Deidentified IRB exempt human patient surgical specimens from bone containing prostate cancer restricted to bone without soft tissue involvement were decalcified and formalin-fixed and paraffin-embedded (FFPE) obtained from the University of Colorado Pathology Shared Resource. Ex vivo mouse tissues were harvested and immediately placed in 10% formalin and fixed for 24 hours. Formalin was replaced with 14% EDTA (Research Products International) for 5-7 days until decalcified and placed in 70% ethanol for 24 hours prior to embedding in paraffin wax. FFPE tissue blocks were sectioned at 5µm thickness with two sections per slide and mounted on plus coated microscope slides. Slides were dewaxed and rehydrated with the following sequence of washes: xylene (StatLab), 100% ethanol, 95% ethanol, 70% ethanol, 50% ethanol, and then phosphate buffered saline (PBS) (Research Products International). Heat-induced antigen retrieval was performed by microwaving slides in citrate pH6 Antigen Unmasking Solution (Vector Laboratories) in a rice cooker. Routine H&E staining was performed in Harris hematoxylin (Vector Labs). Primary antibodies for pSMAD1/5/8 (Millipore 1:100) and pSmad3 (Abcam 1:200) were used for immunohistochemistry (IHC) staining. Signal was detected by ImmPRESS polymer secondaries to appropriate host and DAB chromogen substrate (Vector Labs) and counterstained with Hematoxylin QS (Vector Labs). All bright field IHC and H&E were scanned at 40X (0.22um/pixel) magnification using a ScanScope XT System (Aperio Technologies). To quantitate IHC staining, a grid of up to five 20X images per slide was captured, avoiding excessive stroma or necrotic tissues, on an Eclipse Ni microscope (Nikon) and imported into ImageJ (U.S. National Institutes of Health) to auto-contrast images then change contrast to blue and then the threshold was set and staining was measured by percent area.

For tumor naïve studies, twenty C57BL6/N (Taconic) males at 8 weeks of age with healthy appearance and weights over 20g were used. Mice were placed into four groups (n=5/group). Mice received a 6-week osmotic pump (Alzet 2006 #0007223), experiments in Figures 3 , 4 used 4-week pumps (Alzet 2004 #0000298) of DMSO or DMH1 (30mg/mL) and two groups received zoledronic acid (ZA) (Sellekchem #S1314) treatment via IP injection (25ug/100µl). Animals were anesthetized using isoflurane 2-5%, site was prepped for a small incision at the base of the shoulders for subcutaneous implantation of the pump. Skin was closed using wound clips (BD 427631). 6 and 12 weeks from initial treatment the animals were analyzed by Dual-energy X-ray absorptiometry (DXA) while anesthetized. At week six, with pump expired, the ZA groups received an additional injection of zoledronic acid. To model bone metastasis in mice, a Caudal Artery Injection (CAI) of PC3 cells in 100ul sterile PBS were injected into the tail via the caudal artery as previously reported (24). NSG (Jackson Laboratories Stock#005557) male mice of at least 8 weeks and 25g weight were warmed and placed into a tube restrainer for CAI procedure where they were placed in a supine orientation to visualize the artery for efficient injections. As another model of bone experimental metastasis in mice, intra-tibial injection of tumor cells, 1x10⁵ MyC-CaP cells in 10µL PBS were injected into the tibia of FVBn (Jax Stock # 001800) male mice no younger than 8 weeks of age. Mice were bred and maintained at Rocky Mountain Regional Veterans Affairs Medical Center (protocol number CD1611M) as well as at the University of Colorado Anschutz Medical Campus (protocol number 00553). All the immune competent mice (FVBn and C57bl6) were housed together and immunodeficient animals (NSG) were house solitarily because of aggression in accordance with ARRIVE guidelines (25). All animal procedures were performed in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committees.

Dual-energy X-ray absorption (DXA) (Faxitron Ultrafocus DXA, Hologic, USA) was performed weekly or biweekly while animals were under anesthesia of inhaled isoflurane of 2-5%. Analysis of the femurs to quantify bone mineral content (BMC) and density (BMD) were performed by manually drawn region of interest surrounding the femur or tibia bilaterally under the bone only option. Bones were scanned with a µCT scanner (Skyscan 1276, Bruker, USA) at 70kV, 114mA with a 0.5mm aluminum filter using a pixel size of 6µm. Images were acquired at 0.6þ through 180þ, reconstructions and analysis were performed using NRecon, DataViewer and CTAn software (Bruker). ROI for trabecular bone (distal femur and proximal tibia), 2mm were analyzed 0.1 mm distal to the growth. For cortical tibial midshaft ROI; 2mm proximally from the tibia/fibula junction, 1mm was analyzed. IVIS Spectrum (Perkin Elmer) was used to image the animals with RediJect 2-DeoxyGlucosone (DG) 750 (Perkin Elmer 760561). IVIS imaging of 750nm 2-DG was performed as directed by using the wizard setting with Ex Filter 745 and Em Filter 800 (26).

The human prostate cancer cell line PC3 cell line was obtained from the American Type Culture Collection (ATCC) and cultured in DMEM (Cat#10-013-CV) Medium (Corning) 1X Antibiotic-Antimycotic (Thermofisher Scientific Cat#15240062) and 10% Fetal Bovine Serum (FBS) (Seradigm). The FVBn MYC-CaP cell line was obtained from ATCC/Dr. Austin Kirschner at Vanderbilt University and cultured in DMEM (Corning), Antibiotic-Antimycotic and 10% FBS (Seradigm). All cell lines were routinely tested for mycoplasma infection by PCR and authenticated by morphology and published growth rates available from ATCC. Additional DNA fingerprint authentication services were performed the CU Cancer Center Tissue Culture Core Facility for the PC3 cell line.

Submandibular blood collection was performed prior to sacrifice, 200µl of blood was drawn using lancets (Goldenrod) (27). Blood was placed directly into a 1.3mL EDTA collection tube (Sarstedt 41.1395.105). Samples were analyzed for a complete CBC with a HemaTrue (Heska) analyzer using manufactures recommend settings.

Statistical analyses were performed using GraphPad Prism (version 9.2.0 for Windows; GraphPad Software Inc.) and Excel (version 2020/office 365 for Windows; Microsoft Corp.). All statistical tests used a cutoff p-value of 0.05 for significance with the Mann-Whitney test (non-parametric, one-tail and unpaired). Multiple group comparisons were analyzed by one way ANOVA in GraphPad with a Kruskal-Wallis post-test. The choice for these statistical tests were to compare significant differences in numerical values originating from similar sample types such as bone radiologic data.

To investigate whether canonical readouts of BMP and TGFβ signaling is altered in TIBD we performed Immunohistochemistry (IHC) on clinical samples containing bone with metastatic prostate cancer. These samples were previously utilized to identify cellular and molecular differences in the tumor microenvironment of TIBD based on their bone pathology, either being enriched for lytic destructive bone lesions or sclerotic and blastic bone features (10). Staining for canonical BMP signaling by hosphor-SMAD1/5/9 (pSMAD1/5/9) proteins was surprisingly low in blastic samples ( Figure 1A ). Osteolytic lesions had very strong nuclear staining in not only the tumor cells engulfing the bone but within the osteocytes and surrounding bone cells ( Figure 1B ). Staining quantitation for the six patients from both bone pathologies revealed that BMP signaling was significantly increased in lytic type disease (p=0.0125) ( Figure 1C ). We next turned to the canonical readout of TGFβ signaling hosphor-SMAD3 (pSMAD3) and were surprised to see enriched IHC staining in blastic features of metastatic prostate cancer in bone ( Figure 1D ). We noted the staining was strongly enhanced within the bone matrix and nucleus of osteoblast and osteocyte-like cells. Osteolytic lesions contained reduced nuclear pSMAD3 staining, which was largely restricted to the tumor cells and less so in the lytic bone ( Figure 1E ). Quantification of six patient samples for blastic and lytic samples revealed pSMAD3 staining was significantly increased in blastic bone lesions (p=0.0011) ( Figure 1F ).

To further examine BMP signaling as a potential targeted therapy for TIBD we first wanted to test the impact on bone health in non-tumor bearing mice. Healthy C57Bl6/j male mice at 8 weeks of age were assigned into one of four groups where they all received a 6-week duration osmotic pump containing either BMP antagonist DMH1 (30ug/ml) or a control DMSO pump alone or in addition to the anti-resorptive bisphosphonate Zoledronic acid (ZA) via intraperitoneal (IP) injection. Mice underwent DXA X-ray at six weeks and representative images of femurs revealed no deficiencies in bone ( Figure 2A ). Using Faxitron software to restrict bone mineral content (BMC) and bone mineral density (BMD) analysis of both femurs bilaterally for the five mice in all four groups we found BMP inhibition improved BMC (p=0.04) and trended towards increased BMD (p=0.11) when combined with Zoledronic acid treatment ( Figure 2B ). In order to determine whether changes to bone health were permanent or temporary after DMH1 treatment ended, we collected the femur and tibia bones for µCT and histological analysis after the mice persisted for another six weeks past the lifetime of the osmotic pumps. Representative images from micro computed tomography (µCT) scans of the femurs indicate the changes in trabecular architecture were not significantly altered and remained healthy ( Figure 2C ). Quantification of trabecular features such as trabecular thickness (Tb.Th), trabecular spacing (Tb.Sp) and trabecular number (Tb.N) were not significantly affected by the addition of DMH1 to bisphosphonate treated mice ( Figure 2D ). Representative H&E histology of tibias highlighted the healthy architecture of mice receiving either Zoledronic acid and/or DMH1 ( Figure 2E ). Cortical image slices of mice tibias also demonstrated healthy bone ( Figure S1A ) and no appreciative changes in cortical BMD in the tibia midshaft were observed with DMH1 addition to bisphosphonate treatment ( Figure S1B ). Because BMP signaling can have a variety of effects on circulating blood we wanted to see if lasting changes to the cell populations in the peripheral blood were altered. Complete blood counts (CBCs) on mice at time of euthanasia at 12 weeks revealed normal levels of white blood cells, lymphocytes, monocytes granulocytes, and red blood cells ( Figure S1C ). However, a significant yet small increase in platelets were observed in mice treated with DMH1.

Modeling experimental metastasis to bone is required with mCRPC as there are limited primary dissemination models from primary mouse prostate cancer (28). Approaches to colonize the bone in mouse models are typically divided into direct injection of a long bone such as a femur or tibia or injections into the heart or femoral artery can avoid the destruction and injury to trabecular bone. However, models of bone metastasis through the circulatory system face challenges for cells to engraft in the bone due to the high pressure of arterial circulation (29). A novel approach to model bone metastasis using a caudal artery injection in the tail of mice is able to utilize this arterial vasculature pressure to specifically flood the hind limbs of mice accurately and without causing an artificial injury to the bone (24). We chose to utilize the PC3 neuroendocrine like osteolytic human metastatic prostate cancer cell line with the caudal artery injection approach as the most reliable method to model vasculature mediated metastasis to the hind limbs of mice. We aimed to analyze the strong PC3 lytic phenotype in mouse TIBD and lack of tibial injury approach for bone metastasis to test the effect of BMP inhibition on bone health and metastatic engraftment. Because our data from Figure 2 suggested DMH1 would work best in concert with ZA and many patients receive anti-resorptive therapies, all mice received an IP injection of ZA. Nine successfully injected male NSG mice received 50,000 PC3 cells in circulation via the tail caudal artery. The next day mice were given an IP injection of Zoledronic acid and subcutaneous implant of a 4 week duration osmotic pump carrying DMSO (n=4) or DMH1 30ug/ml (n=5) then were euthanized upon pump expiration after 4 weeks ( Figure 3A ). Mice were scanned by DXA the day prior to euthanasia and revealed no significant changes to bilateral BMC or BMD by Faxitron analysis ( Figure 3B ). Representative images were able to identify small lesions areas present in some animals ( Figure 3C ). Analysis ex-vivo by µCT for femoral bone disease revealed no significant changes to bone volume, Tb.Th, Tb.Sp or Tb.N with DMH1 treatment ( Figure 3D ). Representative femoral images from µCT revealed some absence of trabecular features equally in both treatment groups ( Figure 3E ). Validation of tumor cell engraftment came from histological confirmation of tumors in bone. Only one of the four DMSO control mice did not show evidence of disease where only one of the five DMH1 treated animals showed evidence of disease revealing that the DMH1 had restricted engraftment to femurs and tibias inspected in all animals grossly and histologically ( Figure 3F ). Representative histology confirmed the lytic nature of PC3 cells that can be seen destroying trabecular bone as well as cortical bone which led to growth outside the periosteum ( Figure 3G ). Cortical areas of bone however were not common engraftment sites and did not lead to significant changes in bone volume ( Figures S2A, B ). Analysis of peripheral blood at the time of euthanasia revealed no significant changes in blood composition from DMH1 treatment ( Figure 3H and Figure S2C ).

The majority of clinical bone metastases present as established lesions upon diagnosis and treatments for TIBD are still required that can target existing bone disease. This led us to test a more injury-like established model of TIBD with the tibial injection of an immune competent syngeneic cell line MYC-CaP into FVBn mice. 100,000 MYC-CaP cells were injected into the left tibia of male eight week old FVBn mice (n=8) and PBS was injected into the right tibia as a control ( Figure 4A ). Four mice received a 4 week duration osmotic pump containing DMH1 30ug/ml 6 days following tibial injection and the remaining four other mice received a DMSO osmotic pump. Mice were scanned by DXA the day prior to euthanasia and revealed significant (p=0.01) changes to BMD by Faxitron analysis of reduced density in DMH1 treated mice ( Figure 4B ). Representative images were able to identify small lesions areas present in tibias injected with MYC-CaP cells ( Figure 4C ). Analysis by µCT of tibia features demonstrated a significant loss of BMD (p=0.03), BV/TV (p=0.02) and moderate reduction in trabecular features of metastatic tibias from the DMH1 treated mice ( Figure 4D ). Analysis of cortical bone did not significantly alter the appearance or BMD in the cortical tibias ( Figures S3A, B ). Interestingly, representative images of tibias revealed significant generation of tibial and cortical bone five weeks post tumor cell injections despite the injury imposed by the intra tibial injection model ( Figure 4E ). To measure tumor burden while avoiding immune mediated rejection to Luciferase or GFP labeled MYC-CaP cells we imaged glucose uptake by IVIS with a 750nm 2-Deoxyglucosone probe and found that tumor engraftment was not significantly changed and possibly moderately increased in the DMH1 treated animals ( Figure 4F ). Histological analyses of tibial lesions revealed the metastatic progression was limited to the trabecular region injected, with no lesions progressing deep into cortical bone ( Figure 4G ) and lungs examined grossly and histologically did not indicate metastasis from the bone (data not shown). In the immunocompetent mouse model of TIBD we observed significant changes to peripheral blood composition at time of euthanasia with DMH1 resulting in significant decreases in white blood cells (p=0.002) and lymphocytes (p=0.001) ( Figure 4H ). Furthermore, peripheral blood was altered in DMH1 treated animals by significant reductions in red blood cells (p=0.001) and platelets (p=0.01) at time of sacrifice ( Figure S3C ). Finally we sought to investigate the canonical BMP and TGFβ Smad signaling response on these tumors as previously performed in clinical samples in Figure 1 . We found that DMH1 did not reduce the canonical BMP-Smad1/5/8 signal but significantly (p=0.04) increased the staining in the bone lesions ( Figures S4A–C ). Canonical TGFβ signaling via pSmad3 was also significantly (p=0.03) elevated in DMH1 treated tumors, potentially highlighting the efficacy of DMH1 in vivo ( Figures S4D–F ).