We have shown that TSGs derived from HRPCa recapitulate characteristics of parent tumors including histopathological features, biomarker expression, and responses to ADT. Our study differs from previous reports [21–26] of generation of prostate tumorgrafts in several ways. First, we implanted precision-cut tissue slices rather than minced tissues as used in previous studies. The ability to reliably determine the presence and grade of cancer prior to implantation is one of the advantages of this methodology, which in turn allows more accurate assessment of therapeutic effects in large-scale animal trials. For example, our study design ruled out the possibility that “bad” tumors were missed in the two cases that showed complete regression after castration. Because we used only slices that were in-between two slices containing high-grade PCa and assigned adjacent tissue slices to control and ADT groups, it is unlikely that the “bad” tumors were missed only in the ADT but not the control group. If minced tissues were used without prior knowledge of histopathology, it would be impossible to know whether the complete regression was just by chance, i.e., the bad tumors were missed or there was no cancer present initially in the ADT group. Second, we focused on high-risk tumors, the major contributor to PCa mortality, rather than benign or low-risk tumors that have been used in most previous studies. Third, we systematically evaluated the effects of ADT, a standard treatment for HRPCa, in our TSGs. Although the number of cases in our study is small, it is the largest cohort of HRPCa evaluated for response to ADT to date in a preclinical model.
HRPCa is the target of adjuvant and neoadjuvant therapies since low-risk PCa is largely curable by surgery or radiation or needs no treatment. A growing inventory of new agents has been discovered that may improve the clinical outcome of HRPCa. Clinical trials evaluating such candidate compounds require a large number of patients, are expensive and time-consuming, and expose patients to certain risks. The TSG model of HRPCa provides a much-needed pre-clinical screening platform that can be used to rapidly narrow down the number of agents or regimens for further investigation in clinical trials. The authenticity of the model in recapitulating the features of the parent tumors increases confidence in the likelihood of similar drug responses in humans. In addition, our study demonstrates the feasibility of generating a relatively large number of TSGs from the same HRPCa specimen with similar cell composition and histology among control and experimental samples in an in vivo setting. This capability is particularly useful since PCa specimens are becoming smaller due to early cancer detection. Our model can be used to test a variety of therapeutic strategies, including potential curative therapies for HRPCa that can either prevent CRPC from arising during ADT or kill CRPC cells after disease progression. Since ADT may be associated with numerous side effects such as increased cardiovascular mortality, other alternative therapies should also be investigated . Finally, our model can be used to better understand the mechanisms of development of CRPC, which will in turn accelerate the discovery of effective therapies.
As proof-of-principle, we have demonstrated that our model closely mimics the response of PCa in humans to ADT. First, ADT decreased cell proliferation and reduced graft weight of TSGs. Second, ADT downregulated the conventional AR target gene PSA while selectively upregulating CCNA and TOP2A in CR-TSGs, as in human PCa , suggesting that the TSG model is a suitable platform for pre-clinical testing of the ever-growing number of new therapeutic agents that aim to better prevent AR activation in CRPC. Third, consistent with recent studies highlighting a role for EMT after ADT in facilitating human PCa progression and metastasis [14, 40, 41], cancer cells in CR-TSGs exhibited EMT by simultaneously expressing both mesenchymal and epithelial cell markers, VIM and K18, respectively. In addition, E-cadherin was mislocalized away from cell membranes into the nuclei in CR-TSGs, presumably disrupting the function of E-cadherin in preventing beta-catenin from entering the nucleus . Such mislocalization was recently observed in a metastatic colorectal cancer model in which E-cadherin nuclear translocation was associated with aggressive focal growth , suggesting that mislocalization of E-cadherin may be a general mechanism of cancer progression. The documentation of ADT-induced EMT in CR-TSGs derived from HRPCa suggests an attractive model for testing novel therapeutics aimed at blocking EMT.
Our findings are the first to link seminal vesicle invasion, positive surgical margin and extracapsular extension to lack of complete pathologic response to ADT by HRPCa. The efficacy of neoadjuvant ADT in the TSG model appears much better than in patients determined by histology [43, 44]. Since the presence or absence of tumor cells in TSGs was evaluated one month after castration, we can’t rule out the possibility that the regressed tumors might relapse at later time points. In addition, most studies show a lower serum testosterone level in castrated mice than in humans [26, 45–47], possibly because unlike in humans, adrenal glands in mice do not produce androgen [48–50]. Thus, castration of mice may more effectively eliminate HRPCa cells in TSGs than does ADT in humans. Further experiments are needed to determine the long-term effects of ADT and to investigate the possibility of serial passage in this model. Mechanisms of resistance to therapy can be explored, such as the role of stem cells in castration-resistance.
It is interesting to note that endothelial and stromal cells in TSGs are mostly of human origin, rather than replaced by their host counterparts. This is consistent with a recent report demonstrating a burst of angiogenesis by endogenous human blood vessels in primary xenografts of benign prostate or PCa tissues that occurred between days 6–14 after transplantation into SCID mice pre-implanted with testosterone pellets . In contrast, DeRose et al. demonstrated that, in human breast tumorgrafts, cancer-associated stroma and endothelial cells from the original tumor were largely replaced by mouse-derived stroma and endothelial cells . This difference may be due to the disparate growth properties of these two types of cancers − PCa is a slow-growing cancer with a long natural history, whereas breast cancer is much more aggressive. The slow-growing nature of prostate TSGs perhaps makes it unnecessary to incorporate host stromal and endothelial cells in the grafts. In our study, human stromal and endothelial cells survived up to 2 months in TSGs derived from HRPCa. It would be interesting to determine whether the human endothelial and stromal cells would eventually be replaced by their mouse counterparts in long-term follow-up of these grafts.