Differential Regulation of Human Bone Marrow Mesenchymal Stromal Cell Chondrogenesis by Hypoxia Inducible Factor‐1α Hydroxylase Inhibitors

Abstract The transcriptional profile induced by hypoxia plays important roles in the chondrogenic differentiation of marrow stromal/stem cells (MSC) and is mediated by the hypoxia inducible factor (HIF) complex. However, various compounds can also stabilize HIF's oxygen‐responsive element, HIF‐1α, at normoxia and mimic many hypoxia‐induced cellular responses. Such compounds may prove efficacious in cartilage tissue engineering, where microenvironmental cues may mediate functional tissue formation. Here, we investigated three HIF‐stabilizing compounds, which each have distinct mechanisms of action, to understand how they differentially influenced the chondrogenesis of human bone marrow‐derived MSC (hBM‐MSC) in vitro. hBM‐MSCs were chondrogenically‐induced in transforming growth factor‐β3‐containing media in the presence of HIF‐stabilizing compounds. HIF‐1α stabilization was assessed by HIF‐1α immunofluorescence staining, expression of HIF target and articular chondrocyte specific genes by quantitative polymerase chain reaction, and cartilage‐like extracellular matrix production by immunofluorescence and histochemical staining. We demonstrate that all three compounds induced similar levels of HIF‐1α nuclear localization. However, while the 2‐oxoglutarate analog dimethyloxalylglycine (DMOG) promoted upregulation of a selection of HIF target genes, desferrioxamine (DFX) and cobalt chloride (CoCl2), compounds that chelate or compete with divalent iron (Fe2+), respectively, did not. Moreover, DMOG induced a more chondrogenic transcriptional profile, which was abolished by Acriflavine, an inhibitor of HIF‐1α‐HIF‐β binding, while the chondrogenic effects of DFX and CoCl2 were more limited. Together, these data suggest that HIF‐1α function during hBM‐MSC chondrogenesis may be regulated by mechanisms with a greater dependence on 2‐oxoglutarate than Fe2+ availability. These results may have important implications for understanding cartilage disease and developing targeted therapies for cartilage repair. Stem Cells 2018;36:1380–1392


INTRODUCTION
Acute lesions to the articular cartilage that do not heal may be painful and can progress to osteoarthritis. Conventional treatments such as microfracture or articular chondrocyte implantation are not always effective in mediating repair [1,2]. Tissue engineering strategies that combine cells with bioactive factors and biomaterial scaffolds may allow for de novo articular cartilage formation and provide an alternative therapy for patients [3][4][5]. However, the provision of cues that can appropriately direct progenitor cell differentiation and tissue formation remain a challenge.
One of the regulatory factors controlling articular cartilage development is the cellular a strongly enhances HIF-mediated transcription of key chondrogenic genes. Nevertheless, DMOG negatively impacted the production of Collagen Type II and glycosaminoglycans (GAGs), which could be alleviated by only exposing cells to the compound during the latter stages of chondrogenesis. Together, these observations highlight the potential importance of mechanisms which utilize 2-OG compared with Fe 21 for the transcriptional control of HIF target genes during hBM-MSC chondrogenesis. They also suggest that 2-OG inhibitors may better promote a chondrogenic transcriptome compared with either DFX or CoCl 2 . These observations may inform on improved, targeted strategies for stimulating cartilage ECM formation in tissue engineering-based therapies.

Isolation and Expansion of hBM-MSC
hBM-MSCs were isolated from bone marrow aspirates collected from the iliac crest of healthy pediatric donors, with informed consent from their parents or guardians. Cells were seeded in CellSTACK (Corning, Sigma Aldrich, UK) culture chambers at 10-25 3 10 6 /636 cm 2 and cultured in aMEM supplemented with human platelet lysate (Stemulate, Cook Medical, USA). At 90%-100% confluency, cells were passaged and seeded at 5,000 cells per cm 2 . For immunophenotyping of hBM-MSCs, the following antibodies were used in conjunction with a FACSCalibur analyzer (BD Biosciences, UK): CD90-FITC, CD105-APC, CD73-PE, CD34-PE, and CD45-FITC (all from BD Biosciences). All human tissue was approved for use by the UK National Research Ethics Service (12/WA/0196) and was collected by the National Institute for Health Research, which is supported by the Imperial College Healthcare Tissue Bank (HTA license 12275). Cultures were found to express CD90, CD105, CD73 and not express hematopoietic markers CD34 and CD45 [26] (data not shown). hBM-MSCs were expanded in growth media (GM; aMEM 1 10% fetal bovine serum [FBS], Thermo Fisher Scientific, UK) under standard conditions (5% CO 2 ).

Chondrogenic Induction of hBM-MSC
hBM-MSCs were expanded to passage 5 in GM under standard culture conditions before cryopreservation in a solution composed of 10% dimethyl sulfoxide, Sigma-Aldrich, 40% FBS, and 50% GM. Cells were stored in liquid nitrogen prior to use. For chondrogenic induction experiments, cryovials of hBM-MSCs were thawed in GM and grown to confluence before plating at 3 3 10 4 /cm 2 into multi-well tissue culture plates. Cultures were incubated for 24 hours in GM prior to induction using standard chondrogenic differentiation media (CDM). See Figure 1C for experimental plan. Cells were differentiated as monolayers to prevent the formation of a local hypoxic microenvironment independent of experimental conditions (physiological or chemically induced hypoxia). Indeed, while pellet/micromass cultures may be more conducive for chondrogenesis, the bioavailability of oxygen may vary between cells at the periphery and center of such cultures. CDM consisted of High Glucose Dulbecco's modified Eagle medium (Sigma-Aldrich) 1 2 mM L-Glutamine (Thermo Fisher Scientific) 1 100 nM Dexamethasome (Sigma-Aldrich) 1 1% Insulin, Transferrin, Selenium Solution (Thermo Fisher Scientific) 1 1% Antibiotic Antimycotic solution (Sigma-Aldrich) 1 50 lg/ml Ascorbic acid-2-phosphate (Sigma- Aldrich) 1 40 lg/ml L-proline (Sigma Aldrich) 1 10 ng/ml TGFb 3 (Peprotech). CDM was supplemented with HIF-stabilizing compounds (Sigma-Aldrich): 100 lM CoCl 2 , 50 lM DFX, and 200 lM DMOG, or incubated in un-supplemented CDM at hypoxia (5%O 2 ) or normoxia. To achieve HIF-1a inhibition, media was further supplemented with 500 nM Acriflavine (ACF; Santa Cruz Biotechnology, USA).

Neutral Red Viability Assay
Neutral red dye (Sigma-Aldrich) dissolved in cell culture medium was incubated with differentiating hBM-MSC for 2 hours before fixation in 0.1% Calcium Chloride 1 0.5% paraformaldehyde (both from Sigma-Aldrich). Dye retained by hBM-MSC was solubilized in 1% acetic acid 1 50% ethanol (both from Sigma-Aldrich). Quantification of solubilized Neutral Red was then performed on an absorbance spectrophotometer at 540 nm.

PicoGreen Assay
Samples were snap-frozen at 2808C and digested in 400 lg/ ml Papain Buffer at 658C for 18 hours. Double stranded deoxyribonucleic acid (dsDNA) content in papain-digested cultures was quantified using a PicoGreen kit (Thermo Fisher Figure 1. Schematics highlighting the role of hydroxylase inhibitors in regulation of HIF-1a-mediated transcription and the study experimental design. (A): Under hypoxic conditions, HIF-1a forms an active transcription complex with HIF-1b and co-factors such as CBP/ p300. This HIF complex then binds to the promoter regions of target genes at the HIF-response element sites, inducing transcription. (B): At normoxia, two hydroxylases-PHD2 and FIH, utilize oxygen and other substrates to hydroxylate HIF-1a which promotes its degradation and inhibits binding by CBP/p300. Here, we aimed to stabilize HIF-1a at normoxia by inhibiting the hydroxylases with CoCl 2 , DFX or DMOG. (C): Experiment design. To produce each biological replicate, hBM-MSCs were thawed and expanded to passage 5 before reseeding at a density of 3 3 10 4 cell per cm 2 in multi-well plates. Each well or set of wells was assigned to a specific condition: 20%O 2 , 20%O 2 1CoCl 2 , 20%O 2 1DFX, 20%O 2 1DMOG, or 5%O 2 . Separate experiments included each HIF-stabilizing compound in the presence or absence of ACF, and a comparison of late with constitutive exposure. In each condition, cultures were chondrogenically differentiated before assays at the time points specified in the legend of each figure. Abbreviations: ACF, Acriflavine; CoCl 2 , cobalt chloride; DFX, desferrioxamine; DMOG, dimethyloxalylglycine; ECM, extracellular matrix; FIH, factor inhibiting hypoxia inducible factor; hBM-MSCs, human bone marrow-derived mesenchymal stem cells; HIF, hypoxia inducible factor; PHD2, prolyl hydroxylase 2.

Sodium Dodecyl Sulfate-PAGE and Western Blotting
Following 24-hours of culture, cells were lysed in sodium dodecyl sulfate (SDS) buffer and protein was quantified using a Bicinchoninic Acid assay (Thermo Fisher Scientific). Lysates were run on polyacrylmide gels (Biorad, UK) and transferred using the Trans-Blot Turbo Transfer System (Biorad). HIF-1a and housekeeping protein b-Actin were bound by primary antibodies (H-206; Santa Cruz Biotechnology and ab8227; Abcam, UK). Signal detection produced between a horseradish peroxidase-conjugated secondary antibody (sc-2004; Santa Cruz) and the Chemiluminescent ECL substrate (Biorad) were detected on a Chemidoc Touch imaging platform (Biorad). HIF-1a and protein levels were generated by densitometric analysis with ImageJ and normalized to that of b-Actin.

Quantitative Polymerase Chain Reaction
RNA was extracted using the RNeasy Mini Kit (Qiagen, DE). Hundred nanograms of RNA per sample was reverse transcribed using M-MLV Reverse Transcriptase (Promega, UK) and cDNA was amplified using quantitative polymerase chain reactions (qPCR) in a CFX384 (Biorad). Brilliant III Ultra-Fast SYBR Green QPCR Master Mix (Agilent, USA) was used in conjunction with primers specific to genes of interest. Primer sequences are shown in Supporting Information Table S1. All primers produced a linear relationship between template concentration and Ct value. Reaction efficiencies were confirmed to lie between 90 and 110%. Raw Ct values were converted to transcript copy number by the relative standard curve method of analysis, and expression levels were normalized to that of RPL13A. Following normalization to the housekeeping gene, expression levels were then normalized to that of the untreated control to determine fold change in expression induced by each treatment.

Immunofluorescence Staining
Cultures were fixed in 4% (wt/vol) paraformaldehyde for 15 minutes. HIF-1a and Collagen Type II were then detected using H-206 (Santa Cruz) and ab34712 (Abcam), respectively, overnight, following blocking with (10%) goat serum (Sigma-Aldrich) for 60 minutes and permeabilization in 0.1% (vol/vol) Triton X-100 solution (Sigma Aldrich) for 60 minutes, both at room temperature (RT). Collagen Type X was detected using ab49945 (Abcam) at a 1:250 dilution overnight. Rabbitderived primary antibodies were visualized with ab150077 (Abcam) after staining for 60 minutes at RT at dilutions of 1:100 and 1:200 for Collagen Type II and HIF-1a, respectively. Mouse-derived primary antibodies were detected with biotin (ab6788, Abcam) and Streptavidin (S11223, Thermo Fisher Scientific) both at 1:350 for 60 minutes. Cultures were counterstained with 0.1 lg/ml DAPI for 60 minutes to visualize nuclei and fluorescence was imaged on an Axiovert200M microscope (Zeiss, DE). The images in Supporting Information Figure S1 confirm that signal was due to each primary antibody and not background fluorescence or nonspecific binding of the secondary antibody.

Alcian Blue Staining
Cultures fixed in 4% paraformaldehyde were stained with 1% Alcian blue solution, pH 1.0 (Sigma-Aldrich) prepared in 0.1N HCl. Hematoxylin (Vector Laboratories, UK) was used to visualize cell nuclei and staining was imaged on an Axiovert200M microscope (Zeiss).

Glycosaminoglycan Quantification
At day 21 of chondrogenesis, cultures were washed in phosphate buffered saline and frozen at 2808C before their digestion in 400 lg/ml Papain buffer (Sigma-Aldrich) supplemented with 0.2M Sodium Phosphate 1 5 mM Ethylenediaminetetraacetic acid 1 5 mM L-Cysteine (all Sigma-Aldrich) at 658C for 18 hours. GAGs were quantified from Papain-digested lysates using a Blyscan GAG assay kit (Biocolor, UK) in which GAGs were dyed with 1,9-dimethyl-methylene blue and subsequently dissociated with Propan-1-ol solution before quantification on an absorbance spectrophotometer at 640 nm. Values were normalized to levels of dsDNA, which were quantified using the PicoGreen assay.

Immunofluorescence Quantification
Immunofluorescence images were captured using identical gain, exposure, and offset for all conditions in each experiment. These were determined with positive controls that expressed the antigen of interest, and negative controls in which the primary antibody was omitted (Supporting Information Fig. S1). The same threshold fluorescence intensity for images of all conditions within an experiment was set, below which the signal produced was negated as background. The signal produced above the threshold was regarded as bona fide protein detection and was used to create a binary representation of each image. The percentage of immunofluorescence staining present within a specified area was then determined.

Statistical Analysis
All statistical analyses were performed in Prism7 (GraphPad, USA) with the Mann-Whitney test used to compare two conditions and Kruskal-Wallis with Dunn's Correction for multiple condition comparisons. Nonparametric tests were used as we were unable to demonstrate normality in all datasets. *marks all differences that were statistically significant (p < .05).

Hypoxia Promotes HIF Stabilization and a More Articular Cartilage-Like Cell Phenotype
It is well established that hBM-MSCs can be chondrogenically differentiated with transforming growth factor b 3 (TGF-b 3 ) ligands. Therefore, we first aimed to determine if chondrogenesis could be further enhanced by culture under hypoxic conditions, as previously reported [11]. Hypoxia increased expression of a selection of known HIF target genes including VEGFA, EGLN, and PGK1 (all p 5 .0286) [27][28][29] compared with that in hBM-MSC cultured under normoxic conditions ( Fig. 2A). These observations were in line with previous studies which have similarly shown a rapid (24 hours) upregulation of HIF and HIF-mediated transcription in response to 4 HIF-1a-Stabilizing Agents for Chondrogenesis  hypoxic conditions compared with that at normoxia (Fig. 2B). COL2A1 and COL10A1 are targets of transcription factors SOX9 and RUNX2, respectively, and are known to be regulated as the chondrogenic differentiation of MSC proceeds [11]. Culture for 21 days under hypoxic conditions did not affect cell viability or proliferation (Fig. 2C). However, as expected, we did observe increased HIF-1a nuclear localization (p < .0001) in hypoxic compared with normoxic cultures (Fig. 2D-2H). Hypoxia also increased Alcian Blue staining of GAGs (Fig. 2I,  2J), but did not affect the immuno-detection of Collagen Type II protein (Fig. 2K, 2L). Nevertheless, we did detect a decrease in staining for Collagen Type X (Fig. 2M, 2N), consistent with hypoxia's inhibitory role to chondrocyte hypertrophy [17]. Together, these observations confirmed that culture under hypoxic conditions in the presence of TGF-b 3 promoted an articular chondrocyte-like phenotype that was conducive for articular cartilage ECM rather than hypertrophic cartilage formation. This effect appeared to not require a corresponding upregulation of SOX9, but instead correlated with increased immunostaining for HIF-1a, upregulation of select HIF target genes VEGFA, EGLN, and PGK1, and increased HIF-1a nuclear localization.
CoCl 2 , DFX, and DMOG Induce HIF-1a Localization, but Only DMOG Strongly Upregulates HIF Targets VEGFA, PGK1, and ELGN Having determined that hypoxia promoted HIF-1a stabilization and expression of VEGFA, EGLN, and PGK1, we next aimed to determine if inhibitors of the hydroxylases PHD2 and FIH would have a similar effect on hBM-MSC cultured under normoxic conditions. We first determined appropriate doses for the hydroxylase inhibitors DMOG, DFX, and CoCl 2 by confirming that concentrations of each used extensively in the literature [22,24,[32][33][34] were nontoxic to hBM-MSC over 21 days of chondrogenic differentiation (Supporting Information Figs. S2, S3). Next, we confirmed that each could stabilize HIF by carrying out Western blots for HIF-1a in whole-cell lysates after 24 hours, as HIF is known to be rapidly induced in response to PHD2/FIH inhibition [22]. Levels of HIF-1a protein were significantly increased in cells cultured under hypoxic conditions (p 5 .0286); however, despite trends for increased levels of HIF-1a after treatment with HIF stabilizing compounds, we failed to detect statistically significant differences (p 5 .314) compared with controls (Fig. 3A, 3B). Nonetheless, nuclear localization of HIF-1a was enhanced compared with controls (p .0001) in response to treatment with all three compounds (Fig. 3C-3K).

DMOG Inhibits Incorporation of Collagen Type II, Type X, and GAGs into the Cell-Secreted ECM
As treatment with DMOG regulated the expression of genes associated with a chondrocyte phenotype, we next asked if this influenced cartilage-like matrix formation. In line with changes in gene expression, hBM-MSC treated with DMOG for 21 days showed little to no staining for Collagen Type X compared with controls (Fig. 5A, 5D). We observed a similar effect in both DFX-and CoCl 2 -treated cultures (Fig. 5B, 5C). However, while CoCl 2 -and DFX-treated cultures showed similar levels of staining for Collagen Type II as controls (Fig. 5E-5G), DMOG-treated cultures showed only sparse staining (Fig.  5H). This was confirmed by quantification of Collagen Type II immunofluorescence both without (p 5 .0286) and with normalization to cell number which indicated reduced Collagen Type II production per cell ( Fig. 5M; p 5 .0286). Alcian blue staining confirmed these observations as DMOG-treated cultures showed fewer GAG-positive areas than the other groups, although quantitative differences in staining on a per cell basis were not significant ( Fig. 5I-5L, 5N). Overall, DMOG appeared to reduce the total amount of cartilage-like ECM that cells formed in their immediate extracellular space.

HIF-1a Mediates DMOG's Induction of an Articular Chondrocyte Transcriptional Profile
As DMOG mediated antithetical effects in terms of chondrogenic transcriptional profile and ECM formation, we next aimed to study its mechanism of action. To accomplish this, we supplemented CoCl 2 /DFX/DMOG-containing CDM with Acriflavine (ACF), an inhibitor of HIF-1a and HIF-1b binding [38]. ACF abolished the DMOG-induced upregulation of established HIF targets, but did not affect total cell number during chondrogenesis (Supporting Information Fig. S4A, S4B). Staining for Collagen Type II in DMOG-treated cultures supplemented with ACF remained sparse (Fig. 6A-6C), but quantitative image analyses showed staining on a per cell basis was no different from controls, while cultures treated with DMOG alone were significantly lower ( Fig. 6D; p 5 .0076). This suggests that the inhibitory role of DMOG on Collagen Type II matrix formation may be partly mediated through HIF-1a activity.
We next asked if DMOG's stimulation of the chondrogenic transcription profile in hBM-MSC was also mediated through HIF-1a. ACF abrogated DMOG-mediated changes in expression of HIF targets, VEGFA (   6G; p 5 .0286). This is consistent with the observation that ACF reduced the ratio of COL2A1/COL10A1 under hypoxic conditions ( Fig. 6J; p 5 .0286) and suggests that hypoxia, via HIF-1a, does indeed regulate basal levels of chondrogenic targets genes, such as SOX9. Overall, these data suggest that HIF-1a mediated DMOG's effect on the transcriptional profile of chondrogenically induced hBM-MSC. Moreover, ACF appeared to have a larger effect on DMOG-mediated transcription than that induced by either CoCl 2 or DFX (Supporting Information Fig. S4C-S4J).

Late DMOG Treatment Enhances Chondrogenesis
As DMOG upregulated chondrogenic transcripts but continuous treatment led to reduced staining for cartilage-like matrix, we next asked if altering either the length/timing of treatment would influence ECM formation. Therefore, we next treated hBM-MSC with DMOG, DFX, and CoCl 2 either continuously (as before) or during late (days 14-21) time periods and analyzed mRNA and protein expression of ECM markers after 21 days. Late DMOG treatment did not negatively affect the secretion of Collagen Type II compared with controls (p 5 .282), as we observed with continuous DMOG treatment (p 5 .0188). This contrasted with treatment with either DFX or CoCl 2 , where both continuous and late treatment had no effect on Collagen Type II secretion (p .9999 for both, Fig.  7A-7H). At the gene expression level, like continuous treatment (p 5 .0023), late exposure to DMOG induced significant upregulation of SOX9 ( Fig. 7I; p 5 .0168). Late DMOG also upregulated expression of P4HA1 ( Fig. 7J; p 5 .0286), and HIF targets VEGFA (p 5 .0358, Fig. 7K) and EGLN (p 5 .0208, Fig. 7L) as with continuous DMOG treatment (p 5 P4HA1: .0313, VEGFA: .0118, EGLN: .0088). In contrast, neither continuous nor late CoCl 2 and DFX treatment significantly affected the expression of these genes, with the exception of continuous DFX treatment on SOX9 (p 5 .0286; Fig. 7I) and P4HA1 (p 5 .0286; Fig. 7J). Taken together, late treatment with DMOG induced a similar expression profile to continuous treatment, but without negatively impacting the formation of cartilage-like ECM.

DISCUSSION
Hypoxic conditions are known to favor articular cartilage development. The pro-chondrogenic effects of hypoxia are thought to be mediated primarily through HIF-1a via the formation of a transcriptionally-active complex at target genes [7,12]. Therefore, we and others hypothesized that compounds that increase HIF-1a availability would promote HIFmediated chondrogenesis. Previous studies have examined the effect of CoCl 2 [12], DFX [39], and DMOG [21,40] in this context. While such studies have cemented the role of HIF-1a in chondrogenesis, to our knowledge no study has yet examined their comparative effects during cartilage formation or the chondrogenic differentiation of precursors. As the inhibitors have differential mechanisms of action, comparatively studying their effects may have important implications for HIF biology and cartilage regenerative medicine. Indeed, instead of utilizing physiological hypoxia for regenerative medicine, stabilizing the HIF complex under normoxic conditions would remove the complex logistics required for spatial organization of oxygen. This may be particularly valuable in engineering constructs for the repair of full osteochondral defects due to the contrasting oxygen requirements of avascular cartilage and vascularized bone [6]. HIF mimetics could also potentially avoid the undesirable HIF-independent effects of hypoxia such as the unfolded protein response and associated cell stress [41], and could preclude the development of a tolerance to the reduced oxygen levels [42,43].
In our control conditions, we defined hypoxia as 5%O 2 to balance its well-described pro-chondrogenic effects against its negative impacts on cell viability [44]. As expected, after 24 hours in culture under hypoxic conditions, we detected upregulation of HIF target genes, as others have described [19,45], as well as increased expression of SOX9 target COL2A1 and downregulation COL10A1 (day 14). We also detected an increase in staining for GAGs and reduced Collagen Type X protein formation. Surprisingly, upregulation of SOX9 was not maintained throughout the 21-day differentiation. This is in keeping with previous reports that continued upregulation of SOX9 expression in mouse MSC under hypoxic conditions does not correlate with upregulation of its target genes [11]. We also observed that SOX9 expression was downregulated in the presence of ACF, perhaps suggesting that hBM-MSC cultures do rely on HIF for physiological hypoxia's downstream effects. Indeed, cells may develop a tolerance to hypoxia following the initial induction [43], and during long-term culture, hypoxia may act to maintain basal levels of expression of chondrogenic genes.
One of our most striking observations was the ability of DMOG, via HIF-1a, to induce hBM-MSC to upregulate expression of HIF target genes and chondrogenic transcripts, and downregulate mRNA encoding hypertrophic chondrocyte markers such as Collagen Type X. In comparison, neither CoCl 2 nor DFX stimulated similar changes, despite their ability to promote HIF-1a nuclear localization. The stability and nuclear localization of HIF-1a is controlled by PHD2, whereas HIF-1a co-factor binding is controlled by FIH; DMOG has been shown to inhibit both hydroxylases [22]. This is unlike the effect of iron chelators which target PHD2, but do not inhibit FIH as potently [24]. Others have shown that FIH requires higher levels of 2-OG than PHD2 to achieve the same levels of enzymatic activity [46], which may suggest an increased sensitivity of FIH than PHD2 to inhibition by 2-OG analogs. HIF-1a activity in hBM-MSC may also be more dependent on FIH inhibition, rather than PHD2, as high levels of HIF-1a mRNA have been observed in these cells [42]. Indeed, high levels of HIF-1a transcription might compensate for decreases in HIF-1a stability due to PHD2-mediated hydroxylation. Taken together, these observations suggest that regulation of HIFmediated transcription that is conducive for hBM-MSC articular chondrogenesis is dependant more on 2-OG-mediated mechanisms than those controlled by intracellular Fe 21 levels. Additionally, previous studies which demonstrate the dependence of FIH on 2-OG availability and the ability of DMOG to inhibit both PHD2 and FIH, suggest that DMOG's potent effect here may be via inhibition of both hydroxylases, whereas CoCl 2 and DFX may inhibit PHD2 only. The ability of DMOG to induce an expression profile that is conducive for articular chondrogenesis, suggests its advantage over CoCl 2 and DFX for use in cartilage-regenerative therapies. However, despite inducing expression of COL2A1 and genes involved in post-translational modifications of collagen, DMOG had a negative effect on cartilage-like ECM production. We showed that this was partly mediated via HIF-1a, however, other mechanisms are likely involved as we were unable to completely rescue cartilage-like ECM formation with Acriflavine. DMOG has been shown to reduce the activity of prolyl-4-hydroxylase, which is required for correct folding and polymerization of collagen fibrils [21]. Correspondingly, both FIH and collagen prolyl hydroxylase (CP4HA1) have similar affinities for 2-OG, as they have similar K m values for this cofactor [47]. Therefore, FIH and P4HA1 are likely equally sensitive to DMOG. This suggests that DMOG-mediated upregulation of HIF target genes via FIH inhibition might be accompanied by a similarly potent inhibition of collagen processing and incorporation into the ECM. Treatment with DMOG for the final 7 days of induction restored the reduced levels of Collagen Type II while upregulating expression of HIF target and chondrogenic genes to similar levels we observed in response to continuous treatment. This response could have been mediated by a lack of continuous inhibition of the collagen prolyl hydroxylase. Taken together, late DMOG treatment, which can stimulate the formation of appropriate ECM, and

CONCLUSION
Hydroxylase inhibitors are potentially valuable in cartilage tissue engineering strategies as they can mimic many of the effects of hypoxia, providing important environmental cues to progenitors, but without many of its potential drawbacks. Here, we show that CoCl 2 , DFX, and DMOG treatment all induced HIF-1a stabilization. However, unlike CoCl 2 and DFX, DMOG treatment strongly regulated HIF targets, and promoted chondrocyte-specific gene expression. This suggests that in hBM-MSC undergoing chondrogenic differentiation, HIF-mediated changes in gene expression may rely more on mechanisms that utilize 2-OG than those that rely on Fe 21 . Our observations also suggest a role for DMOG in cartilage tissue engineering strategies. For example, scaffolds that spatially and/or temporally control the release of DMOG could target the articular cartilage to aid in the repair of focal defects. However, the maintenance of cartilage ECM in late treatment-only conditions suggests the use of this 2-OG analog would need to be optimized with regard to dosage/treatment time. Alternatively, knowledge that DMOG inhibits both FIH and PHD2 may suggest that dual and specific inhibition of these hydroxylases during de novo cartilage formation may result in HIF-mediated transcription that is conducive for articular chondrogenesis.