Editors' ChoiceCancer

Oncogenic Driver Mutations: Neither Tissue-Specific nor Independent

See allHide authors and affiliations

Science Translational Medicine  04 Dec 2013:
Vol. 5, Issue 214, pp. 214ec200
DOI: 10.1126/scitranslmed.3008075

Large-scale cancer genomics projects offer the promise of informing genetically targeted therapies according to a molecular, rather than histopathologic, taxonomy of cancer. Over the past several years, the Cancer Genome Atlas (TCGA) research network has published detailed molecular analysis of multiple tumor types, creating a critical mass of data with which to undertake the comparative analysis of genetic alterations underlying a variety of cancers. Over the past month, a consortium of TCGA researchers has published a collection of “Pan-Cancer” analysis papers, reporting the analysis of more than 3000 tumors of 12 different types, each profiled with up to six molecular data modalities.

Kandoth et al. have presented the landscape of somatic mutations across 12 tumor types based on whole-genome and exome-sequencing data from the Pan-Cancer data set. Average mutation frequencies varied significantly across tumor types, ranging from 0.28 mutations per megabase in acute myelogenous leukemia to 8.15 mutations per megabase in lung squamous cell carcinoma. Statistical analysis within and across tumor types identified only 127 significantly mutated genes (SMGs), after filtering out those with low expression levels. Nearly all (93%) samples contained nonsynonymous mutations in at least one SMG, averaging between two and six mutations of SMGs per tumor.

TP53 was by far the most frequently mutated gene (42% of samples), with PIK3CA a distant second (17.8% of samples). All other genes were mutated in less than 10% of samples. Most SMGs displayed selectively high mutation frequencies in a particular tumor type and elevated mutation frequencies in several other tumor types. Mutations of several genes, including TP53, were correlated with poor prognosis. Last, analysis of the variant allele fraction distribution suggested a sequence of driver mutations in clonal evolution, with mutations in TP53, DNMT3A, and PIK3CA being early tumor-initiating events, and mutations in KRAS and NRAS contributing to later-stage progression.

The vision of personalized, molecularly targeted anticancer therapies relies on a future in which the molecular drivers of individual tumors can be routinely characterized. The first step toward this future is to catalog the “parts list” of driver mutations and their prevalence across tumor types. The work of Kandoth et al. demonstrates that nearly all highly recurrent mutations show increased frequencies across multiple tumor types. Therefore, a critical translational question is to characterize the extent to which such mutations correlate with therapeutic response across tissues. More generally, the relatively low number of SMGs is perhaps a surprising finding of such large-scale cancer genome surveys. However, it is possible that many more genes may be implicated in cancer based on additional mechanisms of dysregulation, such as methylation, or that low-frequency events may constitute a “long tail” of mutations with collectively high importance. A detailed molecular taxonomy of cancer will require sustained efforts to characterize the full menu of cancer-related molecular alterations and determine how the vast combinatorial space of such mutations converges to a discrete number of molecular subtypes linked to therapeutic strategies.

C. Kandoth et al., Mutational landscape and significance across 12 major cancer types. Nature 502, 333–339 (2013). [Full Text]

Navigate This Article