Digital spatial profiling of α-PD-1 treated breast cancer bone metastases reveals region-specific signaling and enrichment of immune-suppressive markers.

1
Introduction
Breast cancer patient survival and prognosis has steadily improved over the years; however, there remain >40,000 deaths a year in the United States due to distant metastatic spread [1]. Bone serves as a preferential site of metastasis for breast cancer, and 70–80% of patients with metastatic breast cancer who succumb to the disease present with bone metastases upon autopsy [2]. Breast cancer cells are thought to engraft in the endosteal, hematopoietic, or perivascular niches in bone, which provide diverse environments primed for survival, colonization, and/or breast cancer dormancy [3]. While genetic drivers and signaling cascades that mediate breast cancer bone metastasis have been previously identified [4], there is still limited understanding of how immune cells are distributed in the breast cancer bone metastatic environment, and whether they are present in the endosteum. There is also a paucity of information of which signaling pathways are activated in the endosteum versus the surrounding marrow in the setting of breast cancer bone metastasis. As such, having a multiplexed way to assess enrichment of signaling pathways in specific geographic regions within the bone will help us understand how breast cancer bone metastases progress and how to prevent bone metastatic progression.
Spatial profiling techniques have recently gained traction as a method to visualize and quantify cells and pathways while preserving the spatial integrity of tissues. Sequencing-based platforms like Visium by 10X Genomics use spatially barcoded spots on a slide to capture RNA from the tissue [5]. Captured RNA is converted to cDNA, sequenced, then mapped back to the spatial location of the barcoded spots. Probe based platforms like Bruker Spatial Biology (GeoMx) combine staining with morphology biomarkers and digital spatial profiling (DSP) probes to pair RNA or protein expression with cell types and regions of interest [5]. The CosMx Spatial Molecular Imager (Bruker) uses cyclic fluorescence in situ hybridization to visualize and quantify RNA and protein expression in tissue sections at a single cell level. Similarly, 10X Genomics offers the Xenium platform which uses specific probes to RNA targets that are amplified to detect and locate individual transcripts within tissue sections at the single cell level. These technologies have improved resolution with each iteration, moving from resolutions of several hundred microns (e.g., GeoMx) to the single cell level (CosMx and Xenium), with a parallel increase in cost. With the rise of spatial profiling, transcriptomic and proteomic analyses have been applied to multiple tissues, including bone [6]; however, there is only one study that has applied this technology to bone metastases [7], and differences between distinct regions within the bone microenvironment were not investigated. We report here that proteomic spatial profiling using the GeoMx DSP platform can be used to identify distinct changes in marrow or endosteal signals within breast cancer bone metastases and at lower cost than true single cell technologies. Our study identifies immune-suppressive markers and proliferation/cell death signaling pathways that cluster in the endosteum or marrow and may be important for progression of breast cancer bone metastases.
The bone microenvironment is composed of an incredibly heterogeneous cell population including hematopoietic stem cells which are important for red and white blood cell production and mesenchymal stem cells necessary for production of bone, fat, and cartilage [8], [9]. The bone marrow houses many types of immune cells (e.g., B and T cells, neutrophils, macrophages, dendritic cells, myeloid-derived suppressor cells, etc.) [10], but the osteoimmunologic microenvironment in the context of bone metastases is poorly understood. Immune checkpoint inhibitors (ICIs) are a groundbreaking class of drugs approved to treat over 17 cancer types, including patients with advanced disease that frequently metastasize to the bone [11], [12]. Through targeting immune checkpoint proteins (e.g., programmed cell death protein 1 (Pd-1) and programmed death ligand 1 (Pd-l1), ICIs can increase immune infiltration and activation in the tumor microenvironment [13]. We previously reported that Pd-1 blockade results in expansion of CD4+ and CD8+ effector T cell populations in the bone marrow in tumor naïve mice [14], but it remains unknown how these cell populations interact with bone and tumor cells in a bone metastatic microenvironment. Our study identifies immune-suppressive markers and proliferation/cell death signaling pathways that cluster in the marrow or endosteum and may be important for progression of breast cancer bone metastases.

2
Materials & methods
2.1
Animals
Experiments were performed under the regulations of the Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals and approved by the Vanderbilt University Institutional Animal Care and Use Committee (IACUC). Female 6-week-old immunocompetent mice (Jackson Laboratory Stock No. 000664, C57Bl/6 background) were injected with 1 × 105 E0771 mouse mammary carcinoma cells in 100 µL by intracardiac injection (n = 10 mice). Since the bone marrow is immune suppressive in breast cancer [15] and breast tumors are often immune-excluded [16], [17], mice were treated with the immune checkpoint inhibitor α-PD-1 (BioXCell #BE0273, Clone 29F.1A12, 200 µg initial treatment followed by 100 µg every 3–4 days) by intraperitoneal injection for two weeks starting 36  hrs after tumor cell injection to enhance immune infiltration.

2.2
Histology and DSP sample preparation (detailed protocol included further below)
Upon sacrifice, mouse tibiae were collected and fixed in 10% formalin for 48 hr and stored in 70% ethanol until decalcification. Bones were decalcified in a 10% Disodium EDTA solution (pH 7.4) for 7 days, processed, and paraffin-embedded for further analyses. Samples were cut to 5 μm-thick sections and hematoxylin and eosin (H&E) stained to identify samples with significant tumor burden. These tissue sections were deparaffinized by heating the slides to 50°C and placed in Citrisolv for 5 min. Next, slides were soaked in 100%, 95%, and then 70% ethanol for 3 min each and then rinsed with water. Slides were immersed into filtered hematoxylin and rinsed with water then immersed in ammonia water. Following a second water rinse, tissue sections were soaked in an eosin solution with phloxine B and orange G. Slides were then rehydrated in 70%, 95%, 100% and placed in xylene for 5 min, then cover slipped with Cytoseal XYL. After identifying the sections for DSP analysis, bone cross sections were stained using the GeoMx Solid Tumor Morphology kit which contains a nuclear stain (Syto13), tumor morphological marker (Pan-Cytokeratin, PanCK), and immune cell morphological marker (CD45) (Bruker #121300304) and sent to the Vanderbilt University Medical Center Genomic Sciences Shared Resource (GSSR) core facility to confirm quality of morphological staining. If morphological staining sufficiently detected tumor burden in the bone, a new sample section was then stained with the same morphological markers and probed for the GeoMx Immuno-Oncology Protein Panels (catalog numbers provided below). These include the mouse modules for immune cell profiling (#121300106), immune activation status (#121300117), immune cell type (#121300118), pan-tumor (#121300119), Mitogen-activated protein kinase (Mapk) signaling (#121302141), PI3k/Akt signaling (#121300128), and cell death panels (#121300127) (Table 1). Following primary antibody overnight incubation, postfixing, and nuclei staining according to manufacturer instructions, slides were taken again to the Vanderbilt GSSR core.

2.3
DSP ROI selection and overview
Multiplex DSP was performed on the proximal tibia following successful staining of select morphological markers and protein panels outlined above (Fig. 1). Slides with tibial sections were scanned and regions of interest (ROI) were selected within the GeoMx DSP software system such that tumor and immune cells were both present within the ROI and located either in the medullary space ≥100 µm from the surface of any visible trabecular or cortical bone (termed “Marrow”) or directly adjacent to bone (≤100 µm from the endosteal surface, termed “Endosteum”). ROIs were stochastically selected within the medullary space or adjacent to the endosteum beginning 200  μm distal to the tibial growth plate and extending 900  μm distal to the growth plate (Fig. 2). Circular ROIs with 100 μm diameter were utilized and manually segmented to uniquely threshold the tumor and immune cells for each area of interest (AOI). Following ROI selection and segmentation, AOIs were illuminated with ultraviolet light to release barcoded oligos corresponding to their described proteins, which were collected within a 96-well plate and quantified by the GeoMx nCounter system. A detailed protocol for DSP can be found in the Supplemental DSP Protocol.

2.4
DSP data analysis
All DSP protein expression outputs were normalized to housekeeping proteins glyceraldehyde-3-phosphate (Gapdh) and Histone H3 protein nCounter read counts in the GeoMx DSP portal prior to data export. Comparison analyses were made to identify 1) differences in protein markers expressed on CD45+ immune cells in the endosteum vs marrow regions in mice with bone metastases and 2) differences in signaling pathway proteins expressed on PanCK+ tumor cells and CD45+ immune cells in the endosteum vs marrow regions of tumor-bearing mice. Murine H&E-stained tissue sections were microscopically assessed to identify those with a balance of visible tumor burden in the bone and intact marrow (i.e., samples in which the tumor burden had not completely colonized the bone to the point that no marrow remained). After evaluating these factors and balancing sample number with the cost to perform DSP, we proceeded with n = 3 mice for DSP analyses. We collected a total of 6 ROIs per sample (3 in the marrow and 3 in the endosteum) which yielded a total of 9 AOI values per region for analysis (3 mice, with 3 AOIs per mouse). These normalized data values (expressed as nCounter reads) were further downloaded into an Excel file and analyzed in GraphPad Prism for statistical analysis as outlined below.

2.5
Statistics
For all studies, n per group is as indicated in the figure legend, bar graphs indicate the mean of each group, and error bars indicate standard error of the mean. All graphs and statistical analyses were generated using Prism software (GraphPad) and data were tested for outliers by ROUT test. All protein targets were analyzed for statistical significance using Student’s unpaired t-test to compare the endosteal and marrow regions of the bone. For all analyses P<0.05 was considered statistically significant, and ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001.

3
Results
To assess whether spatial profiling can be used to analyze immune and cancer cells residing in distinct geographic regions in the bone marrow, we performed proteomic DSP using the GeoMx platform on three C57Bl/6 female mice injected with E0771 breast cancer cells by intracardiac injection to model bone metastasis. Since immune cells do not robustly infiltrate breast cancer bone metastases [16], [17], we treated the mice with α-PD-1 for four weeks (200 µg initial dose, followed by 100 µg 3-4x/week for 2 weeks) to improve immune infiltration in the bone marrow, as we have previously described [14], and ensure adequate levels of CD45+ immune cells for analysis. Upon sacrifice, tibiae were formalin-fixed, decalcified, and paraffin embedded (FFPE). The bones were H&E stained to confirm the presence of robust tumor burden in the bone (Fig. 3).
DSP analysis relies on the accuracy of morphological markers staining positive for the cell populations of interest. We therefore first performed a quality control assay on bone tissue sections, which were probed with Syto13 (nuclear stain), PanCK (to detect cytokeratin+ E0771 cancer cells), and CD45 (to detect CD45+ immune cells) morphological markers to identify the tumor and immune cell populations. Stained bone sections were scanned into the GeoMx DSP system in the Vanderbilt University Medical Center Genomic Sciences Shared Resource (GSSR). Qualitative analysis of the morphological marker staining showed PanCK+ staining for E0771 cells, detectable by morphology and positive staining in the proximal metaphysis and extending distally through the medullary cavity (Fig. 4A, dashed white outline), with modest GFP auto-fluorescent background, which is common in bone and the surrounding remnant muscle tissue [18]. While the breast cancer bone metastases remained predominantly immune-excluded (Fig. 4A, white arrows indicate CD45+ cells residing adjacent to the tumor), we observed CD45+ immune cells within the breast cancer bone metastases (yellow arrows), allowing for DSP analysis of immune populations within the tumor-bearing bone marrow and endosteum. A similar staining pattern of immune exclusion and gradient of immune infiltration was observed across all mice bearing breast cancer bone metastasis (Fig. 4B&C).
Following successful completion of the quality control step, we combined our validated morphological markers Syto13, PanCK, and CD45 with antibodies coupled to photocleavable oligonucleotide probes, which hybridize to the FFPE bone sections for the DSP assay. We simultaneously stained each bone section with multiple immune-oncology protein panels that included immune cell profiling, immune activation status, a pan-tumor panel, as well as panels for Mapk signaling, Phosphoinositide 3-kinase-protein kinase B (PI3k/Akt) signaling, and cell death (Table 1); the panels are designed for ultra-high multi-plexing, allowing for detection of all panels on one tissue section. Stained tissue sections were scanned into the GeoMx instrument and 9 uniform ROIs of 100 µm diameter were selected across each bone tissue section (Fig. 5A&B) as described in the methods and Fig. 2. Approximately half of the ROIs were selected in the endosteum (which we defined as ≤100 µm from the surface of trabecular or cortical bone) and the other half selected in the marrow (defined as ≥100 µm from the endosteum).
In order to capture staining for the selected protein panels in each sample, we segmented each ROI into a tumor segment (based on positive PanCK morphological staining) and an immune segment (based on positive CD45 staining) (Fig. 5C&D). These segments were termed AOIs. The thresholding for each segment was performed manually for each ROI to ensure there was no overlap between the two AOIs. Some AOIs included a small amount of the bone matrix depending on how the ROI was placed, but since the bone matrix does not stain for PanCK or CD45, the bone matrix was not included in either segment analysis (as in Fig. 5D).
After AOI segmentation, the protein probes for each segment were UV-cleaved and read counts were quantified for each protein in the multiplexed panels (∼70 proteins) on the GeoMx nCounter instrument. This captures the proteomic profiles from each AOI in the bone (PanCK+ tumor or CD45+ immune cells). We normalized all nCounter reads to the Gapdh and Histone H3 housekeeping nCounter reads then exported the data from the DSP Control Center to the Analysis Suite (raw data included in Supplemental Table 1).
Our analyses revealed modest but significant changes in protein expression within the CD45+ segmented AOI between the marrow and endosteum (Fig. 6A). In the marrow, we observed significantly higher expression of the immune receptor CD127/interleukin 17 receptor A (IL17ra) compared to the endosteum (Fig. 6B). In the endosteum, we observed significant enrichment of markers associated with T cell function and activation, including T-cell surface glycoprotein CD3 epsilon chain (CD3e), cluster of differentiation 27 (CD27), forkhead box P3 (Foxp3), cluster of differentiation 163 (CD163), and cluster of differentiation 31 (CD31) compared to the marrow (Fig. 6C–G). Additionally, we observed elevated expression of fibronectin, androgen receptor (AR), Pd-1, and Pd-l1 expression in the CD45+ segmented AOI in the endosteum (Fig. 6H–K), suggesting immunosuppressive (e.g., Foxp3) and immune exhaustion (e.g., Pd-1) markers are enriched within the immune cell fraction that resides near the endosteum within breast cancer bone metastases.
Similar to the CD45+ AOI, we observed modest but significant changes in protein markers between the marrow and endosteum in the PanCK+ segmented AOI (Fig. 7A). We observed significantly higher poly(ADP-ribose) polymerase (Parp), B-cell lymphoma-extra large (Bclxl), epithelial cell adhesion molecule (Epcam), and epidermal growth factor receptor (Egfr) in the marrow compared to the endosteum (Fig. 7B–E), suggesting apoptosis resistance in breast cancer cells localized to the marrow. There was also elevated expression of p38 Mapk, pan-Ras, mitogen-activated protein kinase kinase 1 (Mek1), and phosphorylated-extracellular signal-regulated kinase 1/2 (phospho-Erk1/2) proteins in the PanCK+ AOI in the marrow compared to the endosteum (Fig. 7F–I), suggesting an increase in Mapk pathway signaling and proliferation in breast cancer cells residing in the marrow (i.e., further away from the endosteum).

4
Discussion
Spatial profiling technology is continuously and rapidly evolving, but its use in pre-clinical bone metastasis models has been limited due to the challenges of working with mineralized tissue. In bone metastatic prostate cancer, multiplexed proteomic DSP analysis has been performed on pre-clinical and patient prostate cancer bone metastasis samples to identify differences in lytic and blastic bone metastases [7], but the technology has not been used to analyze different regions within bone metastases. Disseminated tumor cells can home to and colonize multiple niches within the bone marrow (which can also overlap with each other), including the endosteum, the hematopoietic stem cell niche, and the endothelial niche [19], [20]. Herein, we assessed proteomic changes in what we broadly termed the marrow and the endosteum in a preclinical mouse model of breast cancer bone metastasis. Utilizing DSP, we analyzed changes in populations and proteins within the immune cell fraction (CD45+) and tumor fraction (PanCK+) within the same ROI to gain insight into where immune cells reside in the bone metastatic microenvironment and where proliferation and apoptosis pathways are enriched within bone metastases.
In the CD45+ immune cell segment analyses, we observed an increase in T cell markers like CD3e, CD27, Foxp3, and CD163 near the endosteum. While the role of T cells in breast cancer bone metastasis remains an active area of investigation [21], tumor infiltrating leukocytes (TILs) can be found in breast cancer bone metastases, where inactivated T cells from 4T1 bone metastases stimulate osteoclastogenesis [22]. T cell activation has been shown to reduce bone metastasis in melanoma, and this effect is reversed by T cell depletion [23], but this has not been investigated in breast cancer. CD163 is expressed in both activated T cells [24] and in monocyte-macrophage lineage cells [25]; our data suggest that CD163+ CD45+ immune cells, including potentially activated T cells which may reduce osteoclastogenesis and bone metastatic progression, cluster in the endosteum. However, a unique population of VCAM1+CD163+CCR3+ monocytes was recently identified as a crucial iron supply that is hijacked by tumor cells in the bone metastatic site, leading to tumor mimicry of erythroblasts, which may ultimately promote bone metastatic progression [26]. Our data suggest that CD163+ immune cells are enriched in the endosteum, which may include VCAM1+CD163+CCR3+ monocytes, but would require validation in future studies.
Our data also suggest that Foxp3 is elevated in the endosteum. There are no publications that have directly looked at whether Foxp3+ regulatory T cells (Tregs) regulate breast cancer bone metastasis, but the finding that this marker is elevated in the endosteum suggests that Tregs may reside in this region. Tregs have been shown to be “bone sparing” by increasing fracture repair [27] and inhibiting osteoclast activity [28] but they also suppress T cell activity and promote immune evasion by tumor cells [29]. Our data suggest these cells may be enriched in the endosteum. In combination with monocytes that may support breast cancer growth, collectively our findings suggest that the majority of immune signals in the endosteum are pro-tumorigenic. This is supported by previous findings that the endosteum acts as a calcium reservoir, which similarly fuels breast cancer growth in bone [30]. Fibronectin, which was also elevated in CD45+ cells in the endosteum, can promote epithelial-to-mesenchymal transition (EMT), cell proliferation, migration, and survival [31], [32], which may also contribute to a pro-tumorigenic environment. We also observed enrichment of the immune checkpoint proteins Pd-1 and Pd-l1 in CD45+ cells in the endosteum, further suggesting this region fosters an immune-suppressive environment, which may promote breast cancer growth. We previously demonstrated that blockade of Pd-1 also promotes bone loss and increases CD8+ T cells in the bone marrow [14] and our data here suggest that Pd-1 may be enriched in the endosteum, where Pd-1+ immune cells are in close proximity with bone remodeling cells like osteoblasts and osteoclasts. Osteoblasts are thought to promote inflammation in bone metastasis and create an immune-suppressive microenvironment that promotes cancer progression [33], [34], [35]; this aligns with our findings that immune-suppressive signals are enriched in the endosteum, where osteoblasts reside, and suggests that osteoblasts may be one potential source of those signals. “Tumor-educated osteoblasts” which have been exposed to cancer cells have also been shown to limit progression of bone metastatic breast cancer cells [36]. Our finding that breast cancer cell proliferation markers are enriched in the marrow rather than endosteum indirectly supports this role for osteoblasts in tumor progression. Furthermore, the Pd-1/Pd-l1 axis in breast cancer may promote inflammatory signaling, give rise to cancer associated fibroblasts, and suppress immune response through secretion of cytokines [37]. While we see elevated expression of Pd-1 and Pd-l1 in the endosteum, it is important to note that since all of the mice were treated with α-PD-1 to enhance immune infiltration, this may have impacted baseline Pd-1 and Pd-l1 expression, and future studies should assess the localization of Pd-1 to the endosteal surface in treatment naïve mice. Overall, these data support that the endosteum harbors immune cell populations and signals that may support breast cancer cell survival in bone.
Within the PanCK+ tumor cell segment, significantly higher Parp, Bclxl, and Epcam expression in the marrow compared to the endosteum suggests that breast cancer cells that are in the marrow may have elevated amounts of DNA damage, inhibition of apoptosis, and increased tumorigenicity. Elevated Parp and Bclxl expression have been associated with apoptosis evasion and favorable metastatic potential [38], [39], and Epcam expression on breast cancer cells has been shown to enhance cancer cell stemness and epithelial-to-mesenchymal transition [40]. These data suggest that breast cancer cells in the bone marrow may have greater sensitivity to Parp inhibitors; however, the Parp inhibitor olaparib has been shown to promote osteoclast differentiation and immune suppression, which increases breast cancer bone metastasis [41]. The enrichment of these markers in breast cancer cells specifically in the marrow and not in the endosteum suggests that breast cancer progression may also occur within the marrow. This is intriguing and suggests that both the marrow and endosteum are favorable to breast cancer growth, but through different cellular and molecular mechanisms. Of note, there has been therapeutic interest in mobilizing bone-disseminated tumor cells out of the endosteum to sensitize them to chemotherapeutics, such as through blocking breast cancer interactions within the perivascular niche, which has been shown to reduce bone metastasis [42]. Our data suggest that these cells would need to be mobilized out of the bone marrow altogether and into the bloodstream, since the marrow milieu contains alternative pro-tumorigenic signals. Moreover, increased expression of p38 Mapk, pan-Ras, Mek1, and phospho-Erk1/2 proteins in the PanCK+ tumor cells in the marrow suggest the Mapk pathway is specifically elevated in breast cancer cells located far from the endosteum. Mapk signaling is up-regulated across many cancer types [43], [44], [45], and our data suggest that the marrow may support proliferation of tumor cells and potentially drive bone metastatic progression.
There are several limitations to our study: First, we broadly investigated immune enrichment in the marrow, but this could be further segmented into other niches within the bone marrow microenvironment with the incorporation of other morphological markers (e.g., targeted to endothelial cells or HSCs). Second, we used bone tissue sections from α-PD-1 treated mice to increase baseline immune infiltration, as we previously reported [14], in order to validate the utility of DSP in profiling immune and tumor changes within the bone marrow microenvironment. It is therefore unclear whether these same signals are enriched in the absence of α-PD-1 treatment. It is also unknown whether the cells and pathways are enriched in these regions due to the presence of tumor cells, or if the signals are present and/or cluster similarly in tumor naïve mice. In future studies, our approach could be expanded to compare immune infiltration to untreated and tumor naïve groups. Additional validation staining for protein targets that are significantly enriched in the marrow or endosteum will also be useful in future studies; however, in support of our findings, we demonstrate a significant enrichment of Pd-1 in the endosteum, which is consistent with staining patterns we previously reported for Pd-1 in tumor naïve bone marrow, using IRDye800-labeled α-PD-1 [14]. The costs associated with DSP limited the number of samples analyzed, but future studies should validate these findings in a second pre-clinical breast cancer bone metastasis model. The E0771 cells used herein share some features with both triple negative breast cancer cells [46], [47], [48] and luminal B cells [49], [50]. It would be useful to compare our findings to other models that more classically represent triple negative and luminal breast cancer subtypes, particularly bone-tropic variants (our E0771 cells were not previously selected as bone-tropic); however, it is worth noting that robust models of breast cancer bone metastasis are somewhat limited [51]. Lastly, since spatial proteomic analyses require the use of pre-selected protein panels, we were unable to perform unbiased analyses to detect enriched signaling pathways; our results are skewed toward the pre-selected signaling pathways. In future studies, spatial transcriptomic approaches could be used to enrich for receptor-ligand interactions, using a similar technical approach for the bone processing. As with other DSP studies, future studies will be required to validate the functional relevance of immune populations or proteins enriched within specific niches, but these profiling experiments are still essential to understanding how the geography of the bone impacts cell distribution and signaling pathways, which are challenging to functionally test.
Our observations may provide a better understanding of the location of different cell lineages and enriched pathways within the tumor-bone microenvironment (Fig. 8). Our findings identify the location of immune-suppressive signals primarily in the endosteum and reveal that pro-tumorigenic markers are enriched in breast cancer cells in the marrow. Collectively, these studies may inform how to better target breast cancer cells in the bone marrow microenvironment.

5
Author statement
The work described has not been published previously except in the form of an abstract.
The article is not under consideration for publication elsewhere.
All authors approve of the publication by the responsible authorities where the work was carried out.
If accepted, the article will not be published elsewhere in the same form, in English or in any other language, including electronically, without the written consent of the copyright-holder.

CRediT authorship contribution statement
Déja M. Grant: Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Gwenyth J. Joseph: Writing – review & editing, Visualization, Methodology, Data curation. Madeline Searcy: Writing – review & editing, Conceptualization. Rachelle W. Johnson: Writing – review & editing, Visualization, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Rachelle W. Johnson reports financial support was provided by US Department of Defense. Rachelle W. Johnson reports financial support was provided by National Institutes of Health. Gwenyth Joseph reports financial support was provided by National Institutes of Health. Deja Grant reports financial support was provided by Howard Hughes Medical Institute. Madeline Searcy reports financial support was provided by National Institutes of Health. RWJ is a member of the editorial board for the Journal of Bone Oncology. Given her role as an editorial board member, RWJ had no involvement in the peer review of this article and had no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to another journal editor. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.