DNA damage and senescence in osteoprogenitors expressing Osx1 may cause their decrease with age

Summary Age‐related bone loss in mice results from a decrease in bone formation and an increase in cortical bone resorption. The former is accounted by a decrease in the number of postmitotic osteoblasts which synthesize the bone matrix and is thought to be the consequence of age‐dependent changes in mesenchymal osteoblast progenitors. However, there are no specific markers for these progenitors, and conclusions rely on results from in vitro cultures of mixed cell populations. Moreover, the culprits of such changes remain unknown. Here, we have used Osx1‐Cre;TdRFP mice in which osteoprogenitors express the TdRFP fluorescent protein. We report that the number of TdRFP‐Osx1 cells, freshly isolated from the bone marrow, declines by more than 50% between 6 and 24 months of age in both female and male mice. Moreover, TdRFP‐Osx1 cells from old mice exhibited markers of DNA damage and senescence, such as γH2AX foci, G1 cell cycle arrest, phosphorylation of p53, increased p21CIP 1 levels, as well as increased levels of GATA4 and activation of NF‐κB – two major stimulators of the senescence‐associated secretory phenotype (SASP). Bone marrow stromal cells from old mice also exhibited elevated expression of SASP genes, including several pro‐osteoclastogenic cytokines, and increased capacity to support osteoclast formation. These changes were greatly attenuated by the senolytic drug ABT263. Together, these findings suggest that the decline in bone mass with age is the result of intrinsic defects in osteoprogenitor cells, leading to decreased osteoblast numbers and increased support of osteoclast formation.


Introduction
Old age is, by far, the most important risk factor for the development of osteoporosis. In bone biopsies from elderly men and women, the agerelated loss of both cancellous and cortical bone is associated with decreased mean wall thicknessthe histomorphometric hallmark of decreased bone formation (Parfitt et al., 1995). Loss of bone mass in aged rodents is associated with a decline in the number of osteoblasts, the cells responsible for the synthesis and mineralization of the bone matrix (Almeida et al., 2007). Because osteoblasts are postmitotic cells with a short lifespan (Weinstein et al., 1998), they need to be constantly replaced with new ones. Osteoblasts arise from progenitors of mesenchymal origin, which express the transcription factors Runx2 and Osterix1 (Osx1) (Park et al., 2012).
The decline in the regenerative capacity of most tissues with old age has led to the idea that aging is due, at least in part, to increased cell senescence causing the loss of functional adult stem/progenitor cells (Rossi et al., 2008). Cellular senescence is a process in which cells stop dividing and initiate a gene expression pattern known as the senescenceassociated secretory phenotype (SASP) (Campisi, 2013;Lopez-Otin et al., 2013). Several stimuli associated with aging promote senescence. Because the number of senescent cells increases in multiple tissues with aging, it has been widely assumed that senescence contributes to aging (Lopez-Otin et al., 2013;Van Deursen, 2014). Importantly, ablation of senescent cells using genetically modified mice prolongs lifespan and delays age-related pathologies in naturally aged mice or progeria models (Baker et al., 2011(Baker et al., , 2016. We have recently shown that senescent cells induced by normal aging or ionizing radiation (IR) can be eliminated by administration of ABT263, a drug that kills senescent cells selectively; and clearance of senescent cells rejuvenates aged tissue stem and progenitor cells (Chang et al., 2016).
DNA damage is a major cause of senescence (Mandal et al., 2011). Stimulation of p53 and its target gene p21 CIP1 by the DNA damage response (DDR) plays a fundamental role in the initiation of senescence by causing cell cycle arrest (Roninson, 2002). An alternative or concurrent barrier to proliferation is the derepression of the INK4a/ARF locus and increase in the cyclin inhibitor p16 Ink4a (Robles & Adami, 1998). Another component of the DDR is the accumulation of the zinc finger transcription factor GATA4, which in turn stimulates NF-jB and the SASP (Kang et al., 2015). GATA4 protein levels increase in tissues of mice treated with senescence-inducing stimuli and in multiple mouse and human tissues during physiological aging. The SASP comprises a multitude of pro-inflammatory cytokines, chemokines, and proteases, which not only reinforce senescence growth arrest in an autocrine manner, but also cause senescence in surrounding cells in a paracrine manner (Coppe et al., 2008;Kuilman et al., 2008). These factors also cause low-grade inflammation.
In both humans and rodents, the reduced osteoblast number in the aging skeleton has been attributed to changes in bone marrow-derived mesenchymal progenitors, including a decrease in the number of mesenchymal stem cells, defective proliferation/differentiation of progenitor cells, increased apoptosis, or increased senescence (Stenderup et al., 2003;Sethe et al., 2006). Studies using bone marrow-derived stromal cells as a surrogate for osteoblast progenitors have suggested that serially passaged cells from aged humans or mice become senescence at earlier passages compared to cells from young individuals (Stenderup et al., 2003;Zhou et al., 2008;Sui et al., 2016). However, serial passaging in and of itself causes replicative senescence (Hayflick & Moorhead, 1961) and bone marrow cell cultures are heterogeneous. Therefore, it remains unclear whether the number of senescent osteoblast progenitors increases with old age. Moreover, the contribution of the decline in osteoblast progenitor number to the decrease in bone formation with age remains unknown because of the lack of methods to specifically identify and isolate mesenchymal progenitors. Therefore, the molecular mechanisms responsible for the decline in osteoblast number have remained elusive. To overcome these limitations, we generated a mouse model in which osteoblast progenitors are labeled with a red fluorescent protein to facilitate their isolation by fluorescence-activated cell sorting (FACS) and examination of the effects of aging in freshly isolated cells. We present evidence that the decline in bone formation with age can be accounted for by a decrease in the number of osteoprogenitors due to DNA damage-induced cell senescence.

Results
The number of osteoprogenitors marked by Osx1-Cre declines with age To determine whether osteoblast progenitors change with age, we crossed Osx1-Cre (Osx1-GFP::Cre) (Rodda & McMahon, 2006) and stop-loxP-tandem dimer red fluorescent protein (Td-RFP) mice (Luche et al., 2007) to generate Osx1-Cre;TdRFP mice. In these mice, osteoprogenitors and all their descendants, including osteoblasts and osteocytes, express TdRFP. Cells targeted by this Osx1-Cre transgene are present in the bone marrow of adult mice and represent bipotential progenitors capable of generating osteoblasts and adipocytes (Ono & Kronenberg, 2015). Furthermore, Osx1-Cre-expressing cells give rise to all osteoblast and osteocytes in the skeleton of young mice (Xiong et al., 2011). Similarly, all osteocytes were TdRFP-positive in femoral cortical bone of 24-monthold Osx1-Cre;TdRFP mice ( Fig. 1A). Cortical bone from TdRFP control mice did not exhibit any red fluorescence; however, some background fluorescence was detected in the bone marrow. Freshly isolated Osx1-TdRFP + bone marrow cells from Osx1-Cre;TdRFP mice, using FACS, had elevated levels of Runx2 and osteocalcin compared to Osx1-TdRFPcells ( Fig. 1B) and formed mineralized nodules when cultured in osteogenic conditions (Fig. 1C).
We counted Osx1-TdRFP + cells in the bone marrow of young adult (4-6 months) mice and found that this population represents about 0.02-0.03% of all cells ( Fig. 2A-C). The number of the TdRFP+ cells decreased by approximately 50% in old (20-24 months) female or male mice (average of four independent experiments, one of which is shown in Fig. 2A). We also compared the number of cells marked by a Prx1-Cre transgene, which targets mesenchymal stem cells at a stage earlier than Osx1, in the bone marrow of young and old Prx1-Cre;TdRFP mice. Prx1-TdRFP + cells represented about 0.05-0.06% of all cells in the bone marrow, and their abundance did not change with age ( Fig. 2D,E).
Cultured bone marrow stromal cells from old mice exhibited lower mRNA levels of the osteoblast markers Runx2, Osx1, collagen 1a1 (Col1a1), and osteocalcin (Ocn) (Fig. 2F) and formed fewer mineralized nodules than cells from young mice (Fig. 2G), despite the fact that both cultures reached confluence. In contrast, the adipocyte marker Pparc was increased (Fig. 2F). The decline in osteoblast generation seen in the cultured cells is probably due to the lower number of Osx1-expressing cells present in the bone marrow of old mice. Because the mineralization assay requires the cells to remain in culture for an extended period of time, it is not possible to attribute the changes in mineralization to a particular cellular event, for example, proliferation, apoptosis, differentiation, or matrix secretion.
Osteoprogenitors in old mice are arrested at G1 and exhibit increased DNA damage and p53 activation We next examined cell cycle distribution of Osx1-TdRFP + cells. The percentage of cells at the G1 phase in old female and male mice was approximately fivefold higher than in cells from young mice (Fig. 3A, B). Because DNA damage is one of the major causes of cell cycle arrest, we quantified cH2AX focia common marker of DNA damage in the nucleus of Osx1-TdRFP + cells. To this end, we performed immunostaining in freshly isolated cells from young or old mice. Approximately 70% of the cells isolated from old mice exhibited c-H2AX foci as compared to 20% in cells from young mice (Fig. 3C,D). The average number of cH2AX foci per cell also increased with age (Fig. 3D).
We next determined whether p16 and/or p21 might mediate the cell cycle arrest seen in osteoprogenitors from old mice. The mRNA levels of p16 were unaffected by age (Fig. 4A). In contrast, both the mRNA and protein levels of p21 were elevated in Osx1-TdRFP + cells from old mice (Fig. 4A,B). In line with the increase in p21, the number of osteoprogenitors exhibiting p53 phosphorylation increased in old mice, as determined by immunostaining with a Ser15 phospho-p53 antibody in freshly isolated Osx1-TdRFP + cells (Fig. 4C). A similar increase in p21 ( Fig. 4D), but unaltered p16 levels ( Fig. 4E), was seen in cultures of bone marrow cells from old mice. These findings suggest that Osx1-TdRFP + cell cycle arrest was caused by DNA damage-induced stimulation of the p53/p21 pathway.
Osteoprogenitors from old mice exhibit increased GATA4, NF-jB, expression of the SASP, and increased osteoclastogenic support capacity We next looked for signs of the SASP. GATA4 protein levels were increased in freshly isolated Osx1-TdRFP + cells from old mice when compared with cells from young mice (Fig. 5A), as determined by Western blot. Similarly, the levels of p-IkB and p-p65 were elevated in cells from old mice, indicating stimulation of NF-jB signaling. Bone marrow-derived osteoblastic cells from old mice cultured in osteogenic medium for 7 days also exhibited higher GATA4 and p-p65 levels than cells from young mice (Fig. 5B). In line with evidence that GATA4/NF-jB signaling is a major inducer of the SASP, the mRNA abundance of common SASP components, such as TNF-a, IL-1a, MMP-13, CXCL12 but not IL-6, was elevated in cultured cells from old mice (Fig. 5C). The expression of the critical osteoclastogenic factor RANKL was also elevated. We also examined the ability of stromal cells from old mice to support osteoclast formation. The number of osteoclasts formed in co-cultures of macrophages with bone marrow-derived stromal cells, in the presence of 1,25 dihydroxy vitamin D3, was higher when stromal cells originated from old mice compared to cells from young mice (Fig. 5D). Similar findings have been reported by others (Cao et al., 2005).
The senolytic drug ABT263 attenuated the expression of SASP and the osteoclastogenic support capacity of bone marrow stromal cells from old mice We have previously shown that ABT263 kills selectively senescent cells (Chang et al., 2016). Culture of bone marrow stromal cells from old mice with ABT263 for 5 days, dose dependently, increased apoptosis (data not shown). ABT263 also caused a decrease in protein levels of p21 and GATA4 (Fig. 6A), as well as the expression of IL-1a, MMP-13, CXCL12, and RANKL (Fig. 6B). TNF-a was not affected. Co-cultures of bone marrow stromal cells from old mice pretreated with ABT263 exhibited lower capacity to support osteoclast formation in co-cultures with macrophages, compared to cells pretreated with vehicle control (Fig. 6C). These findings suggest that ABT263 can eliminate senescent osteoprogenitor cells.

Radiation-induced senescence replicates the effects of aging in cultured osteoblastic cells
Finally, we determined whether the effects of aging are similar to those seen with other forms of DNA damage. Exposure to ionizing radiation (IR) causes DNA damage and senescence (Shao et al., 2014). To determine whether IR could replicate the effects of aging in primary osteoblastic cells, we subjected cultured newborn calvaria cells or bone marrow-derived stromal cells to 10 Gy of c-radiation. After a few passages, the irradiated cells exhibited increases in senescence-associated b-galactosidase (SA-b-gal) activity ( Fig. 7A) and mRNA expression of p21 and IL-1a (Fig. 7B), while the levels of p16 were unaltered. Western blot analysis confirmed the increased levels of p21 and further revealed that irradiation-induced senescent calvaria and bone marrow-derived osteoblasts had a 2.5-and twofold increased expression of GATA4, respectively (Fig. 7C). Phospho-p65 was also increased in both cell types. These findings indicate that induction of DNA damage in cultured osteoblastic cells by IR causes DNA damage responses and senescence similar to the ones seen in osteoprogenitors from old mice.

Discussion
In the studies described herein, we isolated a homogeneous population of osteoblast progenitors from the bone marrow and determined the effects of old age. We have found that osteoblast progenitors from old mice exhibit several markers of senescence. These senescence features are associated with a decline in the number of osteoblast progenitors and a decrease in matrix synthesizing osteoblast number and bone formation rate, shown previously (Almeida et al., 2007). The decline in cell number is specific for the population expressing Osx1. Cells expressing Prx1, which include a broader group of cells such as mesenchymal stem cells, fibroblasts, CXCL12-abundant reticular (CAR) cells, adipocytes as well as osteoprogenitors, did not decline with age. A decline in Osx1-expressing cells with age was also noted, but not quantified, by Kusumbe et al. (2014) using immunofluorescent staining and confocal microscopy of bone sections. Our findings suggest that the osteoprogenitor cell population marked by Osx1-TdRFP is susceptible to the damaging effects of aging.
The number of cH2AX foci was increased in osteoprogenitors from old mice, strongly suggesting that DNA damage is elevated in these cells. Several other mammalian tissues show an age-related increase in foci of cH2AX Wang et al., 2009). This may result from the decreased tendency of double-strand breaks to be repaired with aging (White et al., 2013), which may contribute to the accumulation of senescent cells, that harbor persistent DNA damage foci (Rodier et al., 2011). Previous studies with mouse models of accelerated aging due to defective DNA damage repair have elucidated that accumulation of DNA damage causes several features of early aging including decreased bone formation and low bone mass (Vogel et al., 1999;Dolle et al., 2011;Saeed et al., 2011;Chen et al., 2013). DNA damage in these models is caused by a deficiency in telomerase, ERCC1 (Excision Repair Cross Complementary group 1), or Ku80. Similar to physiological aging, the low bone mass in these accelerated aging models is associated with low bone formation. Although freshly isolated osteoprogenitors were not examined, bone marrow stromal cell cultures from ERCC1 À/À or Tert À/À mice also exhibit markers of senescence (Saeed et al., 2011;Chen et al., 2013). Further support for the contention that DNA damage is responsible for osteoprogenitor cell senescence and the decline in bone formation with age are our findings that DNA damage induced by irradiation causes changes in osteoprogenitors that are similar to the ones seen with aging. In addition, DNA damage, due to focal irradiation in long bones, causes senescence in osteoblast lineage cells, decreases bone formation, and leads to bone loss in mice (Chandra et al., 2017). Phosphorylation of p53 and the levels of p21 were greatly increased in osteoprogenitors from old mice, and these changes were associated with a higher number of cells arrested at the G1 phase. An increase in p21 levels due to p53 activation is a critical mediator of cell cycle arrest and senescence following DNA damage in several cell types (Herbig et al., 2004;Passos et al., 2010). In osteoblastic cells, p21 inhibits proliferation and differentiation (Bellosta et al., 2003). Studies using p53 À/À mice, or mice with osteoblast specific p53 gain-or loss-offunction have provided compelling evidence that p53 inhibits osteoblastogenesis (Lengner et al., 2006;Wang et al., 2006). Furthermore, transgenic mice with global gain-of-function mutation of p53 rapidly accumulate senescent cells and exhibit accelerated aging phenotypes, including low bone mass (Tyner et al., 2002;Maier et al., 2004). Nevertheless, it remains unknown whether gain-of-function of p53 causes osteoprogenitor senescence and whether this contributes to the skeletal defects seen in the mice. Interestingly, p16 Ink4a levels were unaffected in osteoprogenitors from old mice and in senescent calvaria cells induced by c-radiation. These results are in line with evidence that p16 and p21 pathways are triggered independently and that activation of the p53/p21 pathway, but not p16, correlates with the presence of c-H2AX foci in response to DNA damage . Furthermore, studies in human cells have shown that some cell types, for example, skin fibroblasts such as BJ cells, do not exhibit p16 upregulation with senescence and activation of p53 alone is sufficient to cause the senescence arrest (Itahana et al., 2003). Human endothelial cells primarily depend on p21 but not p16 for senescent induction . In contrast, lung-derived fibroblasts require both p53 and p16 to achieve senescence arrest. Similar to osteoprogenitors, osteocytes of aged mice exhibit markers of senescence and expression of SASP. However, in difference to Osx1-expressing cells, osteocytes from old mice exhibit elevated p16 (Farr et al., 2016). While the reasons for the differences in the expression of p16 and p21 remain unclear, it has been proposed that distinct cell types might utilize different cyclin-  dependent kinase inhibitors to induce senescence, due to activation of different stress response pathways Wang et al., 2016). Together, these findings support the idea that stimulation of the p53/p21 component of the DDR is responsible for cell cycle arrest in osteoprogenitors and contributes to the decline in osteoblast number in the aging skeleton.
GATA4 levels increased with age in osteoprogenitors, as well as with radiation in cultured osteoblastic cells. These observations are in line with evidence that DNA damage promotes GATA4 accumulation (Kang et al., 2015). Importantly, stabilization of GATA4 is sufficient to induce senescence and the SASP in cultured human cells. GATA4 increases the SASP, at least in part, via stimulation of NF-jB activity (Kang et al., 2015). Consistent with this evidence, NF-jB activity was increased in osteoprogenitors; and SASP components such as TNF-a, MMP13, CXCL12, and IL-1a were also elevated. All these proteins have osteoclastogenic properties (Pfeilschifter et al., 1989;Wright et al., 2005;Fu et al., 2016) and most probably contributed, along with RANKL, to the increased osteoclast formation seen in co-cultures with stromal cells from old mice. Support for the contention that senescent cells contribute to osteoclastogenesis is provided by the findings that senescent osteoblastic cells, resulting from overexpression of the cyclin inhibitor p27KIP1, increase osteoclast formation in a co-culture system in vitro and osteoclasts number in vivo . It is, therefore, plausible that the SASP from osteoprogenitors and osteocytes contributes to the increased bone resorption that occurs in cortical bone with age (Jilka et al., 2014;Ucer et al., 2016). In addition, our work suggests that ABT263 can attenuate the SASP and osteoclastogenesis in bone marrow stromal cell cultures from old mice. These results are in line with our previous work showing that ABT263 abrogates the SASP response in bone marrow stromal cells caused by irradiation in mice (Chang et al., 2016). Taken together, these findings suggest that ABT263 can eliminate senescent osteoblast progenitors in aged mice. GATA4 stimulation might also contribute to the age-related decline in bone formation. Indeed, overexpression of GATA4 in primary calvaria cells attenuates Runx2 promoter activity and inhibits osteoblastogenesis (Song et al., 2014). In contrast, GATA4 in osteoblasts might contribute to bone mineralization during development (Guemes et al., 2014).
Several cell types within the osteoblast lineage have been proposed to implicated in the decline of bone formation with age (Kassem & Marie, 2011;Almeida & O'Brien, 2013). Our present findings indicate that the age-related decrease in bone formation and increase in bone resorption could be accounted for by intrinsic defects in osteoblast progenitors that lead to a decrease in their number combined with an increase in their capacity to support osteoclast formation because of SASP. Together with evidence that removal of senescent cells via genetic or pharmacologic means attenuates the age-dependent degeneration of several tissues (Baker et al., 2011(Baker et al., , 2016Chang et al., 2016), the present work supports the notion that removal of senescent cells may represent a therapeutic approach to the prevention and treatment of involutional osteoporosis.

Bone histology and fluorescence imaging
Freshly dissected bones were fixed in 4% paraformaldehyde overnight, washed in PBS, decalcified in 14% EDTA pH 7.1 at 4°C for 2 weeks, and then stored in 30% sucrose solution. Bones were embedded in Cryo-Gel (Electron Microscopy Sciences, Hatfield, PA, USA) and sectioned using CryoJane tape-transfer system (Instrumedics Hackensack, NJ, USA) with 15 lm thickness. Frozen sections were rinsed with PBS and cover-slipped with Vectashield mounting medium containing DAPI (Vector Laboratories Burlingame, CA, USA). Fluorescent images were acquired using Olympus BX53 fluorescence microscope (Center Valley, PA, USA) and appropriated filter set (excitation; 540/10 nm band pass filter; emission: 600/50 nm band pass filter) fluorescence microscope using a 209 lens objective.

Isolation of bone marrow Osx1-TdRFP + cells
The tibiae and femurs were dissected from mice immediately after death. Total bone marrow cells were flushed from the bones, using a 23-gauge needle and syringe, into ice-cold FACS buffer containing CaCl 2 -and MgCl 2 -free 1X PBS (Thermo Fisher Scientific, Carlsbad, CA, USA) and 2% FBS. Cells from individual mice in each group were centrifuged at 450 g for 6 min at 4°C. After the red blood cells were removed with RBC lysis buffer (0.9% NH 4 Cl with 20 mM Tris base, pH 7.4), bone marrow cells were suspended in ice-cold FACS buffer. Cells were then incubated with biotin-conjugated rat antibodies specific for mouse CD45 (eBioscience, San Diego, CA, USA; 14-0451, 1:100). The labeled hematopoietic cells were depleted 3 times by incubation with anti-rat IgG Dynabeads (Invitrogen, Grand Island, NY, USA) at a bead:cell ratio of approximately 4:1. Cells binding the Dynabeads were removed with a magnetic field. The negatively isolated CD45 À cells were washed twice and suspended with ice-cold FACS buffer at 1-2 9 10 6 cells mL À1 . Osx1-TdRFP + cells were sorted in an Aria II cell sorter (BD Bioscience, San Jose, CA, USA) using the PE-A fluorochrome gate.

Osteoblast differentiation
Freshly sorted Osx1-TdRFP À or Osx1-TdRFP + cells (approximately 0.1 9 10 6 /well) pooled from six mice from each group were immediately cultured with feeder layer cells (approximately 0.8 9 10 6 /well), 20% FBS, 1% PSG, and 50 lg mL À1 of ascorbic acid in 12-well plates for 7 days. Half of the medium was replaced every 3 days. Cells were then cultured with 10% FBS, 1% PSG, 50 lg mL À1 of ascorbic acid (Sigma), and 10 mM b-glycerophosphate (Sigma) for 21 days. For bone marrow-derived osteoprogenitor cells, total bone marrow cells pooled from three to five mice from each group were cultured with 20% FBS, 1% PSG, and 50 lg mL À1 of ascorbic acid in 10-cm culture dishes for 5 days. Half of the medium was replaced every 3 days. Mineralized matrix was stained with 40 mM alizarin red solution. To remove senescent cells selectively, bone marrow-derived osteoprogenitor cells were collected as described above and incubated with 5 lM ABT263 (Selleckchem #S1001) in the presence of 50 lg mL À1 of ascorbic acid in 10-cm culture dishes for 5 days, followed by removal of the drug. Medium was replaced every 2 days.

Osteoclast differentiation
For co-culture assays, red blood cell-free bone marrow-derived macrophages (300 000 cells cm À2 ) and stromal cells (25 000 cells cm À2 ) were seeded in 48-well tissue culture plates with 10 À8 M 1a,25(OH) 2 D 3 (Sigma-Aldrich, St. Louis, MO, USA) and 10 À7 M PGE 2 (Sigma-Aldrich) in a-MEM containing 10% FBS for 7 days. Medium was replaced every 3 days. The cells were fixed with 10% neutral buffered formalin for 15 min, and mature osteoclasts (with >3 nuclei) were enumerated after staining for TRAP using the Leukocyte Acid Phosphatase Assay kit (Sigma-Aldrich).  of more than 100 cells per slide were counted in >30 random fields on a slide to determine the percentage of p-p53 + cells and the number of cH2AX foci for calculation of the average number of cH2AX foci/cell.

Induction of cellular senescence by IR
To induce cellular senescence by IR, bone marrow-derived osteoprogenitor or calvarial cells about 60-70% confluence were exposed to 10 Gy of IR in a J.L. Shepherd Model Mark I 137 Cesium c-irradiator (J.L. Shepherd) at a dose of 1.080 Gy/min. Three days after irradiation cells were passaged once at a 1:3 dilution and became fully senescent 7 days after irradiation, as confirmed by SA-b-gal staining.

SA-b-galactosidase staining
Senescent status of cells was verified by staining for SA-b-gal activity using a SA-b-gal staining kit (Cell Signaling Technology) according to the manufacturer's instructions. Senescent cells were identified as bluestained cells by standard light microcopy.

Statistical analysis
For the animal experiments, no specific blinding trial was used, but mice in each group were selected randomly. The experimental sample size (n) of each group is described in each corresponding figure legend. Student's t-test (independent samples, two-sided) was used to detect significant aging and treatment effects, after determining that the data were normally distributed and exhibited equivalent variances. All experiments were repeated at least twice. Statistical significance was set at a P < 0.05. Error bars in all figures represent SD.