The complications and high mortality of prostate cancer (PCa) are primarily due to the development of distant metastases [1
]. Although it is well established that PCa cells frequently metastasize to the bone [3
], the causes for its preferential homing to the bone organ are not fully understood. This may be related to limitations of current preclinical PCa metastasis models which frequently fail to recapitulate tumor formation at the orthotopic site, and hence overlook the influence of the cellular and extracellular primary tumor microenvironment on PCa growth and metastatic priming, prompting the development of alternative animal models recapitulating the full metastatic cascade.
Of late, the local tissue microenvironment has gained interest as an important niche for primary and metastatic tumor growth and development [4
]. PCa cells have been shown to interact with the surrounding cellular and extracellular microenvironment in both the local and metastatic milieu [7
]. In particular, human cancer-associated fibroblasts (CAFs), as well as lymphatic and blood vessel endothelial cells (BVECs) have been reported to stimulate the proliferation and metastasis of malignant epithelial cells in vivo [10
], indicating the significance of the tumor microenvironment in disease modelling.
We have previously demonstrated the importance of human specific cell-cell interactions within a murine host by generating a humanized tissue-engineered bone construct (hTEBC). The hTEBC was employed as a platform to study the interaction of human cancer cells with a human bone environment following direct intraosseous injection [14
], or to study species-specific homing to the bone organ following intracardiac injection of cancer cells [16
]. However, a humanized orthotopic xenograft model of PCa bone metastasis that incorporates the full metastatic cascade has been a quest from a clinical point of view.
Current in vivo mouse models of PCa bone metastasis are typically rooted in cancer cell injection into the blood circulation. This bypasses the first critical steps in the metastatic cascade, where establishment of the primary tumor is followed by local invasion, intravasation of the circulatory system and extravasation to distant metastatic sites, such as bone [1
]. Furthermore, from a clinical and therapeutic perspective, mouse models of PCa metastasis generally do not account for the important interaction of cancer and stromal cells in the prostate gland or of bone metastases. Therefore, we hypothesized that a preclinical model could be established by applying tissue engineering principles and that this would allow studies of species-specific bone metastasis from cancer cells primed with human stromal cells at the orthotopic tumor site. In the current study, we report spontaneous metastasis of human PC3 cells to a humanized bone from an orthotopic xenograft model of human PCa. This original model provides an important platform to study species-specific metastasis of PCa to the bone and will have important implications in preclinical PCa metastasis studies as well as the development and testing of therapeutic strategies related to personalized medicine.
Bone is the most common site of PCa metastasis. Understanding bone-PCa interactions is essential for dissecting the organotrophic homing mechanisms and developing novel therapies [1
]. To date, several in vivo models of PCa bone metastasis have been established, including patient-derived xenograft models and more recently, humanized animal models (reviewed by [1
]). Previous works from our team and others have shown that PC3 cells home to an hTEBC in an experimental metastasis scenario [16
]. Implantation of human PCa cells intravenously or directly into the human bone, developed by transplantation of fetal human bone fragments, showed tumors only established in the human bone and not in the murine bones. It was suggested that molecules such as growth factors and their receptors, adhesion molecules, chemotactic factors, proteinases, that regulate bone metastatic cascade are species-specific and contribute in different results in xenograft models of bone metastasis [30
Furthermore, the efficacy of tissue-engineering approaches to study species-specific cancer-bone interactions was investigated in a humanized hematochimeric mouse model of breast cancer metastasis from an orthotopic site to a humanized bone organ [20
]. Others have failed to develop tissue-engineered bone construct models for cancer metastasis, as the engineered tissues have only randomly organized mineralized matrix and do not contain a bone marrow, mature matrix embedded osteocytes or human cells [29
]. Interestingly, Seib et al. (2015) demonstrated metastasis of PC3 cells from an orthotopic site to a TEBC utilizing a BMP-functionalized silk scaffold [33
]. In this instance, the randomly organized scaffold architecture was populated with host (murine)-derived cells and did not account for species-specific metastasis as in the current study. Furthermore, the authors manipulated the silk surface to contain receptor activator of nuclear factor K-B ligand (RANKL), which increased PC3 metastasis to the TEBC [33
]. In contrast, to our knowledge we are the first to report a model of PCa bone metastasis that recapitulates fully humanized metastasis from a prostate microenvironment to a bone organ.
Recent studies highlight the critical role of stromal cells during primary tumor development and metastasis [5
]. CAFs are one of the major stromal cell components of the tumor microenvironment and reciprocal feedback loops between CAFs and cancer cells have been suggested. Furthermore, CAFs may induce resistance of cancer cells to therapy [6
]. Giannoni et al. (2010) showed the enhancement of tumor growth and development of spontaneous lung metastases when PC3 cells were co-injected subcutaneously with CAFs isolated from PCa patients [35
]. Moreover, a role for CAFs in the development of the metastatic niche through remodeling of the tumor microenvironment and secretion of soluble factors has been reported, whereby they promote the metastatic spread of tumor cells and de novo angiogenesis [35
]. Although the impact of CAFs in cancer progression has been shown, their functional contribution to the metastatic process to a humanized bone remains to be investigated. Here, our results showed that humanization of the murine prostate tumor microenvironment, created by co-injecting PC3 prostate cancer cells with CAFs and BVECs, did not influence PC3 primary prostate tumor growth (Figure 2
). Unexpectedly, PC3 cells preferentially metastasized from the non-humanized primary tumors to the hTEBC (Figure 3
). Furthermore, PC3 cells disseminated to the murine skeleton at the same frequency and intensity from both the humanized and non-humanized prostate microenvironments (Figure S2
Interestingly, we discovered that PC3 cells metastasized to the soft tissue organ systems, and that liver and GI tract metastases were enhanced from the humanized prostate microenvironment compared to the non-humanized prostate tumors (Figure 4
E,F,I,J). There are reports showing that CAFs may promote or inhibit cancer cell metastasis [38
]. The contradicting results might derive partly from the method used for humanization. Additionally, the differences could be because of the origin of fibroblast and cancer cell types they used. While previous studies employed CAFs or BVECs for humanization, in the current study humanization was performed by co-injection of CAFs and BVECs and PC3 cells. Further investigation is needed to determine the impact of each cell in cancer metastasis. Clearly, these data suggest that the tumor cancer-associated stroma may affect cancer metastasis.
Stromal cells communicate with the surrounding tumor tissue and influence metastasis to distant sites in a phenomenon termed ‘metastatic organotropism’ [26
]. Zhang et al. (2013) reported that breast cancer stroma, rich in CXCL12 and IGF-I secreting mesenchymal cells, selects for a sub-population of cancer cells that have a predisposition to metastasize to bone [22
]. Moreover, stromal-produced CXCR4/CXCL12 has been suggested to be crucial for priming breast cancer cells to metastasize to the liver [42
]. Co-injection of fibroblasts has been shown to enhance the engraftment of mammary epithelial cells in vivo [11
], which contrasts with a previous report that fibroblasts inhibited the growth of malignant breast tissue [43
]. Tuxhorn et al. (2002) shed further light on these conflicting results and described the influence of stromal cells co-engrafted with LNCaP PCa cells as patient-specific; some patient-derived stroma enhanced LNCaP tumor formation and host vascular recruitment, while other patient-derived stroma had no measurable influence [7
]. Overall, the cancer-associated stroma is important in priming tumor cells for bone metastasis, yet the specific stromal factors which enhance bone organotropism are still to be identified.
Orthotopic models are a cornerstone in in vivo cancer models as they recapitulate most of the physiologically relevant steps of cancer development and metastasis. Experimental metastatic animal models developed by tail-vein injection or intra-cardiac implantation of cancer cells cannot be used as ideal models to investigate the primary tumor or the early steps in the metastatic cascade. To date, numerous orthotopic PCa mouse models have been reported [44
]. Intraprostatic implantation of cancer cells allows characterization of molecular and cellular events at the primary tumor and the cross-talk between tumor and microenvironment. However, these models have rarely demonstrated successful bone metastases from the orthotopic niche without serial in vivo passaging [47
]. Disseminated cancer cells from murine prostate tumors frequently home to the lymph nodes, lungs and liver [48
], as well as to the spleen and kidneys [24
], similar to our findings (Figure 4
C,D,G,H). Here, we injected PC3 cells into the dorsal lobe of the murine prostate to simulate a primary tumor in the native site and to investigate bone metastases. We chose PC3 cells as they were derived from a PCa bone metastasis. We found that the humanized prostate microenvironment did not change primary tumor growth or metastasis to the murine skeleton, but there was preferential metastasis from the non-humanized prostate to the hTEBC. One of the limitations of our model was that the study needed to be terminated 4 weeks after cancer cell injection due to the development of lethal metastases in the murine organs, a known restriction in in vivo bone metastasis studies [49
]. Surface colonization, which did not appear to have penetrated into the organ, was detected on the kidneys and the GI tract (Figure 4
G,I,K). These organs are located adjacent to the prostate tumor microenvironment and may have been colonized by PC3 cells due to their proximity to the primary prostate tumor. In future studies, we aim to create an orthotopic prostate tumor with a less inherently invasive cell line to recapitulate tumor-stroma interactions over a longer timeframe than 4 weeks. Lastly, understanding the mechanisms involved in reduced homing of PC3 to hTEBC needs further investigation.