NEW YORK (GenomeWeb) – By analyzing data from nine different tumor types in The Cancer Genome Atlas, a Cancer Research UK-led team has confirmed that therapeutically actionable mutations often occur not only in the early evolution of a tumor, but also later in smaller subpopulations of a tumor's cells.
This new comprehensive look at how mutations occur throughout the evolution of cancers, published today in Science Translational Medicine, could help explain why some treatments targeting a particular mutation or molecular pathway don't work for all the patients who test positive for the target in question, or only work for a limited period.
"The proof of the pudding is yet to be seen. But I think that the fact that we can easily detect these driver events in a proportion of cells but not the whole population really tells you that in all likelihood it's going to be important," Charles Swanton, the first author of the study and a lab leader at the Francis Crick Institute, told GenomeWeb this week.
In the study, Swanton and his colleagues analyzed TCGA data for more than 2,500 patients with nine different cancer types including bladder, breast, lung, colon, head and neck, kidney cancers, and melanoma, to determine the pattern of mutations' occurrences through the growth of a tumor from one cell to many.
To do this, the investigators estimated the mutation copy number and cancer cell fraction of each recorded single nucleotide variant using SNP array and exome sequencing data from the TCGA datasets for each tumor type.
The group classified mutations as either clonal, meaning present in all the sequenced tumor cells, or subclonal, meaning present in only a subset. They also classified mutations as occurring late or early in a cancer's evolution based on the cancer cell fraction and the mutation copy number.
While the team did see many known driver mutations that they determined had occurred early, or clonally, they also identified so-called "actionable" mutations that emerged later, or subclonally, in almost every gene for which there are either currently used or experimental targeted therapies, including BRAF IDH1, PIK3CA, EGFR, and KRAS.
According to the authors, the preponderance of these actionable subclonal mutations could explain a lack of efficacy of targeted treatments for some patients, either due simply to a treatment being effective only in a subset of a tumor's cells, or because efficacy against mutated cells sparks a proliferation of wild-type, treatment-resistant tumor cells, causing acquired resistance.
Although this suggested impact of subclonal driver mutations on the sensitivity or resistance to targeted therapies is not fully proven, there have been some real-world indications, according to the authors. For example, subclonal RAS mutations have been shown to precipitate resistance to cetuximab in colorectal cancer, the authors wrote.
According to Swanton, by providing a first pan-cancer census of driver events in the context of tumors' cellular heterogeneity, the team's TCGA analysis offers more persuasive evidence than previously existed of the potential impact of this complexity on the success of targeted treatments.
It also provides some important first hints at how future therapeutic strategies could overcome this complexity.
Interestingly, Swanton said, he and his colleagues saw in the study that different subclones in a tumor appeared to follow parallel tracks of evolution, acquiring different individual mutations, but ones that seemed to affect the same genes or pathways overall. This suggests that the number of evolutionary paths a tumor can take may be finite.
The researchers also saw that the timing and clonality of various mutations appeared to frequently follow a similar pattern across different tumour types.
For example, they found that in four of the nine cancers, cells developed mutations driving expression of the protein APOBEC later in their development. Higher APOBEC activity appears to encourage higher mutation rates and a resulting increase in genetic diversity in a tumor, and thus a higher likelihood of treatment failure or resistance.
Ideally, an increasing understanding of evolutionary processes like this one, especially if they can be understood to follow a predictable pattern, could inform the development of therapeutic strategies to preempt future mutations, essentially blockading them from the proliferative options available.
"It's turning out to be remarkably constrained. There aren’t a limitless number of ways a tumor can evolve," Swanton said. "We are seeing each mutation constrains the next and the one after that, and if we understood that better the idea is that we could potentially initiate therapies that would stop a tumor taking that next path."
According to the study authors, the results also have implications for the drug development process. When patients don't respond to targeted treatment despite testing positive for a particular mutation, it might not necessarily mean that an experimental drug has failed, it could be effective in those with clonal mutations but ineffective in those with subclonal mutations.
Thus, dividing patients with clonal versus subclonal mutations could tease out whether there is a sensitive subgroup of patients who do benefit from treatment.
Admittedly, Swanton said, the practicality of looking so deeply at tumor heterogeneity in the context of drug development or in clinical practice may be far off.
Pharmaceutical developers have only recently embraced molecular subtyping of patients in clinical trials. Further stratifying patients based on the clonality of their genomic alterations is a whole other story.
"This is not easy. It's turning out to be more complex than we'd hoped, but that's not a reason to back away," Swanton said.
Moving forward, he and his colleagues are now working to collect more clear evidence of the effect of tumor heterogeneity on drug response in several current and planned studies.
Cancer Research UK is funding the ongoing TRACERx study in which Swanton's group is analyzing multiple biopsy samples in lung cancer patients from their primary disease diagnosis and adjuvant treatment through progression and metastasis if it occurs.
Unlike in the TCGA study, in which the group was only able to analyze single biopsies — likely underestimating the presence of subclonal mutations — in TRACERx the researchers are looking at multiple samples to gain a more accurate picture of heterogeneity, and overcome what Swanson called the "illusion of clonal dominance."
"A particular mutation can look clonal in one region, but in another region that mutation is completely absent," he said.
Then, in a set of follow up studies called DARWIN I and II, recurrent TRACERx patients will be treated, if possible, with targeted drugs based on their genomic profile. The group will investigate whether patients with targeted mutations in the majority of their tumor cells indeed do better on targeted therapy than those with mutations in only a subset of cells, Swanton said.
The investigators will also look for particular mutations present in subclonal populations that appear to confer resistance to particular drugs.
According to Swanton, his team is also involved in or planning studies of tumor diversity and evolution in renal cancer, esophageal cancer, and breast cancer.