|Home | About | Journals | Submit | Contact Us | Français|
Papillary differentiation is one of the most common histological features of ovarian cancer, although the underlying mechanism that leads to such differentiation is not known. We hypothesized that human ovarian surface epithelial cells can be transformed into carcinoma with papillary differentiation by overexpressing HER2/neu in these cells. Mice were injected either subcutaneously or intraperitoneally with two immortalized human ovarian surface epithelial cell lines after enforced expression of HER-2/neu. Mice subcutaneously injected with tumor cells from either the T29Nt or T80Nt developed undifferentiated carcinomas. In contrast, mice injected intraperitoneally with T29Nt cells developed papillary carcinoma, and those injected intraperitoneally with T80Nt cells developed undifferentiated carcinoma. Our results demonstrate that ovarian surface epithelial cells can develop into papillary carcinoma in mice, and that the induction of papillary differentiation depends not only on specific genetic modifications but also on the tumor microenvironment and epithelial cell type from ovary from different patients.
Ovarian cancer is the most lethal form of cancer among women in the United States.1 Epithelial ovarian cancer encompasses a heterogeneous group of cancers with several distinct histologies. Serous carcinoma is the most common histotype of epithelial ovarian cancer and is often accompanied by papillary differentiation. The development of specific histotypes of human ovarian cancer is thought to be influenced by multiple factors, including genetic and epigenetic changes in epithelial cells. The systemic microenvironment may further dictate the formation of specific histotypes. The expression of lineage-specific homeobox genes has also been shown to contribute to the pattern of histologic differentiation,2 but the processes controlling the expression of these genes remain elusive. In addition, it has been difficult to retrospectively identify the cell of origin in clinical samples because the original cell type changes from its original state during tumor development. Because of these complicating factors, the molecular basis for the histological differentiation of epithelial ovarian cancer is still unclear, despite the fact that these morphological features have been recognized for decades.
In a previous study, we successfully transformed human ovarian surface epithelial cells using a set of combined genetic elements, including the SV40 early genomic region, the catalytic subunit of human telomerase reverse transcriptase (hTERT), and hHRAS or KRAS.3 In that study, RAS-transformed human ovarian surface epithelial cells developed into tumors when they were subcutaneously or intraperitoneally injected into nude mice. However, these tumors remained undifferentiated to a large extent.
HER2/neu gene encodes a 185 kd transmembrane protein with intrinsic tyrosine kinase activity. Amplication of this gene has been found in 25–30% breast and ovarian cancer and correlate with poor prognosis.4 HER2/neu has been shown to be involved in the progression of ovarian cancer through multiple mechanisms.5 High expression of HER2/neu is associated with advanced cancer stages, higher frequency of surgical residual larger tumor, poor histological grade, higher recurrence frequency, shorter survival time and lower sensitivity to platinum-based chemotherapy.6 We hypothesized that overexpression of HER2/neu in ovarian surface epithelial cells may involve in the initiation of tumor growth and tumor developed from individual patient may have different histological differentiations. Our results show that HER2/neu converts one of the two cell lines tested into papillary carcinoma in the peritoneal microenvironment.
A schematic of the experimental design is shown in Figure 1. T29 and T80 cells were first transduced with HER2/neu, and the resulting cells were injected subcutaneously into mice to observe tumor growth. To generate the T29Nt and T80Nt cell lines, cells were isolated from the T29N- and T80N-derived xenografts and reinjected subcutaneously or intraperitoneally. The resulting tumors were analyzed for histopathologic characteristics and immunophenotype. The introduction of HER2/neu cDNA into two immortalized nontumorigenic human ovarian surface epithelial cell lines (T29 and T80) resulted in increased expression of the HER2/neu protein (Fig. 2). The resulting cells overexpressing HER2/neu were designated T29N and T80N. HER2/neu overexpression increased the number of anchorage-independent colonies that grew in soft agar in both the T29N and T80N cell lines (Fig. 2B). The subcutaneous injection of T29N and T80N cells into nude mice led to slow-growing tumors that reached 1.0 cm in size in approximately 4 months (Fig. 2C). We did not observe any tumors in the six mice injected intraperitoneally after 6 months observation (Table 1). Tumor cells from the xenografts were cultured, and the isolated tumor cell lines were designated T29Nt and T80Nt.
The T29 cells have mesenchymal morphologic characteristics, and the T80 cells have an epithelial phenotype consistent with that of the parental cells, IOSE-29 and IOSE-80, as previously described.7,9 However, the tumor cells isolated from T29N- and T80N-derived xenografts (T29Nt and T80Nt) all had an epithelial phenotype (Fig. 2D). T29Nt and T80Nt cells were reinjected into the subcutis and peritoneum of different nude mice, and the mice were observed for 6 months. Three of the six mice subcutaneously injected with the T29Nt and T80Nt cell lines respectively developed subcutaneous tumors. Histopathologic examination of these tumors revealed undifferentiated carcinomas (Fig. 3A–D). Three of the ten mice injected intraperitoneally with T29Nt and five of the nine mice injected with T80Nt developed tumors. Intraperitoneal inoculation of mice with T29Nt cells resulted in extensive tumor growth along the omentum (Fig. 4A). Interestingly, histologic examination revealed papillary carcinoma (Fig. 4B–G) that was indistinguishable from human papillary serous ovarian carcinoma. These tumors also invaded the liver (Fig. 4H) and pancreas (Fig. 4I). Tumors derived from the T80Nt cells grew along the peritoneal fat (Fig. 5A) and the pancreas (Fig. 5E and F) but remained undifferentiated (Fig. 5B–D).
To characterize the immunophenotype of the T29Nt-, T80Nt-derived tumors, we evaluated their immunoreactivity for several common human ovarian cancer markers. Immunohistochemical staining for large T antigen was positive (Fig. 6A and B), demonstrating that the tumors were derived from the immortalized T29 and T80 cells rather than having developed spontaneously in the mice. The tumor cells were also positive for cytokeratin (Fig. 6C and D), verifying their epithelial cell origin, and positive for p53 (Fig. 6E and F), demonstrating that their p53 stability was increased and its function was disabled due to its binding by T antigen. The cells were also positive for CA125 (Fig. 6G and H), WT-1 (Fig. 6I and J), which are common markers expressed in human ovarian cancer.
In this study, we successfully generated papillary ovarian carcinoma by overexpressing HER2/neu in normal ovarian surface epithelial cells. In a previous study, we transformed human ovarian surface epithelial cells with RAS, hTERT, and SV40 T/t, but the resulting tumors remained either largely undifferentiated or poorly differentiated.3 Here, we find that human ovarian surface epithelial cells can be induced to transform into high-grade carcinoma by overexpressing HER2/neu, which demonstrates that HER2/neu is involved in the initiation of tumor growth. In addition, that HER2/neu can lead to papillary differentiation of immortalized human ovarian epithelial cells; this ability of human ovarian surface epithelial cancer to undergo such differentiation has been inferred for years but has never been shown experimentally.10 Interestingly, we also found that the peritoneal microenvironment and the genetic background of the cells play a critical role in the differentiation of papillary carcinoma, suggesting that genetic modifications alone are not sufficient for development of this carcinoma; rather, a specific ovarian epithelial cell type and an appropriate tumor microenvironment are required.
Our results demonstrated that T29Nt cells but not T80Nt cells can form tumors with papillary differentiation. Since the SV40 T/t antigen was introduced to the IOSE29 cells at a later passage after the ovarian surface epithelial cells had undergone the epithelial-mesenchymal transition, the T29 cells retained a mesenchymal phenotype (Fig. 2D). The precursor cell line for IOSE-29, T29 has always had mesenchymal characteristics, especially in its shape and growth pattern,4 while the T80Nt cells have an epithelial cell phenotype. The T80Nt-derived tumors remained undifferentiated under all conditions, suggesting that ovarian surface epithelial cells that have undergone the epithelial-mesenchymal transition may contribute to the papillary differentiation of the tumor. These results are not surprising, as different cell types have been shown to give rise to different carcinomas in breast can cer.11 In addition, other reports have shown that HER2/neu can confer a stem cell-like property to cells,12 and the epithelial-mesenchymal transition may contribute to this property.13 However, additional experiments are needed to confirm this hypothesis.
Several murine models of different histologic types of ovarian cancer have been generated Depending on the genetic background and the genetic elements introduced, murine ovarian epithelial cells can result in ovarian tumors of different histologic types.8,14–17 Previous reports have demonstrated that poorly differentiated carcinoma can be generated by introduction of the SV40 T/t antigen into ovarian epithelial cells or the ovarian epithelial cell—specific knockout of either p53 retinoblastoma protein.14,16 Furthermore, endometrioid carcinoma has been successfully modeled in mice by knocking out either K-RAS and PTEN or PI3K/PTEN and Wnt/β-catenin.8,15 These studies demonstrated that a specific combination of genetic modifications in murine ovarian epithelial cells dictated the development of a specific histotype of ovarian carcinoma. Likewise, our results demonstrate that overexpression of HER2/neu along with SV40 T/t antigen and hTERT is also required to regulate the cancer histogenesis of human ovarian epithelial cells. In addition, our results show that the development of specific histotypes depends on the peritoneal microenvironment and the cell type. Even with identical genetic modifications these tumor cells can remain undifferentiated when they are placed into an unnatural microenvironment such as the subcutis.
Although it is widely believed that ovarian cancer is derived from a single layer of ovarian surface epithelial cells,10,18,19 alternative cellular origins have been proposed, including fallopian tubal epithelial cells20 and rete ovarii.21 In support of this idea, a similar p53 mutation has been observed in fallopian tube in situ carcinoma and pelvic serous carcinoma.20,22 Similar gene expression profiles have also been reported in tubal lesions and ovarian cancer.23 Our findings provide cause-and-effect evidence that certain types of ovarian surface epithelial cells can give rise to papillary carcinoma in a peritoneal microenvironment when modified with specific genetic elements, suggesting that ovarian surface epithelial cells can be an alternative original of human ovarian carcinoma.
In summary, we describe the development of a novel murine model of papillary carcinoma using genetic modification of a human ovarian surface epithelial cell line in combination with a peritoneal microenvironment. With this model, we have demonstrated that some human ovarian surface epithelial cells can give rise to high-grade serous carcinoma with papillary differentiation. This murine model should be useful for studying the mechanisms involved in the development and histologic differentiation of human ovarian carcinoma. It may also serve as a prototype human ovarian cancer model, allowing us to generate other ovarian epithelial cancer models with additional or different genetic modifications. This model should be helpful for defining pathways and identifying novel targets to improve the treatment and early detection of ovarian cancer.
Two ovarian epithelial cell lines, T29 and T80, were derived from normal human ovarian surface epithelium that had previously been immortalized by transformation with SV40 large and small T antigen7 and then with hTERT.3 These cell lines were maintained in regular complete medium (450 mL of basic media, 50 mL of heat-inactivated fetal bovine serum, and 5 mL of penicillin-streptomycin mix; Invitrogen, Carlsbad, CA). Retroviral packaging cells (Phoenix amphotropic cells) were purchased from American Type Culture Collection (Manassas, VA) and maintained at 37°C in Dulbecco’s modified Eagle’s medium, which contained 10% fetal bovine serum, 1 mM sodium pyruvate, 100 units/mL penicillin, and 100 μg/mL streptomycin (all from Sigma-Aldrich, St. Louis, MO). The T29N and T80N cell lines, HER2/neu-expression cell lines T29N and T80N, the tumor-derived cell lines T29Nt and T80Nt cells were maintained in complete medium (500 mL of regular complete media and 2 mg/mL suspension of epidermal growth factor; Nalgene, Rochester, NY).
HER2/neu cDNA was transfected into the Phoenix amphotropic cell line, and the retrovirus-containing HER2/neu cells were harvested for the experiments. T29 and T80 cells were cultured in 60-mm dishes and incubated for 24–48 h with 5 mL of a 1:4 mixture of virus supernatants and fresh culture medium containing 4 μg/mL polybrene (Sigma-Aldrich). After a 24-h recovery from virus infection, the cells were selected by growing them at 37°C in a medium containing neomycin (1 mg/mL) for 48 h and then grown in the medium without neomycin and used for western blot analysis and other in vitro assays and tumor growth in nude mice. The percentage of infected cells among the 1 × 106–5 × 106 target cells was determined using the retrovirus-encoded green fluorescent protein. The infected cells were directly visualized under a fluorescence microscope or analyzed by fluorescent-activated cell sorting using a FACStation flow cytometer. The same number of cells (1 × 106–5 × 106) from each cell line was used for virus infection.
HER2/neu expression was detected using the SC-8334 rabbit polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) at a 1:2,000 dilution. The secondary antibody used was an NA9340 anti-rabbit immunoglobulin horseradish peroxidase-linked F2 fragment from donkey (Amersham Biosciences UK Limited, Little Chalfont, United Kingdom).8 Western blot reagents were obtained from an electrochemiluminescence kit (Amersham Biosciences UK Limited).
Cells (1 × 104) were suspended in 2 mL of growth medium with 0.35% agarose (Invitrogen), and the suspension was placed on 5 mL of solidified 0.7% agarose. Triplicate cultures of each cell type were maintained for 14 days at 37°C in a 5% CO2 atmosphere, and fresh medium was added at 7 days. The number of colonies that were larger than 50 μm in diameter was counted on the 14th day. These experiments were repeated twice.
For these experiments, 5 × 106 cells from each cell line (T29, T80, T29N, T80N, T29Nt or T80Nt) were harvested by trypsinization, washed twice with 1x phosphate-buffered saline, and resuspended in 0.1 mL of saline. The cell suspensions were then injected subcutaneously into 4- to 6-week-old BALB/c athymic nude mice (Jackson Laboratory, Bar Harbor, ME), with the T29- and T80-control cells injected on one side of the mouse and the T29N and T80N or the T29Nt and T80Nt cells on the other side. The same number of cells from each cell line was injected intraperitoneally into different nude mice. The mice were kept in a pathogen-free environment and checked every 2 days for 4 months. The date on which grossly visible tumors first appeared and the size of the tumors were recorded. The mice were killed when tumors reached 1.5 cm in diameter. The mice given intraperitoneal injections were observed for increased abdominal size, lethargy and jaundice and were killed when these signs occurred. All animal experiments were approved by the Animal Care and Use Committee at The University of Texas M.D. Anderson Cancer Center.
Tumors from mice were fixed in 10% formalin and stained with hematoxylin and eosin. The histopathologic characteristics were examined by microscopy by a board-certified pathologist specializing in gynecologic cancer (Liu J or Malpica A). The following antibodies were used for immunohistochemical analysis: monoclonal anti-HER2/neu (NeoMarker; 1:100 dilution), monoclonal anti-SV40 T antigen (Santa Cruz Biotechnology; 1:200 dilution), monoclonal anti-p53 antigen (Santa Cruz Biotechnology; 1:50 dilution), monoclonal anti-pan cytokeratin antigen (Biocare, Walnut Creek, CA; 1:50), and mouse monoclonal anti-CA125 (DakoCytomation, Carpinteria, CA; 1:20). Secondary antibodies against each host were biotinylated.
J.L. is supported by R01 CA131183-2 grant from NIH, and by M.D. Anderson SPORE in Ovarian Cancer (IP50CA83638). This work is also supported in part by Cancer Center Core grant (CA016672) from the National Cancer Institute.