Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Cancer

Drivers and passengers

Studies that have provided the first unbiased, large-scale analyses of DNA mutations across an array of cancers also have lessons for the proposal to annotate the entire cancer genome.

Cancer results from an accumulation of mutations and other heritable changes in susceptible cells. So far, abnormalities in about 350 genes have been implicated in human cancers, but the true number of 'cancer genes' is unknown. On page 153 of this issue, Greenman and colleagues1 build on their previous analyses of breast, lung and brain tumours2,3,4 by identifying mutations in the genes encoding all known protein kinases — enzymes that regulate other proteins through the addition of a phosphate residue — across various types of cancer. Together with a whole-genome resequencing analysis of a smaller number of breast and colorectal cancers published by Sjöblom et al.5 in October 2006, this study presents a largely unbiased overview of the spectrum of mutations in human cancers.

Greenman and co-workers1 undertook a comprehensive sequencing of 518 protein-kinase-encoding genes in 210 cancers. Kinases have been implicated in many aspects of tumorigenesis and several have now been validated as targets for drug therapy. The spectacular success of the drug imatinib (Gleevec) in treating chronic myeloid leukaemia stems from its suppression of a kinase known as BCR–ABL, which is the product of a gene located within a cancer-specific, translocated chromosome6. In their analysis of the collection of cellular kinases, the 'kinome', Greenman et al. identified 1,000 mutations. Mutations were relatively common in cancers of the lung, stomach, ovary, colon and kidney, and rare in cancers of the testis and breast, and in carcinoid tumours, which are usually found in the gastrointestinal tract. Tumours with defects in DNA-mismatch repair harboured large numbers of mutations, whereas other types of tumour revealed no detectable mutations. Specific patterns of nucleotide substitution differed among cancers from various tissue types, possibly reflecting the effects of external mutagens or defects in DNA repair.

By the time a cancer is diagnosed, it comprises billions of cells carrying the DNA abnormalities that initiated malignant proliferation and many additional genetic lesions acquired along the way. Some of these secondary mutations emerge owing to selective pressure during tumorigenesis (drivers); others may be incidental (passengers), resulting from mutational exposures, genome instability or simply the large number of cell doublings that leads from a single transformed cell to a clinically detectable cancer. To distinguish driver from passenger mutations, Greenman et al. used a statistical model comparing the observed-to-expected ratio of synonymous (no amino-acid change) mutations with that of non-synonymous (altered amino acid) mutations. An increased proportion of non-synonymous mutations implies selection pressure during tumorigenesis. Overall, 158 predicted driver mutations were identified in 120 genes encoding kinases. In contrast to the recurrent mutations in the gene encoding a kinase known as BRAF and previously identified by this group in malignant melanomas7, most kinase mutations identified across different tumour types were 'single hits'.

Sjöblom and co-workers5 used a different strategy, but reached similar conclusions. By initially sequencing about 13,000 genes in 11 breast and 11 colorectal cancer cell lines, they identified 1,307 validated nucleotide changes in 1,149 genes, of which 189 met their criteria for significance. Few overlapping driver mutations were identified between the kinase genes analysed in these two studies, highlighting the requirement for a very large number of samples to capture the full repertoire of genetic heterogeneity in cancer.

Indeed, the US National Cancer Institute and the National Human Genome Research Institute have proposed the Human Cancer Genome Project8, which aims to sequence 12,500 tumour samples (250 specimens from 50 different cancers), focusing on at least 2,000 genes implicated in tumorigenesis. The scale of this proposed analysis is such that it should reveal recurrent mutations in subsets of cancer, an important clue to identifying truly significant drivers. Nonetheless, small nucleotide changes within genes that are detectable by sequencing constitute only a subset of abnormalities underlying human cancer — gene amplification or deletion, inactivation of genes through epigenetic silencing, and chromosomal translocations also contribute to alterations in cancer genes (Fig. 1), and these would not be detectable by sequence analysis of known genes.

Figure 1: The variety of defects underlying human cancer.
figure1

DNA-sequencing strategies, such as those described by Greenman et al.1 and Sjöblom et al.5, are aimed at detecting small nucleotide changes within genes (intragenic mutations), and identify these genes as the target of a mutational event (highlighted in green). But various abnormalities may be transmitted from a cancer cell to its progeny, some of which activate specific genes (gain-of-function mutations), whereas others inactivate them (loss-of-function mutations). Alternative strategies, such as comparative genomic hybridization12, are required to scan for deletion or amplification of chromosome fragments, which often contain many genes, making it more difficult to identify the specific gene(s) targeted by these events. Epigenetic silencing involves heritable modifications of nucleotides and histones in regulatory regions of genes, leading to suppression of gene expression in the absence of DNA mutations. Translocations lead to the fusion of DNA fragments from different chromosomal regions, either creating an abnormal fusion protein or leading to aberrant expression of a normal gene.

The Human Cancer Genome Project is likely to be costly at this time of crisis in US funding for biomedical research, and its ultimate value will depend on identifying crucial genetic lesions that point to more effective therapies from among the many driver mutations identified in such a screen. The exceptional dependence of chronic myeloid leukaemia on the translocated BCR-ABL kinase6, and the dependence of some 10% of non-small-cell lung cancers on a mutated epidermal growth factor receptor kinase9,10, are both correlated with dramatic responses to small-molecule inhibitors. These and, it is to be hoped, other genetically defined, highly drug-responsive subsets of cancer that are yet to be identified, exemplify the phenomenon of 'oncogene addiction', whereby only one of the many genetic lesions in a tumour proves to be its Achilles' heel11. Mutant kinases and other critically altered proteins in cancer cells may thus prove to be good drug targets; however, many other driver mutations that occur with tumour progression may not be essential for tumour maintenance. Therefore, complementary functional screens will be needed to validate the potential of suspected cancer genes as therapeutic targets.

Together, these initial sequencing studies1,5 provide insight into mutational profiles in various cancers that might help to identify the molecular mechanisms responsible for tumour initiation and progression. But, in considering the application of sequencing strategies to a vast number of specimens from cancer patients, these studies also give a cautionary traffic signal — each cancer genome carries many unique abnormalities, and not all mutations identified contribute equally to the manifestation of the associated cancers. A combination of genetic and functional approaches will be essential to correctly identify true drivers from the many passengers on the road to tumorigenesis.

References

  1. 1

    Greenman, C. et al. Nature 446, 153–158 (2007).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Stephens, P. et al. Nature Genet. 37, 590–592 (2005).

    CAS  Article  Google Scholar 

  3. 3

    Stephens, P. et al. Nature 431, 525–526 (2004).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Hunter, C. et al. Cancer Res. 66, 3987–3991 (2006).

    CAS  Article  Google Scholar 

  5. 5

    Sjöblom, T. et al. Science 314, 268–274 (2006).

    ADS  Article  Google Scholar 

  6. 6

    Kantarjian, H. et al. N. Engl. J. Med. 346, 645–652 (2002).

    CAS  Article  Google Scholar 

  7. 7

    Davies, H. et al. Nature 417, 949–954 (2002).

    ADS  CAS  Article  Google Scholar 

  8. 8

    von Eschenbach, A. & Collins, F. http://cancergenome.nih.gov/about/TCGA_executive_summary.pdf (2005).

  9. 9

    Lynch, T. J. et al. N. Engl. J. Med. 350, 2129–2139 (2004).

    CAS  Article  Google Scholar 

  10. 10

    Paez, J. G. et al. Science 304, 1497–1500 (2004).

    ADS  CAS  Article  Google Scholar 

  11. 11

    Weinstein, I. B. Science 297, 63–64 (2002).

    CAS  Article  Google Scholar 

  12. 12

    Pinkel, D. & Albertson, D. G. Annu. Rev. Genomics Hum. Genet. 6, 331–354 (2005).

    CAS  Article  Google Scholar 

Download references

Author information

Affiliations

Authors

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Haber, D., Settleman, J. Drivers and passengers. Nature 446, 145–146 (2007). https://doi.org/10.1038/446145a

Download citation

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing