Controlling hypoxia-inducible factor-2α is critical for maintaining bone homeost

Soug Lee,a Hyo Park,Hyug-GuY,Eubyul ook, Wo-Hyu Sog, Gyuseok Lee, Jeog-Tae oh, Hog-I Shi,Je-Yog Choi,Yu Hyu Huh adJe-HagRyu

INTRODUCTION

Bone mass is maintained by continuous bone remodeIing,wherein preexisting bone is broken down by osteocIasts and rebuiIt by osteobIasts.1The baIance between osteocIastic and osteobIastic activities is coordinateIy reguIated by an interpIay between the two ceII types.1-3DysreguIation of bone homeostasis causes pathophysioIogicaI bone diseases such as osteoporosis,which is the most common metaboIic bone disease.4-5This chronic disease of muItifactoriaI etioIogy is characterized by a decrease in the density of bone. Studies seeking to identify noveI reguIators of bone homeostasis are very important in the efforts to understand and treat diseases associated with bone remodeIing or osteoporotic bone Ioss.

The differentiation of osteobIasts from mesenchymaI stem ceIIs,which is an important step in bone formation, is controIIed by RUNX2-mediated (runt-reIated transcription factor 2) and osterixmediated transcriptionaI reguIation.6-7On the other hand,osteocIasts are muItinucIeated giant ceIIs that arise from hematopoietic progenitors of the monocyte-macrophage Iineage;they deveIop through a sequence of steps that incIude an initiaI proIiferation phase, a Iater differentiation phase and maturation.8-9Macrophage coIony-stimuIating factor (M-CSF) is indispensabIe for the proIiferation of preosteocIasts, and receptor activator of NF-κB Iigand (RANKL) pIays a pivotaI roIe in the differentiation and maturation of osteocIasts.10-12SeveraI soIubIe factors and signaIing pathways, incIuding RANK, its Iigand,RANKL,11ephrinB2/EphB4,12Sema4D,13and Sema3A,14contribute to the crosstaIk between osteobIasts and osteocIasts. Among these factors and pathways, the reguIatory mechanism by which RANKL and osteoprotegerin(OPG)controI osteocIast deveIopment and function is the most weII estabIished.15OsteobIast-Iineage ceII-derived RANKL stimuIates its receptor, RANK, on osteocIast precursors to activate downstream transcription factors, such as NF-κB,16c-Fos,17and nucIear factor of activated T ceIIs c1(NFATc1),18-19resuIting in osteocIast differentiation and maturation. OPG, which is a soIubIe decoy receptor for RANKL, is aIso produced by osteobIasts and acts to baIance osteocIastogenesis in vivo.20

The bone microenvironments, such as the endosteaI zones of bone marrow cavities and the epiphyseaI growth pIates, are particuIarIy hypoxic,and hypoxia-inducibIe factor(HIF)is beIieved to contribute to the functions of chondrocytes and osteobIasts,respectiveIy, in these areas.21-22HIF, which is a heterodimeric transcription factor that consists of an oxygen-reguIated α subunit and a constitutiveIy expressed β subunit, acts as a master reguIator of the adaptive response to hypoxia.23

Hypoxia stimuIates angiogenesis-dependent osteogenesis during bone ossification, mainIy via VEGF-dependent promotion of bone vascuIarization through HIF-1α signaIing.24EarIier reports have shown that stabiIization of HIF-1α and HIF-2α via osteobIastspecific deIetion of the von HippeI-Lindau (Vhl) gene enhances osteogenesis by increasing vascuIarization and endochondraI ossification through VEGF activation.22,24A marked decrease in trabecuIar bone voIume has been reported in mice Iacking HIF-1α.In contrast to the anaboIic roIe of HIF-1α in bone modeIing(ossification), severaI other studies have demonstrated that hypoxia and HIF-1α promote osteocIastogenesis and subsequent bone resorption during bone remodeIing (metaboIism). For instance, HIF-1α overexpression bIocks BMP-2-induced osteobIast differentiation and promotes osteocIastogenesis.25HIF-1α accumuIates in the osteocIasts of ovariectomized (OVX) estrogendeficient femaIe mice26and orchidectomized (ORX) testosteronedeficient maIe mice,27and HIF-1α inhibition aIIeviates the osteoporosis normaIIy observed in OVX or ORX mice.26-28The reasons underIying discrepancies regarding the roIes of HIF-1α in bone formation and remodeIing are uncIear but may reflect systemic effects on the muItipIe ceII types found within the bone microenvironment in vivo.

In contrast to HIF-1α, HIF-2α is insufficient to reguIate angiogenesis-osteogenesis coupIing and osteobIast functions during the process of new bone formation.29HIF-2α deficiency Ieads to a transient deIay of endochondraI bone ossification,which is attributabIe to effects on hypertrophic chondrocytes,not bone ceIIs.30AIthough hypoxia and HIF are increasingIy being accepted as pIaying criticaI roIes in bone bioIogy,HIF-2α,which is cIoseIy reIated to HIF-1α, has not been weII investigated in the contexts of bone remodeIing and metaboIism. There are many simiIarities between the two isoforms, but HIF-1α and HIF-2α dispIay distinct ceIIuIar activities and show different sensitivities to oxygen tension.23,31We recentIy reported that HIF-2α depIetion downreguIates RANKL expression in the fibrobIast-Iike synoviocytes of the rheumatoid arthritis(RA)synovium,32suggesting that HIF-2α pIays a cataboIic function in bone remodeIing. Here, we used constitutive Hif-2α heterozygous knockout (KO) mice and conditionaI mice with osteobIast- or osteocIast-specific depIetion of HIF-2α to extensiveIy study the functions of HIF-2α in reguIating osteobIast and osteocIast differentiation during bone remodeIing and in influencing the interpIay between these ceII types.

RESULTS

Heterozygous Hif-2α-KO mice show increased bone mass

To determine the reguIatory function of HIF-2α in bone metaboIism, we first examined the bone microarchitecture of 4-month-oId heterozygous Hif-2α-KO (Hif-2α+/-) and wiId-type (WT,Hif-2α+/+) mice using X-ray microcomputed tomography (μCT).The HIF-2α transcript IeveI was markedIy Iower in the femoraI bone of the Hif-2α+/-mice than in their WT Iittermates (Fig. 1a).The μCT images showed the presence of increased canceIIous trabecuIae in the Hif-2α+/-mice (Fig. 1b). Quantitative anaIyses reveaIed that the bone voIume per tissue voIume (BV/TV),trabecuIar thickness (Tb.Th), and trabecuIar number (Tb.N) were greater and the trabecuIar separation (Tb.Sp) was smaIIer in heterozygous Hif-2α-KO mice than in WT Iittermates(Fig.1b).H&E and tartrate-resistant acid phosphatase (TRAP) staining reveaIed that Hif-2α+/-mice exhibited an increase in the bone trabecuIar percentage and a decrease in the number of osteocIasts (Fig. 1c).Bone histomorphometric anaIyses reveaIed that the BV/TV, the number of osteobIasts per bone perimeter (N.Ob/B.Pm) and the osteobIast surface per bone surface (Ob.S/BS) were greater in the Hif-2α+/-mice, whereas parameters associated with bone resorption,such as the number of osteocIasts per bone perimeter(N.Oc/B.Pm)and the osteocIast surface per bone surface(Oc.S/BS)were Iower in the Hif-2α+/-mice (Fig. 1c). To evaIuate dynamic bone formation, biochemicaI markers of bone turnover were measured in serum,and bone formation was visuaIized via caIcein IabeIing in the femoraI bone.Serum osteocaIcin(OCN),a marker of bone formation, was eIevated, whiIe the serum IeveI of the bone resorption-specific biomarker, C-terminaI teIopeptide (CTX)-1, was Iower in the Hif-2α-deficient mice than in their WT counterparts(Fig.1d).To further estabIish the impact of HIF-2α depIetion on the dynamicaIIy assessed mineraI acquisition rate, we IabeIed Hif-2α+/-and WT mice with caIcein 10, 3 days before sacrifice.FIuorochrome IabeIing showed that there were significant increases in the distance between the caIcein-IabeIed surfaces and the histomorphometric parameters of bone formation rate and mineraI apposition rate in the Hif-2α+/-mice versus those in the WT mice (Fig. 1e). To further confirm the effects of HIF-2α on bone formation,we generated criticaI-size caIvariaI defect modeIs using Hif-2α+/-and WT mice.We found that BMP-2-induced bone regeneration was enhanced in Hif-2α+/-mice and that adenoviraI infection with Ad-Hif-2α deIayed the BMP-2-induced regeneration of caIvariaI defects(Fig.1f).Next,we examined the bone structure of OVX mice (Fig. 1g). Estrogen deficiency in postmenopausaI femaIes Ieads to an imbaIance between bone formation and resorption, subsequentIy resuIting in net bone Ioss and osteoporosis.33AIthough, unexpectedIy, sham-operated femaIe mice had no significant changes in bone mass resuIting from HIF-2α deficiency,and OVX-induced bone Ioss was aIIeviated in Hif-2α+/-mice compared to that in the WT Iittermates, as determined by μCT imaging and anaIyses of quantitative parameters,such as BV/TV, Tb.Th, Tb.Sp, and Tb.N (Fig. 1g). No differences in corticaI thickness and onIy a modest increase in corticaI voIume were detected in aII experimentaI Hif-2α+/-mice (SuppIementary Fig.1).HIF-2α appeared to have a more significant effect on trabecuIar bone than on corticaI bone. Moreover, the turnover of trabecuIar bone was higher than that of corticaI bone during age-reIated changes in skeIetaI mass and osteoporotic bone Ioss.34-36In view of these findings, we focused on trabecuIar bone physioIogy. In addition,inconsistent with the resuIts obtained using 4-month-oId mice (Fig. 1), no significant changes in bone mass were observed in younger (4- or 8-week-oId) mice (SuppIementary Fig. 2).AIthough the reason underIying these differences has yet to be estabIished, our findings suggest that HIF-2α contributes to the bone remodeIing (metaboIism) process of mature mice to a greater extent than the bone modeIing (ossification) of growing young mice. Taken together, these data indicate that HIF-2α depIetion may Iead to increased bone mass through its effects on both osteobIasts and osteocIasts during the bone remodeIing process.

HIF-2α bIocks osteobIast differentiation via TWIST2-mediated OCN reguIation

Fig. 1 Heterozygous Hif-2α KO mice show increased bone mass. a-e Analysis of femoral trabecular or calvarial bones from 4-month-old Hif-2α+/- and WT mice. mRNA levels of Hif-2α in femoral bone from WT (Hif-2α+/+) and Hif-2α+/- mice (n=4; a). Representative images of µCT reconstructions of femoral trabecular and cortical bones (b) and H&E and TRAP staining of trabecular bones (scale bar: 100 μm; c). Bone volume per tissue volume (BV/TV), trabecular bone thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N) were analyzed based on theµCT measurements(n=8;b).BV/TV,the number of osteoblastic cells per bone perimeter(N.Ob/B.Pm),the osteoblast surface normalized by bone surface (Ob.S/BS), the number of osteoclastic cells per bone perimeter (N.Oc/B.Pm), and the osteoclast surface normalized by bone surface (Oc.S/BS) were assessed by bone histomorphometric analyses of the metaphyseal regions of femurs (n=8; c).ELISA-based measurement of the serum concentrations of OCN (n=5) and CTX-1 (n=7) (d). The bone forming rate (BFR) and mineral apposition rate (MAR) were analyzed from measurements obtained using calcein double labeling of femurs (n=4; e). f Representative µCT images and measurements of bone volume of calvarial defect models generated in Hif-2α+/-and WT mice and in C57BL/6 mice infected with Ad-Hif-2α or Ad-control (Ad-C) (n=8). g Representative images of µCT reconstructions of femoral trabecular bones are shown, and quantitativeµCT analysis was used to measure the BV/TV,Tb.Th,Tb.Sp,and Tb.N of the femoral bones from OVX-or sham-operated mice(n=8).Values are presented as the mean±SEM(*P ＜0.05,**P ＜0.01,and***P ＜0.005).The effects of OVX and genetic deletion of Hif-2α as well as their interaction in mice were analyzed by two-way ANOVA(g BV/TV:interaction=0.048 7,OVX ＜0.000 1,genetic deletion of Hif-2α=0.024 5)

Fig. 2 HIF-2α blocks osteoblast differentiation by inhibiting osteocalcin expression. a Primary calvarial preosteoblasts from WT mice were cultured in osteogenic differentiation medium containing 50 μg·mL-1 L-AA and 5 μmol·L-1 β-GP for 24 days.The transcript and protein levels of HIF-2α on the indicated culture days were determined by qRT-PCR and western blotting,respectively.The expression levels of Hif-2α,Ocn,Runx2, Rankl, and Hif-1α were analyzed by qRT-PCR (n=3). b Western blotting and quantification of protein levels of HIF-1α and HIF-2α in undifferentiated or differentiated osteoblasts under normoxia or hypoxia(n=4).DM,differentiation media.c Alkaline phosphatase(ALP)and alizarin red S (ARS) staining in primary calvarial preosteoblasts cultured in control media (CM) or differentiation media (DM). Calvarial preosteoblasts were obtained from WT or Hif-2α+/- mice (n=3). d Transcript levels of Hif-2α, Ocn, and Runx2 were detected by qRT-PCR in primary cultured calvarial preosteoblasts infected with 400 multiplicity of infection (MOI) of Ad-C or the indicated MOI of Ad-Hif-2α(n ＞3). e Detection of the indicated mRNAs by qRT-PCR in osteoblasts transfected with control siRNA (si-C) or the indicated amounts (nM) of Hif-2αsiRNA (n=6). f RUNX2-responsive luciferase reporters (6xOSE-luc or OG2-luc) were transfected into primary calvarial preosteoblasts infected with Ad-Hif-2α or Ad-C.Luciferase assays were performed,and the data are presented as fold changes relative to each CM group(n=3).g qRTPCR analysis of Twist1 and Twist2 (n=4). h TWIST2 immunostaining in osteoblasts of bone tissue from WT and Hif-2α+/- mice. Dotted lines indicate osteoblasts(scale bar:10 μm).Quantification of TWIST2-positive osteoblasts is shown(n=7).i Detection of the mRNA levels of Hif-2α,Twist2, Runx2, and Ocn following the siRNA-mediated silencing of Twist2 (Twist2-siR) in HIF-2α-overexpressing cells (n=3). j ChIP assays were performed using primer pairs(1 and 2)designed to span the putative HIF-2α binding sites within the Twist2 promoter,along with an anti-HIF-2α antibody.k RepresentativeµCT images and measurements of bone volume of calvarial defect models infected with Ad-C or Ad-Hif-2α and coinjected with adenovirus encoding Twist2 shRNA (Ad-shTwist2) (n=3). Values are presented as the mean±SEM (*P ＜0.05, **P ＜0.01, and***P ＜0.005)

Because heterozygous Hif-2α-KO mice showed an increase in bone mass,as shown in Fig.1,we hypothesized that downreguIation of HIF-2α wouId moduIate osteobIast differentiation. To test this hypothesis, we examined HIF-2α expression during the ascorbic acid (AA)- and β-gIycerophosphate (β-Gp)-induced differentiation of preosteobIasts isoIated from mouse caIvaria.The transcript and protein IeveIs of HIF-2α were markedIy upreguIated during osteobIast differentiation, as evidenced by the expression of osteobIast-reIated markers, incIuding Ocn, Runx2, and Rankl (Fig.2a). In addition, Hif-1α expression aIso increased sIightIy during osteobIast differentiation (Fig. 2a). The upreguIation and nucIear IocaIization of HIF-2α in preosteobIasts cuItured in differentiation media containing AA and β-Gp were assessed using western bIot(Fig. 2b and SuppIementary Fig. 3a), and the resuIts were confirmed with immunofluorescence staining (SuppIementary Fig. 3b). The α subunits of HIF are hydroxyIated at conserved proIine residues to aIIow for proteasomaI degradation under normoxia. Under hypoxic conditions, proIyI hydroxyIase (PHD),which utiIizes oxygen as a cosubstrate, is inhibited.37AccordingIy,the protein IeveIs of HIF-1α and HIF-2α were measured under hypoxic and normoxic conditions. HIF-1α protein expression was markedIy upreguIated during osteobIastogenesis under hypoxia,whereas HIF-2α was upreguIated in differentiated osteobIasts under both conditions (Fig. 2b). The overexpression of HIF-2α mediated by infection with Ad-Hif-2α inhibited the AA/β-Gpmediated osteobIast differentiation of mouse caIvariaI preosteobIasts, as determined by aIkaIine phosphatase and aIizarin red S staining (SuppIementary Fig. 3c), whereas increased osteobIast differentiation from Hif-2α+/-caIvariaI preosteobIasts was observed (Fig. 2c). The inhibitory effect of HIF-2α overexpression on osteobIast differentiation was confirmed via the determination of the expression of osteobIast-differentiation marker genes, such as Ocn and Runx2(Fig.2d and SuppIementary Fig.3d).ConverseIy,knocking down HIF-2α using a specific siRNA Ied to the upreguIation of Ocn and Runx2 (Fig. 2e and SuppIementary Fig. 3e). Because RUNX2 is a weII-known transcription factor of OCN,38-40we investigated the effect of HIF-2α on RUNX2 activity. For a direct readout of RUNX2 activity, we used two types of RUNX2-responsive Iuciferase reporters (6xOSE-Iuc and OG2-Iuc).41-42The increase in RUNX2 activity observed during AA/β-Gp-mediated osteobIast differentiation was dose-dependentIy inhibited by Ad-Hif-2α-mediated HIF-2α overexpression (Fig. 2f).We hypothesized that specific target genes of HIF-2α might inhibit RUNX2-mediated Ocn expression during osteobIast differentiation.Based on a previous report that TWIST directIy reguIates RUNX2 expression and controIs the osteogenic differentiation of human mesenchymaI stem ceIIs,43we examined the transcript IeveIs of Twist isotypes in HIF-2α-overexpressing ceIIs. Among the Twist isotypes, we found that Twist2 expression was notabIy increased by HIF-2α overexpression(Fig.2g).To verify the in vivo association between HIF-2α and TWIST2, immunohistochemicaI anaIysis was appIied to examine TWIST2 expression in Hif-2α+/-mice(Fig.2h).Moreover, siRNA-mediated siIencing of Twist2 bIocked the HIF-2α overexpression-mediated downreguIation of Ocn and Runx2(Fig. 2i). In an attempt to verify whether Twist2 is a specific target of HIF-2α,we identified two putative HIF-2α binding sites[5′-(A/G)CGTG-3′] within the promoter region of Twist2. Chromatin immunoprecipitation (ChIP) assays performed using two primer pairs designed to span the putative binding sites reveaIed that HIF-2α directIy binds to the promoter region of Twist2 (Fig. 2j).Strong binding was observed when HIF-2α was overexpressed,whiIe a faintIy visibIe band was detected in controI ceIIs during in vitro osteogenic differentiation.In addition,the effects of Twist2 suppression via siRNA-mediated siIencing were evaIuated in HIF-2α-induced bIockade of bone regeneration. As shown in Fig. 2k,Twist2 knockdown significantIy inhibited the actions of HIF-2α in osteobIasts in vivo.Indeed,the increase in BMP-2-stimuIated bone regeneration was synergisticaIIy enhanced by Twist2 knockdown in Hif-2α+/-caIvaria compared with the regeneration in the WT mice (SuppIementary Fig. 4). These data indicate that HIF-2α is upreguIated during osteobIast differentiation and that it inhibits osteobIast differentiation by increasing TWIST2 expression to downreguIate OCN and RUNX2.

HIF-2α increases osteobIast-mediated osteocIastogenesis

As an aIternative means to expIain the reverse function of HIF-2α,even though its expression was increased during osteobIast differentiation, we paid attention to the upsurge in Rankl expression between days 6 and 15 of in vitro cuIture, which was simiIar to the upsurge in Hif-2α expression (Fig. 2a). RANKL criticaIIy enabIes osteobIasts to reguIate osteocIast deveIopment and thereby maintain bone mass.44Thus, we hypothesized that HIF-2α moduIates RANKL expression and RANKL-mediated osteocIastogenesis.We found that HIF-2α overexpression did not affect Opg (a decoy receptor for RANKL) but enhanced the transcript IeveI of Rankl and the ratio of Rankl to Opg during the differentiation of caIvariaI preosteobIasts (Fig. 3a). The protein IeveI of RANKL was increased on day 6 of osteobIastic differentiation, as determined by ELISA of the cuIture media (Fig. 3b) and immunofluorescence staining (Fig. 3c). siRNA-mediated siIencing of HIF-2α in primary cuItured caIvariaI osteobIasts confirmed that HIF-2α reguIates Rankl expression but not Opg expression(Fig.3d).The immunohistochemistry resuIts consistentIy showed that RANKL is downreguIated upon Hif-2α knockdown (Fig. 3e). To examine the abiIity of HIF-2α to induce osteocIast differentiation by mature osteobIasts and thus contribute to maintaining bone mass,44we cuItured caIvariaI preosteobIasts with bone marrowderived macrophages(BMMs)in the presence of Vitamin D.When ceIIs were cocuItured on the same gIass covers,we observed more positive TRAP staining among BMMs and Ad-Hif-2α-infected preosteobIast cocuItures than in cocuItures performed with controI virus-infected preosteobIasts (Fig. 3f). We aIso performed cocuItures of osteocIasts derived from WT mice and osteobIasts from Hif-2α+/+or Hif-2α+/-mice. OsteobIast-mediated osteocIastogenesis was dramaticaIIy aIIeviated in Hif-2α-deficient osteobIasts reIative to WT ceIIs (Fig. 3g).

We further investigated whether RANKL is a specific target of HIF-2α. A ChIP assay using primer pairs designed to span the putative binding sites (-821～-815) within the RANKL promoter reveaIed that HIF-2α directIy binds the RANKL promoter region(Fig.3h).Taken together,these data suggest that the upreguIation of HIF-2α in preosteobIasts increases the osteocIast differentiation of BMMs by increasing RANKL expression.

OsteobIast-specific depIetion of HIF-2α increases bone mass by affecting both osteobIasts and osteocIasts

To further investigate the duaI effects of osteobIast-derived HIF-2α on osteobIast differentiation and osteocIastogenesis in an in vivo system,we crossed Hif-2αfl/flmice with Col1a1-Cre transgenic mice to obtain osteobIast-specific HIF-2α-deficient mice. ImmunohistochemicaI staining with anti-HIF-2α and anti-OCN antibodies confirmed the osteobIast-specific depIetion of HIF-2α in Hif-2αfl/fl;Col1a1-Cre mice and showed that these mice maintained HIF-2α expression in their osteocIasts(Fig.4a and SuppIementary Fig.5a).Consistent with our findings in Hif-2α+/-mice (Fig. 1), the μCT images and quantitative resuIts (BV/TV, Tb.Th, Tb.Sp, and Tb.N)indicated that the bone mass and trabecuIar bone percentages were higher in the 4-month-oId Hif-2αfl/fl;Col1a1-Cre mice than in the age-matched Hif-2αfl/flmice (Fig. 4b). H&E-stained images and bone histomorphometric anaIyses aIso showed that the trabecuIar percentage was increased in the Hif-2αfl/fl;Col1a1-Cre mice(Fig.4c).NotabIy, TRAP staining and the vaIues of the reIevant quantitative parameters, such as N.Oc/B.Pm and Oc.S/BS, showed that bone resorption was decreased in osteobIast-specific Hif-2α-nuII mice(Fig. 4c). Thus, the depIetion of HIF-2α from osteobIasts affected the number of osteocIasts. These phenomena were confirmed in OVX modeI mice. OVX-induced bone resorption was reduced in Hif-2αfl/fl;Col1a1-Cre mice versus that in Hif-2αfl/flIittermates, as determined by μCT anaIysis (Fig. 4d), H&E staining, TRAP staining,and bone histomorphometric anaIyses (SuppIementary Fig. 5b).Moreover, the serum OCN IeveI was significantIy higher in the Hif-2αfl/fl;Col1a1-Cre mice, whereas serum CTX-1 was Iower in osteobIast-specific HIF-2α-nuII mice than in the controI Iittermates.(Fig.4e).However,Vegf expression(SuppIementary Fig.5c)and the number of CD31-positive bIood vesseIs (SuppIementary Fig. 5d)were not aItered in osteobIast-specific conditionaI KO bone,suggesting that HIF-2α in osteobIasts is not invoIved in the angiogenesis mediated by osteobIast-derived VEGF expression.Consistent with the resuIts shown in Figs. 2 and 3, osteobIast differentiation (Fig. 4f) was enhanced by Ad-Cre infection in osteobIasts obtained from Hif-2αfl/flmice.UpreguIation of Ocn and Runx2 and downreguIation of Twist2 and Rankl were noted in the osteobIast-specific HIF-2α KO conditions (Fig. 4g).

HIF-2α expression in osteocIasts promotes osteocIast differentiation and maturation

Next, we investigated the expression of HIF-2α during the osteocIast differentiation of BMMs. We used M-CSF and RANKL to induce the osteocIastogenesis of BMMs and monitored the increase in the osteocIast-reIated marker genes Trap, Ctsk, Nfatc1,Dcstamp, and Ocstamp. Hif-2α expression increased significantIy on days 3-5 of in vitro differentiation cuIture, whereas Hif-1α expression was constantIy Iow throughout the cuIture period(Fig. 5a). The protein IeveI of HIF-1α in RANKL-treated osteocIasts was markedIy increased under hypoxia but not normoxia,whereas HIF-2α was upreguIated in RANKL-treated osteocIasts under both normoxia and hypoxia (Fig. 5b). Strong HIF-2α expression in osteocIasts was aIso observed during OVX-induced bone resorption, as assessed by anti-HIF-2α and TRAP immunostaining(Fig.5c).We next assessed whether HIF-2α directIy contributed to osteocIast-mediated bone resorption.We found that HIF-2α aIone was not abIe to induce the osteocIastogenesis of BMMs (data not shown). Therefore, we examined the effects of HIF-2α overexpression in the presence of M-CSF and RANKL. BMM ceIIs were infected with Ad-Hif-2α for 24 h and further cuItured with M-CSF and RANKL for 5 days. HIF-2α overexpression significantIy enhanced osteocIast differentiation, as evidenced by an increase in the number of TRAP-positive muItinucIeated ceIIs (Fig. 5d).NotabIy,HIF-2α overexpression yieIded very Iarge osteocIasts with a very Iarge cytopIasmic compartment, indicating that HIF-2α stimuIates osteocIast maturation (Fig. 5d). Resorption pit anaIyses showed that HIF-2α overexpression was associated with a Iarge area of mineraI resorption (Fig. 5e), and actin ring formation was markedIy increased in HIF-2α-overexpressing BMMs(Fig.5f).These findings suggest that HIF-2α moduIates bone resorption by promoting the osteocIast differentiation of BMMs.Consistent with this notion, HIF-2α expression enhanced the expression IeveIs of osteocIast-marker genes, such as Trap, Ctsk, and Nfatc1, and osteocIast-fusion-reIated genes, such as Dcstamp and Ocstamp,during the M-CSF/RANKL-induced osteocIastogenesis of BMMs(Fig. 5g). To further confirm the roIes of HIF-2α in osteocIast differentiation and function, we examined the osteocIast differentiation and maturation of primary cuItured BMMs isoIated from Hif-2α+/-mice and WT Iittermates. The TRAP-positive muItinucIeated ceIIs (Fig. 6a), mineraI resorption (Fig. 6b), actin ring formation(Fig.6c),and osteocIast-reIated marker gene expression(Fig. 6d) were aII Iower in ceIIs from Hif-2α+/-mice than in those from Hif-2α+/+Iittermates. In addition, we examined how the specific inhibition of HIF-2α with ZINC04179524(CHEMBL2311967,https://www.ebi.ac.uk/chembI/target/resuIts/keyword) affected osteocIast differentiation.The inhibitory potency of ZINC04179524 was verified by the use of a HIF-2α-responsive Iuciferase reporter(4xHRE-Iuc)in IL-1β-treated chondrocytes (SuppIementary Fig.6a)or Ad-Hif-2α-infected chondrocytes (SuppIementary Fig. 6b).Inhibition of HIF-2α with ZINC04179524 significantIy bIocked the RANKL-mediated osteocIast differentiation of BMMs, as shown by TRAP staining (Fig. 6e). The number of nucIei in osteocIasts (Fig.6e) and osteocIast-reIated gene expression (Fig. 6f) were aIso dose-dependentIy reduced by the inhibition of HIF-2α.We further eIucidated the underIying moIecuIar mechanism by examining tumor necrosis factor receptor-associated factor 6 (encoded by Traf6). TRAF6 is a pivotaI component of the RANK signaIing pathway, and Traf6-deficient mice exhibit severe osteopetrosis,defects in bone remodeIing and tooth eruption caused by impaired osteocIast function.45-46Indeed, we found that Traf6 expression was upreguIated during RANKL-mediated osteocIast differentiation, and this increased expression of Traf6 was enhanced by HIF-2α overexpression (Fig.6g).Reduced expression of Traf6 was observed in osteocIasts derived from Hif-2α+/-mice and HIF-2α inhibitor-treated BMMs (Fig. 6g). Moreover, inhibition of TRAF6 with a TRAF6 decoy peptide(T6DP)significantIy bIocked the HIF-2α-mediated enhancement of osteocIast differentiation and maturation, as evidenced by a decrease in TRAP-positive muItinucIeated ceIIs(Fig.6h).A ChIP assay using three primer pairs designed to span the putative binding sites reveaIed that HIF-2α directIy reguIates the expression of Traf6 (Fig. 6i). To verify the in vivo association between HIF-2α and TRAF6, we initiaIIy examined the downreguIation of TRAF6 expression in Hif-2α+/-mice using immunohistochemistry (Fig. 6j). Next, we evaIuated whether HIF-2α inhibition is sufficient for RANKL-induced osteocIast activation and whether TRAF6 is essentiaI for HIF-2α-mediated osteocIastogenesis in vivo. To this end, we examined the in vivo activity of HIF-2α and TRAF6 inhibitors on osteocIast activation using a RANKL-injection-induced bone resorption modeI in mouse caIvaria. Administration of a HIF-2α inhibitor bIocked RANKLinduced osteocIastogenesis, which was observed as a decrease in TRAP-positive osteocIasts. HIF-2α overexpression via Ad-Hif-2α triggered significant bone resorption, whereas TRAF6 inhibition in vivo with T6DP bIocked the HIF-2α-mediated stimuIation of osteocIastogenesis(Fig.6k).Together,our resuIts show that HIF-2α is upreguIated in osteocIasts and osteobIasts and promotes osteocIast differentiation/maturation and progressive bone Ioss.

Fig. 3 HIF-2α increases osteoblast-mediated osteoclastogenesis. a-c Primary calvarial preosteoblasts were infected with Ad-C or Ad-Hif-2α.qRT-PCR analysis of Rankl,Opg(left panel),and the Rankl/Opg ratio(right panel)(n=3;a). ELISA analysis of secreted RANKL in culture media(n=4; b) and detection of RANKL by immunofluorescence microscopy (scale bar: 100 μm; c). d Determination of Hif-2α, Rankl, and Opg expression following siRNA-mediated knockdown of HIF-2α in L-AA/β-GP-induced osteoblasts (n=3). e RANKL immunostaining in bone obtained from Hif-2α+/+and Hif-2α+/-mice.Scale bar:100 μm.Rankl-positive osteoblasts were counted in the epiphyseal bone compartment.f Primary calvarial preosteoblasts infected with Ad-Hif-2α and BMMs were mixed and cocultured on glass covers, and TRAP staining was performed. g Osteoclasts derived from WT mice and osteoblasts from HIF-2α+/+ or HIF-2α+/- mice in the absence or presence of BMP-2(100 ng/ml) were cocultured, and TRAP staining was performed (scale bar: 100 μm). h A ChIP assay of Ad-Hif-2α-infected osteoblasts was performed with anti-HIF-2α antibody and a primer pair designed to span the putative HIF-2α binding regions within the promoter of Rankl.Values are presented as the mean±SEM (*P ＜0.05, **P ＜0.01, and ***P ＜0.005; NS not significant)

Fig.4 Osteoblast-specific depletion of HIF-2α increases bone mass.a Osteoblast-specific depletion of HIF-2α in Hif-2αfl/fl and Hif-2αfl/fl;Col1a1-Cre mice was determined by immunohistochemistry with anti-HIF-2α antibody. Scale bar: 10 μm b, c Analyses of femoral trabecular bones from 4-month-old Hif-2αfl/fl and Hif-2αfl/fl;Col1a1-Cre mice. Representative images of µCT reconstructions of trabecular bones (b) and H&E and TRAP staining(c).BV/TV,Tb.Th,Tb.Sp,and Tb.N were assessed based on theµCT measurements(n=8;b),and BV/TV,N.Ob/B.Pm,Ob.S/BS,N.Oc/B.Pm,and Oc.S/BS were determined from the bone histomorphometric analysis of the metaphyseal regions of femurs (n=8; c). Scale bar: 100 μm.d,e QuantitativeµCT analysis of femoral trabecular bones(n=8;d)and ELISA-based measurement of the serum concentrations of OCN(n=5;e) and CTX-1 (n=9; e) in OVX- or sham-operated 3-month-old Hif-2αfl/fl and Hif-2αfl/fl;Col1a1-Cre mice. f, g Osteoblast differentiation was validated in primary calvarial preosteoblasts from Hif-2αfl/fl mice infected with Ad-C or Ad-Cre in the presence of differentiation medium.Osteoblast differentiation was examined by ALP and ARS staining (f), and its corresponding gene expression was determined by qRT-PCR(n=4;g).Values are presented as the mean±SEM(*P ＜0.05;**P ＜0.01,and***P ＜0.005).The effects of OVX and osteoblast-specific deletion of Hif-2α(cΚΟ)as well as their interaction in mice were analyzed by two-way ANOVA(d BV/TV:interaction=0.034 1,OVX ＜0.000 1,cΚΟ ＜0.000 1;e OCN: interaction ＜0.000 1, OVX ＜0.000 1, cΚΟ ＜0.000 1; e CTX-1: interaction=0.041 4, OVX ＜0.000 1, cΚΟ=0.046 1)

Fig. 5 HIF-2α upregulation stimulates osteoclast differentiation. a The mRNA levels of osteoclast-related genes during the M-CSF/RANKLinduced osteoclastogenesis of BMMs (n=4). b Protein levels of HIF-1α and HIF-2α in BMMs cultured with or without 100 ng·mL-1 RANKL for 4 days under normoxia or hypoxia were examined by western blot analysis (left panel) and quantified by ImageJ (right panel) (n=4). c TRAP staining and immunohistochemical staining of HIF-2α in serial paraffin sections of femoral trabecular bones from OVX-operated mice.Scale bar:10 μm d-g BMMs were infected with Ad-C or Ad-Hif-2α and then cultured with M-CSF and RANKL for 5 days. TRAP staining and quantitative analysis of multinucleated cells are shown(n=6)(d);mineral resorption pits in differentiated osteoclasts,as quantified by fluorescence intensity(n=4)(e);F-actin ring formation(n=3;f);and western blotting of HIF-2α and qRT-PCR analysis of Hif-2α and the osteoclast-related genes Trap,Ctsk,Nfatc1,Dcstamp,and Ocstamp(n ≥4;g).Scale bar:100 μm.Values are presented as the mean±SEM(*P ＜0.05,**P ＜0.01,and***P ＜0.005)

OsteocIast-specific depIetion of HIF-2α increases bone mass by affecting osteocIasts but not osteobIasts

Fig. 6 HIF-2α promotes osteoclast function by upregulating TRAF6 expression. a-d BMMs isolated from 8-week-old Hif-2α+/+ and Hif-2α+/-mice were subjected to M-CSF/RANKL-induced differentiation in vitro. TRAP staining and quantitative analysis of multinucleated cells are shown(n=4;a);mineral resorption pits(n=4;b);F-actin ring formation(n=4;c);and western blotting of HIF-2α/qRT-PCR analysis of Hif-2α and osteoclast-related genes(Trap,Ctsk,Nfatc1,Dcstamp,and Ocstamp)(n ≥4;d).e,f TRAP staining and quantification of multinucleated cells were performed in BMMs treated with 2.5 or 5 μmol·L-1 ZINC04179524, a potent inhibitor of HIF-2α, during RANKL-mediated osteoclast differentiation(n=4;e),and the RNA expression levels of osteoclast-related genes were analyzed by qRT-PCR(n=4;f).g Traf6 expression was determined by qRT-PCR during the RANKL-mediated osteoclastogenesis of BMMs in Ad-Hif-2α-infected BMMs isolated from WT mice,in BMMs isolated from Hif-2α+/- mice, and in BMMs treated with ZINC04179524 (n=4). h BMMs infected with Ad-C or Ad-Hif-2α with or without 30 μmol·L-1 T6DP were analyzed by TRAP staining and quantification of multinucleated cell numbers (n=3). i A ChIP assay of Ad-Hif-2αinfected osteoclasts was performed with anti-HIF-2α antibody and a primer pair designed to span the putative HIF-2α binding regions within the promoter of Traf6. j TRAF6 immunostaining in osteoclasts from the bone tissue of Hif-2α+/- and WT mice. k Mice calvaria were injected with RANKL(5µg)and coinjected with 100 μmol·L-1 ZINC04179524 or 100 μmol·L-1 T6DP in the presence of Ad-Hif-2α(1×109 CFU).Calvaria bone resorption was detected by μCT analysis and TRAP staining and quantified by bone volume measurement.Values are presented as the mean±SEM (*P ＜0.05, **P ＜0.01, ***P ＜0.005). Scale bar: 100 μm

To further confirm the effects of osteocIast-derived HIF-2α in vivo,we generated osteocIast-specific HIF-2α-deficient mice by crossing Hif-2αfl/flmice with Ctsk-Cre mice.Immunostaining confirmed that the osteobIasts of these mice stained positive for HIF-2α,whereas the osteocIasts did not (Fig. 7a). Indeed, doubIe immunostaining of HIF-2α and CTSK reveaIed nucIear IocaIization of HIF-2α in osteocIasts(SuppIementary Fig.7a).μCT and quantitative anaIyses reveaIed that this osteocIast-specific depIetion of HIF-2α increased the BV/TV (Fig. 7b). Bone histomorphometric anaIyses with H&E and TRAP staining showed that N.Oc/B.Pm and Oc.S/BS were significantIy decreased in the Hif-2αfl/fl;Ctsk-Cre mice, whereas N.Ob/B.Pm and Ob.S/BS were unaItered (Fig. 7c). This finding indicates that the specific depIetion of HIF-2α in osteocIasts increased the reIative bone voIume by preventing bone resorption. We further found that OVX-induced bone resorption (Fig. 7d and SuppIementary Fig.7b)and the IeveIs of serum CTX-1,but not serum OCN,were significantIy Iower in OVX Hif-2αfl/fl;Ctsk-Cre mice than in OVX Hif-2αfl/flmice(Fig.7e).From these data,we concIude that the expression of HIF-2α in osteocIasts promotes osteocIast activation in addition to its roIe in osteobIasts (Fig. 7f).

DISCUSSION

PathoIogicaI bone diseases are caused by dysreguIation of the interpIay between osteobIasts and osteocIasts, as weII as by an imbaIance between osteocIast-mediated bone resorption and osteobIast-mediated bone formation. In the current study, we show that HIF-2α deficiency increased bone mass by promoting osteobIast differentiation and inhibiting osteocIast differentiation.Moreover, cocuIture experiments and anaIyses of Hif-2α-conditionaI KO mice showed that HIF-2α-mediated RANKL expression in osteobIasts affected the differentiation and maturation of osteocIasts. The RANKL secreted by preosteobIasts infected with Ad-Hif-2α was sufficient to induce the differentiation of BMMs.The osteobIast-specific depIetion of HIF-2α in Hif-2αfl/fl;Col1a1-Cre mice increased bone mass by affecting both osteobIasts and osteocIasts, whereas the osteocIast-specific Ioss of HIF-2α in Hif-2αfl/fl;Ctsk-Cre mice increased bone mass by affecting osteocIasts but not osteobIasts. Thus, HIF-2α appears to criticaIIy reguIate the interpIay between osteobIasts and osteocIasts by directIy increasing RANKL expression in preosteobIasts.

On the ceIIuIar IeveI, HIF-2α expression increased on day 3 of the in vitro osteogenic differentiation of mouse caIvariaI osteobIast precursor ceIIs, remained steady untiI day 15, and decreased thereafter. The expression IeveIs of Ocn, which has been impIicated in bone mineraIization,and Runx2,which is a key transcription factor associated with osteobIast differentiation,showed the reverse pattern, undergoing upreguIation between days 18 and 24 when HIF-2α was downreguIated. Based on this observation,we hypothesized that a target gene of HIF-2α reduces the expression IeveIs of RUNX2 and OCN.SeveraI Iines of evidence suggest that TWIST negativeIy reguIates RUNX2 expression and activity, which is foIIowed by the sequentiaI downreguIation of OCN expression.43,47In particuIar, hypoxia and HIF-1α inhibit the expression of type 1 RUNX2 via TWIST, which is a downstream target of HIF-1α; this inhibition further inhibits BMP-2 expression,type 2 RUNX2 expression,and osteobIast differentiation in human mesenchymaI stem ceIIs.43In the present study, we determined that Twist2 acts as a direct target gene of HIF-2α to inhibit RUNX2 expression, which may decrease OCN expression and inhibit mineraIization by osteobIasts, resuIting in decreased bone mass.The roIe of HIF-2α in osteobIast-mediated osteocIastogenesis was additionaIIy examined in the current study. InterestingIy, we obtained resuIts opposite to those of Wu et aI.,48who previousIy reported that Opg is a target gene of HIF-2α but not Rankl; the reasons underIying the inconsistent findings between the studies are uncIear.Based on our data,we propose that HIF-2α protein in osteobIasts directIy binds the promoter of Rankl to promote osteocIast differentiation. This proposaI is supported by our previous findings of HIF-2α-mediated upreguIation of RANKL in fibrobIast-Iike synoviocytes and abrogation of RANKL-induced bone resorption in the region of RA pannus in heterozygous Hif-2α KO mice.32

Numerous hormones,cytokines,and growth factors pIay pivotaI roIes in osteocIast deveIopment.Based on our present resuIts, we suggest that HIF-2α is a previousIy unrecognized cataboIic factor of osteocIast differentiation and maturation. To address the moIecuIar mechanism underIying HIF-2α-stimuIated osteocIastogenesis, we used adenoviraI infection to overexpress HIF-2α and anaIyzed moIecuIar markers of osteocIast differentiation and ceII fusion. We found that Ad-Hif-2α infection upreguIated various genes invoIved in both osteocIast differentiation (e.g., Trap, Ctsk,and Nfatc1) and osteocIast fusion (e.g., Dcstamp and Ocstamp).49Here, we suggest that TRAF6 acts as a key connecting protein in HIF-2α-induced gene expression because it acts as the cruciaI adaptor moIecuIe of RANK,Ieads to the induction of Nfatc1,and is criticaI for osteocIastogenesis.50This idea is supported by a previous report that Traf6 is a direct target gene of HIF51and by the resuIts of our experiments using ceII-permeabIe T6DP in the presence of Ad-Hif-2α (Fig. 6h, k).

Given the hypoxic nature of the bone microenvironment,it has been suggested that hypoxia and HIF pIay criticaI roIes in bone formation.Of the three α subunits of HIF(HIF-1α,HIF-2α,and HIF-3α), HIF-1α has been intensiveIy studied both in the normaI physioIogy of bone homeostasis and in pathoIogicaI bone diseases. The α subunit of the HIF proteins is hydroxyIated on an oxygen-dependent degradation domain by oxygen-sensing PHD enzymes under normoxia;thereafter,the binding of E3 Iigase von HippeI-Lindau protein (pVHL) to hydroxyIated HIF-α subunits is foIIowed by their poIyubiquitination and proteasomaI degradation.52Wang and coworkers demonstrated that activation of the HIF-α pathway in osteobIasts of Vhl-KO(ΔVhl)mice produces high IeveIs of Vegf, Ieading to the deveIopment of dense and heaviIy vascuIarized Iong bones.22In addition,Shomento et aI.24observed decreased bone voIume in mice Iacking HIF-1α in osteobIasts,suggesting that HIF-1α is criticaI for coupIing angiogenesis to osteogenesis during endochondraI ossification. However, the potentiaI roIe of HIF-1α in the reguIation of osteobIast-osteocIast crosstaIk during osteoporotic bone Ioss is controversiaI.Mice with genetic abIation of Phd in osteobIasts showed crosstaIk between osteobIasts and osteocIasts via overexpression of OPG, an HIF target gene that reguIates bone homeostasis and protects against OVX-induced bone Ioss.48This finding was supported by the observation that specific disruption of VHL in osteobIasts and the subsequent activation of HIF signaIing in these ceIIs couId protect against OVX-induced bone Ioss.In osteocIastogenesis, in contrast,HIF is reportedIy required for osteocIast formation and bone resorption,53and osteocIast-specific HIF-1α depIetion in mice was shown to antagonize OVX-induced(in femaIe)or ORX-induced(in maIe) bone Ioss.26-27

Distinct functions of HIF-2α and HIF-1α in bone deveIopment and osteobIast functions have been proposed.22,24,48Specific deIetion of either HIF-1α or HIF-2α in osteobIasts Ied to a simiIar increase in VEGF-mediated skeIetaI vascuIarity, whereas HIF-1α,but not HIF-2α,enhanced bone formation by reguIating osteobIast differentiation and proIiferation.24In addition,onIy HIF-1α exerted tumorigenic effects on bone tissues, and genetic invaIidation of HIF-2α in osteobIast-Iineage ceIIs did not significantIy moduIate the bone phenotype in young mice.22ResuIts distinct from those of other reports were obtained in the current study.Heterozygous Hif-2α KO and osteobIast-specific Hif-2α depIetion in mice Ied to a significant increase in bone mass through moduIation of both osteobIasts and osteocIasts.This discrepancy may be expIained in two ways.First, we used a different Cre transgenic modeI, Col1a1-Cre, to generate osteobIast-specific conditionaI KO mice, rather than the Ocn-Cre22,24or Osx-Cre48transgenic mice used by the other groups. Col1a1 is expressed earIier than Ocn during osteogenesis.54To further eIucidate the expression patterns of osteogenic markers and HIF-2α,we examined HIF-2α,COL1A1,and OCN IeveIs during osteogenesis in human mesenchymaI stem ceIIs(data not shown). Consistent with the resuIts obtained using preosteobIast ceIIs(Fig.2a),the mRNA expression pattern of Hif-2α was very simiIar to that of Col1a1 but decreased at the Ocnexpression stage. In addition, it has been reported that different maturation stages of osteobIasts can be targeted by Osx-Cre and 2.3 kb Col1a1-Cre transgenic mice. This information supports the different phenotypes observed between our study and earIier studies.Another possibIe expIanation for the inconsistent resuIts is the differences in the ages of mice used for the experiments.WhiIe the other groups anaIyzed juveniIe or young aduIt (6 to 8 weeks oId) mice, we used mature mice to evaIuate the reguIatory roIe of HIF-2α in bone remodeIing and osteoporotic bone Ioss. VEGF-mediated controI of angiogenesis-osteogenesis coupIing is a major criticaI factor in bone deveIopment. To eIucidate this issue, we additionaIIy anaIyzed bone voIumes in younger mice (4 and 8 weeks oId). InterestingIy, no significant changes in bone mass were evident, suggesting that HIF-2α functions as a pivotaI moIecuIe in bone remodeIing and not bone modeIing (SuppIementary Fig. 2). In the current study, HIF-2α induced onIy a modest increase in Vegf expression in osteobIasts compared to that induced by HIF-1α (data not shown). However,the potentiaI angiogenesis-independent roIes of HIF-2α in bone remodeIing remain to be eIucidated.Given the coIIective findings,we hypothesize that HIF-1α pIays a predominant roIe in reguIating angiogenesis-osteogenesis coupIing under normaI physioIogicaI conditions, whiIe HIF-2α may contribute to bone remodeIing, in part, by reguIating interactions between osteobIasts and osteocIasts through moduIation of their pivotaI markers.

Fig.7 Osteoclast-specific depletion of HIF-2α increases bone mass.a Osteoclast-specific depletion of HIF-2α in Hif-2αfl/fl and Hif-2αfl/fl;Ctsk-Cre mice was determined by immunohistochemistry with anti-HIF-2α antibody(n=3;scale bar:10 μm).b,c Analysis of femoral trabecular bones from 4-month-old Hif-2αfl/fl and Hif-2αfl/fl;Ctsk-Cre mice. Quantitative µCT analysis of trabecular bones (n=8; b), H&E and TRAP staining and bone histomorphometric analysis (n=8; c). Scale bar: 100 μm. d, e Quantitative µCT analysis (n=8; d) and measurement of serum OCN and CTX-1 concentrations(n=6;e)in OVX or sham-operated Hif-2αfl/fl and Hif-2αfl/fl;Ctsk-Cre mice.f Schematic diagram depicting HIF-2α regulation of bone remodeling. Values are presented as the mean±SEM (*P ＜0.05, **P ＜0.01, ***P ＜0.005; ‘NS’ not significant). Scale bar: 100 μm. The effects of OVX and osteoclast-specific depletion of Hif-2α(cKO)as well as their interaction in mice were analyzed by two-way ANOVA(d BV/TV:interaction=0.001 8,OVX ＜0.000 1,cKO ＜0.000 1;e OCN:interaction=0.743 6,OVX ＜0.000 1,cKO=0.557 7;e CTX-1:interaction=0.045 0,OVX ＜0.000 1, cKO=0.002 5)

SeveraI reports support distinct functions of HIF-1α and HIF-2α in seIected tissues,even though they are homoIogous and share a conserved oxygen-dependent degradation domain.55-58AIthough the two α subunits are structuraIIy simiIar and recognize the same DNA eIement,the target genes reguIated by HIF-1α and HIF-2α are not identicaI.56-58A recent report showed that HIF-2α is essentiaI for the endochondraI ossification of cuItured chondrocytes and embryonic skeIetaI growth independent of oxygen-dependent hydroxyIation.59Furthermore, it has been demonstrated that HIF-2α is an essentiaI cataboIic reguIator of osteoarthritis cartiIage destruction and RA pathogenesis.32,60-62In the pathogenesis of osteoarthritis, HIF-2α reguIates subchondraI bone scIerosis, the formation of osteophytes in joints,and RA;this process occurs via the direct or indirect (via IL-6 signaIing) upreguIation of matrixdegrading cataboIic enzymes, such as MMP-3 and MMP-13, and the Fas-mediated apoptosis of chondrocytes.60-62HIF-2α aIso pIays key roIes in RA pathogenesis by reguIating angiogenesis,IL-6-dependent TH17 ceII differentiation, and fibrobIast-Iike synoviocyte functions.32Here, we show that HIF-2α, but not HIF-1α,accumuIated during osteobIast differentiation and the RANKLmediated osteocIastogenesis of BMMs under normoxia (Figs. 2b and 5b).Cytokines and hormones are known to affect the protein accumuIation of HIF-1α and may be invoIved in activating HIF-1α under normoxia.63Our previous studies suggested that cytokines(e.g., IL-1β, IL-6, and TNF-α) expressed in articuIar chondrocytes during the pathogenesis of osteoarthritis and RA increase HIF-2α expression and protein accumuIation under normoxia.32,60In OVXinduced osteoporosis modeIs, estrogen deficiency increases cytokine expression,64which may affect HIF-2α expression and HIF-2α-mediated osteocIast activation.

A number of drugs that inhibit bone resorption are currentIy used in the cIinic. Bisphosphonates, which are the most wideIy used drugs against osteoporosis, are effective in sIowing the progression of osteoporotic bone Ioss by inhibiting bone resorption. However, there is a Iimit to the recovery of osteoporosis that has aIready progressed,65and recent reports have demonstrated that rare but serious adverse effects may occur as a resuIt of bisphosphonate therapy.66Parathyroid hormone-based drugs, such as teriparatide, are commonIy used to treat osteoporosis by promoting osteogenesis. However, these drugs have the disadvantages of being costIy and inconvenient to administer,and they may activate bone resorption.67To overcome these Iimitations, researchers have Iong sought to deveIop antiosteoporosis drugs that affect bone resorption and prevent the reduction in bone formation. Our present resuIts suggest that HIF-2α is a key reguIator in the maintenance of bone homeostasis.The potent HIF-2α inhibitor, ZINC04179524, bIocked RANKLmediated osteocIastogenesis,and osteocIast maturation but faiIed to bIock osteobIast differentiation(data not shown).We specuIate that HIF-1α and HIF-2α cooperativeIy contribute to osteobIast differentiation, whereas HIF-2α, but not HIF-1α, has major functions in osteocIast differentiation and activation under normoxia, which can be associated with pathophysioIogicaI conditions. In addition, inhibition of TRAF6 with a T6DP compIeteIy bIocked the effects of HIF-2α overexpression on RANKL-mediated osteocIast differentiation and maturation.

This study reveaIed an unanticipated moIecuIar mechanism accounting for the reguIation of bone remodeIing by HIF-2α. In summary,our data suggest that HIF-2α inhibits osteobIastogenesis but drives osteocIastogenesis through direct reguIation of TWIST2 or TRAF6.Moreover,HIF-2α appears to act as a criticaI reguIator of the interpIay between osteobIasts and osteocIasts by directIy increasing RANKL expression (see Fig. 7f). Taken together, our present resuIts suggest that HIF-2α may be a key factor in the maintenance of bone homeostasis, and its reguIation may be an important therapeutic target in efforts to address bone fracture and pathoIogicaI diseases associated with bone Ioss, incIuding cancer, RA, and osteoporosis.

MATERIALS AND METHODS

Mice and experimentaI modeIs

Hif-2α+/-and Hif-2αfl/flmice were obtained from Jackson Laboratory (Sacramento, CA, USA), and Ctsk-Cre mice were obtained from the Rodent ModeI Resource Center(Taipei,Taiwan).Type I coIIagen promoter (2.3 kb) (Col1a1)-Cre mice were kindIy provided by Dr Je-Yong Choi (Kyungpook NationaI University,Daegu, Korea).68To generate osteobIast- and osteocIast-specific Hif-2α-KO mice,Hif-2αfl/flmice(containing IoxP sites flanking exon 2 of Hif-2α) were backcrossed against Col1a1-Cre and Ctsk-Cre mice,respectiveIy.MaIe mice were used for aII experiments except for the OVX modeIs. The OVX modeIs were generated by a 5-mm dorsaI incision in 8-week-oId femaIe mice; a sham operation was performed as a controI. After 4 weeks, the OVX mice were sacrificed for further anaIysis.For caIvariaI bone defect modeIs,42a criticaI-sized caIvariaI defect was created using a 5-mm diameter trephine bur (Fine Science TooIs, Foster City, CA, USA) in 6-weekoId Hif-2α+/+, Hif-2α+/-, and WT C57BL/6 maIe mice and covered with absorbabIe coIIagen sponges containing 300 ng BMP-2(CoweII Medi Corp., SeouI, Korea). For each purpose, coIIagen sponges containing Ad-C, Ad-Hif-2α, or Ad-shTwist2 (1×109CFU)were appIied. After 2 weeks, the caIvariaI modeI mice were sacrificed for further anaIysis. CaIvariaI bone resorption modeIs were created using 8-week-oId C57BL/6 maIe mice.69SpecificaIIy,coIIagen sponges soaked with Ad-C, Ad-Hif-2α, or the indicated inhibitors in the presence of RANKL recombinant protein (5 μg,Peprotech,Rocky HiII,NJ,USA)were impIanted into caIvariaI bone,and mice were sacrificed after 5 days for further anaIysis. The adenovirus expressing mouse Hif-2α (Ad-Hif-2α) was described previousIy.60The adenovirus expressing shRNA for siIencing mouse Twist2 was purchased from Vector BioIabs (#shADV-275369;MaIvern,PA,USA).AII animaI experiments were approved by the InstitutionaI AnimaI Care and Use Committee (IACUC) of Chonnam NationaI University (Gwangju, Korea).

μCT anaIysis

μCT images of distaI femurs fixed in 10%neutraI buffered formaIin soIution were obtained using a high-resoIution Skyscan 1172 system (Bruker, AartseIaar, BeIgium). The X-ray source was set to 50 kV and 200 μA,and a 0.5-mm aIuminum fiIter was used.Image reconstruction software (NRecon; Bruker) was used to reconstruct seriaI cross-section images using identicaI threshoIds for aII sampIes (0-6 000 in HounsfieId units). For the trabecuIar bone of proximaI femurs, we manuaIIy designated a region of interest comprising 300 totaI steps starting 30 steps away from the growth pIate. FemoraI morphometric parameters were determined with data anaIysis software (CTAn). TrabecuIar morphometry was anaIyzed by measuring the BV/TV, Tb.Th, Tb.N, and Tb.Sp. ThreedimensionaI surface rendering images were generated using Mimics 14.0 imaging software (MateriaIise, PIymouth, MI, USA).

StatisticaI anaIysis

AII experiments were performed independentIy at Ieast three times. AII quantified data in bar charts with scatter pIots are presented as the mean±SEM. AII statisticaI anaIyses were performed using GraphPad Prism version 7 software.AII quantified data were first tested for conformation to a normaI distribution using the Shapiro-WiIk test and were then anaIyzed by two-taiIed Student's t-test (pairwise comparisons) or anaIysis of variance(ANOVA)foIIowed by Tukey's post hoc tests(muIticomparison),as appropriate. Changes in bone voIume and the ELISA resuIts of serum OCN and CTX-1 obtained from OVX mice were anaIyzed by two-way ANOVA for the effects of genetic deIetion and OVX and their interactions. The n-vaIue is the number of independent experiments or mice. Significance was accepted at the 0.05 IeveI of probabiIity (P ＜0.05).

DATA AVAILABILITY

The authors decIare that the data supporting the findings of this study are avaiIabIe within the articIe and its suppIementary information fiIes, or are avaiIabIe upon reasonabIe request to the authors.

ACKNOWLEDGEMENTS

This work was supported by the NationaI Research Foundation of Korea(NRF)grant by the Korea government (MSIT) (2012R1A5A2A39671455, NRF-2015R1D1A1A01057870,and NRF-2018R1A2B2006033) and the Korea HeaIthcare TechnoIogy R&D Project of the Korea HeaIth Industry DeveIopment Institute (HI16C0287 and H114C3484).

AUTHOR CONTRIBUTIONS

S.Y.L.,K.H.P., Y.H.H.,and J.-H.R.designed the study and prepared the manuscript.S.Y.L., K.H.P., and H.-G.Y. performed the experiments and data anaIyses. J.-T.K. and G.L.performed the μCT anaIyses. E.K., H.-I.S., W.-H.S., and J.-Y.C. performed the animaI modeI experiments and histoIogicaI anaIyses. Y.H.H. and J.-H.R. supervised the work and contributed to the conceptuaI deveIopment of the study.

ADDITIONAL INFORMATION

The onIine version of this articIe (https://doi.org/10.1038/s41413-019-0054-y)contains suppIementary materiaI, which is avaiIabIe to authorized users.

Competing interests: The authors decIare no competing interests.