Therapy-Induced Senescence Drives Bone Loss

Chemotherapy is commonly administered as adjuvant therapy in patients with breast cancer to reduce the risk of recurrence after surgical resection (1, 2). Although efficacious at prolonging disease-free survival (3, 4), chemotherapy has debilitating side effects including bone loss (5) that can exceed 7% per year in women with breast cancer, which is significantly more than the 1%–2% observed in postmenopausal women. Chemotherapy can induce 4%–5% bone loss in men with prostate cancer as well (6).

Estrogen is a key anabolic agent in the bone and its loss disrupts bone homeostasis. This is best illustrated in postmenopausal women, where low estrogen contributes to 1%–2% bone loss per year and in many cases the subsequent onset of osteoporosis (7–9) that cost the U.S. health care system over 20 billion dollars in 2008 (10). Thus, bone loss has a substantial impact on a cancer survivor's quality of life and on the cost to the health care industry. Estrogen's role in bone homeostasis is complex. Indeed, estrogen exerts proosteoblastic and antiosteoclastic actions on the bone (9, 11) and regulates T-cell–induced inflammation and bone loss (12).

While estrogen's role in bone homeostasis is well established, the degree to which its loss contributes to chemotherapy-induced bone loss remains controversial (6, 13–16). Furthermore, it is unclear what other mechanisms contribute to bone loss following chemotherapy. Some clinical studies suggest that chemotherapy only induces bone loss in premenopausal women who experience amenorrhea, whereas patients that retain ovarian function display negligible changes in their bones (17). Other studies reported that adjuvant chemotherapy induces substantial bone loss in postmenopausal patients in the absence of additional estrogen loss (15). Furthermore, in many instances chemotherapy-induced bone loss is more rapid and aggressive than menopausal bone loss, and more aggressive than that observed in patients subject to complete hormone deprivation through treatment with aromatase inhibitors (6). Together, these findings suggest that chemotherapy has a direct effect on bone that results in bone loss and is independent of estrogen levels.

All mice were obtained from JAX laboratories and were housed in accordance with Washington University in St. Louis Animals Studies Committee. All animal procedures were approved by Washington University's Institutional Animal Care and Use Committees. Wild-type, 6- and 16-week-old female FVB/NJ or C57BL/6 mice were used in experiments as indicated in figure legends, unless otherwise noted. The INK-ATTAC-transgenic mice were obtained from Mayo Clinic (a kind gift from D. Baker and J.M. van Deursen).

INK-ATTAC mouse genotyping was carried out as follows: genomic DNA was extracted from mouse tissue using the RED Extract N Amp Kit (XNAT, Sigma). Primer sequences used to determine the presence of the INK-ATTAC transgene were as follows: 5′-GCAAGGTCTGTTGAATGTCGTGAAGGAAGC-3′ (forward) and 5′-CGACTAAACACATGTAAAGCATGTGCACCG-3′ (reverse). To amplify DNA of interest, 1 μL DNA extract was combined with 9 μL of master mix (XNAT kit with 0.5 μmol/L appropriate forward and reverse primers) in PCR tubes and amplified for one cycle at 94°C for 1 minute followed by 35 cycles at (94°C for 60 seconds, 63°C for 90 seconds, and 72°C for 2.30 seconds) and then one cycle at 72°C for 5 minutes followed by a 4°C hold. PCR products were resolved on a 2% agarose/TAE (Thermo Fisher Scientific) gel at 100 volts for 45 minutes, and visualized with UV light.

Femurs were isolated from euthanized mice and fixed with formalin overnight. The bones were washed in PBS and placed in 70% ethanol. The femurs were then suspended in 2% agarose and scanned using microCT [μCT40 or vivaCT (Scanco Medical)] at 70 kVp, 114 μA, and 20 μm resolution. For the trabecular compartment, contours were traced on the inside of the cortical shell using 2D images of the femoral metaphysis. The end of the growth plate region was used as a landmark to establish a consistent location for starting analysis, and the next 30 slices were analyzed. The following trabecular parameters are reported for all μCT experiments: bone volume over total volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), trabecular separation (Tb.Sp), and volumetric bone mineral density (vBMD). For the cortical compartment, contours were traced on the outside of the cortical shell using 2D images of the femoral mid-diaphysis and 100 slices were analyzed (as noted in figure legends). 3D reconstructions were generated using OsiriX, of 0.9 mm thick sections of femur right below the growth plate.

Mice were anesthetized with isofluorane and trabecular volume in the distal femoral metaphysis of the right leg was measured using in vivo μCT (vivaCT 40). For the trabecular compartment, contours were traced on the inside of the cortical shell using 2D images of the femoral metaphysis. The end of the growth plate region was used as a landmark to establish a consistent location for starting analysis, and the next 30 slices were analyzed. The following trabecular parameters are reported for all μCT experiments: BV/TV, Tb.Th, Tb.N, Tb.Sp, and vBMD.

Doxorubicin (PHR1789, Sigma) was prepared in water and administered at 5 mg/kg via intraperitoneal injections for 4 weeks, once weekly. Mice in the vehicle group were given water. Paclitaxel (Fresenius Kabi) was purchased in suspension, each milliliter containing 6 mg paclitaxel, USP; 527 mg polyoxyl 35 castor oil NF; and 49.7% (v/v) dehydrated alcohol, USP. We prepared a 1:3 dilution in PBS and administered it via retroorbital injections for 6 weeks, once weekly.

The vehicle and AP20187 (batch no.: A113-20, Chemveda Life Sciences) were prepared with 100% ethanol: polyethylene glycol 400: 2% Tween-20 in molecular water at 4:10:86. INK-ATTAC transgenic mice were administered 2 mg/kg vehicle or AP20187 through intraperitoneal injections twice weekly for 4 weeks.

Bilateral ovariectomy was performed on mice anesthetized using isofluorane. Mice were either 6-week-old or 16-week-old at the time of the surgery, depending upon the experiment (indicated in the figure legends). The ovary was approached by a single skin incision (∼5 cm) made in the “waist” region of the mouse's flank in the dorsal–ventral direction. The ovary was gently pulled through the muscle opening, a hemostat was clamped at the boundary between the ovary and uterus, and a single cut was made to detach the ovary from the uterus. A cauterizer was used to stabilize any bleeders. The muscle layer was closed with absorbable 4.0 vicryl sutures. The skin was closed with sterile sutures or wound clips and betadine applied to the incision area to avoid infection. The same procedure was repeated on the other side of the mouse. Sham-operated mice had similar incisions to the skin and peritoneum on each side, but the ovaries were not removed. To minimize discomfort post-surgery, 0.1 mg/kg buprenorphine was administered subcutaneously.

Mouse femur bones were isolated and fixed in 10% neutral buffered formalin for 24 hours. Bones were decalcified in 14% EDTA for 14 days at 4°C, embedded in paraffin, and sectioned at the histology core of the Washington University Musculoskeletal Research Center. Standard hematoxylin and eosin and tartrate-resistant acidic phosphatase (TRAP) staining techniques were used for all bone sections. Images were collected using the Zeiss Axio Scan Z1 Brightfield slide scanner, Washington University Center for Cellular Imaging (WUCCI). To quantitate the TRAP-positive cells and determine how much of their surface area covered the bone, we used the BioQuant software to quantitate osteoclasts per bone surface (Oc.S/BS and N.Oc/BS). We carried out these measurements by defining a region of interest with the stamp tool window. We then chose a diagonal line type and set the distance to the growth plate and measured the Oc.S/BS and N.Oc/BS.

Calcein (10 mg/kg, Sigma) and alizarin red (30 mg/kg, Sigma) were intraperitoneally injected into mice 9 and 2 days prior to sacrifice, respectively. Dissected long bones were fixed with 10% formalin overnight, then embedded in methymethacrylate for sectioning. Sections were left unstained and images were collected using the Zeiss Axio Scan Z1 Fluorescence slide scanner, WUCCI. Histomorphometric analysis was performed using BioQuant OSTEO 2010 software (BioQuant Image Analysis Corporation) and standard parameters to quantify mineralizing surfaces (mineral apposition rate, MAR) and bone formation rate per bone surface (BFR/BS) were performed.

Bones were isolated from sacrificed mice, and the marrow cavity was flushed out by centrifuging at 10,000 rpm for 10 seconds at 4°C. Bones devoid of marrow were lysed and homogenized in TRIzol (Life Technologies) containing RINO BulletBlender Navy Bead Lysis tubes (MidSci) and pulverized using a Bullet Blender Homogenizer at 4°C at speed 10 for 5 minutes. Once the bones were completely crushed, the tubes were centrifuged at 10,000 rpm for 10 minutes. The supernatant was transferred to a new tube and RNA was isolated using the RiboPure RNA Purification Kit (AM1924, Thermo Fisher Scientific). cDNA was synthesized from RNA (1 μg) using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time PCR was performed using the TaqMan PrimeTime Gene Expression Master Mix kit and gene-specific primers and probes. All reactions were performed in duplicate. P16 and Il6 relative mRNA levels were calculated by the comparative cycle threshold (C_t) method [2^{(reference C_t-gene of interest C_t)}] using GAPDH as an internal control. The rest of the genes' mRNA were normalized to the mean of five housekeeping genes (Hprt, Tbp, Cyclophilin, Tubulin, and Actin). Primers included: Il6: (Forward) 5′-AGC CAG AGT CCT TCA GAG A-3′, (Reverse) 5′-TCC TTA GCC ACT CCT TCT GT-3′. p16: (Forward) 5′-AAC TCT TTC GGT CGT ACC CC-3′, (Reverse) 5′-TCC TCG CAG TTC GAA TCT G-3′. Ccl3: (Forward) 5′-CGATGAATTGGCGTGGAATC-3′, (Reverse) 5′-CCTTGCTGTTCTTCTCTGTACC-3′. Mcsf: (Forward) 5′-GGAAGATGGTAGGAGAGGGTA-3′, (Reverse) 5′-AGGATGAGGACAGACAGGT-3′. Mmp12: (Forward) 5′-GCTCCTGCCTCACATCATAC-3′, (Reverse) 5′- GGCTTCTCTGCATCTGTGAA-3′. Hmgb1: (Forward) 5′-AATGGCGGTTAA AGGAGAGTC-3′, (Reverse) 5′-TGGCAAAGCAAGGAGTGAT-3′. Igfbp4: (Forward) 5′-GAT CCA CAC ACC AGC ACT T-3′, (Reverse) 5′-CTC TTC ATC ATC CCC ATT CCA-3′. Cxcl2: (Forward) 5′-CTT TCC AGG TCA GTT AGC CTT-3′, (Reverse) 5′-CAG AAG TCA TAG CCA CTC TCA AG-3′. Gapdh: (Forward) 5′-AATGGTGAAGGTCGGTGTG-3′, (Reverse) 5′-GTGGAGTCATACTGGAACATGTAG-3′. Hprt: (Forward) 5′-GTTGTTGGATATGCCCTTGAC-3′, (Reverse) 5′-TCAACAGGACTCCTCGTATT-3′. Tbp: (Forward) 5′-CCCTTCACCAATGACTCCTATG-3′, (Reverse) 5′-CAGCCAAGATTCACGGTAGAT-3′. Cyclophilin: (Forward) 5′-TTCACCTTCCCAAAGACCAC-3′, (Reverse) 5′-CAAACACAAACGGTTCCCAG-3′. Tubulin: (Forward) 5′-TCTTGTCACTTGGCATCTGG-3′, (Reverse) 5′-CGCGAAGCAGCAACCAT-3′. Actin: (Forward) 5′-TTTCCAGCCTTCCTTCTTGG-3′, (Reverse) 5′-GGCATAGAGGTCTTTACGGATG-3′. Ptprc: (Forward) 5′- TTT CCA ATG TGC TGT GTC CT-3′, (Reverse) 5′- TGA AGA AGA GAG ATC CAC CCA-3′. Col1a1: (Forward) 5′- CAT TGT GTA TGC AGC TGA CTT C-3′, (Reverse) 5′- CGC AAA GAG TCT ACA TGT CTA GG-3′. Col2a1: (Forward) 5′- AGC CTT CTC GTC ATA CCC T-3′, (Reverse) 5′- GCA GAG ATG GAG AAC CTG GTA-3′. Pecam: (Forward) 5′- TGT TGG AGT TCA GAA GTG GAG-3′, (Reverse) 5′- TCA TTG GAG TGG TCA TCG C-3′. Fabp4: (Forward) 5′-CCT TTC ATA ACA CAT TCC ACC AC-3′, (Reverse) 5′- AAA TCA CCG CAG ACG ACA G-3′. p21: (Forward) 5′-GAA GAG ACA ACG GCA CAC T-3′, (Reverse) 5′-CAG ATC CAC AGC GAT ATC CAG-3′.

Bones (femurs and tibias) were harvested from euthanized mice and placed in cold DMEM/F12 (Gibco). Muscle and connective tissues were removed. Bones were gently ground and then cut into small fragments. Fragments were transferred and digested with fresh 2 mg/mL collagenase (Sigma, C0130) in DMEM/F12 on a rotating water bath at 37°C for 30 minutes. Following the first digestion, the released cell suspension was filtered through a 70-μm nylon mesh into a collection tube and placed on ice. A second digestion of the remaining fragments was performed and the cell suspension was filtered into the same collection tube. Reactions were quenched with FACS buffer (PBS plus 0.5% BSA) with 2 mmol/L EDTA. Following FACS buffer wash and RBC lysis, dissociated cells were labeled with anti- CD45 magnet microbeads (Miltenyi Biotec) and enriched into a CD45⁺ fraction (hematopoietic cells) and CD45⁻ fraction (bone stromal cells) by magnetic-activated cell sorting (MACS, Miltenyi Biotec). Cells were stained with fluorochrome antibodies for 20 minutes on ice for cell sorting. We used CD45 (BioLegend, 103126), CD11b (BD Biosciences, 557397), CD3e (BD Biosciences, 553062), CD19 (eBioscience, 17-0193-80), Sca-l (Ly6A/E) (BD Biosciences, 563288), and c-kit (BioLegend, 105813) to isolate independent populations from the CD45⁺ fraction and CD45 (BD Biosciences, 553080) and CD31 (eBioscience, 48-0311-80) for cells from the CD45⁻ fraction. Using a BD FACSAria III Cell Sorter (BD Biosciences), indicated cell populations were directly sorted into 650 μL RLT containing 2-β mercaptoethanol lysis buffer and processed for RNA extraction with RNeasy Micro Kit (Qiagen). cDNA was synthesized using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). RT-PCR was used to validate the cell populations and test the senescence signatures. Primers used are listed in RT-PCR section.

Mice were anesthetized with 2% isoflurane prior to irradiation to the right leg by shielding the rest of the body with lead. The mice were irradiated with a single dose of 20 Gy using an RS-2000 (Rad Source) irradiator at a dose rate of 1.81 Gy/minute at 160 kVp X-rays.

Senescence-associated β-galactosidase (SA-β-Gal) staining in femurs of mice was performed on frozen sections that were fixed in 0.2% glutaraldehyde for 10 minutes at 4°C. Slides were washed in PBS and then submerged in X-gal solution (1 mg/mL 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside, 150 mmol/L NaCl, 2 mmol/L MgCl₂, 5 mmol/L K₃Fe(CN)₆, 5 mmol/L K₄Fe(CN)₆, 40 mmol/L NaPi pH 6.0, in H₂O). The X-gal solution was passed through a 0.45-μm filter before use to remove particulate. Slides were kept at 37°C in the dark until the stain developed (∼6 hours). After staining, the slides were fixed for 10 minutes in 0.2% glutaraldehyde at room temperature, washed in PBS, and then the nuclei were counterstained using Nuclear Fast Red (Sigma).

SA-β-Gal staining on fresh visceral fat was performed after fixation in 4% paraformaldehyde (PFA) at 4°C for 20 minutes. Fat tissues were then immersed in X-gal solution and kept at 37°C in the dark for approximately 6 hours. After staining, fat tissues were fixed in 4% PFA, washed with PBS, and images acquired.

Neutral buffered formalin-fixed, paraffin embedded mouse bones that had been decalcified in 14% EDTA for 14 days were cut in serial sections (5 μm). IHC staining was performed on a BOND RXm Fully Automated Research Stainer from Leica Biosystems. Phosphorylated-MK2 was stained using the Bond Polymer Refine Detection Kit (catalog no: DS9800) according to the manufacturer's recommendations for rabbit primary antibodies. In brief, tissue sections were baked for 15 minutes at 60°C, dewaxed, then heat-induced epitope retrieval was performed with ER2 (EDTA) for 20 minutes at 100°C followed by 5 minutes of blocking of endogenous peroxidase activity, primary antibody phosph-MK2 (Cell Signaling Technology, 1:100; catalog no. 3041S) was incubated for 1 hour at room temperature. A no primary negative control was simultaneously stained. Next, slides were subject to Bond washes and then incubated with the polymer anti-rabbit Poly-HRP-IgG for 8 minutes. Finally, DAB was applied and counterstaining was done with hematoxylin. Tissue was then dehydrated in ethanol followed by xylene, and slides were mounted with xylene-based mounting media. For quantification, whole-tissue slide scans were obtained using a Zeiss Axio Scan Z1 Brightfield slide scanner. Images were analyzed with HALO software and the whole tissues were defined as regions of interest (ROI). We adjusted and detected the threshold of positive p-MK2 staining on a single nuclear basis to quantify the percentage of positive cells in a given ROI.

The p38MAPK small-molecule inhibitor CDD111 (p38i) (Aclaris Therapeutics, Inc.) was compounded as described previously (18). The p38MAPK/MK2 small-molecule pathway inhibitor ATI-450 (MK2Pi) was compounded at 1,000 ppm (Aclaris Therapeutics, Inc.). Female C57BL/6 mice were fed ad libitum. Mice were randomized onto inhibitor containing or regular chow on the same day as their first doxorubicin dose.

All statistical analyses were carried out using GraphPad Prism. Numerical data are expressed as mean ± SEM. All qRT-PCR data are presented as mean ± SD. Between-group comparisons of mouse results were performed by Student t test or one-way ANOVA.

Chemotherapy is a potent inducer of DNA damage (19–22) that can robustly induce cellular senescence (21–24). Recent work demonstrated that senescence drives age-related bone loss (25), raising the possibility that it contributes to chemotherapy-induced bone loss. To study the role of senescence in chemotherapy-induced bone loss, we established a clinically meaningful mouse model by employing doxorubicin (DOXO), an anthracycline-based chemotherapeutic agent that is widely used to treat patients with early and advanced breast cancer (26, 27), and results in adverse effects on bone in adults (28) and children alike (29). Having established this model, we compared it with a bone loss model that is completely dependent on estrogen, namely ovariectomy (OVX), which induces robust bone loss in the absence of senescence (30).

First, we compared the effects of chemotherapy versus OVX-induced bone loss in a cohort of 6-week-old, wild-type, female FVB/NJ mice. Subsets of the cohort were subjected to OVX or SHAM surgery, followed by once weekly treatments of vehicle or a biologically relevant dose of doxorubicin (5 mg/kg; refs. 31, 32) for 4 consecutive weeks (Fig. 1A). As expected, OVX mice gained weight throughout the time course, while doxorubicin-treated mice lost only a negligible amount of weight that did not correlate with bone loss (Fig. 1B). At endpoint, we assessed estrogen levels in mice by measuring uterine weight, which is used as a surrogate of biologically active estrogen levels (33). Mice undergoing OVX experienced a 73.76% loss in uterine weight compared with SHAM mice (Fig. 1C), while doxorubicin-treated mice lost 48.91%. When OVX and doxorubicin treatments were combined, the uterine weights were identical to uteri from OVX-treated mice demonstrating that OVX led to the lowest levels of estrogen.

We next used μCT to measure trabecular and cortical bone volumes in the femurs of mice in each experimental group. μCT analysis on the femoral metaphysis and mid-diaphysis of doxorubicin-treated mice revealed a significant decrease in trabecular bone volume (−70.41%) compared with vehicle-treated SHAM mice, as measured by the ratio of BV/TV (Fig. 1D) and corroborated by 3D reconstructions (Fig. 1E). Doxorubicin treatment decreased all bone parameters including trabecular bone thickness (Supplementary Fig. S1A), vBMD (Supplementary Fig. S1B), and trabecular bone number while increasing the separation between the trabecular bones (Supplementary Fig. S1C and S1D). In addition, the cortical thickness was decreased by 23.98% in the doxorubicin group compared with vehicle-treated SHAM mice (Fig. 1F).

Next, we compared the impact of doxorubicin versus estrogen deficiency alone on bone density. As expected, μCT analysis of the femoral metaphysis showed a reduction in trabecular bone volume in OVX mice (black symbols, −34.35%) relative to the SHAM group (blue symbols; Fig. 1D). However, doxorubicin-treated mice (purple symbols) showed a more striking decrease in trabecular bone volume relative to the OVX control group (black symbols), indicating that doxorubicin treatment induced bone loss beyond the impact of estrogen loss alone. A similar pattern of decrease in cortical bone volume was observed in OVX (−5.48%) and doxorubicin-treated mice (−23.98%), relative to the SHAM mice, respectively (Fig. 1F). Remarkably, even though the uterine weights (indicative of estrogen levels) of OVX mice were significantly lower than doxorubicin-treated mice (Fig. 1C), the latter group experienced far more bone loss. If estrogen deficiency were the sole perpetrator of bone loss following chemotherapy, we would expect to see the same degree of bone loss in OVX and doxorubicin-treated mice. These data suggest that chemotherapy-induced bone loss is due, at least in part, to an estrogen-independent mechanism.

To exclude strain-dependent effects, we performed the identical experiment in 6-week-old C57BL/6 mice and observed similar trabecular bone loss following chemotherapy (Supplementary Fig. S1E). Similar results were obtained when C57BL/6 mice were treated with a second commonly used chemotherapeutic agent, paclitaxel (Fig. 2A). μCT analysis of the trabecular bone volume showed a striking decrease (−41.6%) in paclitaxel-treated mice relative to age-matched controls (Fig. 2B and C) with associated changes in trabecular number, separation, thickness, and vBMD (Supplementary Fig. S2A–S2D). Paclitaxel treatment was also associated with a reduction in uterine weight (Supplementary Fig. S2E). Finally, because estrogen is a chief regulator of bone metabolism in both males and females (7, 34–38), we treated male mice with 4 weekly doses of doxorubicin (5 mg/kg). Chemotherapy induced a significant decrease (−62.59%) in trabecular bone volume (Fig. 2D) and a decrease (−6.59%) in cortical bone volume (Fig. 2E) in doxorubicin-treated male mice relative to vehicle-treated mice, similar to the changes observed in female mice.

To create a clinically relevant scenario of postmenopausal chemotherapy treatment where estrogen levels are already low, we performed OVX in 6-week-old FVB/NJ mice, and then measured bone mass (measured as BV/TV) longitudinally in the same mice at 4 and 6 weeks post-OVX by in vivo imaging (VivaCT; Supplementary Fig. S3A). We found that bone loss ceased by 4 weeks post-OVX and the same BV/TV was maintained at 6 weeks (Supplementary Fig. S3B). To determine whether chemotherapy induced additional bone loss in the postmenopausal scenario, 7 weeks post-OVX, mice were treated with vehicle or doxorubicin once weekly for 4 weeks (Fig. 2F). At the end of chemotherapy treatment, bone volume changes were evaluated via a single μCT scan. As observed in postmenopausal women treated with chemotherapy, doxorubicin-treated mice lost more trabecular bone volume relative to vehicle-treated mice over the 4-week period after OVX-induced bone loss (Fig. 2G). The mice also lost cortical bone (Fig. 2H). This demonstrates that chemotherapy induces bone loss in a low-estrogen setting and establishes the clinical relevance of our model.

To ask whether bone loss could be induced in an estrogen-sufficient setting, we turned to radiation, which induces marked bone loss in humans and mice (39–41). To avoid the ovaries, we used single-limb irradiation and shielded the remainder of the animal (Fig. 2I). We selectively irradiated the right hind limb of 6-week-old female FVB/NJ mice with a single dose of 20 Gy (40, 41). Uterine weights were measured at endpoint to ensure that the radiation had no impact on estrogen levels. The uterine weights from irradiated mice were stable and comparable with untreated, wild-type 10-week-old female FVB/NJ mice (Supplementary Fig. S4A). μCT analysis of the femurs revealed a significant decrease in trabecular bone volume (−46.28%) in the radiated limbs compared with the contralateral control limbs (Fig. 2J). As expected, other measures of bone loss including changes in trabecular vBMD, trabecular thickness and number, and increased separation of trabecular bone, were also observed only in the irradiated femurs (Supplementary Fig. S4B–S4F).

Both chemotherapy and radiation can induce robust senescence, and recent work demonstrated that age-related increases in senescent cells drive bone loss (25). Thus, we hypothesized that senescence contributes to chemotherapy- and radiation-induced bone loss. To assess the presence of senescent cells in bone resident cells, we isolated the irradiated and contralateral tibias from mice subjected to 20 Gy single-limb radiation, flushed the bone marrow, and extracted RNA from bone resident cells and performed qRT-PCR. To assess for senescence-related alterations we assayed for changes in the senescence markers p16INK4a (p16) and Hmgb1 and senescence-associated secretory phenotype (SASP) factors. Upon qRT-PCR analysis, we observed a 3.4-fold increase in p16 levels and a 0.71-fold drop in Hmgb1 in the irradiated tibias relative to the nonirradiated contralateral tibias (Fig. 3). These changes correlated with increased SASP expression including Il6 (1.4-fold), Ccl3 (1.6-fold), Mcsf (2.0-fold), and Mmp12 (2.2-fold; Fig. 3). We next asked whether there was similar activation in the bones of mice treated with chemotherapy. We isolated tibias from 6-week-old FVB/NJ mice at 48 hours or 7 days following a single dose of vehicle or doxorubicin (25 or 10 mg/kg). As above, bone marrow was removed from the tibias and RNA was extracted from bone resident cells. Similar to irradiated bones, we observed a significant upregulation of p16 (2.46-fold) and a concordant decrease in Hmgb1 (0.78-fold) in the bones from doxorubicin-treated mice compared with vehicle (Fig. 4A). These changes were accompanied by increases in SASP factors including Il6 (7.5-fold), Ccl3 (1.7-fold), Mcsf (1.8-fold), and Mmp12 (2.5-fold) in the tibias from doxorubicin-treated mice (Fig. 4A). Importantly, we failed to observe increases in these factors in mice subject to OVX (Supplementary Fig. S5), demonstrating that estrogen loss is not sufficient to induce senescence in the bone.

To determine which cells were senescing within the bones of chemotherapy-treated mice, we isolated femurs and stained vehicle- and doxorubicin-treated bones for SA-β-Gal, a well-established marker of senescence (42–44). We found SA-β-Gal increased by approximately 40% in femurs from doxorubicin-treated mice relative to vehicle-treated mice (Fig. 4B). Upon histologic analysis, we found that SA-β-Gal–positive cells were present in the marrow space, as well as along the trabecular bone of chemotherapy-treated mice. To identify which cells were senescing in the bone following chemotherapy, we isolated cells by crushing whole bones and digesting them with collagenase to release bone resident cells. Following digestion, we used MACS CD45 beads to separate CD45⁺ and CD45⁻ cells. The CD45⁺ fraction was further stained with antibodies against CD3e, CD19, and CD11b to isolate T cells, B cells, and myeloid cells, respectively. We also stained cells with antibodies against Sca-I and cKit to isolate hematopoietic stem cells (HSC). The CD45⁻ fraction was further stained with anti-CD31 to separate endothelial cells from bone resident cells (gating strategy can be found in Supplementary Fig. S6A and S6B). Following cell sorting, we performed qRT-PCR to first confirm the identity of our sorted populations. As shown in Supplementary Fig. S6C, we successfully isolated populations as indicated in our flow sorting. To demonstrate the induction of senescence in the different populations, we assessed p16, p21, Hmgb1, and SASP expression by qRT-PCR. We found that only the CD45⁻CD31⁻ population showed a significant increase in the senescent marker P16 (Fig. 4C). Consistent with the increased P16 levels, we found that the CD45⁻CD31⁻ population expressed increased levels of p21 and decreased levels Hmgb1 (Fig. 4D; Supplementary Fig. S7). Finally, the CD45⁻CD31⁻ population also expressed a significant increase in SASP factors including Il6 (Fig. 4C), Igfbp4, Cxcl2, and Mmp12 (Fig. 4D; Supplementary Fig. S7). Together this work suggests that the only population robustly induced to senesce was the bone resident cells but not CD31⁺ endothelial cells or CD45⁺ cells.

Having found evidence of senescence in the bones following irradiation and chemotherapy, next we asked whether senescence drove bone loss in these scenarios. To establish a model that minimized the impact of estrogen on bone loss and mitigated any effects on bone development, we subjected 16-week-old C57BL/6 mice to SHAM or OVX surgery. In agreement with earlier work (45, 46), we found that over the 4-week time course (Supplementary Fig. S8A), 16-week-old mice failed to lose bone following OVX surgery (Supplementary Fig. S8C–S8H) despite being deprived of estrogen, as evidenced by a significant reduction in uterine weight (Supplementary Fig. S8B). It should be noted that after 3 months, OVX-treated mice maintained low estrogen levels as evidenced by reduced uterine weight (Supplementary Fig. S8I) and eventually lost cortical bone but not trabecular bone (Supplementary Fig. S8J and S8K). This demonstrated that estrogen contributes to bone homeostasis as reported previously (30), just not within the 4-week time course under study. In contrast, treating 16-week-old C57BL/6 mice with four weekly doses of doxorubicin (Supplementary Fig. S8B–S8H) or subjecting these mice to single-limb irradiation (Supplementary Fig. S9A and S9B) resulted in robust bone loss in the same 4-week time course. Together, these studies establish that, unlike estrogen loss alone, chemotherapy and radiation induce bone loss within 4 weeks in a fully mature 16-week-old mice.

To establish a role for senescence in chemotherapy-induced bone loss with minimal effects of estrogen, we turned to the INK-ATTAC transgenic mouse model (47, 48). In this model, the p16INK4a minimal promoter drives expression of a conditional suicide gene that kills senescent cells upon administration of the dimerization drug AP20187 (AP). Using this model, it was shown that senescent cells drive age-related bone loss (25) while having no impact on OVX-driven bone loss (30). We postulated that if senescence mediates chemotherapy-induced bone loss, then the clearance of senescent cells would protect against this bone loss. To test our hypothesis, we treated 16-week-old female INK-ATTAC (C57BL/6) transgenic mice with vehicle or a single dose of 10 mg/kg doxorubicin. These mice were then randomized to receive vehicle or three doses of AP20187 (2 mg/kg) and sacrificed 7 days post- doxorubicin injection (Fig. 5A), a time at which we observed increased senescence in the bones of doxorubicin-treated wild-type animals (Fig. 4B). We confirmed that AP20187 could reduce doxorubicin-induced senescence in the bones of INK-ATTAC mice, as illustrated by a decrease in p16 (0.47-fold) and Il6 levels (0.41-fold), and a concordant increase of Hmgb1 (1.33-fold) levels (Fig. 5B). Next, we extended the experimental timeline to 4 weeks, wherein the mice received four weekly doses of doxorubicin (5 mg/kg) to induce senescence and four cycles (twice weekly) of AP20187 (2 mg/kg) to kill senescent cells (Fig. 5C). At 4 weeks, mice were sacrificed, and we analyzed visceral fat for evidence of senescence and found, as expected, that doxorubicin treatment (purple symbols) increased SA-β-Gal, but we failed to see a rescue of senescence in fat upon AP20187 treatment (gray symbols; Supplementary Fig. S10A and S10B). Strikingly, μCT analysis of femoral metaphysis revealed that mice receiving the combination treatment (DOXO+AP, gray) had significantly higher trabecular bone volume (+75.84% increase) compared with mice receiving doxorubicin alone (purple symbols; Fig. 5D), and a similar trend was noted in the cortical bone (Fig. 5E). Thus, despite the fact that we failed to reduce senescence in the fat following doxorubicin treatment, we did rescue the bone loss in this model.

Bone homeostasis is maintained by a balanced coupling of bone formation and resorption, and investigation into age-related bone loss suggested that senescent cells within the bone led to an uncoupling of these activities (25). To ask whether a similar decoupling occurs following chemotherapy-induced senescence, we performed histomorphometric analysis on the femurs of mice treated with DOXO ± AP (Fig. 6A). Quantification of osteoclasts by TRAP staining (Fig. 6B) revealed a significant decrease in osteoclast-covered surface per bone surface (Oc.S/BS) and osteoclast number per bone surface (N.Oc/BS; Fig. 6C and D) in DOXO + AP–treated mice (gray symbols, surface = 0.041 ± 0.004 and number = 3.39 ± 0.46) relative to those administered doxorubicin alone (purple, symbols, surface = 0.07 ± 0.01 and number = 5.06 ± 0.47). To quantify osteoblast mineralization activity, mice were administered calcein and alizarin 7 days apart and bone formation was assessed at endpoint (Fig. 6E). Analyses of the femurs from mice treated with doxorubicin alone (purple symbols) revealed a BFR/BS and a MAR of 0.42 ± 0.06 and 0.86 ± 0.08, respectively (Fig. 6F and G). Elimination of senescent cells in doxorubicin-treated mice by treating with AP20187 (gray symbols) increased the BFR/BS to 0.87 ± 0.15 and MAR to 1.06 ± 0.10. These data show that chemotherapy-induced senescence drives bone loss by increasing osteoclast activity while simultaneously decreasing osteoblast activity. This agrees with our previous work demonstrating that senescent osteoblasts fail to mineralize bone and increase osteoclastogenesis (49).

Senescent cells express a wide array of factors collectively referred to as SASP factors that have wide-ranging effects on tissue homeostasis (50). We have shown that the p38MAPK-MK2 pathway regulates the expression of many SASP factors (18) and, in the setting of bone metastasis–specific inhibitors of these kinases, preserves bone density (51). Given our findings here that chemotherapy was sufficient to induce senescence, we next asked whether the p38MAPK-MK2 pathway was activated in the bones of treated mice. Indeed, we found that the p38MAPK pathway was activated in the bone following chemotherapy as indicated by the increased appearance of phosphorylated MK2 (Fig. 7A). Thus, we next asked whether targeting SASP factors would protect against chemotherapy-induced bone loss. To address this, we administered a p38MAPK inhibitor (p38i) or MAPKAPK2 pathway inhibitor (MK2Pi) ad libitum (compounded into chow) to mice treated with four weekly doxorubicin doses (5 mg/kg) and assessed bone density by μCT (Fig. 7B). These analyses revealed that the inhibitors had only minimal effects on estrogen levels as assessed by uterine weight (Supplementary Fig. S11A and S11B) but both p38i and MK2Pi treatment reduced SASP factors (Fig. 7C) and mitigated chemotherapy-induced bone loss (Fig. 7D and E). Indeed, we found that the inhibitors had no effect on trabecular bone volume compared with vehicle-treated controls, while doxorubicin treatment induced bone loss (−66.97%) as expected. Strikingly, mice that were administered p38i or MK2Pi in combination with doxorubicin had attenuated trabecular bone loss compared with mice receiving doxorubicin alone (Fig. 7D and E). The DOXO+p38i–treated group increased trabecular bone volume by 103.21%, while the bone in the DOXO+MK2Pi–treated group increased by 69.89%, relative to the doxorubicin only group (Fig. 7D and E). Interestingly, the protective effect seemed restricted to the trabecular compartment because we failed to observe significant differences between the doxorubicin-treated and DOXO+p38i or +MK2Pi treatment groups in the cortical compartment (Supplementary Fig. S11C and S11D). Together, these data suggest p38MAPK or MK2 inhibition are viable therapeutic strategies to protect against chemotherapy-induced bone loss.

Chemotherapy-induced bone loss has been attributed to estrogen deficiency resulting from ovarian failure (13, 17). Although several clinical studies have hinted that other mechanisms impact bone loss, they have remained largely unexplored. Our results challenge the view that chemotherapy-induced bone loss is primarily estrogen driven and demonstrate that senescence significantly contributes to chemotherapy-induced bone loss independent of estrogen. Furthermore, our findings that inhibition of p38MAPK or MK2, which target the SASP, rescues chemotherapy-induced bone loss suggest that they are viable therapeutic targets that could benefit any patient treated with chemotherapy.

Bone loss is associated with an increased risk of fracture that significantly impacts a survivor's quality of life. Currently, the available treatment options for patients suffering from chemotherapy-induced bone loss are bisphosphonates, denosumab, or hormone replacement (52), which can limit osteoclast differentiation and activity (53, 54). Unfortunately, estrogen and bisphosphonates are not without sequelae. Indeed, bisphosphonates can persist in the skeleton for long periods of time and can result in long-term complications including osteonecrosis of the jaw, hypocalcemia, and kidney toxicity (53, 55). Denosumab use can result in musculoskeletal pain, hypercholesterolemia, and cystitis and its discontinuation has been associated with a rapid decrease in bone density and increased vertebral fracture risk (56). Furthermore, because osteoclast and osteoblast functions are tightly coupled, by killing osteoclasts bisphosphonates ultimately limit osteoblast activity, which can negatively impact bone quality. Estrogens, on the other hand, are not used in cancers that respond to hormones, including estrogen receptor–positive breast cancer (∼70% of patients with breast cancer) or ovarian cancer. Given there is an exponential relationship between bone loss and increased fracture risk (6), the extensive bone loss observed in early stage, nonmetastatic patients argues that safe and potent inhibitors of bone loss are needed. Our data demonstrate that targeting senescent cells and their downstream effector molecules, p38MAPK or MK2 preserves bone in the face of chemotherapy. These data combined with our recent findings that inhibiting p38MAPK and MK2 can significantly reduce metastatic breast cancer growth (51) further suggests that in the metastatic cancer setting, inhibition of these key signaling pathways will have added benefit for patients treated with chemotherapy.

S.A. Stewart reports receiving a commercial research grant from Aclaris. J. Monahan is an executive vice president R&D (paid consultant) at Aclaris Therapeutics, Inc. D. Baker holds ownership interest (including patents) in Unity Biotechnology. No potential conflicts of interest were disclosed by the other authors.

Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the Department of Defense. This publication is solely the responsibility of the authors and does not necessarily represent the official view of National Center for Research Resources or the NIH.

Conception and design: Z. Yao, B. Murali, R. Faccio, S.A. Stewart

Development of methodology: Z. Yao, B. Murali, X. Luo, J.M. van Deursen, D. Baker, J.K. Schwarz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Yao, B. Murali, J. Monahan, J.M. van Deursen, D. Baker, J.K. Schwarz, S.A. Stewart

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Yao, S.A. Stewart

Writing, review, and/or revision of the manuscript: Z. Yao, B. Murali, D. Thotala, J. Monahan, D. Baker, J.K. Schwarz, S.A. Stewart

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Murali, X. Luo, D.V. Faget, T. Cole, B. Ricci, D. Thotala, S.A. Stewart

Study supervision: B. Murali

Other (provided technical support and important feedback): Q. Ren

This work was supported by the Cancer Biology Pathway Molecular Oncology Training Grant NIH T32CA113275 (to B. Murali), NIH grants R01 CA130919 (to S.A. Stewart), and R01 CA181745 (to J.K. Schwarz), and an American Cancer Society Research Scholar Award (to S.A. Stewart). The U.S. Army Medical Research Acquisition Activity, 820 Chandler Street, Fort Detrick, MD 21702-5014, is the awarding and administrating acquisition office, and this was supported in part by the Office of the Assistant Secretary of Defense for Health Affairs, through the Breast Cancer Research Program, under award No. W81XWH-16-1-0728. This work was also supported by the Siteman Cancer Center Investment Program (NCI Cancer Center Support Grant P30CA091842, Fashion Footwear Association of New York, and the Foundation for Barnes-Jewish Hospital Cancer Frontier Fund to S.A. Stewart), a Siteman Cancer Center grant to R. Faccio, NIH grants R01 AR053628 and R01 AR066551 to R. Faccio, and Shriners Hospital grant 85100 to R. Faccio, and J.K. Schwarz is supported by R01 CA181745. Finally, the Glenn Foundation for Medical Research (to J.M. van Deursen and D. Baker) and the NIH (AG057493 to J.M. van Deursen and AG053229 to D. Baker) also supported this work. We thank Deborah Veis and Matthew Silva for their valuable suggestions. We thank Crystal Idleburg and Samantha Coleman at the Musculoskeletal Histology core for their expert technical assistance with bone tissue sectioning and staining. In addition, we thank Daniel Leib, Yung Kim, and Michael Brodt in the Structure and Strength Musculoskeletal core for μCT imaging; Washington University Musculoskeletal Research Center (NIH P30 AR074992). Finally, we thank Lorry Blath and Judy Johnson for their constant support, enthusiasm, and critical assessment of our work and its impact on patients with breast cancer. Imaging and analysis of bone biopsy slides were performed using Zeiss Axio ScanZ.1 at the Washington University Center for Cellular Imaging, which is supported by Washington University School of Medicine, The Children's Discovery Institute of Washington University and St. Louis Children's Hospital (CDI-CORE-2015-505), and the National Institute for Neurological Disorders and Stroke (NS086741). The Centers are partially supported by NCI Cancer Center Support Grant #P30 CA91842 to the Siteman Cancer Center. Finally, we thank Melissa Stauffer of Scientific Editing Solutions for editing the article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.