In order to establish new cellular models of EOC, all samples of ovarian tissues collected for our tumor banks are processed to derive primary cell cultures. Occasionally, some of the primary cultures evolve to become immortal cell lines. So far, we have been successful in establishing seven EOC cell lines , three of which are described here. All cell lines were derived from patients who were never exposed to chemotherapy. This is of importance, since the majority of available cellular models originate from samples obtained following neoadjuvant therapy that could introduce genetic events not related to the biology of the disease.
Early in culture, cell lines presented heterogenous populations of cells that eventually progressed toward a more uniform population of cells with cobblestone-like morphology typical of epithelial cells. All cell lines grew as adherent monolayers without evidence of cellular piling. However, we noted that TOV-2223 cells were more adherent to the petri dish when compared to TOV- and OV-1946. Indeed, a longer trypsinization time is required to detach these cells from the petri dish. Also, when fixed in methanol, a large portion of both 1946 cell lines detached from the solid support while 2223 cells did not (data not shown). The homogeneity of cells observed at passage 0 of OV-1946 could reflect the fact that these cells came from a mass in the patient's ascites that is more homogenous then a tumor mass often composed of, among others, epithelial tumor cells, stromal cells and endothelial cells.
Cell lines were also evaluated on their capacity to survive and grow in low serum conditions. It is worth noticing that although in normal growth conditions both TOV and OV-1946 cell lines present similar growth rates, in low serum conditions the later is almost two times slower. This could be in line with the fact that the OV-1946 cell line comes from ascites which is rich in growth factors and in culture can replace FBS . This cell line could hence be more dependent on growth signals provided by the environment as opposed to cells originating from solid tumors where cells may be predicted to be more self-sufficient.
In epithelial cells, intermediate filaments are composed of keratins that vary according to the differentiation of the cells. Both OSE and EOC tumors are characterized by the expression of different keratins such as KRT7, KRT8 and KRT18. In order to determine if our cell lines presented these EOC markers, we monitored protein expression of Krt7, Krt8 and Krt18 by western blot. When antibodies were appropriate for immunohistochemistry on paraffin embedded tissues, we also compared keratins expression in the original tumor tissues. It is also well described that monitoring expression of Krt7 and Krt20 could distinguish between ovarian and gastro-intestinal tract tumors (reviewed in ). Both the 1946 and the 2223 patients presented different epithelium specific Krt expression (Figure 2A and 2E) but did not express Krt20. It is worth noting that the Krt patterns were not all the same between different cell lines. Stronger expression of Krt7 is observed in TOV-1946 and TOV-2223G was the only cell line that expressed Krt18 and Krt8. This underlines the fact that these cell lines are biologically different even though they originate from tissues representing the same type of serous EOC disease. Moreover, Krt7 is also differentially expressed between the TOV-1946 and OV-1946 cell lines that not only originate from the same type of disease but also from the same patient and differences are related to solid tumor versus ascites.
In EOC, 25 to 30% of the tumors present an amplification of the HER2 gene leading to the overexpression of the protein Her2 (reviewed in [2, 61–63]). The overexpression of this growth factor receptor alone was shown to be sufficient to induce malignant transformation and is implicated in ovarian cancer as well as many other types of cancer [64–67]. In EOCs of advanced stage, over 50% of the tumors were shown to be mutated in the TP53 gene (reviewed in [2, 61–63]). All three cell lines showed Her2 protein expression and contained TP53 gene mutations.
Differences in growth rates, migration, invasion and spheroid formation between the TOV-1946 and TOV-2223 underline the diversity of phenotypes that can be observed within the same type of serous EOC disease. The results from the comparison of the TOV-1946 and OV-1946 growth characteristics, demonstrate that these cell lines present with unique phenotypes. Indeed, TOV-1946 cells had a better capability to invade and migrate although their capacity to form spheroid is reduced compared to the OV-1946 cells. One might speculate that cells derived from solid tumor conserve their migration and invasion property but in ascites, these characteristics are less vital. It has been previously shown that cells present in ascites form spheres or aggregates that can adhere to different extracellular matrices as well as to normal human mesothelial cells  but are not always invasive. The role of spheroids in ascites of ovarian cancer patients remains undefined. This is the first report of cell lines derived both from solid tumor and ascites cells of the same patient and further studies on these cell lines may provide useful insights into the biological progression of EOCs.
As both the TOV-1946 and OV-1946 injected mice developed ascites we hypothesize that the ability for cell lines to induce ascites formation in mice is intrinsic to the given cell line and independent of their origin (solid tumor versus ascites). It has been previously shown that the in vivo tumorigenicity can usually be predicted by the ability of the cells to grow in soft agar . The OV-1946 cell line, which was the most tumorigenic, also formed the largest and most numerous colonies when seeded in soft agar (Table 2). In previous studies we have also observed that the capacity of the cells to form compact spheroids is related to their ability to form tumors in vivo . Consistent with this notion, here we show that the OV-1946 cell line formed the most compact spheroids among the new cell lines and formed the greatest number of tumors with the shortest latency in xenograft experiments. However, the TOV-2223 cell line does not even form an aggregate in hanging droplets, only formed small colonies in soft agar and produced no tumors in SCID mice. The tumorigenicity results for TOV-2223 are also consistent with the relative indolent disease observed in patient 2223 who survived a relatively long period post diagnosis without treatment. Indeed, we have previously isolated the TOV-81D cell lines, also from an indolent disease, which failed to form tumors in immuno-compromised mice  suggesting inherent qualities of the tumor that are reflected in its clinical behavior can affect this biological parameter.
The unique features of the described cell lines provide complementary models for different aspects of the disease. For example, while TOV-2223 and the TOV-1946 cell lines are both derived from solid tumors of the same type of serous EOC disease, they vary in important aspects including their aggressiveness and karyotype. They may prove useful for comparative studies to uncover molecular events that distinguish very aggressive from more indolent serous disease in ovarian cancer. Although the TOV-1946 and OV-1946 lines were derived from the same patient, they were collected from the solid tumor and the corresponding ascites respectively. Although common genome modification are shared between these two cell lines (see examples of modifications occurring on chromosome X, 1, 2, 7, 8, 9, 11, 12, 13, 15 and 16) (Table 4), others are unique either to the solid tumor or the ascites derived cell line (chromosome 2, 4, 5, 6, 7, 9, 16, 17, 18, 19, 20 and 22) (Table 4). These observations suggest that the two cell lines acquired common modifications during the earlier steps of tumorigenesis and during cancer progression different rearrangements were selected for, depending on the microenvironment.
The results obtained by SKY and G-banding assays reflect previously published karyotype studies on EOC where high genomic instability is observed . Recently, it as been shown that recurrent rearrangements resulting in the formation of new fusion genes could be identified using genomic and bioinformatic tools [72–75]. So far such studies have been difficult to conduct on solid tumors and hence the importance of appropriate cellular models representing different types of tumors. The future fine cartography of recurrent lesions in these cell lines may provide insights into the molecular events that contribute to EOC initiation and progression.