Translating insights into tumor evolution to clinical practice: promises and challenges

Abstract

Accelerating technological advances have allowed the widespread genomic profiling of tumors. As yet, however, the vast catalogues of mutations that have been identified have made only a modest impact on clinical medicine. Massively parallel sequencing has informed our understanding of the genetic evolution and heterogeneity of cancers, allowing us to place these mutational catalogues into a meaningful context. Here, we review the methods used to measure tumor evolution and heterogeneity, and the potential and challenges for translating the insights gained to achieve clinical impact for cancer therapy, monitoring, early detection, risk stratification, and prevention. We discuss how tumor evolution can guide cancer therapy by targeting clonal and subclonal mutations both individually and in combination. Circulating tumor DNA and circulating tumor cells can be leveraged for monitoring the efficacy of therapy and for tracking the emergence of resistant subclones. The evolutionary history of tumors can be deduced for late-stage cancers, either directly by sampling precursor lesions or by leveraging computational approaches to infer the timing of driver events. This approach can identify recurrent early driver mutations that represent promising avenues for future early detection strategies. Emerging evidence suggests that mutational processes and complex clonal dynamics are active even in normal development and aging. This will make discriminating developing malignant neoplasms from normal aging cell lineages a challenge. Furthermore, insight into signatures of mutational processes that are active early in tumor evolution may allow the development of cancer-prevention approaches. Research and clinical studies that incorporate an appreciation of the complex evolutionary patterns in tumors will not only produce more meaningful genomic data, but also better exploit the vulnerabilities of cancer, resulting in improved treatment outcomes.

Fig. 1

Sampling decisions required for comprehensive and evolutionary description of tumors. Tumor genomic sampling can be considered to fall into three separate domains. a Sampling of tumor material, either directly from a tumor mass or shed into the circulation. Samples from the tumor mass can either be pooled as a bulk specimen or disaggregated into single cells. b Only portions of genomic material are sampled and assessed; either targeted panels of a few hundred genes can be used or the whole exome or whole genome can be profiled. c Bulk DNA extractions may contain millions of DNA molecules. These are contributed by different parental alleles from both tumor and normal cells. Samples frequently contain 10–80% normal cells. Library preparation and sequencing only samples a tiny fraction of the available DNA fragments. The schematic shows a representation of sampling at two different sequencing depths (100X and 6X) and illustrates how higher sequencing depths allow more accurate determinations of the frequencies of specific mutations and their clonal or subclonal status. ctDNA circulating tumor DNA

Fig. 2

Evolutionary therapy strategies. Schematics of tumor populations in which each different color implies a new subclonal population. Therapies are denoted by segmented ovals, in which the targeted populations are indicated by the segment shading. a Targeting a clonal mutation that developed in or prior to the most recent common ancestor (MRCA). Resistance may emerge because a (rare) subclone with intrinsic resistance to that therapy (for example, an ESR1-activating mutation) existed prior to therapy. b Targeting of multiple drivers is more likely to lead to tumor extinction. c In adaptive therapy, treatment is discontinued before sensitive cells (pink) are eliminated, allowing them to grow back and suppress resistant cells (red). The resistant subclone would be expected to have an intrinsic survival disadvantage that is related to its resistant phenotype, for example, it may have lost the targeted driver mutation