Moderate tibial loading and treadmill running, but not overloading, protect adult murine bone from destruction by metastasized breast cancer

Abstract

Osteolytic bone lesions, which develop in many metastatic breast cancer patients, impair bone integrity and lead to adverse skeletal related events that are difficult to treat and sometimes fatal. Moderate mechanical loading has been shown to suppress osteolysis in young mice with breast cancer. In this study, we aimed to investigate the dose-dependent effects of mechanical loading on protecting the integrity of adult skeletons with breast cancer. Localized tibial loading and aerobic treadmill running with three levels of varying intensity were tested in a syngeneic mammary tumor bone metastasis model. Adult C57BL/6J female mice (14-week-old, N = 88 mice) received intra-tibial injections of Py8119 triple-negative murine breast cancer cells or PBS and underwent 4 to 5 weeks of exercise or acted as sedentary/non-loaded controls. The bone structure was monitored longitudinally with weekly in vivo micro-computed tomography imaging, while the cellular responses in bone and marrow were examined using immunohistochemistry. Moderate treadmill running (16 m/min, 50 min/day, 5 days/week, and 5 weeks) and tibial loading (4.5 N, 630 με, 4 Hz, 300 cycle/day, 5 days/week, and 4 weeks) suppressed tumor-induced bone destruction, as evaluated by full-thickness perforation of tibial cortex and the volume of osteolytic lesions in the cortex. In contrast, tibial loading at higher magnitude (8 N, 1100 με) induced woven bone and accelerated bone destruction, compared with the sedentary and non-loaded controls. The three exercise regimens differentially affected osteocyte apoptosis, osteocyte hypoxia, osteoclast activity, bone marrow vasculature, and tumor proliferation. In conclusion, the relationship between exercise intensity and the risk of breast cancer-induced osteolysis was found to follow a J-shaped curve in a preclinical model, suggesting the need to optimize exercise parameters in order to harness the skeletal benefits of exercise in metastatic breast cancers.

Graphical abstract

1. Introduction

Breast cancer is the most common cancer among women in the United States [1]. Although the five-year survival rate is as high as 90% owing to early screening and progress in treatments, it decreases dramatically (30%) once the tumor metastasizes to distant sites including bone [1]. Osteolytic bone lesions, driven by the “vicious cycle” between tumors and osteoclasts in the bone environment [2, 3], develop in many metastasized breast cancer patients. During the disease course, these patients can suffer more than two adverse skeletal related events (SREs) including bone pain, spinal cord compression, and pathologic fractures [4]. Anti-resorptive agents such as bisphosphonates are used in treating cancer bone metastasis [2], but severe events such as osteonecrosis of the jaw have been reported after high-dose use [5]. Thus, safe, adjuvant strategies are needed to protect skeletal integrity in metastatic breast cancer patients.

Clinical and epidemiological studies have demonstrated that physical activity improves skeletal health and reduces cancer risk. Mechanical stimuli associated with locomotion and exercise increase bone strength [6, 7], reduce hip fracture risk for all ages including the elderly [8], and reduce risk of breast cancer [9]. A 2012 review [10] of 27 observational studies concluded that physical activity reduces all-cause and breast cancer-specific mortality. Even for metastatic cancer patients, physical activity has been demonstrated to be safe and beneficial [11]. Still it remains a challenge to choose exercise parameters regarding the type (strengthening vs. aerobic training), intensity (moderate vs. intensive), frequency, and duration to ensure effectiveness and safety [10, 12].

To understand the mechanisms by which mechanical loading regulates skeletal integrity in breast cancer, murine xenograft models have been used and subjected to well-controlled exercise. Lynch et al. (2013) demonstrated that a cyclic compression regimen (4.1 N peak load, ~600 με surface strain, 4 Hz, 5 min per day, 5 days per week, and 6 weeks) successfully inhibited osteolysis and breast cancer growth in tibiae of immune-compromised (SCID) mice injected with human MDA-MB-231 breast cancer cells [13]. The protection was attributed mainly to downregulation of osteoclast activity and promotion of osteogenesis induced by mechanical loading [13]. Similarly, after 2-week tibial loading (1 N peak load, ~260 με surface strain, 2 Hz, 5 min daily), suppressed osteolysis in the trabecular bone compartment was confirmed by Fan et al. (2020) in immune-competent C57BL/6 and BALB/c mice inoculated with murine mammary tumor EO771 and 4T1.2 cells, respectively [14]. Interestingly, these authors discovered that the skeletal protective effects were abolished as the loading magnitude increased to 2 N and osteolysis worsened under 5 N loading, possibly due to altered TGF-beta signaling and osteopontin secretion in osteocytes [14], the primary mechano-sensitive cells in bone [15]. These excellent studies reveal the magnitude-dependent loading effects on breast cancer bone metastasis. However, the age of animals in these studies was < 8 weeks old with growing skeletons [16, 17], while human breast cancer patients are adults with mature skeletons [1]. Furthermore, while tibial loading allowed application of well-controlled loading on a specific bone in anesthetized animals, it did not accurately simulate physiological exercises involving the whole body, such as running. Although treadmill running has been tested in mice and found to inhibit breast cancer growth [18, 19], its effect on skeletal integrity in metastasized breast cancer has yet to be determined.

The present work aimed to investigate how mechanical loading regulates the structural integrity of adult murine skeletons in metastatic breast cancer. Localized loading of one bone (tibial loading) and whole body exercise (treadmill running) with three levels of varying loading intensity were tested in a syngeneic mammary tumor bone metastases model. The immune system plays important roles in not only tumor growth but also bone adaptation to mechanical loading [20]. The bone structure was monitored longitudinally using in vivo micro-computational tomography (μCT). Our hypothesis was that mechanical stimulation regulated skeletal integrity in adult mice bearing breast cancer in an intensity-dependent manner.

2. Methods

2.1. Animals

Female C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME, USA) were housed at 4-5 mice per cage with free access to water and food under standard conditions in an accredited animal facility at the University of Delaware. All animal experiment protocols were approved by The Institutional Animal Care and Use Committee (IACUC).

2.2. Tumor cells and inoculation in mice

Mus musculus mesenchymal-like triple-negative Py8119 cells (ATCC, Manassas, VA, USA, CRL-3278™) were maintained and expanded in F-12K cell culture medium (ATCC, 30-2004™) supplemented with 5% FBS (Corning™ 35015CV) at 37°C degree and 5% (v/v) CO₂ in an incubator. At 14 weeks of age, the mice, anesthetized under 2-4% (v/v) isoflurane, received Nair^® hair removal gel, disinfectant iodine, and 70% alcohol on both knees. With the knee flexed, 1000 tumor cells suspended in 20 μl PBS were injected through intact skin and the tibial end into the marrow cavity (~1 mm below growth plate) using a syringe (BD^® U-100 Syringes Micro-Fine 28G-1/2"). The fine needles were chosen to minimize tissue damage from injection because of their small outer diameter (0.36 mm). One hour prior to intratibial injection, mice received one dose of buprenorphine (0.05 mg/kg body weight) intraperitoneally. Similarly, control mice received an injection of 20 μl PBS. We chose to inject both tibiae to reduce the total number of animals used in the study. During the 4-5 weeks of study duration, the small number of tumor cells injected per tibia did not result in tumor metastasis from bone to other organs such as the lungs. Injecting both tibiae and exposing them to similar levels of tumor-derived factors allowed us to apply loading to one tibia and use the contralateral tibia as the non-loaded control. This design was expected to increase the statistical power to detect the loading effects.

2.3. Tibial compliance measurement via strain gauging

To determine the loading magnitude used in our in vivo experiments, the compliance of murine tibia was measured using strain gauging as previously published by us [21] and others [13]. Two 14-week-old female B6 mice were sacrificed, and their tibiae were exposed. While still being attached to the body, the tibia was positioned axially in a Bose LM1 TestBench® loading system (Bose Corporation, Framingham, MA) (Fig. 1A). A single-element gauge (EA-06-015DJ-120; Measurements Group, Inc., Raleigh, NC) was fixed on the relatively flat anterior-medial surface (~35% distal from the proximal end of the tibia, Fig. 1B). The tibia was axially compressed with a gradually increasing load (0.4 N/s to 10 N), while the voltage outputs from the strain gauge were recorded using the LM1's data acquisition unit and converted to strain values using a calibrated conversion coefficient. From the strain vs. load curves, the compliance of the axially loaded tibiae (n = 4) was measured as 139.3 ± 8.5 με/N. Thus, we chose 4.5 N (~630 με) and 8 N (~1100 με) for our in vivo tibial loading experiments. The former value was similar to that used by Lynch et al. (~600 με) [13]. The higher load (8 N) was found to induce bone formation in mid-shafts [22] and activate anabolic mechanotransduction pathways in adult mice [23].

Fig. 1.

Study design and methodology. Two types of physical activity of different intensities were examined. (A) Tibial loading setup. (B) Tibial compliance (139.3 ± 8.5 με/N) was measured on four tibiae of two C57BL/6J female mice (14 weeks of age), to which a strain gauge was attached to the anterior-medial surface at 35% distal from the proximal end. A loading magnitude of 4.5 or 8 N was expected to induce moderate level (~630 με) or higher level of strain (~1100 με). (C) Treadmill running setup. (D) Experimental groups and tibial loading and treadmill running regimens. (E) Micro CT imaging and analysis of the VOI allowed detection of voids in the bone cortex including the normal nutrient foramen and tumor-induced lesions (perforation). Immunohistochemistry was performed on samples collected at varied time points.

2.4. In vivo tibial loading

PBS- or tumor-injected mice underwent axial tibial loading with either 8 N or 4.5 N peak load one day after the intratibial injection (Fig. 1D). Five PBS-injected and 20 tumor-injected mice were used in the 8N loading experiment, and 5 PBS-injected and 24 tumor-injected mice used in the 4.5N loading experiment. Sample sizes of 5, 20 or 24 for the tumor-injected group were predicted to detect effect sizes of 1.7, 0.7, or 0.6, respectively, from loading at 80% power and 0.05 significance level (Student’s t test for dependent groups, G*Power software). Under isoflurane anesthesia, the left tibia was subjected to cyclic uniaxial compression (Fig. 1A) using our published protocol with slight modifications [24]. The loading was applied at either 4.5 or 8N peak load, 4 Hz, 300 cycles per day, 5 days/week for a duration of 4 weeks. We reduced the number of loading cycles from 1200 to 300 to minimize potential knee damage induced by long-term loading [25], which was shown to preserve anabolic effects on bone [26]. To alleviate pain and stress, animals received one dose of buprenorphine (0.05 mg/kg body weight) one hour prior to the first loading session each week. To monitor bone structure, mice underwent weekly in vivo μCT imaging on the days when no loading was scheduled. To examine cellular changes, 2 PBS-injected mice and 3 tumor-injected mice were sacrificed at Week 2. Mice were weighted weekly and their stress level was assessed by mobility, posture, and social behaviors. Among the tumor-injected mice, one mouse in the 8 N loading group experienced a bone fracture on Week 4 and was sacrificed ahead of the scheduled time. Micro CT data prior to facture were included in the analysis.

2.5. Moderate treadmill running

Thirty four mice were divided into 4 groups as PBS nonrunners (4 mice, 8 tibiae), PBS runners (4 mice, 8 tibiae), tumor nonrunners (13 mice, 26 tibiae), and tumor runners (13 mice, 26 tibiae). Sample sizes of 8 or 26 were expected to detect effect sizes of 1.5 or 0.8, respectively, from running at 80% power and 0.05 significance level (Student’s t test for independent groups, G*Power software). A customized plastic divider was placed on top of the running belt of a speed-adjustable treadmill (TF100 ERPEUTIC®, Paradigm Health and Wellness, Inc. City of Industry, CA) to allow eight mice to run simultaneously in individual lanes (Fig. 1C). Prior to the experiments, all mice went through a two-week training. For 3 days per week for 2 weeks, the mice ran at a speed of 10.7 m/min with the duration gradually increasing from 10 to 40 min. If a mouse lagged and stayed within the yellow taped zone at the end of the running lane, a gentle push with a soft brush would encourage it to run forward (Fig. 1C). After three reminders within 5 min, the mouse was deemed exhausted and removed from running, and the running session was marked as incomplete. Most mice (22/34) could complete the 40 min running at the end of the training period. Ranking the mice by increasing number of incomplete training sessions, the top 50% were assigned as runners and the rest as nonrunners. At 14 weeks of age, the runners and nonrunners were randomly assigned to receive intratibial injections of either PBS or tumor cells at the start of the experiment (day 0, Fig. 1D). Starting from Day 1, runners were allowed to run on the treadmill 5 days per week for 5 weeks. Each running session was 50 min long, consisting of a 5-min warmup at 10.7 m/min, 40-min at 16.1 m/min, and 5-min cool-down at 10.7 m/min. The mechanical strain induced by this running protocol was expected to be no greater than that induced by 4.5 N tibial loading (630 με). A previous study using strain gauges reported only ~200 and ~400 με tensile strains on the tibial mid-shaft of 12-week-old C57BL/6J females during locomotion and a 30 cm jump, respectively [27]. The speed in our running protocol was lower than that typically used for fatigue testing [28]. To avoid overloading, running mice in our experiments were allowed to stop after showing signs of exhaustion. PBS-injected mice were able to complete all running sessions during the 5-week period. From Week 1 to Week 5, the percentage of tumor-injected mice that completed the running sessions was 100%, 100%, 90%, 84%, and 64%, and the average running time was 50, 50, 48.2, 46.7, and 42.3 min per session, respectively. Nonrunners were placed in a cage near the treadmill, experiencing similar levels of noises and handling. The mice underwent in vivo μCT imaging weekly on the days when no running session was scheduled (Fig. 1D). Stress level was monitored as described above and two mice were sacrificed ahead of schedule due to excessive tumor growth. Randomly selected mice were also sacrificed at Week 1 (1 PBS nonrunner, 1 PBS runner, 3 Tumor nonrunners, and 3 Tumor runners) and Week 3 (1 PBS nonrunner, 1 PBS runner, 5 Tumor nonrunners, and 5 Tumor runners) for immunohistochemistry analysis (detailed below).

2.6. In vivo μCT scanning and data analysis

The microstructure of mouse tibiae was monitored by weekly in vivo μCT scanning using SkyScan^® 1276 (Bruker, Kontich, Belgium). Baseline scanning prior to intratibial injection (termed Week 0) was performed on mice in the 4.5 N loading and treadmill running experiments, while the first scanning for the 8 N loading group was on Week 1. For each scan, mice were anesthetized with 3% (v/v) isoflurane and held in the built-in holder, which was rotated 180-degree with a step of 0.8 degrees. One frame was taken per step with the following settings: 900 ms exposure time, X-ray of 200 mA current and 50kVp, and a 0.5mm Al filter. The in vivo scanning lasted only 4-5 min per animal and the calculated radiation exposure (<600 mGy) was low, which did not affect bone property and tumor growth [29]. The 3D reconstruction of the murine hind limbs was obtained using the NRecon^® software (Bruker) with a voxel size of 7 μm. The left and right tibiae were separated into individual volumes using Python codes (available upon request). The first scan of each tibia was aligned, where its long axis was positioned vertically and the anterior-posterior and medial-lateral axes were arranged orthogonally. This baseline scan served as the image registration reference for subsequent weekly scans using the SimpleITK package in Python [30]. Our volume of interest (VOI) consisted of 300 slices (2.1 mm) of bone metaphysis below the growth plate at the proximal tibia (Fig. 1E), which was generated using 3D Slicer^® [31]. This VOI was chosen because it contained nearly all bone perforation found during the study period. Cortical and trabecular bones were segmented using CT Analyzer 1.17^© (Bruker). A global threshold value (73/256) was applied and the processed images were found to agree well with the gray-scaled images. Cortical polar moment inertia (Ct.pMOI), bone mineral density (Ct.BMD), and tissue mineral density (Ct.TMD) as well as trabecular bone volume fraction (Tb.BV/TV), thickness (Tb.Th), separation (Tb.Sp), bone mineral density (Tb.BMD), and tissue mineral density (Tb.TMD) were reported using CTan^® 3D analysis software. By overlaying the subsequent scans with the first available scan (Week 0 or Week 1) of the same tibia, which was assumed to be intact, we could identify the osteolytic lesions (Fig. 1E), where bone tissue in the initial scan was replaced with non-bone tissues (low x-ray attenuation), and quantify the volume of the osteolytic lesions as Ct. VoL. The tibia was also classified into one of two classes (“intact” vs. “perforated”) based on the following criteria: (1) there was one or more “penetrating holes” (i.e., osteolytic lesion through the full thickness of the cortex), and (2) the linear dimension of the lesion was greater than 35 μm/5 pixels (i.e., the size of nutrient foramen, Fig. 1E). Using the cortex perforation as a general read-out of bone integrity, the Kaplan-Meier survival curve was calculated for the probability of tumor-injected tibiae free of perforation with or without exposure of exercise using the free software R (ggplot2-package, version 3.5.2, R Foundation) [32].

2.7. Immunohistochemistry (IHC) staining and quantification

To understand the cellular responses to mechanical loading during the cancer-induced osteolysis process, immunohistochemistry and histological assays were performed to examine the activities of osteocytes, osteoclasts, tumors, and marrow vasculature. For the 8N tibial loading study, since periosteal woven bone formation occurred more likely prior to perforation (detailed in Results section) and elevated hypoxia expression peaked at day 7 in fatigue loading-induced woven bone formation [33], we examined HIF-1α staining in cortical osteocytes and periosteum at Week 1. Matrix metallopeptidase 9, 14 (MMP-9, MMP-14), and Gm-CSF (granulocyte macrophage–colony stimulating factor) were also examined at Week 1 because of their roles in regulating bone remodeling, osteoclasts, and osteolytic bone metastasis [34, 35]. At Week 4, functional outcomes such as osteoclast activity and tumor proliferation were examined using TRAP and Ki67 staining [36], respectively. For the 4.5 N loading study, staining of HIF-1a, TRAP, and Ki67 was performed as in the 8N loading study with a slight modification. The assays were performed at Week 2 when loading effects on cortex perforations started to show on in vivo μCT scans. Additionally, TUNEL and endomucin staining were performed because previous studies have shown that osteocyte apoptosis regulated osteoclast activity during mechanical loading [37], that endothelial cells was essential for mechanically stimulated osteocytes in regulation of breast cancer migration [38], and that endomucin positive H-type vasculature was critical for bone homeostasis [39]. For the five-week treadmill running study, we performed TRAP, Ki67, TUNEL, endomucin. HIF-1a, Gm-CSF staining at Week 3, when running-induced changes of functional outcomes emerged but the bone mass was not overtly lost. Additionally, at Week 3 LOX (lysyl loxidase) expression in bone marrow was examined due to its role in driving tumor growth, osteoclastogenesis, and bone metastases [40]. Similar to the 8 N loading experiments, HIF-1a expression in osteocytes was also examined at Week 1 as an indicator of early response in the presence of tumor.

Sample preparations for the IHC and histology assays were as follows. Both hind limbs were removed from the body without disturbing the surrounding soft tissues. The samples were fixed in 4% (w/v) paraformaldehyde (PFA) for 48 hours, decalcified in 14% (w/v) ethylenediaminetetraacetic acid (EDTA, PH 7.4) for 3 weeks at 4°C, and embedded in paraffin. Sequential sagittal sections (thickness of 5 μm) were collected, cleaned of paraffin, and rehydrated through xylene (2 x 10 min), ethanol (100 %, 95%, 70%, and 50%), and deionized water (5 min each). Antigen retrieval was performed by incubating the slides in a citric acid buffer at 63°C for 16 hours. Rehydrated sections were blocked for one hour with 2.5% (v/v) serum that matched the host species of secondary antibodies and then incubated with primary antibodies at 4°C for overnight. The following primary antibodies (in alphabetical order) were used with dilutions according to the vendors’ suggestion: endomucin (1:500, Invitrogen, eBioV.7C7), Gm-CSF (1:200, Proteintech, 17762-1-AP), HIF-1α (1:200, Invitrogen, PA1-16601), Ki67 (1:500, Invitrogen, SolA15), lysyl oxidase (1:200, Invitrogen, PA1-16953), MMP-9 (1:200, Invitrogen, PA5-79689), MMP-14 (1:50, Invitrogen, PA5-13183). Secondary antibodies with a DAB detection kit (IMMPRESS HRP horse anti-rabbit and goat anti-rat detection reagents) were purchased from Vector Laboratories (Burlingame, CA). In addition, a TUNEL detection kit (VitroVivo Biotech, Rockville, MD) and a TRAP staining kit (Sigma-Aldrich) were used according to the manufacturers’ protocols, with hematoxylin and fast green as counter stains, respectively.

Two sections 100 μm apart were stained per assay and quantified for each bone sample. The metaphyseal bone and bone marrow below the growth plate up to 2.1 mm distally were captured with multiple images using a 20x objective. Quantification performed for each image was pooled for each section and the average from both sections was reported for each bone sample. Osteoclastic activity was quantified by the percentage of TRAP stained (+) bone surface over the total bone surface. Hypoxia and apoptosis in osteocytes were evaluated by the percentage of HIF-1α stained (+) or TUNEL stained (+) osteocytes over the total number of osteocytes, respectively. ImageJ (U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.nih.gov/ij/) was used for tracing and counting. The RGB images of Ki67 stained sections were smoothened by a Gaussian filter with a window size of five pixels, decomposed into the DAB/hematoxylin color space by color deconvolution, and then segmented by k-means clustering into three clusters—DAB nuclei (positive staining), hematoxylin nuclei (negative staining), and background—using custom Python codes. Tumor cell proliferation was quantified by the ratio of positive staining pixels to the total nuclear pixels (e.g., summation of positive and negative staining) in the Ki67 staining. Images of endomucin stained sections were processed similarly, and the density of the endomucin stained (+) blood vessels was evaluated by the ratio of the positive staining area to the total tissue area.

2.8. Statistical Analysis

JMP (JMP^® Pro 14, SAS Institute Inc., Cary, NC) was used for statistical analysis. For the μCT data, a mixed model was used to analyze the repeated (weekly) measures of μCT, with time and experimental groups being the fixed effects, in which random effects of individual animals were nested. A default Restricted Maximum Likelihood (REML) with unbounded variance components was chosen in JMP so that normality of the data was not required. Pairwise comparisons were performed using Tukey HSD post hoc tests. For histology and IHC at specific time points, we performed ANOVA and Tukey post hoc tests for multiple comparisons or Student’s t tests for two-group comparisons. Significance was indicated by different letters or marked with a black bar (-: p<0.05). Survival analysis of tibiae free from perforation was performed in the free software R (the R Foundation, Survival-package version 2.4.3) [41].

3. Results

3.1. Loading at 8 N did not prevent but accelerated the destruction of bone by breast cancer

Mechanical loading at the 8 N peak load did not protect bone integrity but instead accelerated bone destruction in the tumor groups (Fig. 2). As shown in the weekly 3D renderings of a mouse injected with tumor cells, the non-loaded tibia showed a perforated cortex and loss of trabeculae on Week 4, while bone destruction was manifested as early as Week 2 in the loaded tibia (Fig. 2A). In the cortical bone compartment, Ct.pMOI, a structural measure of bone rigidity, varied within groups and between time points (p < 0.0001 and p = 0.0016). It was notably increased by 8 N loading in the tumor groups since week 2 (p < 0.05, Fig. 2B). On the other hand, the trabecular bone volume fraction, Tb.BV/TV, decreased with time (p < 0.001), showed no overall difference among the four groups (p = 0.55), but increased by loading on Week 4 in the tumor groups (Fig. 2C). By performing imaging registration on the weekly scans, we detected and quantified the volume of the osteolytic lesions in the cortex (Ct.VoL). As expected, Ct.VoL in the PBS groups remained negligibly small due to the limited mismatch in registration. In contrast, Ct.VoL was elevated in the tumor groups (p = 0.018), which drove the temporal increase of the overall Ct.VoL (p = 0.006, Fig. 2D). Using the cortex perforation as a final outcome, the Kaplan-Meier survival curves showed that the probability of tibiae free from perforation was significantly lower in tumor-loaded vs. non-loaded groups (p = 0.043, Fig. 2E). The seemingly contradictory loading effects such as increased Ct.pMOI (Fig. 2B) vs. accelerated perforation (Fig. 2D) and decreased survival (Fig. 2E) prompted us to examine the bone cortex in detail. From the sequential scans, we noticed that 8 N loading induced periosteal woven bone formation (WvB) in both PBS and tumor groups, which occurred at a much higher rate in the tumor group (19/20 vs. 2/5, p = 0.02, Fisher’s exact test, Fig. 2F). The periosteal woven bone occurred on or after Week 2, which corresponded to the manifestation of elevated Ct.pMOI (Fig. 2B). It was also more likely to appear prior to (8/19) or at the same time of cortical perforation (9/19) than after perforation (2/19, Fig. 2F). Other μCT-derived bone parameters showed that the 8 N loading led to increased Ct.TMD in the tumor group and increased Tb.Th and Tb.TMD in both tumor and PBS groups, while no effects were observed in other parameters (Supplemental Fig. 1S). Detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Fig. 1S.

Fig. 2.

The 8 N tibial loading did not protect bone integrity, but worsened bone destruction in the presence of breast cancer. (A) Examples of 3D reconstructions: loading accelerated bone destruction by tumor. (B) Ct.pMOI varied within groups and between time points (p < 0.0001, p = 0.016 for the group or time effect, respectively). Ct.pMOI was increased by loading in tumor groups since Week 2. (C) Tb.BV/TV decreased with time but showed no group effect. It was increased at Week 4 in tumor-loaded vs. non-loaded groups. (D) Ct.VoL (volume of osteolytic lesions in cortex) increased with time and was elevated in the tumor-loaded group. (E) The Kaplan-Meier survival curves of tibiae free from perforation: survival was significantly lower in tumor-loaded vs. non-loaded groups. (F) Periosteal woven bone (WvB): representative images, incident rates, and temporal sequences relative to perforation (Perf). Statistical analysis: above the bar graphs, the effects of groups and times are shown with p-values and pairs with different letters are significantly different; loading-induced significant differences in PBS or tumor groups are indicated with black bars. Detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Supplementary Fig. 1S. (B-D): Weekly data were presented (left to right) for PBS-NL, PBS-L, Tumor-NL, and Tumor-L groups.

At the cellular level, relative to the non-loaded tibiae, 8 N loading resulted in increased HIF-1α+ osteocytes on Week 1 (~20% vs. ~12%, p = 0.009, Fig. 3A), increased TRAP+ bone surface on Week 4 (~57.3% vs. ~34.6%, p = 0.02, Fig. 3B), and no effect on Ki67 staining of the tumor mass on Week 4 in the tumor groups (Fig. 3C). Loading did not appear to alter the HIF-1α and TRAP staining in the PBS groups (Fig. 3A and 3B). Elevated HIF-1α, MMP-9 and MMP-14 staining was found on the periosteum in the tumor group just one week after 8N loading, suggesting that the presence of tumor cells promoted loading-induced periosteal reactions (Fig. 3D).

Fig. 3.

Immunohistochemistry changes in bone under 8 N tibial loading. (A) HIF-1a response of osteocytes in PBS- or tumor-injected bones. (B) TRAP staining. TRAP+ labeled bone surface (red dotted lines). TRAP− surface (green dotted lines). (C) Ki67 staining of tumor nuclei. (D) Early periosteal positive responses in HIF-1a, MMP-9, MMP-14, and GM-CSF were found in tumor-injected group. P = periosteum; B = bone.

3.2. Loading at 4.5 N suppressed and delayed bone destruction by breast cancer

Mechanical loading at 4.5 N protected bone integrity and delayed bone destruction in the presence of breast cancer (Fig. 4). 3D renderings from a mouse injected with tumor cells showed a perforated cortex and lost trabecular bone in the nonloaded tibia on Week 2, while the manifestation of bone destruction was delayed till Week 3 in the contralateral loaded tibia (Fig. 4A). In the cortical bone compartment, Ct.pMOI did not changed with time until Week 4 when it significantly decreased (p < 0.001); it was lower in the tumor groups (p = 0.01) and partially rescued by the 4.5 N loading on Week 4 (p < 0.05, Fig. 4B). Meanwhile, Tb.BV/TV consistently declined with time since Week 2 (p < 0.001), and it tended to be lower in tumor groups; however, no loading effect was observed over the four weeks within either the tumor or PBS groups (Fig. 4C). As for the volume of osteolytic lesions in the cortex, Ct.VoL was significantly higher in the non-loaded group (p = 0.0006) and increased with time in the tumor groups (p < 0.001), which was significantly suppressed by loading on Week 4 (Fig. 4E). The Kaplan-Meier survival curves showed that the probability of tibiae free from perforation was significantly higher in the tumor-loaded vs. non-loaded groups (p = 0.036). No periosteal woven bone formation was observed in either PBS or tumor-loaded tibiae (data not shown). Within either the PBS or tumor groups, the 4.5 N loading did not affect other μCT-derived bone parameters (e.g., Ct.BMD, Tb.BMD, Ct.TMD, Tb.TMD, Tb.Th, Tb.N, and Tb.Sp) and detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Supplemental Fig. 2S.

Fig. 4.

The 4.5 N tibial loading protected bone integrity and delayed bone destruction in the presence of breast cancer. (A) Examples of 3D reconstructions: loading delayed bone destruction by tumor cells. (B) Ct.pMOI did not change with time until Week 4 when it dropped; it decreased in the tumor groups relative to the PBS groups, and the decrease was partially rescued by loading. (C) Tb.BV/TV declined with time since Week 2 and no loading effect was detected overall within either tumor or PBS groups. (D) Ct.VoL was significantly higher in the tumor nonloaded group and increased with time in the tumor groups. Loading suppressed tumor lesion volume on Week 4. (E) The Kaplan-Meier survival curves of tibiae free from perforation: survival was significantly higher in tumor loaded vs. non-loaded groups. Statistical analysis: on the tope of bar graphs, the effects of groups and times are shown with p values and pairs with different letters are significant different; Loading induced significant difference in PBS or tumor groups is indicated with a black bar. Detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Supplementary Fig. 2S. (B-D): weekly data were presented (from left to right) for PBS-NL, PBS-L, Tumor-NL, and Tumor-L groups.

On the cell and tissue level, the 4.5 N loading appeared to counteract the adverse effects induced by tumor cells (e.g., altered osteocyte hypoxia, marrow vasculature, and TRAP activity, Fig. 5). Tumor cells induced ~ 30% increases in HIF-1α+ osteocytes in the non-loaded tibiae relative to the PBS non-loaded controls; and the 4.5 N loading abolished the increase and reduced the percentage of HIF1a+ osteocytes to levels comparable to the PBS groups (Fig. 5A). As for osteocyte apoptosis, there was an increasing trend of TUNEL+ osteocytes in the tumor non-loaded group vs. PBS controls, and the 4.5 N loading significantly lowered the osteocyte apoptosis in the tumor-loaded group (Fig. 5B). In the bone marrow, the 4.5 N loading appeared to counteract the increasing trend of TRAP+ bone surface in the tumor group and returned it to the PBS levels (Fig. 5C). While tumor cells induced a ~20% decrease in the endomucin+ vessel area, the 4.5 N loading reversed the decrease and returned it to levels comparable to the PBS groups (Fig. 5D). Loading did not alter Ki67 staining on the few macroscopic tumors found on Week 2 (Fig. 5E).

Fig. 5.

Immunohistochemistry changes in bone under 4.5 N tibial loading. (A) HIF-1a response of osteocytes in PBS- or tumor-injected bones: tumor cells increased the % of HIF-1a+ osteocytes and loading returned it to the level of PBS groups. (B) TUNEL staining: loading significantly reduced the % of TUNEL+ osteocytes in the tumor group. (C) TRAP staining: no loading effect was detected. TRAP+ bone surface (red dotted lines). TRAP− surface (green dotted lines). (D). Endomucin staining: tumor reduced the % of endomucin+ blood vessel areas in the bone marrow, and loading returned it to the level of PBS groups. (C) Ki67 staining of tumor nuclei did not show apparent differences due to loading. Pairs with different letters (A, B, C) indicate significant differences (p < 0.05, ANOVA with Tukey HSD post hoc tests).

3.3. Moderate treadmill running suppressed and delayed bone destruction by breast cancer

Treadmill running protected bone integrity and delayed bone destruction in the presence of breast cancer (Fig. 6). The 3D renderings showed a slower progression of cortex perforation and lost trabecular bone in a running mouse than in a sedentary mouse (Fig. 6A). In the cortical bone, Ct.pMOI was maintained during the first three weeks but decreased on Week 4 and Week 5 (p < 0.005), and running prevented its decline in the tumor groups on Week 3 and Week 4 (p < 0.05, Fig. 6B). In the trabecular bone compartment, Tb.BV/TV declined with time (p < 0.001) and was lower in the tumor nonrunners than other groups (p = 0.02). Running increased the Tb.BV/TV on Week 3 (Fig. 6C). As for the volume of osteolytic lesions in the cortex, Ct.VoL was significantly higher in tumor nonrunners (p < 0.001) and increased with time in tumor groups (p < 0.001), which was suppressed by running on Week 3 to Week 5 (p < 0.05, Fig. 6D). The Kaplan-Meier survival curves showed that the probability of tibiae free from perforation was significantly higher in the tumor runners than nonrunners (p = 0.018). No periosteal woven bone formation was observed in either PBS or tumor runners (data not shown). Other μCT-derived bone parameters showed that running led to increased Tb.N and decreased Tb.Sp in the tumor group, while no running effects were observed in Ct.BMD, Ct.TMD, Tb.TMD, Tb.BMD, and Tb.Th within either the PBS or tumor groups (Supplemental Fig. 3S). Detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Fig. 3S.

Fig. 6.

Moderate treadmill running protected bone integrity and delayed bone destruction in the presence of breast cancer. (A) Examples of 3D reconstructions: treadmill running delayed bone destruction by tumors. (B) Ct.pMOI was maintained on Weeks 1-3 and dropped on Week 4 and Week 5. Running prevented its decline in the tumor groups at Week 3 and Week 4. (C) Tb.BV/TV declined with time at a faster rate in the tumor nonrunners than other groups. It was higher in tumor runners vs. nonrunners at Week 3. (D) Ct.VoL, the volume of osteolytic lesions in the cortex, was significantly higher in tumor nonrunners and increased with time in tumor groups. Running suppressed tumor lesion size from Week 4 to Week 5. (E) The Kaplan-Meier survival curves of tibiae free from perforation: survival was significantly higher in the Tumor Runner vs. Nonrunner groups. Statistical analysis: above the bar graphs, the effects of groups and times are shown with p-values, and pairs with different letters are significantly different; loading-induced significant differences are indicated with black bars. Detailed analysis of Group effect, Time effect, and Time*Group interactions can be found in Supplementary Fig. 3S. (B-D): Weekly data were presented (left to right) for PBS-NR, PBS-R, Tumor-NR, and Tumor-R groups.

On the cell and tissue levels, treadmill running counteracted the adverse effects of tumor metastasis on osteocyte hypoxia and apoptosis, TRAP activity, and H-type endomucin+ marrow vasculature, as well as reduced tumor burden and proliferation in the bone marrow (Fig. 7). While tumors significantly increased the percentage of HIF+ osteocytes on Week 1, running returned it to a level comparable to the PBS groups (Fig. 7A). Similar effects were seen in the TUNEL staining (Fig. 7B). After 3 weeks of running, the percentage of TRAP+ bone surface in the tumor group was significantly lower than that in the nonrunners (Fig. 7C). Running also rescued the decline of the endomucin+ vessel area induced by the tumors and returned it to a level comparable to the PBS groups (Fig. 7D). These positive effects may underline the decreased tumor burden and size seen in the bone marrow (H&E staining) as well as decreased tumor proliferation (%Ki67+ tumor nuclei, p < 0.006, Fig. 7E). Additionally, running decreased the staining of HIF-1α, lysyl oxidase (LOX), and Gm-CSF in tumor mass within bone marrow (Fig. 7F).

Fig. 7.

Immunohistochemistry changes in bone under treadmill running. (A) HIF-1a response of osteocytes in PBS- or tumor-injected bones: tumors increased the % of HIF+ osteocytes, while running counteracted the tumor effect and returned it to the level of the PBS groups. (B) TUNEL staining: tumors increased osteocyte apoptosis and running returned it to the level of the PBS groups. (C) TRAP staining: running reduced the % of TRAP+ bone surface in the tumor groups. TRAP+ labeled bone surface (red dotted lines). TRAP− labeled surface (green dotted lines). (D). Endomucin staining: tumor reduced the % of endomucin+ vessel areas in the bone marrow and running returned it to the level of the PBS groups. (E) Running reduced the tumor burden and size (H&E staining) and significantly decreased the % of Ki67+ tumor nuclei (p < 0.006). (F) Responses in tumors: tumors cells increased the HIF-1a, LOX, and Gm-CSF staining, which was reduced by running. B = bone; T = tumor; BM = bone marrow. Pairs with different letters (A, B) indicate significant differences (p < 0.05, ANOVA and Tukey HSD post hoc test).

4. Discussions

While physical activity is commonly prescribed to cancer patients owing to its anti-tumor and anabolic effects, it remains challenging to determine the type, intensity, frequency, and duration of exercise regimens that optimize the benefits while minimize unintended injury [10, 12, 42]. The motivation of the present study was to understand the relationship between the dose/type of physical activity and the risk of osteolysis in breast cancer. Using a preclinical cancer bone metastasis model, controlled loading/exercise modalities, and longitudinal in vivo micro CT monitoring, we confirmed the hypothesis that mechanical loading suppressed breast cancer-induced osteolysis of adult skeletons in a magnitude-dependent manner, as previously tested in younger animals [13, 14]. We demonstrated the efficacy of both localized bone loading and aerobic running in delaying bone destruction in the presence of breast cancer when the loading intensity was moderate. Our data also revealed adverse effects including accelerated osteolysis when loading/exercise intensity was too high. Findings from this and previous studies [13, 14] suggest that physical activity modulates the risk of bone destruction by breast cancer in a J-shape manner (see Graphical Abstract). A similar relationship has been found between exercise and infection risk, where moderate exercise offers better prevention of infection than either sedentary lifestyles or strenuous exercise [43].

The underlying mechanisms for such a relationship could be multifactorial and remain to be further elucidated. In the classic “vicious cycle,” tumor cells, stromal cells (including osteoblasts), and osteoclasts are major players that drive cancer bone metastasis [2]. Recent studies suggest that osteocytes, the most abundant and mechanical sensitive cell type in bone [15], also play an essential role [44, 45]. It is well established that osteocytes are a major source of local RANKL [46] and sclerostin [47] in adult bone, by which osteocytes regulate osteoblast and osteoclast functions to orchestrate bone remodeling [15, 48]. Interestingly, osteocytes’ responses to mechanical loading appear to follow a J-shaped relationship with loading magnitude: disuse increases osteocyte apoptosis, RANKL secretion [49], and hypoxia (HIF-1α) [50]; physiological loading during normal locomotion or exercise has more beneficial effects on osteocytes (i.e., decreased apoptosis and reduced RANKL and sclerostin expression) [51, 52]; overuse or fatigue loading results in more detrimental effects (i.e., elevated osteocyte apoptosis, RANKL, and HIF-1α) [53-55]. In our studies, treadmill running and 4.5 N tibial loading were shown to counteract the adverse effects induced by breast cancer, such as the increase of osteocyte hypoxia, apoptosis, and osteoclast activity, as well as the loss of endomucin+ vasculature in the bone marrow. Previous studies have shown that osteocyte hypoxia and apoptosis could activate osteoclasts for bone resorption. [49, 50], while reduction of the endomucin+ blood vessels, critical for angiogenesis during osteogenesis [39], might contribute to osteocyte hypoxia. Our results showed that treadmill running was particularly effective in maintaining normal marrow environments: it not only inhibited the proliferation of breast cancer cells but also reduced the hypoxia and matrix crosslinking (via lysyl loxidase) within tumor masses [40]. Although not tested in our studies, treadmill running was also found to mobilize neutral killer cells and other immune surveillance cells, as well as regulate pro- and anti-inflammation factors, all of which could inhibit tumor growth [18] and promote bone health [20].

However, there exists an upper limit for mechanical stimulation to provide anti-osteolysis and anti-tumor benefits in vivo. Lynch et al. [13] showed that physiological loading (~600 με, 4 Hz) successfully delayed bone destruction in immunocompromised mice inoculated with human breast cancer after 6 weeks. Recently, Fan et al. [14] demonstrated the skeletal protection effects of ~300 με loads at 2 Hz in two immunocompetent murine strains (BALB/c and C57BL/6). In our studies, skeletal benefits were observed in C57BL/6 mice with Py8119 murine breast cancer after 4 weeks of 4.5 N tibial loading (~630 με, 4 Hz) or 5 weeks of treadmill running (10.7-16.1 m/min, 50 min). This running protocol was expected to be moderate because (1) previously < 400 με tensile strain was recorded on the tibial mid-shaft of 12-week C57BL/6J females during locomotion and a 30 cm jump [27]; (2) our running speed and duration were lower than those used in fatigue testing of mice [28]; (3) the normal PBS injected mice showed no difficulty in completing all the running (100% completion rate for 5 weeks); and (4) treadmill running did not alter cortical and trabecular bone structures in previous studies, in which the running protocols were 5 degree incline, 12 m/min, 30 min, 3-5 days/week, and 6 weeks [56] or 13.8 m/min, 30 min, 3 days/week, and 2 weeks [57]. Despite its moderate nature, treadmill running significantly suppressed tumor-induced osteolysis in adult mice. In contrast, the higher loading intensity in this study (8 N, ~1100 με, 4 Hz) and in Fan et al. (5N, ~ 1300 με, 2 Hz) showed detrimental effects on bone integrity. Because the reported strains were obtained in bones without tumor inoculation, actual strains experienced by mice inoculated with breast cancers were likely higher as bone lesions developed. The fact that woven bone occurred in nearly all mice with tumor in our study was a morphological indicator of overuse. We further confirmed it by the detection of HIF-1α upregulation in osteocytes, a cellular response to damaging loading as demonstrated by Silva’s group [55]. It was possible that overuse-induced apoptosis and RANKL expression in osteocytes drove the accelerated bone destruction in our study, similar to the scenario of targeted remodeling of damaged bone matrix [53]. Furthermore, sustained strenuous exercise and cancer metastasis in bone may shift the innate and adaptive immune systems, promoting a chronic pro-inflammation environment [20], which could alter the sensitivity of bone cells in response to mechanical loading. For example, Wang et al found that condition media from breast cancer cells promoted dendrite formation in osteocyte-like MLO-Y4 cells and decreased OPG expression in osteocytes under fluid shear [58]. These interesting in vitro findings call for further in vivo investigations.

This study has some noteworthy strengths. The application of high-resolution (7 μm voxel) and fast μCT imaging allowed us to identify and track osteolytic lesions in the bone metaphysis, a common site for breast cancer bone metastasis [59]. For the first time, we were able to identify bone lesions by comparing the baseline and subsequent registered images. We also injected a smaller number (1000) of murine mammary tumor Py8119 cells, ~100 times fewer than those used in previous studies [13, 14, 60]. The use of adult immunocompetent syngeneic female mice was adopted to better mimic the time course of tumor growth from micro to macro metastasis, while the contribution of immune responses was taken into account [18]. In this model, the temporal process of bone destruction was consistent, and the duration (4 or 5 weeks) of the osteolysis process permitted interventions like exercise. Given the proven association of increasing cortical porosity with fracture risk in aging and diseased skeleton [61], the full-thickness lesions can serve as morphological markers and real-time in vivo readouts of bone fragility using sequential μCT imaging as demonstrated herein.

There were also several limitations in this study. Only one running protocol was tested and the running intensity and duration could be varied in future study. Although intratibial injection of cancer cells is a well-accepted bone metastasis model [62], it limited our study to only late stages of metastasis. Although our injection method did not expose the knee [13], the local damage at the injection site may, at least partially, contribute to the woven bone formation observed in the PBS-injected mice under 8 N loading, because such responses were not reported in intact tibiae of naïve female C57BL/6J mice [54]. A newly developed inoculation method via tail caudal artery injection could alleviate this concern [63]. Immunohistochemistry assays provided semi-quantitative readouts on bone cells, marrow, and tumors; however, causal-effect relationships could not be established based on these readouts. Also, due to the complexity of the in vivo system, we were unable to isolate the effects of exercise on multiple cell types in bone, nor the interactions among these cells. Future studies using our organ-on-chip systems [38] containing various cell types will be helpful.

In summary, this study demonstrated that mechanical loading affects bone residing cells that regulate breast cancer-induced osteolysis. We found that the relationship between exercise intensity and the risk of tumor-induced bone destruction followed a J-shaped curve in a preclinical breast cancer model. The findings from our study and others [13, 14] suggest that there may exist an effective window of exercise intensity within which exercise could reduce bone loss and adverse skeletal related events in metastatic breast cancer patients.

Highlights.

• Mechanical regulation of bone integrity studied in adult mice bearing breast cancer

• Localized and systemic exercises tested in syngeneic tumor models

• Moderate running and tibial loading (4.5 N) suppressed bone destruction by tumors

• High tibial loading (8 N) induced woven bone and accelerated bone destruction

• Exercise protocols differentially affected bone cells, marrow, and tumors