementary material Fig. S1A-D,F-I; and data not shown). The viability of the cells 48 hours after TM administration provided us with the opportunity to study the role of HIF1α in the hypoxic growth plate. All the studies described hereafter were performed on E17.5 embryos, 48 hours after TM administration.

To study the effect of HIF1α on growth plate chondrocytes, we first analyzed histological sections of the hypoxic central region of Hif1α-deficient growth plate. Analysis revealed that chondrocytes in this region lost their columnar organization and their number was reduced (Fig. 1E,F). BrdU assay confirmed this observation by demonstrating reduced cell proliferation in that region (supplementary material Fig. S1E,J). Growth plate hypocellularization was accompanied by reduced Hematoxylin and Eosin staining of the matrix around the chondrocytes (Fig. 1E,F), indicating abnormal matrix composition.

In order to examine more closely matrix distribution and secretion in Hif1α-deficient growth plate, we performed electron microscopy (EM) analysis. Chondrocytes in Hif1α-deficient growth plates exhibited dilated and fragmented ER and excessive intracellular vacuoles containing ECM-like material (Fig. 2A-D), implying impaired ECM secretion.

To verify the suspected accumulation of ECM components inside Hif1α-deficient chondrocytes, we analyzed the material retained within these cells. First, we examined by EM sections of Hif1α-deficient growth plates that were fixed in the presence of ruthenium hexamine trichloride (RHT). RHT binds to proteoglycans, retains them within the cartilaginous tissue and prevents cell shrinkage and detachment from the pericellular matrix upon fixation (Hunziker et al., 1983; Nuehring et al., 1991). Examination of RHT-fixed chondrocytes showed that, in control growth plates, the pericellular matrix was intact, forming a ring-like shape that surrounded the cells. By contrast, chondrocytes in the hypoxic region of Hif1α-deficient growth plate exhibited a dramatic reduction in ECM content, particularly in the pericellular zone (Fig. 2E,F). Comparison between RHT-treated and untreated tissues revealed a significant increase in proteoglycans inside vacuoles at the Golgi area in Hif1a-deficient chondrocytes, but not in control cells (Fig. 2G-J).

To support the EM observation, we immunostained Hif1a-deficient growth plates using antibodies for two major cartilage proteoglycans: aggrecan and perlecan. In agreement with the EM results, aggrecan and perlecan accumulated inside chondrocytes in the hypoxic central region of Hif1a-deficient growth plates, unlike in control growth plates (Fig. 2K-N″). Moreover, immunostaining for collagen type II revealed that it also accumulated in these cells (Fig. 2O-P″). Intracellular accumulation of ECM components concomitantly with a reduction in ECM extracellularly and loss of cell-matrix attachment imply that in the hypoxic growth plate, HIF1α is necessary for proper ECM secretion by chondrocytes.

It is well established that abnormal folding may interfere with protein secretion and lead to the accumulation of unfolded proteins, a condition termed ER stress (Lin et al., 2008). We therefore hypothesized that the impaired ECM secretion by Hif1a-deficient chondrocytes was a consequence of defective protein folding. Consistent with our hypothesis, the ER in those cells appeared distended and fragmented (Fig. 2B), a typical structure under ER stress.

Another well-known feature of ER stress is the activation of the unfolded protein response (UPR) (Ron and Walter, 2007). To determine whether Hif1a-deficient chondrocytes were under ER stress, we analyzed the expression of two key factors activated during UPR: BiP (HSPA5 – Mouse Genome Informatics) and CHOP (DDIT3 – Mouse Genome Informatics). In situ hybridization demonstrated increased Bip and Chop expression in Hif1a-deficient growth plate, compared with the control (Fig. 3A-D); qRT-PCR and western blot analysis verified this elevation (Fig. 3E,F). To further validate these observations, we examined the expression of Bip and Chop upon Hif1a depletion in chondrocyte primary culture. Under hypoxic conditions, Hif1a depletion led to an increase in Bip and Chop expression, when compared with control cells (Fig. 3G). Western blot analysis showed a compatible increase in CHOP protein level in Hif1a-depleted cultures incubated under hypoxia (Fig. 3H). Finally, we analyzed mRNA splicing of Xbp1, another typical marker for ER stress (Calfon et al., 2002; Yoshida et al., 2001). Although in control cells there was a mild activation of Xbp1 splicing under hypoxia, Hif1a depletion led to increased transcription of Xbp1 and elevated levels of Xbp1 splicing (Fig. 3I).

Taken together, these results strongly suggest that Hif1a loss of function in chondrocytes induces ER stress, implying that protein folding in these cells is defective.

Our finding that Hif1a-deficient chondrocytes are under ER stress and accumulate collagen type II suggested that HIF1α plays a role in collagen folding. Hydroxylation of proline residues is absolutely necessary for proper folding and secretion of collagen (Berg and Prockop, 1973; Cooper and Prockop, 1968; Juva et al., 1966; Lamandé and Bateman, 1999; Uitto et al., 1975). To test the hypothesis that HIF1α is required for collagen proline hydroxylation in chondrocytes, we examined the level of hydroxylation in Hif1a-deficient growth plates. Amino acid analysis on whole growth plate extracts demonstrated a 9% reduction in hydroxyproline/proline ratio in Hif1a-deficient growth plates, when compared with the control (Fig. 4A). Considering that Hif1a-deficient growth plates contained hydroxylated collagen that had been secreted prior to the depletion of Hif1a in chondrocytes, it is most likely that this result represents reduction in the hydroxylation of collagen that was formed post-Hif1a depletion and was retained within the cells (Fig. 2P-P″).

To support this possibility, we studied collagen hydroxylation and secretion in control and Hif1a-depleted chondrocyte primary cultures. Under hypoxic conditions, immunostaining for collagen type II in control cells demonstrated its presence both intracellularly and extracellularly. By contrast, in Hif1a-depleted chondrocytes it was detected only intracellularly (Fig. 4B,C). This result supported our in vivo finding that Hif1a depletion arrests collagen secretion. Next, we compared the amount of hydroxyproline in control and Hif1a-depleted cells under hypoxia. Although collagen type II was clearly produced in Hif1a-deficient cells, hydroxyproline level in those cells was undetectable, unlike in control cells. Importantly, the levels of proline were comparable between Hif1a-depleted and control cells (Fig. 4D).

Together, these results indicate that HIF1α is required for collagen hydroxylation and secretion under hypoxia, and therefore supports the possibility that collagen folding is impaired in its absence, resulting in its intracellular accumulation and in ER stress.

Our finding that collagen hydroxylation is impaired in Hif1α-deficient chondrocytes prompted us to investigate the mechanism by which HIF1α regulates this reaction. First, we examined the effect of HIF1α on the bioavailability of oxygen, which is the substrate of the reaction. To this end, we analyzed chondrocyte oxygenation in Hif1a-deficient and control growth plates by using hypoxyprobe, a marker of hypoxic cells (Arteel et al., 1998). Staining revealed a significant increase in the hypoxic signal in Hif1a-deficient growth plates, compared with the control (Fig. 5A-B″).

HIF1α can affect oxygen level in the growth plate either by increasing the number of flanking blood vessel, or by reducing oxygen consumption in chondrocytes through inhibition of the mitochondrial tricarboxylic acid (TCA) cycle (Fukuda et al., 2007; Kim et al., 2006; Papandreou et al., 2006; Semenza, 2007). As it was recently demonstrated that elevated Vegf expression in Hif1a-deficient growth plates had little effect on oxygen bioavailability (Maes et al., 2012), we focused on the effect of HIF1α on oxygen consumption. HIF1α can directly regulate oxygen consumption through induction of pyruvate dehydrogenase kinase 1 (PDK1), which, in turn, inhibits the activity of pyruvate dehydrogenase and thereby blocks the TCA cycle. qRT-PCR analysis revealed that the expression of Pdk1 in Hif1a-deficient growth plates was dramatically reduced, relative to the control (Fig. 5C). Previously, it has been shown that HIF1 directly regulates human PDK1 gene expression by binding to several HIF1 consensus binding sites, also referred to as hypoxia response elements (HRE) (Kim et al., 2006; Wenger et al., 2005). In order to determine whether HIF1α directly regulates mouse Pdk1 expression in vivo in chondrocytes, a chromatin immunoprecipitation (ChIP) assay was performed on chondrocyte lysate extracted from wild-type growth plates. We found two active HIF1α binding areas on the mouse Pdk1 gene, each containing several HRE sequences (Fig. 5D; supplementary material Table S1). This result provides a mechanistic explanation for the reduction in oxygen levels and for the consequent reduction in collagen hydroxylation we observed in Hif1a-deficient growth plates.

Having found an effect of HIF1α on the hydroxylation substrate oxygen, we proceeded to explore its involvement in regulation of the enzyme cP4H. As mentioned, cP4Hs catalyze collagen hydroxylation in the ER. These enzymes are tetramers and consist of two β subunits and two α subunits. While cP4HαI is the main α subunit in most cell types, in chondrocytes it is cP4HαII (Myllyharju, 2003; Myllyharju and Kivirikko, 2004). Previous in vitro studies have raised the possibility that HIF1α may regulate cP4Hs in chondrocytes and other cell types (Copple et al., 2011; Grimmer et al., 2006; Hofbauer et al., 2003; Takahashi et al., 2000). Therefore, the expression of cP4H subunits was first analyzed in chondrocyte primary culture. Under hypoxia, mRNA levels of P4ha1, P4ha2 and P4hb were elevated relative to cells under normoxic conditions (Fig. 6A). This indicated that cP4H subunit genes are hypoxia responsive, consistent with the hypothesis that under hypoxic conditions, an elevation in their expression is necessary to maintain collagen hydroxylation.

Next, we analyzed the expression of cP4H subunits in cells where Hif1a expression was blocked. As can be seen in Fig. 6A, the elevation in the expression of cP4H subunits was lost in Hif1a-depleted cells, suggesting an involvement of HIF1α in their expression by chondrocytes under hypoxia. To validate this hypothesis, we examined in vivo the expression levels of these subunits in Hif1a-deficient growth plates. A significant reduction in P4ha1 and P4hb mRNA levels was detected in Hif1a-deficient growth plates, when compared with the control (Fig. 6B). Interestingly, the expression level of P4ha2, the main α subunit in chondrocytes, was dramatically reduced by 60% in the absence of Hif1a, when compared with control growth plates. These results imply that in chondrocytes, HIF1α regulates the expression of cP4H subunits under hypoxic conditions.

In order to examine whether HIF1α directly regulates cP4H expression, we searched the mouse cP4H subunits promoters for HRE sequences. We identified several potential binding sites on P4ha1, P4ha2 and P4hb genes (Fig. 6C; supplementary material Table S1). To determine whether HIF1α binds to these putative sequences, a ChIP assay was performed on chondrocyte lysate extracted from wild-type growth plates. We found two active HIF1α-binding sites on P4ha1 promoter (sites A and B) and two active binding sites on P4ha2 gene (sites A and B). Moreover, one area on P4hb promoter that contained three potential binding sites (sites A-C) also demonstrated active binding to HIF1α (Fig. 6D). Analysis of those sequences, using the conservation track of the UCSC genome browser (Fujita et al., 2011), revealed conservation among species for P4ha1 site B, P4ha2 site A and P4hb site C (Fig. 6E). This suggests that the direct regulation of HIF1α on the expression of cP4H subunits is evolutionarily conserved.

Collectively, these results provide a molecular mechanism for the regulation of collagen hydroxylation by HIF1α by showing that this factor controls both the bioavailability of the substrate oxygen and the expression levels of cP4H enzymes.

Post-translational modifications (PTM) are essential for protein secretion and function both during development and through the life of the organism. In humans, collagens are the most abundant proteins and hydroxylation of proline residues is the most common PTM during collagen production. As collagen proline hydroxylation is dependent on bioavailability of molecular oxygen, variation in tissue oxygen levels dictates the need for a mechanism that would secure the robustness of this pivotal process. In this study, we identify HIF1α as a central component in a mechanism that allows collagen hydroxylation in hypoxic cells by controlling the level of both the catalyst and the substrate of the reaction.

The evolutionary origin of HIF1α coincides with the divergence of metazoans (Loenarz et al., 2011; Rytkönen et al., 2011). During the evolution of multicellular organisms, oxygen supply to all cells has become a major challenge owing to environmental constraints and an increase in organism complexity and size. Metazoan cells had to overcome two fundamental problems. The first was to adapt their metabolism in order to ensure survival under varying oxygen levels. The second necessity was to produce under these conditions collagen, which serves as a ‘glue’ that holds the cells together and guarantees the integrity of connective tissues (Towe, 1970). The need to overcome these problems fits well with the emergence of HIF1α as a major regulator of oxygen homeostasis. The role of HIF1α in metabolic adaptation to hypoxia was extensively studied and described. By contrast, its involvement in collagen production has thus far been neglected.

Here, we demonstrate in vivo the ability of HIF1α to regulate directly the transcription of the cP4H subunits cP4HαI, cP4HαII and cP4Hβ in hypoxic growth plate chondrocytes. We also show that, concurrently, HIF1α maintains cellular levels of the substrate oxygen, most likely by controlling the expression of Pdk1, a key component in the mechanism that regulates cellular oxygen level (Fukuda et al., 2007; Kim et al., 2006; Papandreou et al., 2006; Semenza, 2007). Hence, HIF1α acts both to increase the level of the enzyme and to maintain the level of the substrate. In the absence of HIF1α, intracellular oxygen levels and cP4H subunits expression are reduced, leading to reduction in collagen type II hydroxylation and, consequently, to an arrest in its secretion and its intracellular accumulation (Fig. 7).

Our finding that HIF1α regulates cP4H subunits corresponds with previous in vitro studies in different cell types, including fibroblasts and hepatoma cells, which suggest that HIF1α regulates cP4HαI and cP4HαII expression (Copple et al., 2011; Grimmer et al., 2006; Hofbauer et al., 2003; Takahashi et al., 2000). This raises the possibility that HIF1α-mediated regulation of cP4Hs is a broad phenomenon that is not restricted to chondrocytes.

Although our results explain collagen hydroxylation under hypoxic conditions, it may be just the tip of the iceberg regarding the involvement of HIF1 in ECM secretion. Several pieces of evidence support this supposition. First, in vitro studies demonstrate the involvement of HIF1α in regulating the expression of procollagen lysyl-hydroxylases (Plod) that catalyze hydroxylation of lysine residues, which is also required for collagen maturation (Hofbauer et al., 2003). Another indication is our finding that HIF1α regulates the expression of cP4Hβ, which is identical to protein disulfide isomerase (PDI). In addition to its role in collagen hydroxylation, cP4Hβ catalyzes the formation and rearrangement of disulfide bonds, which are required for proper folding of collagen and other proteins (Feige and Hendershot, 2011; Forster and Freedman, 1984; Hatahet and Ruddock, 2009; Koivu and Myllylä, 1987). Moreover, it was previously reported that HIF1α regulates the expression of ERO1Lα, which is also involved in disulfide bond formation (May et al., 2005). As the formation of disulfide bonds is also oxygen-dependent, HIF1α may also regulate this process indirectly by maintaining oxygen homeostasis.

Further support for the central role of HIF1α in ECM secretion is our finding that blockage of Hif1a expression also leads to accumulation of proteoglycans, such as aggrecan and perlecan, within chondrocytes. It was previously reported that interference with proteoglycan glycosylation in the Golgi inhibits their secretion (Hameetman et al., 2007; Ratcliffe et al., 1985). HIF1α is known to regulate uptake and consumption of glucose, which is necessary for proteoglycan glycosylation (Mobasheri et al., 2002; Peansukmanee et al., 2009; Ren et al., 2008). This suggests that HIF1α may regulate not only collagen hydroxylation, but also proteoglycan glycosylation. It is also possible that HIF1α regulates a battery of genes that are necessary for the secretion of a variety of ECM components under hypoxia. Moreover, previous in vitro studies suggested that HIF1α is involved in regulating proteoglycans and collagen expression (Pfander et al., 2003). Collectively, these reports and the findings we present here may indicate a major involvement of HIF1α in ECM production and secretion under hypoxic conditions.

ER stress is known to have severe deleterious effects that significantly contribute to the pathology of a variety of diseases (Bateman et al., 2009; Boot-Handford and Briggs, 2010; Lisse et al., 2008; Rajpar et al., 2009; Tsang et al., 2010). However, several recent studies indicate that during bone development, ER stress in hypoxic chondrocytes plays a role in the normal sequence of their differentiation (Patra et al., 2007; Saito et al., 2009). Our findings that link among hypoxic conditions, PTMs, HIF1α and ER stress may implicate HIF1α as an essential molecule in this mechanism. Further support for this notion comes from a recent finding that HIF1α can induce ER stress in chondrocytes (Yang et al., 2008). In light of this, changes in HIF1α levels may determine the level of ER stress in those cells.

Finally, the connection we found between HIF1α and ER stress may provide a new explanation for the massive chondrocyte cell death observed upon Hif1a inactivation, which was previously attributed to impaired glycolysis and to the role of HIF1α as a regulator of cell metabolism (Maes et al., 2012; Schipani et al., 2001). Another possible explanation is our finding that impaired matrix secretion in Hif1a-deficient chondrocytes results in loss of the vital cell-matrix attachment (Aszodi et al., 2003; Hirsch et al., 1997; Pulai et al., 2002; Raducanu et al., 2009; Svoboda, 1998; Woods et al., 2007).

This work describes a novel role for HIF1α as a central regulator of ECM secretion, which ensures the function of chondrocytes as professional secretory cells in the hypoxic growth plate. Focusing on the secretion of collagen type II, a major ECM component in cartilage, we demonstrate that HIF1α controls a two-armed mechanism to maintain collagen PTM in chondrocytes under hypoxia. It operates by regulating both the expression of cP4H subunits, which catalyze collagen hydroxylation, and the availability of oxygen – the substrate of the reaction. The identification of HIF1α as a pivotal regulator of ECM secretion under hypoxia significantly increases the extent of its known involvement in cellular homeostasis and may shed new light on metazoan evolution.

We thank N. Konstantin for expert editorial assistance, S. Krief for expert technical support, and all members of the Zelzer laboratory for advice and suggestions. We thank Prof. R. S. Johnson for kindly providing us with floxed-Hif1a mice. We appreciate the collaboration and fruitful discussions with Prof. D. R. Eyre, M. Weis and Prof. E. B. Hunziker. We thank Prof. C. W. Pugh for kindly providing us with the anti-sera to HIF1α. We are grateful to R. Kramer, Dr A. Tishbee and Dr V. Shinder from the Department of Chemical Research Support, and to Dr S. Ben-Dor from the Biological Services Unit at the Weizmann Institute of Science. We also acknowledge the help and advice of S. Viukov, K. B. Umansky and Dr O. Meir. The electron microscopy studies were conducted at the Irving and Cherna Moskowitz Center for Nano and Bio-Nano Imaging at the Weizmann Institute of Science.