Phenotypic Variation in Cancer Cells Often Not Due to Mutations

ABOVE: Micrograph of colonic adenocarcinoma © iStock.com, OGphoto

Most models of how tumors evolve have assumed that the process is based predominately on cancer cells’ genetics, and many cancer treatments are specifically targeted to mutations associated with disease. But comparing whole genome sequence data with RNA-seq data from samples of colorectal tumors revealed that the vast majority of gene expression differences among cancer cells cannot be explained by genetics, researchers report in Nature today (October 26).

See “Cancer Researchers Use Evolution to Target Drug Resistance”

“So far, a lot of the work that’s been done exploring cancer evolution in cancer development has focused very much on just the genetics,” says Nicholas McGranahan, a computational cancer researcher at the University College London Cancer Institute’s CRUK UCL Centre who was not involved in the research. But according to the new study, “there’s lots of these alterations that they can’t identify a clear [genetic] underpinning for. . . . It’s a nice study because it highlights some of the limitations of what we’ve been doing before.”

Computational biologist Andrea Sottoriva says that it’s those limitations that led him and his colleagues to consider the transcriptome in cancer evolution. “Basically, looking at the genetic evidence was not explaining everything that we were seeing,” says Sottoriva, a group leader at the Institute of Cancer Research, London, and the head of Computational Biology Research Centre at Human Technopole in Milan, Italy. For example, he explains, “if you just look at . . . mutations in genes that are involved in cancer, it’s not very easy to distinguish a benign cancer from a malignant cancer: In a benign cancer, there are as many cancer driver mutations as in a malignant cancer.”

See “New Understanding of Metastasis Could Lead to Better Treatments”

To gain a better understanding of how cancer cells vary at the gene expression level, the team carried out whole-transcript RNA-seq as well as whole genome sequencing on samples from 27 surgically removed human colorectal tumors, eight of which yielded sufficient data for comparisons between the two types of sequencing. In those eight tumors, out of 8,368 differentially expressed genes included in the analysis, the differences in transcript levels could be traced to underlying genetics in only a median of 166. “A large proportion of the [cancer research] community always believed that everything was genetically controlled,” Sottoriva says, but these results suggest that “the answer is not strictly genetics.”

Sottoriva says that while some of the variation may be due to transcriptional noise, as gene expression is constantly fluctuating in cells, he suspects that the cells’ microenvironments also play a big role, with factors such as hypoxia or the presence of macrophages influencing transcriptional programs. Cancer cells “can adapt plastically to a lot of different environments,” he notes.

This is likely to have implications for the appearance of treatment resistance, Sottoriva adds. Oftentimes when cancers fail to respond to chemotherapy or targeted therapies, there are transcriptional mechanisms at play that are influenced by the tumor microenvironment, he explains, such as those underlying autophagy, senescence, or quiescence. “So that means that a lot of evolutionary models need to be adapted to this, because a lot of phenotypic diversity that could drive drug resistance is not [based in] genetics.”

McGranahan agrees that the apparent large role for nongenetic variation among cancer cells could influence the effectiveness of treatments, especially if a targeted therapy is chosen based on a specific mutation found in the cancer. “Because we see it in the genetics, we make the assumption, therefore, [that] it’s driving the tumor. But what they are also suggesting is that in certain cases . . . what we would call a driver [is] actually a passenger.”

See “Who Benefits from Herceptin and Other Anti-HER2 Cancer Therapies?”

One caveat, McGranahan adds, is that the study involved bulk analyses of the tumor samples, meaning that some immune cells and other healthy cells were included. “Ultimately . . . I think single-cell [analyses] will really, really help.” Sottoriva agrees that single-cell techniques could provide additional insight, but notes that researchers would need much greater numbers of samples to yield sufficient data, as gene expression is variable over time and the genome sequences obtained from individual cells with current assays is “almost invariably” incomplete.

Another extension of the research would be to repeat this sort of study using more patients with different types of cancers, McGranahan says. “What they did here is great, because they go really deep, but necessarily that means they’re only looking at [a small number of] patients.”

Still, he adds, this study is an important first step in considering variation in cancer beyond the genome. “Given that evolution operates on the phenotype, not the genotype, we need to start to understand . . . how much of this genetics is functionally relevant for tumor evolution.”

Sottoriva echoes that sentiment. “Whereas the genetics set the stage for cancer development, it doesn’t write a script.”