Bone marrow adipocytes induce cancer‐associated fibroblasts and immune evasion, enhancing invasion and drug resistance

Abstract Bone metastasis occurs frequently in cancer patients. Conventional therapies have limited therapeutic outcomes, and thus, exploring the mechanisms of cancer progression in bone metastasis is important to develop new effective therapies. In the bone microenvironment, adipocytes are the major stromal cells that interact with cancer cells during bone metastasis. However, the comprehensive functions of bone marrow adipocytes in cancer progression are not yet fully understood. To address this, we investigated the role of bone marrow adipocytes on cancer cells, by focusing on an invasive front that reflects the direct effects of stromal cells on cancer. In comprehensive histopathological and genetic analysis using bone metastasis specimens, we examined invasive fronts in bone metastasis and compared invasive fronts with adipocyte‐rich bone marrow (adipo‐BM) to those with hematopoietic cell‐rich bone marrow (hemato‐BM) as a normal counterpart of adipocytes. We found morphological complexity of the invasive front with adipo‐BM was significantly higher than that with hemato‐BM. Based on immunohistochemistry, the invasive front with adipo‐BM comparatively had a significantly increased cancer‐associated fibroblast (CAF) marker‐positive area and lower density of CD8+ lymphocytes compared to that with hemato‐BM. RNA sequencing analysis of primary and bone metastasis cancer revealed that bone metastasized cancer cells acquired drug resistance‐related gene expression phenotypes. Clearly, these findings indicate that bone marrow adipocytes provide a favorable tumor microenvironment for cancer invasion and therapeutic resistance of bone metastasized cancers through CAF induction and immune evasion, providing a potential target for the treatment of bone metastasis.


| INTRODUC TI ON
Bone metastasis occurs frequently among cancer patients. [1][2][3] The most common primary sites for bone metastasis are breast, prostate, and lung cancers. These cancers are also the most common malignancies in female and male patients or the most frequent site for primary disease in cancer deaths. 4 Along with increasing the prevalence of these cancers, the total numbers of patients with bone metastasis are increasing annually. [5][6][7] Bone metastasis causes specific symptoms such as intractable bone pain, pathological fracture, hypercalcemia, anemia, and paralysis due to spinal cord compression. 8,9 These symptoms significantly worsen the quality of life of cancer patients. Therefore, effective treatments are urgently needed for patients with bone metastasis.
At present, in addition to surgery for pathological fracture, chemotherapy, radiotherapy, and targeted therapy for osteoclast activity have been used to treat patients with bone metastasis. 2,[10][11][12][13] Although these therapies indeed reduce skeletal related events, no significant reduction of tumors or improvement of overall survival has been observed. [14][15][16] Therefore, investigating novel crucial mechanisms of cancer progression in bone metastasis is important to provide new effective therapies for bone metastasis.
Stromal cells interact with cancer cells, providing a favorable microenvironment for cancer progression. [17][18][19] Elucidating the mechanisms of tumor growth in bone metastasis requires comprehensive understanding of tumor-stromal interactions in the bone microenvironment. The major stromal cells in this microenvironment are hematopoietic cells, fibroblasts, adipocytes, endothelial cells, immune cells, chondrocytes, osteocytes, osteoclasts, osteoblasts, and neural cells. 20,21 Among these stromal cells, the number of adipocytes increase with aging, diseases, and lifestyle-related factors. [22][23][24] After adolescence, adipocytes occupy more than 50% of the bone marrow cavity; this increases to more than 70% at middle age. 23 Bone marrow adipocytes are therefore the primary stromal cells interacting with cancer cells in bone metastasis.
The role of adipocytes in tumor growth has been extensively investigated. [25][26][27] Adipocytes secrete several growth factors, cytokines, and fatty acids [28][29][30] that promote tumor growth and invasion, immune evasion, and premetastatic niche formation. 25 Focusing on the invasive front when studying cancer celladipocyte interactions makes it possible to assess the precise role of stromal cells on tumor growth. The invasive front reflects the direct effects of stromal cells on cancer, and has therefore been widely studied to evaluate the role of stromal cells in tumor growth. [31][32][33] Comparing invasive fronts with different stromal components can reveal stroma type-specific functions in tumor growth.
In this study, we examined invasive fronts in bone metastasis, comparing invasive fronts with adipo-BM to those with hemato-BM as a normal counterpart of adipo-BM, focusing on morphological and IHC differences. We evaluated the role of adipo-BM on cancer invasion, tumor immunity, cancer proliferation, and angiogenesis.
To analyze the cancer-side contribution on the phenotypes observed in the invasive front with adipo-BM, we undertook RNA-seq analysis comparing primary cancers to bone metastasis and elucidated the contribution of acquired functions in cancer cells after bone metastasis.

| Case selection
This study was carried out according to the guidelines of the Note: Data are shown as number or mean (range).

| Hematoxylin-eosin staining
Each FFPE tissue specimen (5 μm thick) was placed on a glass slide and stained with H&E.

| Immunohistochemistry
Bone metastasis samples were fixed in 10% neutral buffered formalin and decalcified with HCl or EDTA. The samples were processed, paraffin-embedded, and cut into sections (

| Stained specimen digitalization and quantification
Stained slides were scanned using an Aperio SC2 scanner and imaged using an Aperio eSlide Manager (both from Leica Biosystems).
The staining of cancer cells using whole slide imaging was quantitatively evaluated using the image analysis tools QuPath 34 and Fiji. 35 Following digitalization, snapshot images of the selected area including the bone marrow stroma more than 1 mm away from the invasive front, stroma at the invasive front, cancer cells at the invasive front, and cancer cells more than 1 mm away from the invasive front, were recorded at 10× or 20× magnification using QuPath. One to three snapshot images were obtained for each area for each slide.
Positively stained areas or cells in the snapshot images were automatically calculated using QuPath and Fiji with a plug-in. 36,37 The stained area (or cells) was divided by the total area (or cells) to calculate the relative staining positivity. For the analysis of αSMA-stained tissue, we defined the 100 μm range from the invasive front line with adipo-BM as "adipo-BM area at the invasive front" and "cancer with adipo-BM area at the invasive front", and the same area with hemato-BM as "hemato-BM area at the invasive front" and "cancer with hemato-BM area at the invasive front".

| RNA extraction from tissue and RNAseq analysis
We selected patients who underwent surgery at the primary site and biopsy of bone metastasis. We then chose cases without decal-

| In silico analysis using public database
We undertook an in silico whole genome analysis using a public database, through the cBioPortal website (cBioPortal for Cancer Genomics, https://www.cbiop ortal.org).

| Statistical analysis
For experimental data analysis, GraphPad Prism 9 software (GraphPad Software) was used. Student's unpaired t-tests were used to detect statistical significance. For the comparison of multiple groups, ordinary ANOVA followed by either the Tukey-Kramer multiple comparison tests or Dunnett's multiple comparison tests were carried out. Statistical significance was set at p < 0.05. Pearson's χ 2test was used to determine correlations between clinical information and incidence of recurrence. In the IPA analysis, the activation of canonical pathways was calculated as the z-score, which indicates the prediction value of pathway activation calculated from the average of DEGs, by IPA's original calculation algorithm (positive value: activated in the bone metastasis tissue; negative value: activated in the primary cancer tissue).

| Selection of appropriate cases for histopathological analyses from 144 bone metastasis specimens
First, we selected 144 bone metastasis cases from the year 2010 to 2022 (Table 1). Next, we microscopically selected the cases with an invasive front of cancer and bone marrow stromal cells in the specimens. We excluded 60 advanced-stage cases with multiple metastases to other organs and with the highest T stage; their cancers occupied the entire bone marrow cavity, with no invasive front remaining. We also excluded cases with small sized tissues where not enough images could be obtained for the analysis of invasive fronts.
Finally, we selected and analyzed 56 qualifying cases for morphological or IHC analysis (Table 2, Figure S1).

| Adipocyte-rich bone marrow increased invasive-front complexity
We carefully observed morphological differences between invasive fronts with adipo-BM (average age, 64.7 years) and those with hemato-BM (average age, 64.0 years) in bone metastasis specimens.
Invasive front complexity correlates with tumor aggressiveness including cancer invasion, decreased cytotoxic T cell density, and poor prognosis. 40,41 We first microscopically compared the adipo-BM and hemato-BM invasive fronts. Cancer cells had irregularly infiltrated into the adipo-BM ( Figure 1A), whereas a similar infiltration pattern was rare in hemato-BM ( Figure 1B). We defined the uneven invasive front as "irregular" and smooth invasive front as "smooth" in the microscopic images. The adipo-BM invasive fronts had significantly higher irregular area than the hemato-BM fronts ( Figure 1C). We then calculated the fractal dimension of the invasive frontline using Fiji, 35 for more objective quantification ( Figure 1D): the average fractal dimension was significantly higher for the adipo-BM invasive fronts than for the hemato-BM fronts ( Figure 1E). These results suggest that adipo-BM provides a better microenvironment than hemato-BM for complex invasive front formation in cancer bone metastasis. In addition, we examined the invasive front complexity of

| Cancer-associated fibroblast markers expressed in adipo-BM at the invasive front
Next, we examined the drivers of cancer invasion at the invasive front comprising adipo-BM and cancer cells. Cancer-associated fibroblasts have been broadly known to support cancer invasion, [42][43][44] and we compared expression of CAF markers between adipo-BM and hemato-BM at the invasive front.
Immunohistochemistry revealed that the adipo-BM invasive front, but not adipo-BM away from the invasive front, was positive for αSMA, one of the most commonly used CAF markers (

| CD8 + T lymphocyte density significantly lower in adipo-BM than in hemato-BM at the invasive front
Tumor immunity is one of the most important factors regulating cancer invasion and migration in the tumor microenvironment.
CD8 + T lymphocytes play a major role in the antitumor immune response. 47,48 In colorectal cancer, their density at the invasive front is closely correlated with tumor invasion and prognosis. 49,50 We therefore examined the density of CD8 + T lymphocyte density in adipo-BM and hemato-BM at the invasive front ( Figure 3A); many CD8 + T lymphocytes were observed in hemato-BM, although few CD8 + T lymphocytes were observed in the adipo-BM at the invasive front. In the statistical analysis, the CD8 + T lymphocyte count and density were significantly lower for adipo-BM than for hemato-BM at the invasive front ( Figure 3B). These findings suggest that adipo-BM is an appropriate microenvironment for evading antitumor immunity.  Figure 4A,B). Interestingly, Ki-67 positivity of cancer cells at the adipo-BM invasive front was significantly higher than that at the tumor center (>1 mm from the invasive front) ( Figure 4B). The same result was observed in the IHC of MCM2, another proliferation marker ( Figure S5). These results suggest that adipo-BM, but not hemato-BM, promotes greater cancer cell proliferation at the invasive front relative to that at the tumor center.

| Stromal composition of invasive fronts did not affect CD31 + blood vessel density in bone microenvironment
Blood vessel density at the invasive front is positively correlated with the progression of cancer invasion. 51,52 We therefore examined the vessel density at the adipo-BM and hemato-BM invasive fronts using

| Contact with bone marrow adipocytes may enhance tumor cell dormancy
Bone marrow is a common site of cancer recurrence after long-term disease-free status, due to the regrowth of dormant cancer cells. [55][56][57] For patients with detailed medical information (n = 58), the average period from primary site surgery to diagnosis of bone metastasis was 50.5 months, with a maximum of 266 months (Table S1, Figure S7).

| Cancer cells acquired drug resistance and immune evasion-related genetic phenotype after bone metastasis, consistent with the histopathological phenotype observed in adipo-BM invasive fronts
Our findings indicate that bone marrow adipocytes promoted CAF formation and suppressed CD8 + lymphocyte infiltration at the invasive front, factors that are both closely related to therapeutic resistance. [60][61][62][63][64] We therefore hypothesized that cancer cells acquire therapeutic resistance in adipo-BM after bone metastasis from the primary site. To test this, we compared cancer cell mRNA expression in bone metastasis with that in the primary site of the same case. Six bone metastasis specimens, including five nondecalcified cases and one case with only 1 day of EDTA decalcification, were subjected to quality control using RNA fragmentation assessment (DV200) ( Table 1). We extracted cancer cell mRNA from bone metastasis specimens with adipo-BM (yellow marrow, detected by MRI; Table S2) and from primary sites ( Figure 5A), then applied RNA-seq analysis and compared mRNA expression. We visualized DEGs by volcano plot using TCG-GUI ( Figure 5B) and heatmap analysis manually ( Figure 5C). Matrix metalloproteinase-9, other MMP families, and SMOC1, which interacts with Tenascin-C, 65-67 expressing CAF were upregulated, whereas CCL21, which promotes T cell activation, [68][69][70] was downregulated following bone metastasis ( Figure 5C). In the results of canonical pathway analysis using IPA, the RHOGDI pathway, which closely relates to cancer migration, invasion, metastasis, and resistance to anticancer agents, 71-73 was upregulated following bone metastasis ( Figure 5D,E). Next, we undertook GSEA related to histopathological phenotypes observed in the adipo-BM invasive fronts (Figures 6, S10, and S11). As expected, following bone metastasis, therapeutic resistance-related gene sets were upregulated, while gene sets related to therapeutic sensitivity and antitumor immunity were downregulated ( Figure 6). Together, these findings indicate that, after bone metastasis, cancer cells in bone marrow existing high fat signals assessed by MRI images may have gene expression phenotypes that promote therapeutic resistance and immune evasion compared to primary cancer.

| DISCUSS ION
In this study, we clarified that bone marrow adipocytes cultivate a favorable microenvironment for the spread of cancer, by inducing CAF and immune evasion in bone metastasis.
To the best of our knowledge, this is the first report examining invasive front morphology in bone metastasis. Invasive front complexity is closely associated with cancer invasion. 40 Moreover, the existence of complex invasive front or isolated cancer cell foci in metastatic tissue is associated with recurrence and poor prognosis. 41,[74][75][76][77] Here, we found that bone marrow adipocytes induced a complex invasive front and the formation of isolated cancer cell foci, generating an aggressive phenotype in bone metastatic cancer. Together, these findings indicate that bone marrow adipocytes promote cancer spread in the bone microenvironment.
Next, to investigate why adipo-BM induces cancer invasion, we applied IHC screening for CAFs, which have a strong ability to induce cancer invasion, rebuild the stromal matrix, and support cancer migration. 78,79 Adipocytes are known to transform CAFs because adipocytes and fibroblasts share the same stromal stem cells. 45,46 Here, the CAF marker-positive area was significantly larger for the adipo-BM invasive front than the hemato-BM invasive front.
Interestingly, immunofluorescence analysis revealed the coexpression of CAF and adipocyte markers ( Figure 2D). These findings suggest that adipocytes might be a good source of CAFs at the invasive front, and identify CAF induction as a potential mechanism by which adipo-BM promotes cancer invasion; some CAFs might even be derived directly from bone marrow adipocytes. There are several reports that CAFs are induced from stem cells preserved in the adipose tissue and resident fibroblasts in the adipose tissue. 80,81 Further study is needed to elucidate the origin of CAFs existing in the adipo-BM fronts.
Immunotherapy is conventionally used to treat cancer, and ICIs have been recognized as a new standard therapy for cancer. [82][83][84][85] Combination therapy, involving chemotherapy and other molecular target reagents such as tyrosine kinase inhibitors, has been used to treat patients with bone metastasis, based on many case reports and retrospective studies. Nonetheless, immunotherapy still has not been shown to have significant therapeutic effects in bone metastasized cancers. [86][87][88] One of the reasons for this is that the mechanisms of cancer progression in the bone microenvironment are not well known; as a result, cases that respond to immunotherapy have not been appropriately selected. Here, we revealed that the CD8 + T cell density was significantly reduced at the adipo-BM invasive front compared to that at the hemato-BM invasive front. This result coincides with the evidence that hemato-BM is a good microenvironment for differentiated CD8 + T lymphocytes. 89 In addition, as shown previously, the adipo-BM invasive front induced CAFs, which attenuate antitumor immune response. 90,91 Together, our findings suggest that adipo-BM promotes immune evasion and ICI resistance.
Tumor dormancy is an important mechanism for recurrence after long periods, and bone is a potential site for cancer stem cell escape. 57,92,93 A recent report showed that complex invasive fronts were closely related to tumor recurrence; 77  Following metastasis, cancer cells acquire more aggressive phenotypes than primary cancers. 94,95 To elucidate this, we examined the acquired functions of cancer cells following bone metastasis in adipo-BM by comparing primary and bone metastasis. We undertook RNA-seq of primary and bone metastasis in same patients.
The heatmap analysis and volcano plot revealed the upregulation or downregulation of specific genes related to histopathological phenotypes observed in the adipo-BM ( Figure 5B,C). Furthermore, activation of the RHOGDI pathway, promoting aggressive cancer phenotypes, was identified by IPA following bone metastasis, coinciding with the histopathological phenotypes of adipo-BM invasive fronts ( Figure 5D). In GSEA analysis, among highly upregulated gene sets, drug resistance and immune evasion-related gene sets were observed. In addition, the TFF pathway, reported as facilitating metastasis, was also upregulated ( Figure S10). We also comprehensively analyzed invasion and EMT-related gene sets using TCC-GUI  sites; however, these cases did not apply the same patient derived tissue. 103 Four cases included genetic mutation or CNA data for both primary cancer and metastasis sites. 103 Therefore, our dataset is highly rare, and it is important to clarify transcriptional differences between primary cancer and bone metastasis in the same patients.

F I G U R E 6
Immune evasion and drug-resistance related gene set enrichment analysis (GSEA) of upregulated (red bar) or downregulated (blue bar) gene sets in bone metastasis compared to primary cancer. The y axis represents enrichment score (ES) and on the x axis are genes (vertical black lines) represented in gene sets. The green line connects points of ES and genes. Significance threshold was set to the false discovery rate <0.05. ICI, immune checkpoint inhibitor.
In future, we intend to analyze additional samples to obtain more statistically relevant data.
In addition to conventional chemotherapy and radiotherapy, targeting therapy for osteoclast and osteoblast is currently used. 2,[10][11][12][13] Osteoclasts induced by metastasized cancer cells cause osteolysis, causing the release from the bone matrix of bone-derived growth factors that promote cancer growth, activating the vicious cycle. Osteoclasts are the key player for activating this cycle, and osteoclast-targeted reagents including bisphosphonate and anti-RANKL Abs are used to treat bone metastasis. 104,105 Although these reagents reduce skeletal-related events, no significant reduction in tumor size or significant improvements of overall survival have been observed. [14][15][16] Our results propose the possibility of bone marrow adipocytes as a new therapeutic targets for improving prognosis of bone metastasized cancer patients.
This study has several limitations. We did not undertake RNA-seq analysis of bone metastasized cancer with hemato-BM invasive fronts, as we could not find any hemato-BM invasive fronts in any of the analyzed bone metastasis biopsies. We consider one of the reasons is that naturally the component of hemato-BM is rare for the age of bone metastasized cases (average age, 64.7 years [ Table 1]; usually the percentage of hematopoietic bone marrow was less than 30% in total bone marrow tissue 23 ), and therefore it is difficult to obtain appropriate biopsy samples of bone metastasized cancer with hemato-BM invasive front. Osteoblastic metastasis is common in cancers, such as prostate cancer 106,107 ; however, in this study, we could not find any osteoblastic cases in cases we chose for investigating cancer invasive fronts. We examined bone metastasis specimens with osteoblastic change; however, these osteoblastic metastasis cases occupied bone marrow lesions with cancer cells and woven bone. Thus, we could not see any invasive fronts with cancer and bone marrow adipocytes or hematopoietic cells. We did not examine the role of hematopoietic cells on cancer invasion, including neutrophils and megakaryocytes, which reportedly promote the effects of cancer progression. [108][109][110] Further investigation is needed to discover the roles of all components in hematopoietic bone marrow. In addition, we could not compare the function of adipo-BM between young patients and older patients, as all cases in this study were elderly patients.
Also, we did not examine the dormant status of primary cancer in each case. By addressing these problems in further studies, more comprehensive interpretation of the role of bone marrow adipocytes would be revealed.
In conclusion, our findings reveal that bone marrow adipocytes support metastatic tumor progression in the bone microenvironment through CAF induction and immune evasion. Based on our findings, which show the important role of adipocytes in the tumor microenvironment, we propose the concept of "adipo-oncology". In future, we intend to comprehensively analyze the role of adipocytes in cancer, and particularly in cancer development, invasion, metastasis, and prognosis. We will further examine the role of bone marrow adipocytes in bone metastatic cancer by in vitro and in vivo studies, and examine the value of adipocyte-targeted therapy for bone metastasis.

AUTH O R CO NTR I B UTI O N S
S.S. was responsible for the study concept, experimental design, interpretation of results, writing, review, and revision of the paper.

ACK N OWLED G M ENTS
We would like to thank Dr. Yutaka Kondo (Nagoya University) for providing cogent and meaningful suggestions. We would also like to thank Yukako Komori and Kumiko Ohrui for helping us with the in vivo and in vitro experiments. Figure 5A was created with Biore nder.com.