Our goal was to develop a suitable system for recapitulating human ovarian carcinoma development and progression. The system was designed such that tumors are induced in adult animals, similar to the occurrence in the majority of naturally developing human tumors. Furthermore, the system was designed to be used for the evaluation of multiple genetic lesions, individually and in combination thus increasing the ability to simulate a wide spectrum of human ovarian tumors without extensive breeding protocols.
We used the TVA receptor-dependent avian RCAS retroviral delivery technique to introduce defined multiple genetic lesions into mouse ovarian cells. This delivery technique has recently been adapted for transferring genes into specific mouse cells from transgenic mice programmed to express the avian TVA receptor in various tissues (Federspiel et al., 1994
; Holland et al., 1998
; Holland and Varmus, 1998
; Doetsch et al., 1999
; Murphy and Leavitt, 1999
). We demonstrated that the RCAS-TVA system allows efficient gene transfer to cultured primary ovarian cells. The genes can be transduced to ovarian cells individually or in combination. The ovarian cells that are susceptible to infection with RCAS viruses can be limited by cell-specific expression of the TVA receptor. β-actin-TVA mice (Federspiel et al., 1996
) were used to infect both ovarian epithelium and stroma, while keratin 5-TVA mice were used to restrict the infection to the cells of the ovarian surface epithelium. The major drawback of this system is the necessity for propagation of ovarian cells in culture in order to assure efficient infection with multiple RCAS vectors. However, since the cells are propagated in culture for only a short period of time it is unlikely that spontaneous mutations would occur. Another deviation from naturally occurring tumors that arise through clonal expansion is that multiple cells in this system are potential tumor precursors. Although there is an experimental advantage to inducing rapid tumor proliferation in a large number of cells with the same genetic lesions, this system does not entirely reflect the natural development of tumors.
Ovarian carcinomas are thought to arise from the epithelial lining that covers the surface of the ovary. Alternatively, ovarian carcinomas may arise from other cell types within the ovary, or from the components of the secondary Müllerian system that are not part of the ovary (Dubeau, 1999
). Our mouse model system is well suited to identify the tissue of origin for the ovarian carcinomas we study. First, we were able to induce genetic lesions in isolated ovaries, thus ensuring that only ovarian cells contribute to tumor formation in our model. Second, we were able to use the keratin 5-TVA system to induce oncogenic changes specifically in the cells of the ovarian surface epithelium. Using the TVA receptor as a marker for the ovarian surface epithelial cells, we have demonstrated that the ovarian tumors in the keratin 5-TVA model system arise from the ovarian surface epithelial cells. Surprisingly, ovarian stromal cells that were transduced with oncogenes after infection of cells from β-actin-TVA transgenic mice did not contribute to the tumor (). There has been much speculation as to why the surface epithelial cells are the ovarian cells that most frequently give rise to ovarian tumors (Auersperg et al., 2001
). Our results with cells propagated in vitro indicate that p53−/−
ovarian epithelial cells proliferate more slowly than stromal cells. However, in the presence of oncogenic lesions, the proliferation of epithelial cells is more rapid than the proliferation of stromal cells. It is possible that the combination of oncogenic lesions that we used in this study specifically induces tumorigenic changes in the surface epithelial cells and not in the cells of the ovarian stroma. Because some residual epithelial components of the secondary Müllerian system may exist within the ovarian hilum and medulla (Dubeau, 1999
), we cannot completely rule out the possibility that the tumors in mice are derived from the secondary Müllerian epithelium instead of the ovarian surface epithelium. However, this is very unlikely since the oncogene-induced proliferation of the morphologically distinct ovarian surface epithelial cells is already apparent in vitro.
Several studies indicate that ovarian tumors are the end result of a complex pathway involving multiple oncogenes and tumor suppressor genes, which include c-myc
, and BRCA1/2
(reviewed in Gallion et al., 1995
; Berchuck and Carney, 1997
; Lynch et al., 1998
; Aunoble et al., 2000
). In our study, we introduced c-myc
oncogenes into cells from p53+/+
mice in an attempt to model the genetic aberrations that characterize many human ovarian carcinomas. A combination of three oncogenes (c-myc
, and Akt
) is not sufficient to induce a tumorigenic state in ovarian cells from p53+/+
mice, indicating the importance of the p53
lesion. Our results suggest that the absence of p53
predisposes ovarian epithelial cells to tumor formation while in the presence of another initiating event, such as a growth signal produced by a mutant or an overexpressed oncogene. We determined that at least two oncogenic changes are required for rapid tumor formation in the p53-deficient ovarian cells (). Our preliminary results suggest that deletion of other tumor suppressor genes, such as INK4a/ARF
, also contribute to ovarian tumor development when combined with oncogenic lesions in c-myc
, or Akt
(S.O., unpublished). Thus, the necessity for a proliferative signal balanced with an anti-apoptotic signal may be a prerequisite for tumor formation in the ovarian cells.
Although changes in the function of p53 are commonly seen in ovarian carcinomas, patients with Li-Fraumeni syndrome rarely develop ovarian carcinomas, suggesting that a mutation in p53 is probably an essential step in ovarian carcinogenesis, but not the initial event. Although we have not addressed the temporal sequence of genetic events in ovarian carcinoma initiation and progression, the retroviral gene delivery system allows for multiple sequential infections with different RCAS vectors. Thus, multiple oncogenes, dominant-negative tumor suppressor genes, or the gene encoding the Cre recombinase, could be sequentially introduced into ovarian cells. Additionally, the use of inducible promoters in the RCAS system could provide the means to study the dependence of tumor growth on sustained production of a tumor-inducing oncogene.
Thus far, we have employed this model system to improve our understanding of the genetic changes leading to the initiation of ovarian cancer and to identify cooperating events in malignancy by using several different combinations of genes that are thought to contribute to cancer development. Unlike other existing mouse models that utilize mouse ovarian cell lines that are spontaneously transformed during prolonged growth in culture (Adams and Auersperg, 1981
; Kido and Shibuya, 1998
; Roby et al., 2000
), our model system uses genetically defined combinations of lesions. We have focused on p53
, and Akt
genes because lesions in these genes are commonly present in human ovarian carcinoma samples. However, we have not tested other genes, such as HER-2/neu
β, met, and BRCA 1
, that have been shown to play a role in ovarian tumorigenesis. The system that we established can also be used to test the oncogenic potential of many other genes, in order to identify those that might contribute to the initiation and progression of ovarian cancer.