The genetic mutations that lead to cancer are caused by diverse, often poorly understood processes, some of which involve exposure to external agents. Excessive ultraviolet light is linked to melanoma, for example, and tobacco smoke to lung cancer. A molecular mechanism called nucleotide excision repair deals with UV- and smoke-induced genetic damage by removing damaged pieces of DNA, preventing mutations from arising. However, this process is complicated by the fact that repair occurs alongside other crucial genetic activities, such as DNA transcription. Two papers1,2 in this issue of Nature demonstrate how interplay between the DNA-repair and transcription-initiation machinery leads to an increased mutation rate in regulatory regions of the genome.
Although most cancer studies have focused on mutations in protein-coding DNA, there is a growing understanding of the importance of the non-coding DNA regions that regulate gene expression3,4,5,6 — promoter sequences, which are located close to genes, and distant elements called enhancers. Binding of these regions by transcription factors modulates the expression levels of associated genes. On page 264, Sabarinathan et al.1 describe the use of whole-genome sequences from human melanoma samples to analyse mutations in regulatory regions. They found that the cores of the regulatory regions, where transcription factors are predicted to bind, have a mutation rate five times higher than the flanking sequences.
Because of the major role of nucleotide excision repair (NER) in fixing UV-induced DNA damage, Sabarinathan and colleagues next analysed the locations of NER activity7. This revealed that the increased mutation rates at transcription-factor binding sites were caused by reduced levels of NER. The authors reasoned that mutations in other cancers that rely on NER should also exhibit this pattern. And indeed, they found increased mutation rates at transcription-factor binding sites in lung-cancer samples, particularly for mutations linked to smoking.
On page 259, Perera et al.2 report the analysis of mutations in regulatory elements in multiple cancer types. They found increased mutation density in the centres of active promoters associated with reduced levels of NER. Moreover, the authors' data suggest that mutation density in regulatory regions is linked not only to transcription-factor binding, but also to the level of transcription initiation.
Thus, two independent studies show that NER at regulatory DNA regions is inhibited by the bound transcription-initiation machinery. This discovery is especially interesting in light of a previous study8 that showed that mutation density is decreased over active regulatory regions as a whole, relative to their flanking sequences. The authors of that paper proposed that this decrease occurred because active regulatory regions are more accessible than most DNA regions to repair proteins — DNA is typically packaged around proteins called histones, but regulatory regions are unwound for binding by the transcription-initiation machinery. This apparent discrepancy with the current studies reflects the fact that, although regulatory regions as a whole are accessible for NER, the repair machinery is unable to access the core sites within those regions at which transcription factors bind (Fig. 1).
Certain mutations are considered to be drivers of cancer, because they provide a growth advantage to tumour cells. Such mutations are generally identified by the high frequency at which they occur across patients. However, the current studies highlight that protein binding can also lead to high mutation frequency — and so can other factors, such as late replication of a region during cell division9. Understanding how these features co-vary with mutation rate is vital for designing accurate computer algorithms to identify driver mutations10.
It is notable that the variables affecting mutation rate differ for cancer types and subtypes. For instance, unlike in skin and lung cancer, NER does not have a major role in colon cancer. Accordingly, the current studies found no increase in mutation density at the centres of active promoters in colon-cancer samples.
Errors introduced by DNA replication in colon cells are normally resolved by a process called mismatch repair, which is most effective in genomic regions that replicate early during cell division. Thus, mutation rates in colon-cancer cells are generally lower in early-replicating than in late-replicating regions11. Mismatch-repair proteins are, however, inactivated in some colon tumours, resulting in the loss of strong correlation between mutation density and replication timing. In fact, the regional 'landscape' of mutation rates can be used to infer the time of mismatch-repair inactivation in the history of a colon tumour.
In the past few years, the complex interplay between DNA-repair mechanisms and genomic properties not originally associated with repair (such as replication timing and DNA accessibility) has become evident, largely thanks to the increasing availability of whole-genome sequences from tumour samples. The need for such sequences from cancer cells has been debated, because they are costly and have limited immediate clinical value3. But the current studies demonstrate the immense potential of whole-genome sequences as a lens through which to examine the cellular processes that shape the cancer genome. Genomic studies such as these lay the groundwork for future diagnostic tools and treatments tailored to individuals.
It remains unclear how many more genomic features that correlate with mutation rate are yet to be found. All mutations are ultimately the result of faulty DNA repair — do we need to know all the details of the many ways in which repair can break down to harness the full power of genomics for cancer care? The increasing number of tumour-genome sequences, coupled with our ever-improving knowledge of the machinery involved in genome function, will hold the answer to this question.