How the TRAMP Model Revolutionized the Study of Prostate Cancer Progression

See related article by Gingrich et al., Cancer Res. 1996;56:4096–102.Visit the Cancer Research 75th Anniversary timeline.FDA approval for prostate-specific antigen (PSA) testing in 1994 as a screening aid led to a dramatic increase in the diagnosis of prostate cancer cases. Yet, researchers had few tools to study this disease beyond a handful of human cell lines (mainly LNCaP, PC-3, and DU145), pedigrees of rat carcinogen–enhanced cell lines developed from autochthonous prostate cancer tumors (including the Dunning, Noble, and Pollard series), in vivo growths of prostate cancer/mesenchyme tissue recombinants, and human prostate cancer xenografts (such as CWR22). None of these models, however, recapitulated all the parameters of prostate cancer progression in human disease, including formation of initial hyperplasia, intraepithelial neoplasia (PIN), adenocarcinoma, metastasis to peripheral organs including local lymph nodes and bone, and recurrence from androgen deprivation therapy (1).The few attempts to produce transgenic (Tg) mouse models of prostate cancer in the early 1990s suffered from a lack of prostate specificity (2, 3) or from use of oncogenes irrelevant to typical human prostate cancer oncogenesis (4). An early attempt to increase prostate specificity, by using the PSA promoter to drive expression of a Ras oncogene, paradoxically did not cause prostate cancer, but rather cancer of the salivary glands and gastrointestinal tract (5). This changed, however, in 1994 when Maroulakou and colleagues (6) described Tg mice expressing the SV40 T-antigen (Tag) driven by the rat prostatic steroid binding protein C3 gene promoter. These mice develop prostate hyperplasia that progress to PIN or adenocarcinoma lesions after about 8 months of age. Only one case of metastasis (to the lung) was reported. Interestingly, females of this stock had more aggressive disease, succumbing to mammary adenocarcinomas, and in a few cases, lung metastases after 6 months of age. The slow prostate cancer progression and lack of metastatic potential, however, made this Tg model suboptimal.Nature abhors a vacuum, and into this void came what is now referred to as the transgenic adenocarcinoma mouse prostate (TRAMP) model, whose ability to form prostate cancer was first described in a 1995 Proceedings of the National Academy of Sciences article (7) and then characterized more extensively for its metastatic potential in Gingrich and colleagues (8). Unlike the model of Maroulakou and colleagues, TRAMP drives SV40-Tag oncogene expression by the rat probasin (rPB) gene promoter. Specifically, the notion that rPB expression was specific to the prostate epithelium was strengthened by a decade of research from Robert Matusik's laboratory, culminating with the observation that the −426 to +28 rPB promoter region was sufficient to drive androgen-responsive expression (9). The clinical relevance for the use of SV40-Tag, which induces oncogenic progression by binding to and inactivating the Trp53 and Rb1 tumor suppressors (10), was based on previous data showing loss of p53 and Rb in human prostate cancer (11, 12).Because transgene chromosomal insertion sites as well as strain background were known to influence transgene expression levels and phenotypic penetrance (13), initial production of TRAMP mice included eight independent founder lines in both C57BL/6 and FVB backgrounds. Line 5666, which exhibited low Tag immunohistochemical staining levels in the prostate epithelium as well as restricted expression to the dorsal and ventral lobes of the prostate (compared with other prostate lobes and body organs, assessed by RT-PCR), developed only epithelial hyperplasia and low-grade PIN after 33 weeks. In contrast, line 8247 exhibited PIN lesions by 10 weeks of age, hyperplasias in all prostate lobes by 22 weeks, and invasive adenocarcinomas by 20 weeks of age (14). The Cancer Research article by Gingrich and colleagues (15), where this model was first named TRAMP, followed mice over roughly 30 weeks for pathologic and immunoreactive changes. Several common features were observed. First, in the initial group of 17, all mice exhibited prostate neoplasia between 18 and 24 weeks, characterized by cribriform structures and frequent apoptotic bodies. Second, neoplastic prostates were marked by stromal remodeling, specifically, hyperproliferation of the fibromuscular stroma. This observation, now described as reactive stroma, is appreciated for the role played by the prostate tumor microenvironment in prostate cancer progression (16). Third, although epithelial Tag expression always preceded neoplastic transformation, individual epithelial cells exhibited variations in expression levels, especially when comparing adjacent acini in the same mouse. This suggested cell-to-cell transgene heterogeneity, which might affect which cells (and acini) might progress oncogenically within a lobe. Whether a similar phenomenon of variation in oncogene expression underlies the fact that a large portion of human prostate-confined disease is multifocal yet marked by shared potential driver mutations (17) is not yet clear.The most profound breakthrough of the TRAMP article was in the frequency and breadth of metastases formed. Although this article studied small cohorts, high frequencies of macrometastases were identified in lymph nodes and lungs (roughly a third of cases at 24 weeks of age, rising to between two-thirds to 100% of cases at 28 weeks), and lower frequencies in the kidney, adrenal glands, and bone (spine). Bone metastasis, marked by characteristic osteolytic histology, was only found in a single [TRAMP × FVB]F1 mouse at 22 weeks of age, whereas no bone lesions were detected in a 34-week-old [TRAMP × C57BL/6]F1 mouse. A subsequent study by Gingrich and colleagues (18) with larger cohorts confirmed that TRAMP mice in the FVB background suffered more aggressive pathobiologies compared with those in the C57BL/6 background. The initial prostate tumors and metastases arising in the 18- to 24-week old mice were identified as either HG-PIN or adenocarcinomas, marked by reduced or absent E-cadherin staining, a hallmark of prostate cancer progression in humans (19). Greater than 40% of TRAMP mice castrated at 12 weeks, which is either prior to overt neoplastic onset or at a PIN stage, developed recurrent tumors that were much more poorly differentiated than the tumors arising in non-castrates (20), and castration increased the incidence of metastasis formation in the liver, salivary glands, and calvaria (21). Of note was the persistence of androgen receptor (AR) expression during disease progression to adenocarcinoma (18).A provocative finding regarding the pathobiology of TRAMP mice is that poorly differentiated tumors arising in >28-week-old mice, or at a higher frequency in castrates, expressed neuroendocrine markers such as synaptophysin (21). Neuroendocrine (NE) prostate cancer is rare, comprising less than 2% of all cases, and likely arises from rare, AR-negative NE cells normally found within the basal epithelial layer in acini (22). However, at least 50% of the synaptophysin-positive TRAMP tumors also expressed AR, suggesting a potential transdifferentiation phenomenon. This phenomenon has been described in patient-derived xenograft models forced into castration-resistant (CR) growth (23), or in CR prostate cancer (PC) patients after prolonged therapy with AR antagonists (24). This also may be influenced by the TRAMP strain background, because CR tumors in the FVB background seemed to be formed by NE lineage cells independent of those forming epithelial cell lesions (25, 26).In the decade after the introduction of TRAMP mice, several groups addressed how the SV40-Tag and the nature of the PB promoter affected the incidence of prostate adenocarcinomas, NEPC, and metastases. For example, Masumori and colleagues produced a Tg model expressing only the so-called “large Tag” (versus the “small t-ag” transcript included in the TRAMP construct) whose expression is driven by a larger, 12 Kb PB promoter that contained additional androgen and growth factor–responsive sequences. This LPB (a.k.a. LADY) model produced PIN lesions, adenocarcinomas and metastases that expressed NE markers starting after 24 weeks, but at a much lower frequency than in TRAMP mice (27). Zhang and colleagues (28) sought to address the fact that the rPB promoter in the TRAMP model had weak prostate epithelial expression, and thus, they produced a composite PB promoter (Arr2pb), which removed the inhibitory sequences from −426 to −287, and then introduced a second copy of the androgen response region (−244 to −96). Tg mice with Cre recombinase driven by the Arr2pb promoter showed strong expression in the lateral lobe, with lesser expression in the ventral, dorsal, and anterior lobes, and very low, but discernable expression in the seminal vesicles (29). This Tg model has been used extensively to conditionally knock out many genes in the prostate, although a caveat is that the expression of the transgene in oocytes means that the Cre allele must be transferred through males only. One such use of the Arr2pb-Cre was to produce a conditional prostate-specific knockout of Tp53 and Rb1 (30) model, which exhibited both adenocarcinomas and NEPC, confirming that the oncogenic transformation induced by Tag in the TRAMP model was likely through the inactivation of these two tumor suppressors.TRAMP mice have been used extensively to dissect the roles played by specific pathway mediators, transcription factors, or metabolic pathways in prostate cancer progression. An early example was the cross of TRAMP to mice lacking the transcription factor, Egr-1 (early growth response protein-1; ref. 31), such that the loss of Egr-1 had no effect on the rate or time to neoplastic transformation, but blocked progression from PIN to adenocarcinoma. Ablation of PKCϵ severely decreased adenocarcinoma and metastasis formation in TRAMP mice (32). Our group addressed the role of the Src family kinases, whose activity is known to increase in prostate cancer progression (33), by crossing TRAMP mice with Src-, Lyn-, or Fyn-null mice (34). Whereas the loss of Fyn or Lyn had little or no effect on adenocarcinoma or NEPC formation, the loss of Src severely impaired the frequency and time-to-onset of adenocarcinomas. A side note of this study is that in C57BL/6 TRAMP mice aged past 30 weeks, a small percent (3%) of mice present as “tumor-free with metastases,” that is, mice with PIN lesions yet paradoxically, exhibit extensive pulmonary metastases that are predominantly adenocarcinomas. This suggests that, at least in some cases, prostate metastases derive from rare, aggressive cells that disseminate early. We found that the loss of Src, Fyn, or Lyn increased the frequency of this population >5-fold, strongly suggesting that these kinases might play a secondary, suppressive role in specific pathways of metastatic progression.In the two decades since the publication describing the TRAMP model, many groups have produced multiple Tg models of prostate cancer, incorporating other well-documented changes found in human prostate cancer, such as Myc upregulation or PTEN loss. However, the TRAMP model has exhibited amazing staying power in the field, even surviving its association with NEPC, which was originally thought to relegate it to a rare prostate cancer pathology, but which is now gaining appreciation as a significant mechanism of disease progression arising from continued AR antagonist therapy.No potential conflicts of interest were disclosed.I wish to thank David Goodrich and Barbara Foster for constructive comments.