Hypoxia-inducible factor-2α mediates senescence-associated intrinsic mechanisms of age-related bone loss

Aging is associated with cellular senescence followed by bone loss leading to bone fragility in humans. However, the regulators associated with cellular senescence in aged bones need to be identified. Hypoxia-inducible factor (HIF)−2α regulates bone remodeling via the differentiation of osteoblasts and osteoclasts. Here, we report that HIF-2α expression was highly upregulated in aged bones. HIF-2α depletion in male mice reversed age-induced bone loss, as evidenced by an increase in the number of osteoblasts and a decrease in the number of osteoclasts. In an in vitro model of doxorubicin-mediated senescence, the expression of Hif-2α and p21, a senescence marker gene, was enhanced, and osteoblastic differentiation of primary mouse calvarial preosteoblast cells was inhibited. Inhibition of senescence-induced upregulation of HIF-2α expression during matrix maturation, but not during the proliferation stage of osteoblast differentiation, reversed the age-related decrease in Runx2 and Ocn expression. However, HIF-2α knockdown did not affect p21 expression or senescence progression, indicating that HIF-2α expression upregulation in senescent osteoblasts may be a result of aging rather than a cause of cellular senescence. Osteoclasts are known to induce a senescent phenotype during in vitro osteoclastogenesis. Consistent with increased HIF-2α expression, the expression of p16 and p21 was upregulated during osteoclastogenesis of bone marrow macrophages. ChIP following overexpression or knockdown of HIF-2α using adenovirus revealed that p16 and p21 are direct targets of HIF-2α in osteoclasts. Osteoblast-specific (Hif-2αfl/fl;Col1a1-Cre) or osteoclast-specific (Hif-2αfl/fl;Ctsk-Cre) conditional knockout of HIF-2α in male mice reversed age-related bone loss. Collectively, our results suggest that HIF-2α acts as a senescence-related intrinsic factor in age-related dysfunction of bone homeostasis.


Introduction
Bone homeostasis is maintained by tight regulation of the interplay between two specific cell types: boneforming osteoblasts and bone-resorbing osteoclasts 1 . Imbalances in bone homeostasis cause severe bone diseases, including osteoporosis, the most prevalent chronic disease in aging adults. Bone homeostasis is affected by various factors, such as heart/vascular disease, estrogen deficiency, diabetes, and aging 2 . Aging is the main risk factor for age-related osteoporosis, which is characterized predominantly by decreased bone mineral density. The consequent bone structure fragility is a primary cause of osteoporotic fracture in elderly individuals 3,4 .
Bone-forming osteoblasts have a short lifespan and are constantly being replaced 5 . Osteoblast differentiation is governed by osteoblast-specific transcription factors, including runt-related transcription factor-2 (RUNX2) 6 and osterix 7 . Differentiation consists of a sequence of events, including lineage commitment (osteoprogenitors), proliferation and matrix synthesis (preosteoblasts), and matrix maturation and mineralization (mature osteoblasts). Mature osteoblasts are responsible for the assembly of matrix proteins, including collagen type I (COL1A1), bone integrin-binding sialoprotein (IBSP), osteopontin (OPN) and osteocalcin (OCN) 8 , and are characterized by an increase in alkaline phosphatase (ALP) activity 9 . Age-related bone loss was shown to be caused by a deficit in bone formation 10 . Osteoblast dysfunction is the main cause of age-related bone loss in humans and results from two distinct types of mechanisms. Extrinsic mechanisms are mediated by changes in the bone microenvironment associated with hormones and growth factors, whereas intrinsic mechanisms are mediated by osteoblast senescence following reductions in the number, lifespan, and differentiation of osteoblasts 11 .
Age-related bone loss has also been associated with increases in the number of bone-resorbing osteoclasts 12 . Osteoclasts differentiate from hematopoietic progenitors of the monocyte-macrophage lineage in response to stimulation by macrophage colony-stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL) 13 . Osteoclasts develop into multinucleated giant cells through a sequence of steps that include proliferation, differentiation, and maturation 14,15 . Importantly, osteoclasts exhibit characteristics of senescence, such as the expression of several different cyclin-dependent kinase (CDK) inhibitors, INK4A (p16) and CDKN1A (p21), during osteoclast differentiation 16,17 . Osteocytes in the bones of aged mice are postmitotic and show elevated levels of p16 12,18 , whereas senescent osteoblast progenitors exhibit high levels of p21 but not p16 19 . Selective elimination of p16-expressing cells in mouse models increases bone mass 18 , and stromal cell cultures obtained from the bone marrow of old mice are better at promoting osteoclastogenesis than cells from young mice 20 , supporting the finding that senescent cells contribute to age-related bone loss.
Hypoxia-inducible factor (HIF), a heterodimeric transcription factor consisting of an oxygen-regulated α-subunit and a constitutively expressed β-subunit, plays important roles in bone formation by regulating the differentiation and activation of osteoblasts and osteoclasts 21 . Under normal oxygen tension, the α-subunit of HIF is hydroxylated by oxygen-sensitive prolyl hydroxylase, which is recognized by the von Hippel-Lindau protein, a component of the E3 ubiquitin ligase complex, and targeted for degradation 22 . Unlike HIF-1α, which is structurally very similar to HIF-2α but has many different functions 23 , HIF-2α can accumulate under normoxic conditions in the context of pathologies such as osteoporosis 24 and arthritis [25][26][27] and can also regulate bone remodeling 24 . HIF-2α depletion downregulates RANKL expression in fibroblast-like synoviocytes from the rheumatoid arthritis synovium 25 . In addition, HIF-2α regulates osteoblast and osteoclast differentiation via the expression of essential transcription factors involved in the differentiation of these cells and influences the interplay between these cell types 24 , suggesting that HIF-2α plays a catabolic role in bone remodeling. In this study, we tried to examine whether age-related upregulation of HIF-2α expression in osteoblasts and osteoclasts contributes to the differentiation of these cells, senescence progression and age-related bone loss.

Micro (μ)CT analysis and bone histomorphometry
Mouse femurs were fixed in 10% neutral buffered formalin and examined using μCT as previously described 24 . Trabecular morphometry was analyzed by measuring bone mineral density (BMD), bone volume/tissue volume (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), and cortical thickness (Ct.th). Three-dimensional surface rendering images were generated using Mimics 14.0 imaging software (Materialise, Plymouth, MI, USA). Histomorphometric parameters were analyzed by OsteoMeasure software (Osteometrics, Inc., Decatur, GA, USA). Paraffin sections were stained with hematoxylin and eosin (H&E) and BV/TV, the number of osteoblasts/bone perimeter (N.Ob/B.Pm), and the osteoblast surface/bone surface (Ob.S/BS) were analyzed. To investigate osteoclast parameters, the sections were stained with tartrate-resistant acid phosphate (TRAP) and the number of osteoclasts/ bone perimeter (N.Oc/B.Pm) and the osteoclast surface/ bone surface (Oc.S/BS) were measured.

Osteoblast differentiation of primary cultured preosteoblasts
For primary culture of calvarial preosteoblasts, calvarial bones from 3-day-old pups were enzymatically digested twice with DMEM comprising 0.1% type II collagenase (Sigma-Aldrich, St. Louis, MO, USA) as described previously 24 . To induce cellular senescence in the proliferation stage, osteoblasts stimulated with doxorubicin on differentiation day 2 were treated with a potent HIF-2α inhibitor, TC-S 7009 (R&D Systems, Minneapolis, MN, USA) on differentiation day 4, and cells were harvested after 3 days. Cellular senescence in the matrix stage was induced by doxorubicin on differentiation day 9, and TC-S 7009 was administered on differentiation day 11 for 3 days. Then, the cells were harvested on day 14 of differentiation. A schematic diagram showing the time point at which each chemical compound was administered is provided in Supplementary Fig. 5. To overexpress Hif-2α in vitro, primary calvarial preosteoblasts were infected with adenovirus expressing Hif-2α at the indicated multiplicity of infection (MOI) on differentiation day 2. Doxorubicin is a commonly used chemotherapy drug that can induce DNA cross-linking and lead to cell apoptosis, autophagy, and/or senescence 28 . The concentration of doxorubicin used in this experiment did not cause cell apoptosis.

Osteoclast differentiation of bone marrow macrophages (BMMs)
Mouse bone marrow cells were isolated and harvested as previously described 24 . Adherent BMMs (4 × 10 4 cells per well in 48-well plates) were maintained in complete α-MEM containing 30 ng/ml M-CSF (PeproTech) for 24 h. Then, the medium was replaced with medium comprising 100 ng/ml RANKL (PeproTech), and the cells culture for 5 days. To ectopically overexpress HIF-2α, adenovirus encoding Hif-2α was applied at the indicated MOI on day 1. TRAP (Sigma-Aldrich) staining was performed on day 5 according to the manufacturer's instructions.

ALP and ARS staining
For ALP staining, cells were fixed with 4% paraformaldehyde for 2 min, washed with deionized water and stained with 5-bromo-4-chloro-3-indolyl phosphate (BCIP®)/nitro blue tetrazolium (NBT) liquid substrate solution (Sigma-Aldrich) for 15 min in a dark room. For mineralization assays, the cells were fixed with 10% formalin for 15 min and stained with 2% Alizarin red S (ARS, Sigma-Aldrich) solution for 45 min at room temperature.
Reverse transcription-polymerase chain reaction (RT-PCR) and quantitative real-time (qRT)-PCR Mouse bone tissues were immediately frozen in liquid nitrogen and stored at −70°C until use. Total RNA was isolated from osteoblasts, osteoclasts, and femoral bone tissues using TRI reagent (Molecular Research, Cincinnati, OH, USA) and reverse-transcribed with TOPscript RT DryMIX (Enzynomics, Daejeon, Korea). cDNA was subjected to PCR using AmpOne TM Tap DNA Polymerase Mix (GeneAll, Seoul, Korea) and appropriate primers. qPCR was performed using a StepOnePlus Real-Time PCR system (Thermo Fisher Scientific, Waltham, MA, USA) and SYBR Premix Ex Taq (TaKaRa Bio, Kyoto, Japan). The individual transcript levels of each individual target were normalized to those of Gapdh and are expressed as the fold change relative to the levels in the indicated controls. The primer sequences and sizes of the amplicons are listed in Supplementary Table 1.

Western blotting
Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM NaF, 1% NP-40, 0.2% SDS, and 0.5% deoxycholate) supplemented with protease and phosphatase inhibitor cocktails (Roche, Basel, Switzerland). Proteins were separated by SDS-PAGE, and the blots were incubated with rabbit anti-HIF-2α (NB100-122; Novus Biologicals, CO, USA) at 4°C overnight. After washing, the membranes were incubated with HRPconjugated anti-rabbit IgG, and images were obtained using ImageSaver6 software via the EZ-capture MG system (ATTO, Tokyo, Japan). The protein bands were quantified using a CS analyzer 4 (ATTO), and protein levels were normalized relative to the level of β-actin (Sigma-Aldrich) in the same sample.
All procedures were performed according to the manufacturers' instructions.

Chromatin immunoprecipitation (ChIP) assay
ChIP assays were performed using a Magna ChIP Kit (Millipore) as previously described 24 . Briefly, cells were fixed with 1% formaldehyde to crosslink the DNA and protein complexes. After the lysates were sonicated, soluble chromatin complexes were incubated with rabbit anti-HIF-2α antibody or anti-IgG at 4°C overnight. The precipitated DNA segment of the p16 or p21 promoter was amplified by RT-PCR. The sequences of the primers used for the ChIP assay are listed in Supplementary Table 1.

Statistical analysis
All experiments were repeated at least three times. The results are reported as the mean ± SD. μCT and histomorphometric parameters are presented in bar graphs with scatter plots. All statistical analyses were performed using GraphPad Prism version 7 (GraphPad, Inc., San Diego, CA). All quantitative data were first tested for normal distribution using the Shapiro-Wilk test. Then, the means of quantitative variables were compared between two groups by two-tailed Student's t-test or between three or more independent groups by analysis of variance (ANOVA) followed by Bonferroni's post-hoc test (multicomparison), as appropriate. n is the number of independent experiments or mice. The results were considered statistically significant when the p-value was less than 0.05.

HIF-2α expression is upregulated in skeletal bones with aging
Hif-2α expression was markedly higher in the aged bones of 20-month-old male mice than in the bones of younger mice and exhibited the same pattern of expression as the aging-related markers p16 and p21 (Fig. 1a). Because bone tissue is a complex system that consists of diverse cell types, including osteoblasts, osteocytes and osteoclasts along with various supporting cells, including hematopoietic cells, the upregulated Hif-2α expression shown in Fig. 1a, b might be associated with various cell types. Here, we focused on a representative bone remodeling process maintained by the resorption of osteoclasts and performed by osteoblasts. The expression of the osteoblast markers Ocn and Runx2 was lower in the bones of aged mice than in those of younger mice, while the expression of Trap, a marker of osteoclasts, was dramatically increased in aged bones (Fig. 1a). Consistent with the level of Hif-2α mRNA, HIF-2α protein levels were significantly higher in the bones aged mice than in those of younger mice (Fig. 1b). To evaluate bone turnover in aged mice, the levels of biochemical markers in the serum were compared between mice at different ages. Serum OCN levels were lower but the level of the bone resorption-specific biomarker serum CTX-1 was higher in 12-and 20-month-old male mice than in 1.5-and 4month-old male mice (Fig. 1c). To confirm the in vivo involvement of HIF-2α expression and bone turnover in aged bone cells, the numbers of HIF-2α-expressing osteoblasts (Fig. 1d) and osteoclasts (Fig. 1e) were quantified histochemically in young (4-month-old) and aged (12-month-old) mice. The number of HIF-2α-positive osteoblasts were higher in the bone tissue of aged mice than in that of younger mice (Fig. 1d). The number of HIF-2α-expressing osteoclasts was increased in aged mice (Fig. 1e). These findings indicated that HIF-2α expression was upregulated in both osteoblasts and osteoclasts upon aging.

Age-induced osteoporotic bone loss is reversed in heterozygous Hif-2α knockout (KO) mice
The association between upregulation of HIF-2α expression and aged-related bone loss was investigated by examining the bone microarchitecture of 12-month-old male heterozygous Hif-2α KO (Hif-2α +/− ) and wild-type (Hif-2α +/+ ) mice using X-ray microcomputed tomography (μCT). Immunohistochemical staining with anti-HIF-2α and anti-OCN antibodies confirmed that Hif-2α was depleted of osteoblasts in Hif-2α KO mice ( Fig. 2a and Supplementary Fig. 1a). Since OCN is expressed only in osteoblasts, osteoblasts can be distinguished from bone lining cells, which do not express OCN 29 . Knockdown of HIF-2α expression in the osteoclasts of Hif-2α KO mice was also verified by immunohistochemistry (Fig. 2b). Consistent with our previous findings 24 , quantitative results of bone parameters indicated that the bone mass and trabecular bone percentages were higher in 4-monthold Hif-2α KO mice than in age-matched WT mice ( Supplementary Fig. 1b). Quantitative μCT analyses showed that trabecular bone volume was significantly decreased in aged mice (12 months old) compared to young mice (4 months old) ( Supplementary Fig. 1b) and that cancellous trabeculae in the femoral bone were better preserved in Hif-2α +/− mice than in WT littermates (Fig. 2c). Trabecular bone volume was higher in Hif-2α +/− mice than in WT mice, as measured by BMD and BV/TV (Fig. 2c). Other bone parameters, including Tb.Th and Tb. N, were higher, whereas Tb.Sp was lower, in Hif-2α +/− mice than in WT mice (Fig. 2c). No differences in cortical thickness were detected between Hif-2α +/− mice and agematched WT mice at any age ( Supplementary Fig 1b). The cause of this discrepancy may be the fact that cortical bone is generally less metabolically active than trabecular bone 30,31 . Therefore, we focused on analyzing trabecular bone physiology. Bone histomorphometric analyses showed that the bone trabecular percentage was higher and that the number of osteoclasts was lower in heterozygous Hif-2α KO mice than in WT mice (Fig. 2d). BV/ TV, N.Ob/B.Pm, and Ob.S/BS were increased in Hif-2α +/− mice, whereas N.Oc/B.Pm and Oc.S/BS, which are parameters associated with bone resorption, were decreased in Hif-2α +/− mice (Fig. 2d). These data indicated that HIF-2α deficiency retarded bone destruction in aged mice.

Upregulation of HIF-2α expression is associated with osteoblast dysfunction upon aging
The effect of increased HIF-2α expression in osteoblasts during age-related bone loss was investigated by assessing the effects of HIF-2α on osteoblast differentiation and senescence. OCN was found to be rarely expressed in HIF-2α-expressing osteoblasts from the bones of 12-month-old mice ( Supplementary Fig. 2), suggesting that increased HIF-2α expression in aged bones might be associated with osteoblast senescence, resulting in an imbalance in bone homeostasis and age-mediated bone loss. To assess this hypothesis, doxorubicin-induced senescence during mouse osteoblast differentiation was analyzed. Two different experimental conditions were tested to determine the stage of osteoblast differentiation affected by doxorubicin-induced senescence. Generally, differentiation from preosteoblasts to mature osteoblasts is divided into two stages, proliferation, which occurs in the first 7 days, and matrix maturation, which occurs over the subsequent 7 days, depending on the duration of culture 32,33 . The protein expression of HIF-2α increased during osteoblast proliferation, peaked during matrix The levels of mRNAs encoded by Hif-2α, osteoblast marker genes (Ocn and Runx2), an osteoclast marker gene (Trap) and senescence marker genes (p16 and p21) were determined by RT-PCR and quantified by qRT-PCR (n ≥ 4). b The protein levels of HIF-2α extracted from mice of the indicated age determined by western blot analysis (left panel) and quantified by a CS analyzer 4 (right panel) (n = 5). c ELISA-based measurements of the serum concentrations of OCN (n = 5) and CTX-1 (n = 5) in 1.5-, 4-, 12-and 20-month-old mice. d, e Immunostaining of osteoblasts (d) and osteoclasts (e) from the bone tissues of young (4-month-old) and old (12-month-old) mice for HIF-2α. The dotted lines indicate osteoblasts, and the arrowheads indicate osteoclasts. Scale bar: 25 μm. Quantification of HIF-2α-positive cells is shown in the right panel (n = 8). M = month. The values are presented as the means ± SDs (*P < 0.05, **P < 0.01, and ***P < 0.005). maturation and was observed until day 14 (data not shown). The addition of doxorubicin during both phases effectively blocked L-ascorbic acid (L-AA) and β-glycerophosphate (β-GP)-induced differentiation of preosteoblasts, as determined by measurement of ALP activity and ARS staining ( Fig. 3a and Supplementary  Fig. 3a). Induction of doxorubicin-mediated osteoblast senescence, as determined by increased p21 expression but not p16 expression, resulted in the upregulation of Hif-2α expression and the downregulation of expression of the osteoblastic marker genes Ocn and Runx2 (Fig. 3b and Supplementary Fig. 3b, c). To evaluate whether overexpression of HIF-2α during osteoblast differentiation affects the progression of senescence, calvarial preosteoblasts were infected with adenovirus expressing Hif-2α (Ad-Hif-2α), which significantly inhibited the expression of Ocn and Runx2 (Fig. 3c and Supplementary   Fig. 3d). However, overexpression of Hif-2α did not affect the expression of the senescence marker genes p16 and p21, indicating that Hif-2α expression did not induce senescence ( Fig. 3c and Supplementary Fig. 3d). This result was additionally confirmed by osteoblast-specific Hif-2α knockout by Ad-Cre infection in osteoblasts obtained from Hif-2α fl/fl mice. Consistent with the alteration of the expression of genes, including Hif-2α, Ocn, and Runx2, the expression of the senescence marker genes p16 and p21 was not affected (Fig. 3d). TC-S 7009 is a potent inhibitor of HIF-2α activity, and its inhibitory effect was confirmed by its ability to suppress IL-1β-and Ad-Hif-2α-induced HIF-2α activation in a reporter gene assay using a 4×HRE-containing luciferase vector construct ( Supplementary Fig. 4). TC-S 7009-mediated inhibition of HIF-2α activity during the osteoblast proliferation phase failed to recover osteoblast differentiation inhibited by senescence ( Supplementary  Fig. 5a, b). However, HIF-2α inhibition during matrix maturation resulted in significant reversal of the doxorubicin-mediated reduction in Ocn and Runx2 expression without altering p16 expression ( Fig. 3e and Supplementary Fig. 5c) inhibition of osteoblast differentiation was reversed by TC-S 7009-induced inhibition of HIF-2α, as determined by ALP and ARS staining ( Fig. 3f and Supplementary Fig.  5d). These findings indicated that increased HIF-2α expression was an effect of aging rather than a cause of osteoblast senescence in aged bones. However, inhibition of HIF-2α was able to restore age-related impaired osteoblast matrix maturation by increasing the expression of genes such as Ocn and Runx2.

Osteoblast-specific depletion of HIF-2α increases bone mass in aged mice
To further investigate the association between HIF-2α expression upregulation in osteoblasts and age-mediated osteoporotic bone loss, male osteoblast-specific Hif-2α conditional KO (Hif-2α; fl/fl Col1a1-Cre) mice were analyzed. HIF-2α was specifically depleted in the osteoblasts of these mice (Fig. 4a). μCT images showed that aging-induced trabecular bone loss in 12-month-old Hif-2α; fl/fl Col1a1-Cre mice ( Supplementary Fig. 6) was much lower than that in age-matched Hif-2α fl/fl mice (Fig. 4b). Quantitative analyses showed that BMD, BV/TV, Tb.Th and Tb.N were higher and that Tb.Sp was lower in osteoblast-specific Hif-2αdeficient mice than in Hif-2α fl/fl mice (Fig. 4b). Consistent with these μCT images, H&E staining and bone histomorphometric parameters, such as BV/TV, N.Ob/B.Pm and Ob. S/BS also showed that the trabecular percentage was increased in Hif-2α; fl/fl Col1a1-Cre mice (Fig. 4c). The results of TRAP staining and measurement of quantitative parameters associated with bone resorption, such as N.Oc/B.Pm and Oc.S/BS, revealed that bone resorption was decreased in aged Hif-2α; fl/fl Col1a1-Cre mice (Fig. 4c). These findings indicated that osteoblast-specific depletion of HIF-2α delayed age-induced osteoporotic bone loss.

Aging-associated upregulation of HIF-2α expression accelerates osteoclast function and senescence
Because age-mediated disruption of bone homeostasis is also caused by increased osteoclast activity 34 , the effect of aging on HIF-2α expression in osteoclasts was assessed. HIF-2α was strongly expressed in CTSKpositive osteoclasts in the bones of 12-month-old mice ( Supplementary Fig. 7a). Hif-2α expression was increased on days 3-5 of in vitro RANKL-induced osteoclast differentiation of BMMs, consistent with the increases in p16 and p21 expression ( Fig. 5a and Supplementary Fig. 7b). Immunohistochemistry also showed that the expression of p16 and p21 was higher in osteoclasts from the bones of 12-month-old mice than those from the bones of younger mice (Fig. 5b). Assessment of the direction contribution of HIF-2α to the aging of osteoclasts showed that osteoclastogenesis was significantly enhanced by transduction with Ad-Hif-2α, as evidenced by an increase in the number of TRAPpositive multinucleated cells (Fig. 5c). Interestingly, p16 and p21 expression was concomitantly enhanced by the overexpression of Hif-2α during RANKL-induced osteoclast differentiation of BMMs ( Fig. 5d and Supplementary Fig. 7c). Conversely, HIF-2α knockdown in BMMs obtained from Hif-2α fl/fl mice by Ad-Cre infection led to the downregulation of p16 and p21 expression (Fig. 5e). To determine whether p16 and p21 are specific targets of HIF-2α, putative HIF-2α binding sites [5'-(A/G)CGTG-3'] were identified within the promoter region of these two genes (Fig. 5f, g). ChIP assays using four primer pairs designed to span these putative binding sites revealed that HIF-2α directly regulated the expression of both p16 and p21 (Fig. 5f, g). These findings indicated that increased HIF-2α expression in aged osteoclasts promoted the expression of p16 and p21 and mediated the senescence-related enhancement of the bone resorption activity of osteoclasts.

Osteoclast-specific depletion of HIF-2α inhibits bone loss in aged mice
The association between HIF-2α expression upregulation in osteoclasts and age-mediated osteoporotic bone loss was further analyzed in osteoclast-specific male Hif-2α conditional KO (Hif-2α; fl/fl Ctsk-Cre) mice. Depletion of HIF-2α in osteoclasts but not osteoblasts was (see figure on previous page) Fig. 3 Upregulation of HIF-2α expression by aging inhibits osteoblast function. a, b Doxorubicin-induced inhibition of osteoblast differentiation. Primary calvarial preosteoblasts from newborn mice were cultured in osteogenic differentiation medium (DM) containing 50 μg/ml L-AA and 5 mM β-GP for 7 days. Before harvesting cells on day 14, 0.01 μg/ml doxorubicin was added to the differentiation medium for 1, 3, or 5 days. Representative images of alkaline phosphatase (ALP) and alizarin red S (ARS) staining in primary calvarial preosteoblasts cultured in control medium (CM) or differentiation medium in the absence or presence of 0.01 μg/ml doxorubicin are shown (a). The expression levels of Hif-2α, Ocn, Runx2, p16, and p21 were analyzed by qRT-PCR (n = 5; b). c Quantification of mRNA levels by qRT-PCR. The transcript levels of the indicated genes were measured in primary calvarial preosteoblasts infected with Ad-C at an MOI of 800 or Ad-Hif-2α at the indicated MOI (n ≥ 4). d The mRNA levels of Hif-2α, Ocn, Runx2, p16, and p21. Osteoblasts isolated from Hif-2α fl/fl mice were infected with Ad-C or Ad-Cre in the presence of differentiation medium (n ≥ 4). e, f Effect of HIF-2α inhibition on osteoblast matrix maturation. Doxorubicin-treated osteoblasts were treated with the potent HIF-2α inhibitor TC-S 7009 and incubated for 3 days. The levels of Ocn, Runx2, p16, and p21 mRNAs were measured by qRT-PCR (n = 5; e). Assessment of osteoblast differentiation and mineral deposition by ALP and ARS staining (f). The values are presented as the means ± SDs (*P < 0.05, **P < 0.01, and ***P < 0.005; NS not significant). verified by immunohistochemical staining (Fig. 6a). Immunohistochemistry also showed that p16 (Fig. 6b) and p21 (Fig. 6c) expression was downregulated in Hif-2αdeficient osteoclasts. Quantitative analyses revealed that 12-month-old mice showed much lower trabecular bone mass than 4-month-old mice (Supplementary Fig. 8).
Osteoclast-specific depletion of Hif-2α prevented this reduction in trabecular bone loss, as determined by μCT imaging and analyses of quantitative parameters, such as BMD, BV/TV, Tb.Th, Tb.Sp, and Tb.N (Fig. 6d). Bone histomorphometric analyses by H&E and TRAP staining showed that N.Oc/B.Pm and Oc.S/BS was significantly decreased in Hif-2α; fl/fl Ctsk-Cre mice, whereas N.Ob/B. Pm and Ob.S/BS were increased (Fig. 6e). These in vivo data showed that specific depletion of HIF-2α in osteoclasts delayed age-induced osteoporotic bone loss.

Discussion
As aging progresses, bone remodeling, which is normally tightly regulated, may be dysregulated due to imbalances between osteoblast-mediated bone formation and osteoclast-mediated bone resorption. This imbalance can result in an increased risk of fracture, impaired bone healing, and the chronic age-related bone metabolism disease osteoporosis. This study showed that senescenceinduced HIF-2α expression upregulation in both osteoblasts and osteoclasts acts as a novel intrinsic mediator of age-related bone loss.
The senescence-related genes p16 and p21 play important roles in the initiation of senescence by leading to cell cycle arrest 11 . Doxorubicin-induced senescence during osteoblastic differentiation of osteoblast progenitor cells resulted in increased expression of p21 but not p16. In the absence of doxorubicin-mediated induction of senescence, however, both genes were highly expressed in mature osteoclasts. Upregulated HIF-2α expression in p21-expressing senescent osteoblasts inhibited osteoblast differentiation and function, as determined by decreases in Ocn and Runx2 expression. Twist2, a gene directly targeted by HIF-2α, has been shown to be expressed during early stages of osteogenic differentiation and to inhibit the expression of RUNX2, a key transcription factor for osteoblast differentiation, resulting in decreased expression of OCN, which has been implicated in bone mineralization 24 . TWIST is a negative regulator of RUNX2 expression and function that downregulates OCN expression 35 , whereas doxorubicin enhances TWIST expression in mouse osteoblastic cells 36 . The current study hypothesized that upregulated HIF-2α expression in osteoblasts in aged bones suppresses RUNX2 expression by inducing TWIST2 expression upregulation, resulting in the sequential downregulation of OCN expression. Senescent cells can cause tissue dysfunction through the development of senescence-associated secretory phenotypes, which are characterized by the production of extracellular matrix- degrading proteins, proinflammatory cytokines and chemokines 37,38 . These aging-associated changes in the bone microenvironment may induce increased HIF-2α expression in osteoblasts. Indeed, unlike that of HIF-1α, the expression of HIF-2α can be enhanced by proinflammatory cytokines, including IL-1β and TNFα, in the absence of hypoxia 26 . In general, the osteoblast differentiation process can be divided into three phases: proliferation, matrix . The values are presented as the means ± SDs (*P < 0.05, **P < 0.01, and ***P < 0.005).
maturation, and mineralization 39 . We previously reported that HIF-2α is highly expressed until the matrix maturation stage and that HIF-2α expression declines during later stages of osteogenic differentiation 24 .The present study showed that doxorubicin inhibited osteoblast differentiation during both the proliferation and matrix maturation stages. In contrast, HIF-2α inhibited the maturation process but not the proliferation, thereby controlling osteoblast differentiation. These unexpected discrepancies may be related to the altered expression of differentiationassociated genes, differences in accessibility of the promoters of HIF-2α target genes during the complicated differentiation process, or the sensitivity of the HIF-2α inhibitor employed. Further studies are needed to assess the catabolic effect of HIF-2α on age-related osteoblast dysfunction in mediating age-related reductions in bone mass.
This study also revealed that the expression levels of HIF-2α, p16, and p21 were increased in osteoclasts and that their expression levels were higher in the bones of aged mice than in those of younger mice. Upregulated HIF-2α expression in osteoclasts promoted the transcription of p16 and p21 by binding directly to hypoxiaresponsive elements (HREs; -(A/G)CGTG-) in the promoters of these genes. Thus, low p16 and p21 expression was detected in Hif-2α; fl/fl Ctsk-Cre mice. Upregulation of HIF-2α expression in senescent osteoclasts predominantly enhances bone-resorbing activity, as evidenced by the fact that inducible elimination of p16-expressing senescent cells in INK-ATTAC transgenic mice is associated with decreases in the number of osteoclasts and bone resorption 40 . In addition, senescent cell-condition medium enhances osteoclastogenesis and bone-resorbing activity and impairs osteoblast mineralization 40 . In our current study, the degree of cell senescence was assessed by measuring the expression levels of p16 and p21 because it is well established that these genes are the most wellknown effectors of senescence and that their expression levels increase with age in several rodent and human cell lines 4,18,41 . In addition, Saeed et al. demonstrated that p16 and p21 levels are significantly increased in bone cells from telomerase knockout mice, which are premature aging model mice 42 . Although these genes act as key effectors of cellular senescence, since the phenomena caused by aging are diverse, additional investigations of other factors are necessary. Bone is affected by weight load. The spine and legs, which support the human body, show marked changes in shape, such as bending of the bones, with aging. However, in mice, since the body is supported by four limbs, the morphological and physiological changes associated with age-related bone loss in the femur might be more serious than those in vertebrae. In addition, in murine models, weight-bearing bone is highly sensitive to running, which is a weight-bearing exercise 43,44 . Furthermore, some studies that observed aged bone showed no differences in age-related bone loss between femurs and vertebrae 45,46 . Therefore, in this study, we measured bone parameters in the femurs, not the spines, of mice.
HIF-2α deficiency in aged male mice markedly reversed age-related bone loss, which is consistent with our previous results showing that HIF-2α depletion in mice increases bone mass by promoting osteoblast differentiation and function and by inhibiting osteoclast differentiation and maturation 24 . Interestingly, both Hif-2α; fl/fl Col1a1-Cre and Hif-2α; fl/fl Ctsk-Cre male mice showed relatively better preservation of trabeculae than age-matched Hif-2α fl/fl mice due to changes in both osteoblasts and osteoclasts. Osteoblastmediated bone formation was higher and osteoclastmediated bone resorption was lower in the bones of 12month-old Hif-2α; fl/fl Col1a1-Cre and Hif-2α; fl/fl Ctsk-Cre conditional KO mice than in the bones of control mice, which may have been due to interactions between osteoblasts and osteoclasts. Osteoblasts influence osteoclast formation and differentiation through OPG/RANKL/RANK and LGR4/RANKL/RANK signaling 47,48 . RANKL expressed by osteoblasts can promote osteoclastogenesis by stimulating the RANK receptor on osteoclasts, a process that is regulated by the ratio of RANKL to OPG, a natural antagonist of RANK 49 . We previously reported that Rankl is a direct target gene of HIF-2α and that RANKL expression induced by HIF-2α overexpression promotes osteoblastmediated osteoclast differentiation 24 . We hypothesized that osteoblast-specific HIF-2α depletion might reduce RANKL secretion by osteoblasts, reducing the RANKL/OPG ratio and the number of osteoclasts. Moreover, osteoclasts can affect osteoblast-mediated bone formation via Atp6v0d2 50 , complement component 3a 47 , semaphorin 4D 51 , or exosomal miR-214-3p 52 . Because TRAF6, which acts as a crucial adaptor protein of RANK signaling in osteoclasts, is a gene that is directly regulated by HIF-2α 24 , we hypothesized that the decreased osteoclastic activity in Hif-2α; fl/fl Ctsk-Cre mice was caused by a decrease in TRAF6 levels. However, the molecular mechanism responsible for the increased number of osteoblasts in Hif-2α; fl/fl Ctsk-Cre mice remains to be determined, although it may be due to osteoblast-osteoclast interactions.
Estrogen deficiency in postmenopausal females is another main cause of age-associated net bone loss and subsequently leads to osteoporosis 34 . We previously used an animal model of OVX-induced osteoporosis to evaluate the catabolic role of HIF-2α in bone remodeling during estrogen deficiency-associated osteoporosis 24 . Although the molecular mechanisms underlying pathological bone diseases caused by estrogen deficiency and aging differ somewhat, our findings suggest that HIF-2α is a crucial regulator of both estrogen deficiency-and agingassociated osteoporosis. While osteoporosis generally occurs in women, approximately 12% of men aged >50 years old are expected to experience an osteoporotic fracture during their lifetime 53 . Almost 30% of patients with hip fractures are men, and mortality rates within 1 year after fracture are higher in men than in women 54 . These findings suggest that some pathophysiological mechanisms of aging-induced osteoporosis are independent of estrogen deficiency. There are two possible mechanisms of estrogen-independent age-induced osteoporosis: vitamin D deficiency and age-related osteoblast dysfunction 55 . Aging reduces the production of vitamin D in the skin to almost 50% of the normal level and reduces the intestinal concentration of vitamin D receptor, reducing calcium absorption 56 . Animal studies have shown that the ability of MSCs to differentiate into osteoblasts is impaired by aging 57 . Furthermore, the senescenceassociated secretory phenotype can cause age-related dysfunction in other tissues. For example, aging and cell senescence in bone can induce dysfunction of osteoblast progenitor cells, induce defective bone formation and increase osteoclastogenesis 11,58 . However, the molecular mechanisms by which senescence affects osteoblasts and disturbs bone homeostasis remain unclear.
HIF-2α may play an important role in senescencemediated osteoporosis. Although several drugs, such as bisphosphonates and parathyroid hormone, are currently used to treat osteoporosis 59,60 , they are ineffective against osteoporosis that has already developed. In addition, few therapeutic targets have been identified for treating osteoporotic bone loss in elderly individuals. Targeting senescence-related intrinsic factors involved in bone homeostatic dysfunction may contribute to the development of a novel class of antiaging therapies. Our current results suggest that HIF-2α is a key regulator of agerelated bone loss, as well as the pathogenesis of osteoporosis. Treatment with a potent HIF-2α inhibitor resulted in the recovery of osteoblast function reduced by doxorubicin-mediated senescence. These findings indicate that inhibition of HIF-2α in aged bones not only results in the reversal of age-related bone loss that has already occurred by restoring osteoblast function but also prevents age-mediated osteoporosis by inhibiting osteoclast differentiation and maturation.