Stromal-Initiated Changes in the Bone Promote Metastatic Niche Development

Abstract

More than 85% of advanced breast cancer patients suffer from metastatic bone lesions, yet the mechanisms that facilitate these metastases remain poorly understood. Recent studies suggest that tumor-derived factors initiate changes within the tumor microenvironment to facilitate metastasis. However, whether stromal-initiated changes are sufficient to drive increased metastasis in the bone remains an open question. Thus, we developed a model to induce reactive senescent osteoblasts and found that they increased breast cancer colonization of the bone. Analysis of senescent osteoblasts revealed that they failed to mineralize bone matrix and increased local osteoclastogenesis, the latter process being driven by the senescence-associated secretory phenotype factor, IL-6. Neutralization of IL-6 was sufficient to limit senescence-induced osteoclastogenesis and tumor cell localization to bone, thereby reducing tumor burden. Together, these data suggest that a reactive stromal compartment can condition the niche, in the absence of tumor-derived signals, to facilitate metastatic tumor growth in the bone.

Figure 1. Transgene activation in osteoblasts results in senescence and SASP expression

A. Characterization of transgene activation in FASST mice. Representative immunofluorescence (IF) staining of frozen bone sections obtained from littermate (LM) control or FASST mice was performed with anti-GFP (green) and anti-Osteocalcin (OCN) (red) antibodies. IF staining revealed mosaic activation with variable penetrance of the p27^Kip1 transgene (GFP+) within OCN (+) osteoblasts (see Figure S1A). Scale bar is 100 μm, magnification 10X and inset is 20X (n=4). B. Quantification of transgene activation (GFP+) in five bone sections from four different FASST mice shows a wide variation in transgene activation. GFP-positive counts from the same mouse are coded as a single color. C. Representative SA-β-Gal staining in FASST versus littermate control bone sections. Quantification of SA-β-Gal staining in littermate control (LM) or FASST mice. Senescent bone lining cells are denoted by a black arrowhead. Data are presented as a percent of total bone-lining cells. Scale bar is 100 μm (n=3) (*SEM, p<0.05, two-tailed, Student t test). D. Representative staining of IL-6 (red) SASP expression in the bones of FASST mice revealed robust activation in GFP+ (green), senescent osteoblasts. No GFP expression was observed in the littermate (LM) mice and no significant IL-6 expression was noted in the LM controls. Scale bar is 100 μm, magnification 10X and inset is 40X (n=4). Secondary antibody panel with existing LM and FASST IL6 stains can be found in Figure S1B for reference. E. Quantification of IL-6 expression in GFP+ osteoblasts from bone sections of FASST mice (n=4 mice, 5 bone sections each). IL-6 was expressed in ~75% of GFP-positive cells while ~25% of GFP-positive cells had no detectable IL-6 expression.

Figure 2. Senescent stromal cells increase bone metastasis

A. FASST mice display increased bone tumor burden. Isogenic breast cancer, NT2.5luc cells were introduced into littermate control (LM) or FASST mice by left intracardiac injection (IC). Four weeks later mice were sacrificed and their leg bones were imaged ex vivo. Significantly higher tumor burden was observed in FASST mice compared to littermate controls (n>20). SEM, *p<0.05, unpaired Student T test with Welch’s correction. Importantly, the distribution of tumor burdens corresponded with the variable activation of senescence noted in FASST mice. B. Tumor morphology was assessed in littermate control (LM) versus FASST mice. Tumors in bone sections were stained with H&E and no gross morphological differences were noted. Scale bar is 100 μm. C. Representative p16 and IL-6 staining in human normal bone reveal mosaic co-expression in bone lining cells. Scale bar is 100 μm and insets are 80X.

Figure 3. Senescence abrogates osteoblast function and drives osteoclastogenesis

A. Primary osteoblasts were obtained from the femurs of 6-week-old FASST mice. To activate the p27^Kip1 transgene and induce senescence, FASST osteoblasts were treated with 10μM tamoxifen (4-OHT) for 3 days. SA-β-Gal staining was carried out on 4-OHT or vehicle (veh) treated osteoblasts on day 3. 4-OHT treatment resulted in a significant increase in enlarged, flattened cells that stained positive for SA-β-Gal (blue) compared to vehicle treated cells (lower left panel), scale bar is 200 μm (*p<0.05, two-tailed, Student t test. one of three representative experiments is shown). The induction of the senescence associated secretory phenotype (SASP) factor IL-6 was also quantitated by reverse transcriptase PCR (qRT-PCR) (lower right panel). Senescent osteoblasts expressed significantly more IL-6 mRNA compared to vehicle treated cells (SEM, *p<0.05, two-tailed, Student t test; one of three representative experiment is shown). B. qRT-PCR expression of alkaline phosphatase (ALP) in vehicle (veh) versus 4-OHT treated osteoblasts grown in basic culture medium (CM) versus osteogenic medium (OM). Senescent osteoblasts displayed a dramatic reduction in ALP mRNA levels when grown in osteogenic medium (SEM, *p<0.05, two-tailed, Student t test; one of two representative experiments is shown). C. ALP and Alizarin Red-S staining in senescent versus non-senescent osteoblasts. FASST osteoblasts were treated with vehicle or 4-OHT for 3 days and then plated in basic culture medium (−) or osteogenic medium (+) for 7 or 21 additional days. On day 7 ALP expression was assessed (upper panel, one of four representative experiments is shown) and on day 21 mineralization capacity was assessed by staining with Alizarin Red-S (lower panel, one of three representative experiments is shown). 4-OHT treated osteoblasts displayed a robust reduction of both ALP expression and mineralization capabilities. D. Senescent osteoblasts increase osteoclastogenesis. As shown on the timeline, osteoblasts were treated with vehicle (veh) or 4-OHT for 3 days. 4-OHT was then removed from the culture medium and on day 12, after robust activation of senescence (data not shown) bone marrow macrophages were plated on non-senescent versus senescent osteoblasts and osteoclastogenesis was determined by TRAP staining on day 16. Representative pictographs are shown, scale bar is 200 μm. TRAP-positive osteoclasts (TRAP+OC) were counted and data are displayed as TRAP+OC/field of view (TRAP+OC/FOV). (SEM, *p<0.05, two-tailed, Student t test; one of three representative experiments is shown). Because we noted that osteoclast precursor numbers were similar in co-cultures containing non-senescent and senescent osteoblasts (data not shown), this increase is due to increased osteoclast differentiation.

Figure 4. Senescent osteoblasts increase local osteoclastogenesis

A. Osteoclast surface to bone surface area was assessed by TRAP staining. Representative images of bone sections from 6-week-old littermate (LM) control or FASST mice are shown and osteoclasts are indicated by red, TRAP+ staining and nuclei are counterstained with hematoxylin. For these analyses TRAP+ cells were quantitated in the tibia near the growth plate using the Bioquant software. Scale bar 100 μm (n=4). B. Quantification of osteoclast surface area (OC.S) per bone surface (BS) was carried out on bone sections obtained from 6-week-old littermate (LM) or FASST mice (n=4, SEM, *p<0.05, two-tailed, Student t test). C. Representative bone sections from 6-week-old FASST mice were co-stained with antibodies against GFP (green), OCN (red) and ELF97 TRAP (blue). Scale bar 100 μm. D. Definition of bining distances for computational spatial analyses that were overlaid on the image presented in c. The lines depict: 1) cyan line outlines the osteoclast; 2) thick line = 200 units of distance (pixels); and 3) dashed line = 100 units of distance (pixels). Scale bar 100 μm. E. Computational spatial analysis of bone sections from FASST mice show increased osteoclastogenesis within 50 units of senescent osteoblasts (within the cyan line in d). Statistical analysis of the frequency of senescent versus non-senescent osteoblasts encountering an osteoclast. Data demonstrate a higher association of osteoclasts is observed around senescent osteoblasts and this associated falls off with distance (n=3, SEM, *p<0.05, two-sided Wilcoxon signed rank test). F. Distribution of osteoclasts “within” 50 units is shown from data in panel e. As we noted in Figure 1, the distribution of osteoclasts increasing near senescent osteoblasts corresponds with the variable activation of senescence noted in FASST mice (n=3, SEM, *p<0.05, two-sided Wilcoxon signed rank test).

Figure 5. Senescent osteoblasts drive tumor cell seeding

A. Immunofluorescent staining to detect tumor cells and senescent osteoblasts in FASST mice. For these analyses sections from the femurs of 6-week-old FASST mice were obtained 3 days after NT2.5luc cells were introduced intracardically. Femurs were completely sectioned and EpCAM staining (red) was used to identify tumor cells and anti-GFP staining (green) was used to identify senescent osteoblasts. Representative images are shown, scale bar 1 mm, red inset is 20X and white inset is 60X (n=3 femurs. Total number of tumor cells analyzed was: 32, 16, and 7 from bones from three different mice). Stains using a secondary antibody alone stain and FASST mice with no tumor cells can be found in Figure S4A. B. Computational spatial analysis of tumor cell localization. The image shown in panel g was masked as follows to determine tumor cell localization. The growth plate, which was the region of the femur where greater than 90% of tumor cells were localized was masked to facilitate statistical analysis. Tumor cells were coded as red stars and senescent osteoblasts as green dots. Scale bar 1 mm. C. The likelihood that a tumor cell (grey dots) would encounter a senescent osteoblast by chance (dotted line at 0) versus the observed frequency was calculated for each osteoblast. Osteoblast locations were sampled instead of tumor cell locations to minimize edge effects associated with tumor cells located near the border of the region of interest. Following the sampling, the mean, standard deviation, median, and 95% confidence interval were computed. To better visualize the difference between the observed and expected distance distributions, the ratio of the observed and expected median distance functions was taken. Where this plot is greater than 1, the distance to the n^th senescent osteoblast from a tumor cell was larger than expected, favoring segregation of the two cell populations. Where this plot is less than 1, the distance from a tumor cell to the n^th nearest senescent osteoblast is smaller than expected, thus favoring association of the two populations. To allow for combination of data across multiple samples and mice, a normalized difference (also referred to as effect size) curve was computed. The difference between the observed distance and mean difference for the n^th nearest osteoblast was taken and normalized by sampling standard deviation. The resulting curve provides a measure of how many standard deviations away from the random mean the observed results lie. To allow the graph to show positive numbers for association, the negative of the normalized difference is plotted. P values are found in Figure S5).

Figure 6. Senescent-derived IL-6 drives increased osteoclastogenesis and bone metastasis

A. Quantification of osteoclasts following co-culture with non-senescent (NS) or senescent (SEN) osteoblasts in the presence of a control (IgG) or IL-6 neutralizing antibody (α-IL-6). Bone marrow derived macrophages were cultured with non-senescent or senescent osteoblasts for 4 days in the presence of 5 pg α-IL-6. The antibody was replaced every other day. TRAP staining was carried out after 4 days co-culture and osteoclast numbers were quantitated and are displayed as TRAP-positive osteoclasts per field of view (TRAP+OC/FOV) (SEM, *p<0.05, two-tailed, Student t test; one of two representative experiment is shown). B. IL-6 neutralizing antibody reduces bone metastasis. Littermate (LM) or FASST mice were treated with 500 μg control antibody (IgG) or IL-6 neutralizing antibody (α-IL-6) on day −1 and twice weekly throughout the time course of the experiment. On day 0, NT2.5luc tumor cells were introduced by intracardiac injection. On day 28, mice were sacrificed and tumor burden was assessed by ex vivo imaging of leg bones. Tumor burden is displayed as photon flux (photons/second) (n>6, SEM, *p<0.05, unpaired Student t test). C. The growth of NT2.5luc cells was monitored by bioluminescent imaging in the presence of conditioned medium obtained from vehicle (veh) or 4-OHT treated osteoblasts. Vehicle and 4-OHT conditioned medium was treated with either a control antibody (IgG) or a neutralizing antibody against IL-6 (α-IL-6). Growth was determined by the MTT assay (SEM, *p<0.05, two-tailed, Student t test; one of two representative experiment is shown). D. Quantification of osteoclast surface area (OC.S) per bone surface (BS) was carried out on bone sections obtained from 6-week-old FASST mice. Treated with IgG or anti-IL-6 (n=3 per condition, SEM, *p<0.05, two-tailed, Student t test). E. Computational spatial analysis of bone sections from FASST mice treated with IgG control antibody or anti-IL-6 show that anti-IL-6 eliminates the increased osteoclastogenesis observed locally around senescent osteoblasts. Mice were treated with IgG or anti-IL-6 two days before receiving tamoxifen. Mice then received biweekly doses of IgG or anti-IL-6 for two more weeks before being sacrificed. Statistical analysis of the frequency of senescent versus non-senescent osteoblasts encountering an osteoclast in mice treated with IgG (upper panel) or anti-IL-6 (lower panel). Data demonstrate a higher association of osteoclasts is observed around senescent osteoblasts in IgG treated mice that is absent in anti-IL-6 treated mice (n=3, SEM, *p<0.05 for IgG treated mice and the p value for anti-IL-6 treated mice was 0.36 (non significant, ns), both two-sided Wilcoxon signed rank test).