The best medicine comes from treating the right patient with the right therapy at the right time. And while that seemingly simple prescription rings true, the path to the three Rs of medical treatment is anything but simple and straightforward—especially in treating the multiplicity of cancers. As scientists struggle to decode and decipher the mountain of cancer cell genomic data currently—or soon to be—available, knowing which data set correctly describes a tumor’s inner workings, is a huge challenge. Such is the case with research on established cancer cell lines as opposed to cells derived directly from a patient’s tumor.
Assessing the Utility of Cell Lines
In a paper published in November 2011 in the Proceedings of the National Academy of Sciences (1), researchers from the National Cancer Institute (NCI) found striking genetic differences between 60 established cancer cell lines and clinical samples comprising 75% or greater drug-naïve cancer cells. Because the 380 genes interrogated were known to be involved in anti-cancer drug resistance, the work raised red flags about how accurately cell lines can determine clinical outcomes of cancer drug treatments.
The bottom line, says first author and NCI researcher Jean-Pierre Gillet, is that “all the cancer cell lines we studied are more related to each other [regardless of tissue of origin] than to the cancer type they are supposed to model.”
Because the in vitro micro-environment that laboratory cell lines are cultured in is different from an in vivo environment, the researchers attempted to control for this by growing the cells lines in 3-D or xenograft settings. But neither of these controls prevented the gene expression profiles from clustering to either a cell line profile or primary tumor profile.
“I don’t want to say cell lines are not useful,” says senior author Michael Gottesman, a chief scientist at the NCI. “They are!” But when looking specifically at multi-drug resistance (MDR) cancer genes, they seem to differ more between cell lines and primary tumor cells than a cancer’s other genes—most likely because MDR genes tend to be exquisitely sensitive to the growth environment.
“We’re doing all these studies and learning information about different genes that may be involved in drug resistance and growth promotion but we haven’t been asking the hard questions: Which genes are actually activated and active in real tumors,” says Gottesman who points to the use of primary tumor samples in the Cancer Genome Project as one of that project’s great strengths..
Many researchers agree that new models for assaying primary tumor cells are needed.
For example, as described in a recent paper in the American Journal of Pathology (2), a new culturing technique allows cells taken directly from clinical samples to be grown rapidly into long-lived cultures. The technique uses a kinase-inhibitor drug called ROCK, to down-regulate a protein involved in cell shape and growth. The new culturing method was developed by Richard Schlegel from the Lombardi Comprehensive Cancer Center at Georgetown University to assess the utility of ROCK inhibitors.
Now, Gottesman is partnering with Schlegel—a colleague and former lab partner—to examine the expression profiles of MDR genes in cells grown with ROCK kinase inhibitors. “We’ll look at this new and very efficient approach to growing cancer cells ex vivo to determine if the expression profiles look like the tumors coming out of patients, or more like established cell lines or something different, something in between,” says Gottesman.
In their campaign to promote the use of primary cancer cells, Gillet, Gottesman, and colleagues have just published a paper in Clinical Cancer Research (3) that examines the MDR gene profiles of primary tumor samples taken from 80 patients with drug-naïve ovarian serous cancer. Evaluating the same set of 380 MDR genes used in their PNAS paper, the researchers identified an 11-gene signature that, when combined with clinical variations, significantly improved overall survival prediction in this cohort of patients.
Surprisingly, the scientists found that high-risk patients with low-expression levels of the 11-gene signature in their tumors had a better survival prognosis than patients ranked as low risk. Conversely, low-risk patients with high expression levels of the products of those 11 genes have a worse overall survival prognosis.
In addition, by using these patient tumor samples, the researchers identified two new genes—S100A10 and APC—that help indicate survival outcomes for ovarian cancer, and their hope is that these genes will provide fruitful targets of new anti-cancer therapeutics.
Re-Booting the System
The goal of the project is to advance personalized medicine by predicting which cell lines are sensitive to which drugs via genomics data. In the study, they compared data for about half of the 1000 lines with drug sensitivity data for 24 anti-cancer drugs. As a result, the authors reported the discovery of several novel genetic biomarkers that predict drug sensitivity.
Additionally, they found that on average the cell lines accurately represented the primary tumors from which they were derived. “Ours was a deeper, broader characterization of cells lines than has ever been done before,” says Nicolas Stransky, the lead computational biologist on the study.
Referring to Gottesman’s PNAS paper, Stransky says that the higher power of their study in terms of cell lines interrogated allowed them to better report actual gene expression patterns. “When you have only 60 cell lines to plumb, then for each cancer type you are looking at you have only a handful of cell lines, which doesn’t represent the diversity of cancers,” Stransky says.
“We didn’t need to wait until November of 2011 to find out that the expression profiles of a cell that is growing on plastic and in culture media is very different from a primary tumor because we already knew the environment was very different,” he says. “What we showed in our paper is that when you subtract that effect, the different environment, the expression pattern that remains is still representative of the primary tumor.”
Another intriguing finding of the Nature paper was that active expression of the gene SLFN11 predicted sensitivity to the anti-cancer drug irinotecan, which is currently used to treat colon or rectal cancer. Correlations showed that all three cell lines for Ewing’s sarcoma—a childhood bone cancer—exhibited both high levels of SLFN11 expression and sensitivity to irinotecan. This suggests that patients with high SLFN11 gene expression might respond well to irinotecan.
Likewise, another paper published in the same issue of Nature (5) gives additional directional markers for the treatment of Ewing’s sarcoma—in this case, finding a translocation between two genes called EWS-FL11 and sensitivity to PARP inhibitors, which are currently under investigation in the treatment of ovarian caner. That collaborative study, born out of the Wellcome Trust Sanger Institute’s Cancer Genome Project, screened 130 compounds that are therapeutically relevant to cancer treatment with a panel of several hundred cancer cell lines. Each drug was applied to each cell line for 72 hours, after which the researchers correlated drug sensitivity to genomic variations.
“A big question we had was to ask if our testing would confirm what was already known from the clinic about drug response—for example, would we see what’s known to be operating in clinical response for BRAF mutations responding to BRAF inhibitors—could we capture known associations?” says senior author Cyril Benes from the Massachusetts General Hospital Cancer Center. “It turns out we did capture those associations extremely well.”
They also found several novel gene-drug associations not currently explainable by known cancer biology and that single genes were rarely responsible for drug sensitivity or resistance. “Context is turning out to be very important,” says Benes. “The pendulum swung from treating cancer by tissue type to genetic profile, and now, we are finding that, perhaps it needs to return to a middle ground. Knowing a couple of genes will not tell the whole story.”
Chaos or Not?
“The NEJM study is absolutely concerning, but in many ways this is not news to the research community,” says Joe Gray, associate director for translational research for the Oregon Health and Science University Knight Cancer Institute whose own research (7–8) has shown strong, positive correlations between breast cancer cell lines and primary human breast tumors. “We’ve known for a long time, through in situ hybridization and immunohistochemical analysis, that tumors are heterogeneous. The question is: How much of that heterogeneity is functional—clinically relevant? The NEJM paper didn’t address that.”
Gray is hopeful that new sequencing techniques will allow scientists to pick their way around a tumor and detect low-frequency cell populations resistant to a drug that would have been missed with older techniques. Finding them earlier translates to treating them earlier.
In many ways, the dispiriting news from the NEJM study has been tempered by the debut of the CCLE. “Quite frankly one will guide the other,” says Gray. “I can now go to the CCLE, find examples of cell lines that have these low-frequency mutations that we now have the tools to detect and ask: What happens when you manipulate them? This will tell us much more quickly whether the mutation is important.”
And in terms of the significant differences found between cell lines and primary tumor samples in terms of MDR genes, the NEJM findings of extreme heterogeneity within a tumor may mean that current cell lines suffer from under-sampling of the primary tumor from which they were established.
“While cell line models were good at providing the MDR gene list, or at least part of it, they are not necessarily good at providing the most relevant information on drug resistance influenced by multiple factors in vivo; different research stages require different tool,” says Benes.
At the end of the day, it is what is happening inside a patient that will ultimately matter. And a limitation of both the PNAS paper and the Nature papers is that neither prove clinical outcomes. Yet thet are signposts to be used in tailoring clinical trials to nexus the right patient with the right treatment at the right time.
While Gottesman is adamant that the past 30 years of cell line research has not been for nought, he emphasizes that “at some point we must ask: What is relevant to actual growth in the patient?”
- Gillet, J.-P., A. M. Calcagno, S. Varma, M. Marino, L. J. Green, M. I. Vora, C. Patel, J. N. Orina, T. A. Eliseeva, et al. 2011. Redefining the relevance of established cancer cell lines to the study of mechanisms of clinical anti-cancer drug resistance. Proceedings of the National Academy of Sciences 108(46):18708-18713.
- Liu, X., V. Ory, S. Chapman, H. Yuan, C. Albanese, B. Kallakury, O. A. Timofeeva, C. Nealon, A. Dakic, et al. 2012. ROCK inhibitor and feeder cells induce the conditional reprogramming of epithelial cells. The American journal of pathology 180(2):599-607.
- Gillet, J.-P., A. M. Calcagno, S. Varma, B. Davidson, M. B. Elstrand, R. Ganapathi, A. Kamat, A. K. Sood, S. V. Ambudkar, et al. 2012. Multidrug Resistance-Linked gene signature predicts overall survival of patients with primary ovarian serous carcinoma. Clinical Cancer Research (April).
- Barretina, J., G. Caponigro, N. Stransky, K. Venkatesan, A. A. Margolin, S. Kim, C. J. Wilson, J. Lehar, G. V. Kryukov, et al. 2012. The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483(7391):603-307.
- Garnett, M. J., E. J. Edelman, S. J. Heidorn, C. D. Greenman, A. Dastur, K. W. Lau, P. Greninger, I. R. Thompson, X. Luo, et al. 2012. Systematic identification of genomic markers of drug sensitivity in cancer cells. Nature 483(7391):570-575.
- Gerlinger, M., A. J. Rowan, S. Horswell, J. Larkin, D. Endesfelder, E. Gronroos, P. Martinez, N. Matthews, A. Stewart, et al. 2012. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 366(10):883-892.
- Heiser, L. M., A. Sadanandam, W.-L. Kuo, S. C. Benz, T. C. Goldstein, S. Ng, W. J. Gibb, N. J. Wang, S. Ziyad, et al. 2011. Subtype and pathway specific responses to anticancer compounds in breast cancer. Proceedings of the National Academy of Sciences 109(8):2724-2729.
- Neve, R. M., K. Chin, J. Fridlyand, J. Yeh, F. L. Baehner, T. Fevr, L. Clark, N. Bayani, J.-P. P. Coppe, et al. 2006. A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes. Cancer cell 10(6):515-527.