Hypoxia negatively affects senescence in osteoclasts and delays osteoclastogenesis

Cellular senescence, that is, the withdrawal from the cell cycle, combined with the acquirement of the senescence associated secretory phenotype has important roles during health and disease and is essential for tissue remodeling during embryonic development. Osteoclasts are multinucleated cells, responsible for bone resorption, and cell cycle arrest during osteoclastogenesis is well recognized. Therefore, the aim of this study was to investigate whether these cells should be considered senescent and to assess the influence of hypoxia on their potential senescence status. Osteoclastogenesis and bone resorption capacity of osteoclasts, cultured from CD14+ monocytes, were evaluated in two oxygen concentrations, normoxia (21% O2) and hypoxia (5% O2). Osteoclasts were profiled by using specific staining for proliferation and senescence markers, qPCR of a number of osteoclast and senescence‐related genes and a bone resorption assay. Results show that during in vitro osteoclastogenesis, osteoclasts heterogeneously obtain a senescent phenotype. Furthermore, osteoclastogenesis was delayed at hypoxic compared to normoxic conditions, without negatively affecting the bone resorption capacity. It is concluded that osteoclasts can be considered senescent, although senescence is not uniformly present in the osteoclast population. Hypoxia negatively affects the expression of some senescence markers. Based on the direct relationship between senescence and osteoclastogenesis, it is tempting to hypothesize that contents of the so‐called senescence associated secretory phenotype (SASP) not only play a functional role in matrix resorption, but also may regulate osteoclastogenesis.

roles in both health and disease. It is often linked to tumor suppression, as cell cycle exiting after malignant transformation prevents tumor growth (Collado & Serrano, 2010), but more recently senescent cells have also been identified during embryological development in mammals and birds (Muñoz-Espín et al., 2013;Nacher et al., 2006;Storer & Keyes, 2014;Storer et al., 2013).
Developmental senescence, which is independent of DNA damage and dependent on the cyclin-dependent kinase inhibitor p21, leads to the recruitment of macrophages to clear the embryo from the senescent cells (Muñoz-Espín et al., 2013). Hence, it is essential in tissue remodeling during embryonic development.
Cellular senescence is also observed in healthy adult individuals and considered to be physiological. Both megakaryocytes, formed by endomitosis without cytokinesis (Besancenot et al., 2010) and placental syncytiotrophoblasts, formed by fusion of cytotrophoblasts (Chuprin et al., 2013) become senescent during their development.
Senescence of these cells is speculated to play an essential, yet largely unexplored, role in their specific function while it limits oncogenic transformation.
However, only Chen et al. (2007) regarded osteoclasts as senescent cells, whereas others described them as being quiescent (Kwak et al., 2005;Kwon et al., 2016;Mizoguchi et al., 2009;Sankar et al., 2004;Takahashi et al., 2010;Zauli et al., 2007), leaving the exact classification of these cells and the role of senescence in their functioning open to debate.
Osteoclasts are closely associated with vessels and play an important role during embryonic and post-natal skeletal development, as well as in pathologic conditions of the skeleton such as periodontitis (Hienz, Paliwal, & Ivanovski, 2015) and rheumatoid arthritis (Harre & Schett, 2017). In all activities hypoxia plays an important role: the hypoxia inducible transcription factor (HIF) stimulates angiogenesis and new bone formation (Shomento et al., 2010;Wang et al., 2015). Furthermore, hypoxia and more importantly, subsequent reoxygenation have a stimulating effect on the differentiation and bone resorbing capacity of osteoclasts (Arnett et al., 2003;Fukuoka, Aoyama, Miyazawa, Asai, & Goto, 2005;Knowles, 2015), possibly mediated by HIF-1α. Mechanistically, hypoxia and reoxygenation activate nuclear factor kappa-lightchain-enhancer of activated B cells (NF-κB) (Rupec & Baeuerle, 1995) and the production of reactive oxygen species (Granger & Kvietys, 2015), which are both linked to the induction of senescence and the SASP (Acosta et al., 2013;Hubackova, Krejcikova, Bartek, & Hodny, 2012;Nelson et al., 2012). Reoxygenation is essential, as culturing osteoclasts under constant hypoxia led to extensive cell death and dramatically reduced numbers of osteoclasts (Knowles & Athanasou, 2009). However, it remains undetermined how oxygen tension and potential senescence are related and how these factors may affect osteoclast function.
The aim of this study was to investigate the profile of osteoclasts to answer the question whether these cells should be considered senescent or not and to assess the influence of hypoxia on osteoclast function and senescence status. We hypothesized that functional osteoclasts have a senescent phenotype that is stimulated by hypoxia.
To verify these hypotheses, we studied osteoclastogenesis and bone resorption capacity of osteoclasts, cultured from CD14+ monocytes under the influence of M-CSF, RANK-L in two oxygen concentrations, that is, normoxia (21%) and hypoxia (5%). Osteoclasts were profiled by using specific staining for proliferation and senescence markers, qPCR of a number of osteoclast and senescence-related genes and bone resorption assay. Purity of the isolated monocyte population was confirmed using flow cytometry on a FACSCanto II cytometer (Becton Dickinson) after incubation with a monoclonal mouse anti-human CD45-FITC/CD14-PE dual-tag antibody (Beckman Coulter, Cat# 6603909, RRID: AB_2665483). Purity was on average >90%.
Subsequently, TRAP staining was performed according to the manufacturer's instruction using a commercially available kit (Leukocyte Acid Phosphatase Staining Kit, Sigma-Aldrich). Staining for senescence was performed at 7 and 14 days by incubating the cells overnight in freshly prepared senescence associated beta galactosidase staining solution at 37°C, according to the protocol of Dimri et al. (1995). Digital images were obtained using an Olympus BX-60 microscope, equipped with a Leica DFC450C camera and LAS 4.7 software. For each time point, oxygen concentration, donor and staining, two chamber slides were analyzed by counting and categorizing all cells present in four standardized sites of the chamber.

| Quantitative PCR (qPCR)
After 1, 7, 14, and 21 days of culture, cells were harvested for RT-qPCR analysis. RNA was extracted using a commercial spin-column kit (RNeasy Micro Kit, Qiagen, Hilden, Germany) according to the instructions of the manufacturer. RNA was quantified using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific). cDNA was made using the iScript cDNA synthesis kit with similar RNA input for all the samples. RT-qPCR reactions were performed using Sybr Green Master mix (Thermo Fisher Scientific) for a total reaction volume of 10 μl. The primers for genes related to the osteoclast phenotype and function and senescence related genes are listed in Supplementary Table S2. These primers were validated by a gradient PCR (for retrieving the melting temperature), followed by sequencing of the amplicon (Supplementary   Table S2). Ct values were normalized with four reference genes (HMBS, B2M, GAPDH, and HPRT), corrected for the RT-qPCR efficiency and further relativized with the norm-first method. The stability of the reference genes was assessed by the use of NormFinder (Andersen, Jensen, & Ørntoft, 2004).

| Bone resorption
Cells were cultured for 3 weeks on bovine cortical bone chips in the presence of M-CSF and RANKL. Thereafter, cells attached to the bone were fixated at room temperature for 10 min in 4% paraformaldehyde in PBS. Non-specific background staining was blocked with 20% normal goat serum (Vector Laboratories) for 60 min.
After washing with PBS, bone chips were incubated with Alexa Fluor 647-labeled bisphosphonates (Coxon, Thompson, Roelofs, Ebetino, & Rogers, 2008;Thompson, Rogers, Coxon, & Crockett, 2006) for 1 hr at room temperature, followed by washing. Nuclei were visualized with propidium iodide (Sigma-Aldrich). Bone slices were stored at 4°C in PBS until they were analyzed by confocal laser scanning microscopy (Leica SPE-II DMI-4000, Leica Microsystems), using a 10× ACS APO (NA 0.3) objective at a pixel size of 733 nm (zoom 1.5, 1024 × 1024 image size) and a quadruple dichroic filter that does not reflect the excitation lines in the detector path. 3D datasets were compiled from 11 slices, spaced in Z by 5 μm.
Bisphosphonate fluorescence was recorded using the 635 nm laser line and emission was detected over the 646-706 nm range using the spectral detector. Buffer control staining was performed to determine background. Propidium Iodide signal was recorded using the 561 nm laser over an emission range of 571-635 nm. Quantification of the Bisphosphonate fluorescence was performed using the surface object wizard of Imaris (version 8.2.0 RRID:SCR_007370). Representative images of the different groups are shown with minor linear intensity adjustments (Red 20/190, Blue 15/230).

| Semi-quantitative HIF-1α analysis by Western blot
In order to investigate whether the cells experienced hypoxia, HIF-1α protein expression was quantified. After washing with cold PBS, cell contents were harvested by scraping the 96-well plates with RIPA buffer containing 0.06 nM phenylmethylsulphonyl fluoride, 17 μg/ml aprotin and 1 mM sodium orthovanadate (Sigma-Aldrich). Cells were lysed on ice for 20 min, to prevent HIF-1α degradation followed by centrifugation for 10 min at 12,000g. The Images were obtained and densities were quantified by using Image Lab software (Bio-Rad Laboratories RRID:SCR_014210).

| Statistical analysis
Data were analyzed using R Studio Statistical software version 3.1.2 (RRID:SCR_001905). The p value threshold was set at 0.05 and a correction for multiple comparisons was done with the False Discovery Rate method of Benjamini and Hochberg (1995). For the RT-qPCR and cell counting analysis, normality of the data distribution was checked graphically and with a non-parametric bootstrapped Shapiro-Wilks test. Asthe RT-qPCR data were not normally distributed and the sample size was low (n = 3), differences between the groups were assessed by a non-parametric bootstrapped permutation test without replacement. The number of permutations for each experimental group and gene was set at 1000 and the differences between the mean values of bootstrapped ΔCt were assessed.
As the cell counting data were not normally distributed, differences between the groups were assessed by a Cox proportional hazard model (coxph), considering donor and the different experiments as random effects. Unless indicated, results are presented as mean ± standard deviation (SD). Confidence Intervals (C.I.) were set at 95%. Effect sizes (ES) were retrieved in all cases as the nonparametric Cliff's delta (Cliff, 1993), after bootstrapping 1000 times by a Monte Carlo simulation.
The interpretation for the present work is the following: <0.11, very small or no effect; 0.11-0.28, small effect size; 0.29-0.43, medium effect size; and >0.43, large effect size. Differences were considered as (biologically) relevant if a p value < 0.05 was found and the effect size was medium or large.   Osteoclasts seem to experience hypoxia as there is a tendency toward increased average quantity (+/− SD) of HIF-1α present at week 2 in 5% versus 21% O 2 (a, n = 3 donors per time point) and corresponding blots of HIF-1a (upper) and tubulin (lower) (c) qRT-PCR results of the HIF target gene NIX (BCL2 Interacting Protein 3 Like) over time (c, n = 3 donors per time point). There were no significant differences between culture conditions

| Osteoclasts heterogeneously express both markers of proliferation and senescence, while hypoxia seems to negatively affect senescence
During osteoclastogenesis in two oxygen concentrations, that is, normoxia (21%) and hypoxia (5%), osteoclasts were profiled by using specific staining for proliferation and senescence markers and

| DISCUSSION
The results of this study suggest that during in vitro osteoclastogenesis, osteoclasts obtain a senescent phenotype and that fusion of precursor cells, an essential initial step in osteoclastogenesis was delayed in the presence of 5% O 2 compared to 21% O 2 without negatively affecting the bone resorption capacity of the cells on the long term.

| Osteoclasts express a senescent phenotype
Cell cycle arrest is a hallmark of osteoclastogenesis and based hereon osteoclasts have always been described as quiescent cells (Kwak et al., 2005;Kwon et al., 2016;Mizoguchi et al., 2009;Sankar et al., 2004;Takahashi et al., 2010;Zauli et al., 2007). However, the present study indicates that they should be considered senescent rather than quiescent. Senescence is not limited to the arrest of proliferation, but also includes the acquirement of the senescence associated secretory phenotype (SASP) (Terzi et al., 2016), characterized by the production and secretion of soluble signaling factors (Coppé et al., 2010). In the current study, osteoclasts indeed expressed typical markers of senescence, including p21 and senescence associated beta galactosidase, in line with a previous report (Chen et al., 2007). In fact, the expression profile of the osteoclasts derived from PBMCs stimulated with M-CSF and RANK-L overlaps with the SASP. We observed increasing expressing of CCL2 and CCL5, which are known to be part of the SASP and have chemotactic properties (Ruhland et al., 2016).
The expression of Cathepsin K, which is an enzyme produced by mature osteoclasts to break down the non-mineralized bone matrix, increased over time in both culturing conditions. Interestingly, besides its matrix degrading characteristics, Cathepsin K can also induce senescence in osteoclasts, possibly to control and limit their number (Chen et al., 2007). Altogether, these findings indeed imply that osteoclasts obtain a senescence-associated secretory phenotype and strongly suggest that the SASP secretome exerts paracrine effects that possibly regulate osteoclastogenesis and bone resorption.

| Oxygen tension influences osteoclast senescence
As determined by senescence and proliferation markers. Contrary to our hypothesis, oxygen concentration was inversely correlated with the percentages of positively stained p21 nuclei in osteoclasts, a marker of senescence. This indicates that senescence was negatively affected by decreased oxygen tension. Cyclin-dependent kinase inhibitor 2A or p16 is another senescence marker that typically has very low expression in young and healthy tissue. However, as it is activated by cellular damage or stress, it is abundant in aged tissues There were no significant differences between culture conditions (Krishnamurthy et al., 2006;Zindy, Quelle, Roussel, & Sherr, 1997) and therefore considered to be a good senescence marker for in vivo studies (Baker et al., 2011;Burd et al., 2013;Yamakoshi et al., 2009).
In the present study, only a very limited number of nuclei were positive for p16, whereas cytoplasmic p16 was present in the majority of the cells, both in single and multinucleated ones. Within the context of osteoclastogenesis, the presence of p16 together with p21 has also been reported in murine monocytes cultured to become osteoclasts (Cong et al., 2017). Cytoplasmic p16 localization has been related to malignancy (McCluggage & Jenkins, 2003;Reid-Nicholson et al., 2006;Zhao et al., 2012), but its biological meaning is still under debate. The fact that nuclear p21 was primarily affected in osteoclasts cultured in hypoxia, may be explained by the differences between cellular senescence in adult cells and in developing cells. Storer et al. (2013) observed that in senescent cells of developing embryos, p21 was present, while they were unable to detect p16 or DNA damage; which are both essential features in replicative and oncogene induced cellular senescence of adult cells. To our surprise, we observed that a considerable percentage of osteoclasts were positive for Ki67. The presence of Ki67 in the osteoclasts can obviously not be related to proliferation, as their cell cycle is arrested. However, Ki67 has also been shown to be present in quiescent cells, possibly associated with ribosomal RNA transcription (Bullwinkel et al., 2006;Rahmanzadeh, Hüttmann, Gerdes, & Scholzen, 2007). Another potential reason for the presence of Ki67 positive nuclei might be the possible occurrence of fission of osteoclasts (Jansen et al., 2012). We did not study this, but it has been described that multinucleated osteoclasts can split into two or more multinucleated daughter cells. This cytoplasmic separation has some resemblance with the last phases of mitosis, which might explain presence of Ki67.

| Hypoxia delays osteoclastogenesis without long term effect on bone resorption capacity
The fact that we observed in general less large osteoclasts and more  and hypoxic (5% O 2 ) culture conditions. The arrows indicate p16 positive nuclei in multinucleated cells. Note that while mononuclear cells were p16 positive, no osteoclasts were present after 1 week of culture. There were no significant differences between culture conditions fusion can influence the number of osteoclasts that eventually form (Cong et al., 2017;Motiur Rahmanet al., 2015). However, these two factors were not different between the hypoxic and normoxic culture conditions in our study and similar numbers of adherent cells were observed after 24 hr of culturing in either culture. Possibly, the lower number of mononuclear cells observed after one week is the result of the inhibiting effect of hypoxia on their proliferation (Naldini & Carraro, 1999). While delayed osteoclastogenesis seems to be in contrast to several other papers reporting positive effects of hypoxia on osteoclastogenesis (Arnett et al., 2003;Fukuoka et al., 2005;Knowles, 2015), it confirms others reporting negative effects of hypoxia on osteoclast formation (Leger et al., 2010;Hulley et al., 2017). These contradictory observations may be related to differences in culture set up and to the pH sensitivity of the medium, where even different brands of fetal calve serum could have an influence. Although 2% O 2 has been reported to be the optimal concentration for bone resorption (Knowles, 2015), culturing osteoclasts under constant hypoxia leads to extensive cell death and dramatically reduced osteoclast numbers (Knowles & Athanasou, 2009). Therefore, we cultured at 5% O 2 and allowed reoxygenation twice a week during culture medium change. This frequency of medium change and intermittent exposure to normoxia is comparable to the setup of Hulley et al. (2017), who also reported decreased osteoclastogenesis. Nonetheless, while osteoclastogenesis was delayed in hypoxic culture conditions, gene expression of CAII, CATK, and MMP9, secreted by osteoclasts to resorb bone did not differ between conditions and resorption capacity at week 3 was comparable in the two culture conditions.

| The study has several limitations
Five percent O 2 did not produce a robust down-regulation of HIF-1α protein expression, but did affect the senescence phenotype and osteoclastogenesis. It remains to be determined whether O 2 concentrations lower than 5% would elicit a distinct response.
Regulations at cellular level induced by long term hypoxia, including mRNA HIF-1a stability (Uchida et al., 2004) and degradation of HIF-1α by the proteasome (Demidenko et al., 2005) may also account for the absence of a distinct response at HIF-1α protein level to the hypoxic stimulus. Donor background information is lacking, which could have influenced our results, as for example a clear relation with age has been shown for the

| CONCLUSION
The present study suggests that osteoclasts can be considered senescent instead of quiescent. Notably, senescence is not uniformly present in the osteoclast population, which may represent different stages in the life of the osteoclast. This may also be the background of the heterogeneous expression of Ki67, which is indicative of augmented ribosomal RNA transcription rather than proliferative activity. The direct relationship between senescence and osteoclastogenesis might mean that contents of the SASP not only play a functional role in matrix resorption but also may regulate osteoclastogenesis in a paracrine manner.
Hypoxia seems to affect the expression of senescence markers negatively and cellular fusion and formation of large osteoclasts is delayed at hypoxic (5% of O 2 ) compared to normoxic conditions (21% O 2 ). This, however, does not affect the resorption capacity of the osteoclasts on the longer term.

ACKNOWLEDGMENTS
Microscopy images were acquired in the Centre for Cellular Imaging (CCI) at the Faculty of Veterinary Medicine Utrecht and the authors thank A.R.J. Bleumink and Dr. R.W. Wubbolts for their technical advice and help with the image analysis. The authors kindly FIGURE 8 Senescence associated beta galactosidase staining is present at week 1 and 2 (a) supporting the senescent phenotype of osteoclasts. RT-qPCR analysis of the senescence associated genes CCL 2, CCL5, p21, and MMP9 (b; n = 3 donors) shows differential gene expression profiles in normoxia vs hypoxia. CCL2, C-C Motif Chemokine Ligand 2; CCL 5, C-C Motif Chemokine Ligand 5; p21, Cyclin Dependent Kinase Inhibitor 1A; MMP9, Matrix Metalloproteinase 9 after 1 day of culturing (indicated with "0") and 1-3 weeks of culturing. *0.01 < p < 0.05; **p < 0.01 hypoxia vs normoxia at the same time point acknowledge Dr. T. Schoenmaker from the Academic Centre for Dentistry in Amsterdam for providing us with the primer sequences of the osteoclast related genes. The authors are very grateful to W.A.M de Jong, S.G.M. Plomp, E.A. van Liere, and S.C. van Essenvan Dorresteijn for their technical assistance.