Bone morphogenetic protein pathway responses and alterations of osteogenesis in metastatic prostate cancers

Abstract Background Prostate cancer is a common cancer in men that annually results in more than 33 000 US deaths. Mortality from prostate cancer is largely from metastatic disease, reflecting on the great strides in the last century of treatments in care for the localized disease. Metastatic castrate resistant prostate cancer (mCRPC) will commonly travel to the bone, creating unique bone pathology that requires nuanced treatments in those sites with surgical, radio and chemotherapeutic interventions. The bone morphogenetic protein (BMP) pathway has been historically studied in the capacity to regulate the osteogenic nature of new bone. New mineralized bone generation is a frequent and common observation in mCRPC and referred to as blastic bone lesions. Less common are bone destructive lesions that are termed lytic. Methods We queried the cancer genome atlas (TCGA) prostate cancer databases for the expression of the BMP pathway and found that distinct gene expression of the ligands, soluble antagonists, receptors, and intracellular mediators were altered in localized versus metastatic disease. Human prostate cancer cell lines have an innate ability to promote blastic‐ or lytic‐like bone lesions and we hypothesized that inhibiting BMP signaling in these cell lines would result in a distinct change in osteogenesis gene expression with BMP inhibition. Results We found unique and common changes by comparing these cell lines response and unique BMP pathway alterations. We treated human PCa cell lines with distinct bone pathologic phenotypes with the BMP inhibitor DMH1 and found distinct osteogenesis responses. We analyzed distinct sites of metastatic PCa in the TCGA and found that BMP signaling was selectively altered in commons sites such as lymph node, bone and liver compared to primary tumors. Conclusions Overall we conclude that BMPs in metastatic prostate cancer are important signals and functional mediators of diverse processes that have potential for individualized precision oncology in mCRPC.


| INTRODUCTION
Bone morphogenetic proteins (BMPs) are a class of signaling molecules within the TGF-β (transforming growth factor-β) family that signal via downstream mediator SMAD proteins. During signal transduction, BMP ligand dimers bind to type I and type II serine/threonine receptor monomers, resulting in the formation of a hetero-tetrameric kinase complex. Following ligand-receptor formation, the constitutively active type II receptors phosphorylate the type I receptors to activate the latter's kinase domains. 1 Activated type I receptors signal via the SMAD family, a group of downstream mediator proteins that are divided into three functional classes. Regulatory SMADs (R-SMADs), which include SMAD1, -5, and -9, are phosphorylated by the activated type I receptors. Phosphorylated R-SMADs subsequently activate the single common-mediator SMAD (Co-SMAD), SMAD4, to form a mature transcription factor. Together, the R-SMAD-Co-SMAD complex translocate to the nucleus to regulate the expression of BMP target genes. 2 The third class of SMADs, the inhibitory SMADs (I-SMADs), which include SMAD6 and -7, are intracellular BMP pathway antagonists; they act by competing with R-SMADs 1, -5, and -9 at the type I receptor binding site and with SMAD4 for complex formation with SMAD1. 3 BMP signaling is associated with a wide variety of human cancers, including prostate, breast, colorectal, lung, ovarian, and others. 4 However, the specific effect of BMPs on tumorigenesis and metastasis can be dependent upon the unique BMP ligand and cancer type. 4,5 For example, in lung cancer, decreased BMP expression from normal tissues and high expression of many BMPs is associated with better overall survival. 6 However, in human oral squamous cell carcinoma, BMP2 driven epithelization was suppressed by TGFβ, illustrating dynamic control of cell behavior. 7 BMPs have shown that their ability as morphogens can help epithelial cells expand toward terminal differentiation, and this can be important in reversing early epithelial to mesenchymal transition and accelerating MET for the outgrowth of epithelial cells at distant sites. 8 In breast cancer, BMP4 signaling to Smad7 was found to suppress metastases by sensitizing cells to anoikis. 9 BMPs as secreted factors also play diverse roles in many unique tissues in the body as non-cell autonomous regulators of cell behavior that can be utilized to facilitate cancer growth. 10 Moreover, BMP ligands have been demonstrated to serve dual antiproliferative and pro-metastatic roles in cancer both in vivo and in vitro. 11 Thus, it is evident that BMPs have the dual potential to act as tumor suppressors and tumor promoters, with the specific effects dependent upon the BMP ligand and cancer type. 12 The BMP pathway has long been studied in prostate cancer due to the role of BMPs in the unique prostate cancer tropism to metastasize in bone. In prostate cancer, BMP signaling has been directly correlated with tumor development and metastasis. For example, BMP6 was detected in over 50% of prostate samples from metastatic adenocarcinoma patients, while expression was not observed in non-metastatic tumors or benign prostate tissue. 13 Moreover, BMP6 was found to promote tumor invasion and migration of prostate cancer cells, potentially through the activation of ID-1 and matrix metalloproteinases (MMPs). 14 BMP7 expression was detected in metastatic prostate cancer, with increased levels found in castration-resistant versus androgen dependent prostate cancer, suggesting that the ligand plays a role in the metastatic process. However, BMP7 also demonstrated microenvironment and tumor-dependent antiproliferative effects both in vivo and in vitro. 15 On the other hand, BMP2 expression is decreased in prostate cancer compared to benign prostate tissue, and BMP receptor expression is often lost during tumor progression. 5 However, in vitro, BMP2 was found to enhance C4-2B prostate cancer cell invasiveness and mediate TNF-α-induced invasion. 16 These studies set the stage for continued understanding of the BMP pathways in metastatic prostate cancer for specific context and therapeutic management.
Metastatic prostate cancers have a unique propensity to colonize bone, where they most often exhibit an osteoblastic phenotype characterized by increased osteoblast activity and bone lesions. The high affinity of prostate cancer for bone is the result of multiple interactions between the tumor and pre-metastatic site in a non-cell autonomous manner. 17 The discovery of aberrant BMP expression in prostate cancer and the recognition of its propensity for bone has led to the hypothesis that BMP signaling promotes bone metastasis via crosstalk between the primary tumor and bone prior to metastasis. 18 For example, WNT signaling was found to induce BMP4 and BMP6 expression in prostate cancer cells, which in turn stimulated osteoblast differentiation. 19 In human prostate LNCaP cells, BMP4 induced the production of Sonic hedgehog (SHH), which in turn stimulated BMP receptor and SMAD1 expression in mouse stromal cells to enhance sensitivity to the original BMP4 ligand; together, BMP4 and SHH stimulation also induced the expression of osteoblastic markers. 20 These findings suggest that BMPs secreted by prostate cancer cells play a role in the induction of osteoblast differentiation and upset the balance of bone remodeling, thus contributing to the osteoblastic phenotype. 18 In our study, we demonstrate the nuanced expression of BMP signaling using recent online data from cBioPortal and the Cancer Cell Line Encyclopedia (CCLE). [21][22][23] Additionally, we present the pharmacologic sensitivity to BMP inhibition in a broad array of human prostate cancer cell lines and their known effects on in the bone microenvironment. As mentioned previously, prostate cancer has a propensity to metastasize to bone, where it primarily exhibits a blastic phenotype in the form of blastic or sclerotic bone lesions that stimulate abnormal growth. 24 A minority of bone metastases display a lytic phenotype, which is characterized by bone destruction, while others appear to be a mix between lytic and blastic phenotypes. 25 In a previous study, we identified distinct differences in immune cell populations and bio signaling pathways between the lytic and blastic microenvironments of bone metastatic sites. 26 Of the six prostate cell lines studied, the C42B, MDA PCA 2B, and VCaP lines exhibited blastic-like phenotypes, while the PC3, 22RV1, and DU145 lines displayed lytic-like phenotypes. By studying the effects of BMP inhibition in both blastic-and lytic-like models, we seek to elucidate the differential roles of BMP signaling in the two metastatic phenotypes and potentially identify therapeutic targets for metastatic prostate cancer.

| Gene expression
The RT 2 Profiler™ PCR Array Human Osteogenesis (Qiagen Catalog #PAHS-026ZD-12) was used to assess the expression of 84 genes associated with osteogenesis. The cDNA was synthesized using the  Figure S6B. The IGF1R primer that was used to validate the RT 2 Profiler™ blastic-like cell line results was purchased from Qiagen (GeneGlobe). All genes were run in technical and biological quadruplicate, with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as the housekeeping gene to normalize gene expression.

| Statistics
Statistical analyses were performed using GraphPad Prism (version 9.4 for Windows; GraphPad Software Inc.) and Excel (version 2016 for Windows; Microsoft Corp.). All statistical tests used a cutoff p value of .05 for significance and were one-tailed Mann-Whitney Students t test with assumed heteroscedasticity (non-parametric).

| cBioPortal analysis of the cancer genome atlas
cBioPortal.org datasets were used for the cancer genome atlas (TCGA) bioinformatics analysis. Prostate cancer data sets were selected for the adenocarcinoma study comprised in the firehose legacy data set. 27 Metastatic prostate cancer samples were analyzed from the Stand-Up-To-Cancer (SU2C) dataset by selecting groups from tissues' source locations (bone lymph node, liver, etc.). 28 Cell line data from the CCLE Broad institute profiling were selected for prostate cancer cell lines and then queried for gene expression and responses to pharmacologic treatments. 23 All samples' gene expression was compared to diploid samples with a ±Z score of 2.0. The cbioportal.org was used as a preferred curated resource of the TCGA datasets for robust harmonized uniformed analyses of prostate cancers. 21,22 3 | RESULTS

| Differential BMP pathway activity in prostate cancer adenocarcinomas and metastatic adenocarcinomas
With the availability of diverse molecular genomic data available to appreciate the nature or molecular players in prostate cancer, we accessed TCGA via the cBioPortal. The BMP pathway is a large collection of secreted ligand, soluble antagonists and receptor signaling partners, which we simplified into these three respective categories-ligands, antagonists, and receptor/mediators ( Figure S6A). We utilized the primary NCBI gene symbol identification for these genes and first queried ligands in primary prostate cancer adenocarcinomas. Unlike most cancers, prostate cancer benefits from also having TCGA data derived from metastatic tissues containing DNA, RNA, and protein changes with clinical outcomes for 444 patients. We queried the SU2C supported data set of metastatic prostate cancer (containing 444 patient samples) and first sorted and ordered the patients by neuroendocrine features and then the anatomic tissue source of the metastatic cancer. Neuroendocrine is of importance in the bone metastatic phenotype as these can be preferentially osteolytic than blastic. 17,29 The amplification of GDF6 seen in primary tumors was elevated to 26% of all metastatic patients followed next by BMP15 at 9% of patients ( Figure 1B). We found that BMP ligands were more often overexpressed than lost or decreased and that their expression could be largely mutually exclusive from one ligand to the other. GDF6, also known as BMP-13, was the highest expressed ligand at 9% of patients with primary prostate cancer followed by BMP 2, 6, and 7 at 6% of patients, with all other ligands being expressed less than 5% of patients ( Figure 1B). Negative soluble secreted antagonists of BMP signaling displayed a much more diverse increase and decrease, with NOGGIN being consistently amplified and overexpressed in 6% of F I G U R E 1 Bone morphogenetic protein (BMP) ligands are differentially expressed in metastatic prostate cancer and primary prostate adenocarcinomas. Metastatic (444 patients) and primary tumors (501 patients) from prostate are profiled in two separate studies available via the cancer genome atlas (TCGA) and accessed through the cBIO.org portal. Oncoprint graphs indicate whether a specific gene is amplified (red) or increased (pink) within the study or deleted (blue) or decreased (light blue). The percent of patients with any modification is indicated to the left of the graph following the gene name.  Figure S1C). BMP specific mediator SMAD9 was most strikingly lost with 15% of patients having unique alterations ( Figure S1D). The alterations of SMDA9 were still present at 10% of metastatic patients ( Figure S1C). BMPR1A in metastatic patients was altered at 10% but comprised a diverse set of alterations from amplification, deletion, increased and decreased gene expression ( Figure S1C).   Figure S3A).

|
Genes increased in these conditions were SP7 (also known as osterix),  Figure 4A). Additional qPCR validation with additional primers only partially confirmed these results; however, 22Rv1 and Du145 cells did not show significant BMP3 elevation ( Figure 4B). The decreased MMP8 mRNA was challenging to analyze as only 22Rv1 cells had detectable levels of mRNA to quantitate under DMH1 treatment, while PC3 and Du145 cells MMP8 was not detected following BMP inhibition ( Figure 4B).

| Distinct mutations and BMP pathway gene expression in metastatic tissue sites of prostate cancer
Prostate cancer is unique in that the metastatic sites of disease beyond the primary site have been investigated when many other cancers with metastasis still rely on data from the primary tumor. The 2019 Su2C prostate metastasis study is unique in that it provides insight into lymph node, bone, and liver metastases. We decided to analyze the prostate cancer by metastatic location in bone (152 unique patients), lymph node (163 unique patients) and liver (63 unique patients) groups, which were non-overlapping to determine first if their mutational landscapes were distinct. We found that the most commonly mutated genes were similar across the three sites such as p53 and AR ( Figure 5A). Gene expression across these three tissues revealed that BMP2 ligand was lowest in lymph nodes and elevated in bone and liver and receptors BMPR1A and BMPR1B were highest in lymph node metastases ( Figure 5B-D). While the type II receptor BMPR2 was broadly expressed across all tissues ( Figure 5E). The majority of these metastatic prostate cancers were of adenocarcinoma with little representing the neuroendocrine phenotype ( Figure 5F). When survival was analyzed by disease location a clear separation of these patients could be seen for superior survival of lymph node and bone only metastatic disease when compared with liver metastases ( Figure 5G). These results suggest that location of metastasis may be correlated with mutations and gene expression within the BMP signaling pathway from either the tumor or its unique tissue microenvironment. Copy number alterations separated by tissue location revealed subtle reductions in applications for the liver disease ( Figure S5A). Looking at the most disparate amplifications or deletions resulted in distinct changes in liver disease such as RB1 deletions occurring at 25% of patients with little to no mutations were found in bone and just over 10% were found in lymph node ( Figure S5B). When metastatic prostate cancer is separated by the location of the disease a clear pattern of distinct genetic profiles emerges that can be successfully distinguished and diagnosed.  Prostate cancer's osteo-tropism is unique for its predilection to the blastic phenotype whereby bone is not lost from a vicious cycle of osteoclast bone destruction but instead promotes a poorly organized woven matrix of mineralized collagen. BMP driving osteogenesis has been proposed to drive this blastic phenotype and BMP inhibition was hypothesized to reduce the osteogenesis program ( Figure 3). In three human blastic-like prostate cancer cell lines we showed not a reduction in osteogenesis, but a convergence with the IGF-IGF1R signaling axis ( Figure 3A). IGF1R is currently being investigated as a potential biomarker in metastatic prostate cancers, as appropriate biomarkers that may be derived in the blastic bone phenotype have not yet been identified. 42 Targeting the IGF pathway in tumor induced bone disease has shown great promise alone or in therapeutic combinations. Yet the roles of IGF and BMPs in skeletal health require further investigation to understand the best patient-precise therapeutic interventions. 43 The osteomimicry seen by BMPs has been demonstrated previously in the conditioned medium or coculture of osteoblasts with prostate cancer cells. 16,19,20,[44][45][46][47] These BMP inhibition of lytic-like lesions may be more preferential than blastic-like lesions if those lesions are driven by BMP-directed osteoclastogenesis as compared to TGFβ-driven sclerosis. 51 It has also been demonstrated in multiple myeloma that BMP inhibition of osteolytic lesions improved bone quality through combined lineages of bone matrix producing cells. 52 Genetic and pharmacologic loss of function studies for BMPR1a have surprisingly resulted in improvement in bone formation as opposed to expected loss in BMP directed osteogenesis. [53][54][55] We found that BMP inhibition in lytic-like prostate cancer cell lines increases the transcription of BMP3 in some of the cell lines ( Figure 4B). BMP3 is not well studied in comparison to with more osteogenic BMPs 2/4/7 or even BMP6 and has unique negative functions in skeletal biology. 56 MMP8 is also unique and interesting as a putative metastasis suppres- sor, yet the role of MMP8 in prostate cancer metastasis in bone remain to be fully elucidated. 59 MMP8, like TGFβ/BMP, has many nuanced and context-dependent roles that may be both tumor promoting and or suppressive, a context that may be important to the distinct process a tumor cell is attempting to achieve in a specific microenvironment. 60 It is unsurprising that the BMP pathway is not largely distinguished for in vitro gene expression ( Figure 2). However, the finding that BMPR1b and not BMPR1a is elevated in blastic-like prostate cancer cell lines is compelling ( Figure 2C). One of the most compelling BMP receptor studies has shown that BMPR1b to be uniquely induced by androgen unlike the other type 1 receptors. 61 Blastic-like cell lines are more commonly androgen sensitive to both stimulation and deprivation which raises unanswered questions about the blastic matrix phenotype and androgen. 30 Metastatic prostate cancer has been a pioneer for metastatic disease research as numerous patients and their progression have provided scientists with unique datasets for in silico analysis such as the TCGA datasets. 28