Human cancer genomes are strewn with mutational patterns that represent encounters with environmental mutagens, endogenous exposures and mutational processes. When viewed at any one point in time after a cancer has developed, it becomes impossible to precisely deconvolute how mutations were introduced by such DNA damage as well as the failure of DNA repair processes to correct these errors during DNA replication. Therefore, Aitken et al. reasoned that a genetically reductive and controlled model system could be used to trace an individual strand of the DNA double helix to which damage occurred and correlate this with mutational patterns to inform upon tumour evolution.
The model system comprised of chemically inducing liver tumours in mice through single applications of the mutagen diethylnitrosamine (DEN). Essentially, the carcinogen induces DNA lesions, sites of damage in the base-pairing of DNA that affects one of the two DNA strands, known respectively as the Watson and Crick strands. The resultant 371 tumours that formed from 104 treated mice were whole-genome sequenced to show that each genome had approximately 60,000 somatic point mutations. Most of these point mutations were thymine (T) to N (where N = any other nucleotide) conversions consistent with DEN acting on T bases. Analysing the patterns of mutations revealed that large sections of the genome, often whole chromosomes, displayed strand asymmetry in the distribution of mutations.
The authors present an explanation for this observation whereby lesions induced by DEN persist unrepaired through multiple rounds of cell division. During the first replication cycle, the two parental Watson and Crick strands carrying independent lesions are segregated to form sister chromatids paired with daughter strands in which an incorrect base is matched with the damaged T base. The resulting two daughter cells then carry distinct mutation profiles, a concept the authors’ term lesion segregation. The subsequent rounds of replication then generate distinct lineages, some of which will have driver mutations that can undergo clonal expansion.
The lesion segregation model generated from tracking individual strands enabled the authors to explore events that have previously been hard to quantify in vivo. For example, DNA lesions on the transcribed strand can be repaired through transcription-coupled repair (TCR). Within their model, as predicted, TCR occurred preferentially on the transcribed strand, and increased transcription, such as observed with highly expressed genes, was linked to higher rates of repair and hence, reduced numbers of mutations. The authors also noted that the strand asymmetry of mutations did not always stretch across entire chromosomes, and frequently, within one chromosome an enrichment of mutations could be seen that had switched over to the other strand (A to N conversions). The authors propose that these events are sister-chromatid exchanges (SCEs) where identical regions of parental strands are exchanged before a cell divides, which arises from homologous recombination-mediated DNA repair. A higher frequency of SCEs was associated with a higher mutation rate, perhaps indicating that the presence of DEN-induced lesions might trigger SCEs. Last, the positive selection of driver mutations associated with cancer could be quantified. The assumption was that either the parental Watson or Crick strand would be inherited equally in tumours, but this was not the case, and often there was a bias for a specific strand containing a driver gene mutation. Identifying these retained strands revealed three tumour-promoting genes, Braf, Hras and Egfr, all known to be involved in human liver cancer progression.
The segregated lesion-bearing DNA strands could serve as templates for DNA replication through subsequent cell divisions. The outcome of this would be that the unrepaired, damaged base of the lesion becomes mispaired with different incorrect bases at each replication, generating different mutations at the same base-pair position, known as multi-allelic variation. Indeed, this was observed in DEN-induced tumour genomes and gave rise to genetic diversity.
Yet, all of these observations were made in mice that had been treated with one specific mutagen and so the authors sought to determine if lesion segregation occurred with other DNA-damaging agents and importantly, in human cancers. Using previously published data in which human induced pluripotent stem cells (iPSCs) were singly exposed to 79 environmental mutagens, strand asymmetry of mutations could be demonstrated with mutagens ranging from sunlight, tobacco smoke and chemotherapeutics. As most human cancers are likely to develop following repeated exposures to various mutagen-causing agents over time, it might have been presumed that any evidence of mutational asymmetry would be concealed as DNA lesions are continuously induced. However, mutational asymmetry was detected in kidney, liver and bile duct human cancers that arose following acute exposure to aristolochic acid.
“strand asymmetry in the distribution of mutations”
The lesion segregation concept sheds light on the many, more complex combinations of mutations that could develop in tumours.
Aitken, S. J. et al. Pervasive lesion segregation shapes cancer genome evolution. Nature 583, 265–270 (2020)