In this follow-up study we show that despite similar phenotypic modulation different cell lines of a similar cancer type display diverse transcriptomic modifications. The acquisition of an increased metastatic phenotype during the cross talk between cancer and stromal cells is concordant with previous report in ovarian as well as other tumor types [5, 6, 9, 10, 15]. Interestingly our approach using two different cell lines associated to a network analysis illustrate that the shift of phenotype will be dependent on the transcriptomic background. This is quite relevant in the context of targeted therapies. The network analysis presented above allows a more detailed interpretation of the relationship between transcriptomic pro-metastatic changes in OCC and differences in cell behavior upon co-culture for the different functionality tested.
Migration, invasion and adherence
Increased adherence and invasion constitute a key step toward peritoneal metastasis . Both cell lines regulated genes related to these processes; interestingly some had opposite regulation. FN1 encoding fibronectin was up-regulated (2-fold) in SKOV3, and down-regulated (5.63-fold) in OVCAR3. As demonstrated by different studies Fibronectin does not trigger similar response in the two cell lines studied. In contrast to SKOV-3, OVCAR-3 cells did not migrate or invade matrigel either with or without FN [16, 17]. OVCAR3 cell line up-regulation of Twist and Zeb indicates an epithelial to mesenchymal transition state. The expression of E-cadherin in ovarian cancers plays a role in the adhesion to the peritoneal surfaces. The ovarian surface epithelium (OSE) does not express E-cadherin, but its expression appears in inclusion cysts of the ovary and is reinforced during tumoral progression . One might argue against simultaneous up-regulation of EMT markers (TWIST, ZEB) and up-regulation of E-cadherin and downregulation of FN1. However recently a subpopulation of cells in transitory Epithelial/Mesenchymal stage presenting increased tumorigenicity has been described . The described cell plasticity appeared to be dependent on external factors such as co-culture with mesenchymal cells. Finally the mixed state might represent a transient state induced by the co-culture setting. SKOV3 cells displayed up-regulation of MMP3 concordant with previous reports . Several pro-metastatic cytokines already described were up-regulated such as CXCR4, and CCL2 [21, 22]. Hence, our strategy to build focused networks for specific functionalities allows understanding their inter-relationship and thereby might help designing better disrupting strategies.
OVCAR3 proliferation relied on NR3C1 node that has already been described associated with liver metastasis of gastric cancer . PRKACB stabilizes elF4F expressed at high level in different carcinomas. RhoB functions as tumor suppressor genes and its down-regualtion promotes cancer cell proliferation . CEBPB is involved in transcriptional up-regulation of SphK1 by LPA in gastric cancers . Another interesting function of CEBPB is its role in the occurrence of mesenchymal phenotype in brain tumors where it is a synergistic initiator of mesenchymal transformation with STAT3 . VEGF has a role in increased proliferation and resistance to anoikis of ovarian carcinomas beside its role in angiogenesis . Finally SPARC modulates cell-cell and cell-matrix interactions and demonstrates de-adhesive and growth inhibitory properties in non-transformed cells. Recently Said et al. showed that SPARC significantly inhibited OVCAR3 basal and LPA-induced interleukin IL6 production and attenuated IL-6-induced mitogenic, chemotactic, and proinvasive properties through ERK1/2 inhibition [28, 29]. Once again interestingly SPARC was up-regulated in SKOV3. Increased SPARC is associated to poor prognosis in pancreatic adenocarcinoma and to lymph node metastasis in gastric cancers . In melanoma cells SPARC promotes proliferation and survival through akt-dependant regulation of p53 . In a three dimensional context the interaction of fibronectin and alphaV integrin has been associated to increased proliferation in ductal pancreatic carcinoma . In a model of lung carcinoma, inhibition of the interaction between fibronectin and alphaV integrin resulted in reduced proliferation .
While many genes for both cell lines are similar to the ones responsible for increased proliferation the central nodes are not always the same suggesting the ability of cancer cells to use complex transcriptomic machinery to respond to cytotoxic stress. OVCAR3 displayed increased GADD. The deficiency of gadd45 increases cell sensitivity to UV irradiation or cisplatin [34, 35]. Fibronectin was found to be central for SKOV3 chemoresistance. Fibronectin ligand integrin alphaVbeta1 plays a role in multicellular aggregation and resistance to paclitaxel . Moreover activation of alphaVbeta1 activates GSK3β pro-survival signaling . Finally fibronectin mediated adhesion promotes Akt phosphorylation in highly metastatic cancer cells A2780 and MDA-MB231, and further induced chemoresistance against docetaxel [38, 39].
Other genes have been described in the literature in ovarian cancer biology. However they shall not be uncovered in our approach if they do not play a role in the co-culture context [7, 40, 41]. It is hard to interpret all the genes described in the networks in a cancer-cell autonomous context. Some genes might not have the expected regulation but might play a role in the cross-talk between OCC and MSC.
As recently illustrated in oncogenomic approaches the role of a single genes might not be as important as pathways up or down-regulated. Our approach uncovered some central nodes (such as fibronectin) and their role will be analyzed in separate targeted functional studies.
The main question raised in this research was if broad transcriptmic changes underlie phenotypic modulation in cancer cells co-cultured with mesenchymal stem cells. We have thus chosen the following experimental approach. We have co-cultured for 24 hours two different cancer cell lines and mesenchymal cells and performed cell sorting and transcriptomic analysis at a time-point where we observed functional effect of the co-culture. We then used targeted network analysis of the most significant shifted genes and determined few networks described above.
While we acknowledge that validation is an important part of such studies but there are many studies where the role of single genes or few genes were validated without any translational relevance. Here we do not want to demonstrate the role of any particular genes but we want to emphasize that the interaction in vitro between stromal and cancer cells induces broad transcriptomic profile that should be carefully considered when experimental plans are designed. Interestingly showing differential expression of genes between two routinely used ovarian cancer cells indicate that one must be quite careful before investigating the particular role of a single gene.
Our observation that one cell line can shift expression to more resemble human tumors is interesting but must be examined more closely. It is interesting because it suggests that culturing MSCs with tumor cells can create a better model system to study carcinogenesis since the co-cultured tumors cells bear more resemblance to human tumors. It is also interesting that this effect is cell line dependent as OVCAR3 shifts farther away from human tumors after co-culture with MSC’s. However, while this result is interesting, it should be seen in the context that it is difficult to compare expression data obtained at different times and with different platforms. Nevertheless, this observation merits further investigation.
Other ovarian cancer cell lines need to be tested co-cultured with different cells to obtain a broader picture of the role of the genetic background in the interaction with the microenvironment. Understanding the differences between ovarian serous adenocarcinoma types and their interaction with the microenvironment would open the path to new personalized therapeutic strategies.