Abstract
Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods References Decision letter Author response Article and author information Metrics Abstract Breast cancer genomes have revealed a novel form of mutation showers (kataegis) in which multiple same-strand substitutions at C:G pairs spaced one to several hundred nucleotides apart are clustered over kilobase-sized regions, often associated with sites of DNA rearrangement. We show kataegis can result from AID/APOBEC-catalysed cytidine deamination in the vicinity of DNA breaks, likely through action on single-stranded DNA exposed during resection. Cancer-like kataegis can be recapitulated by expression of AID/APOBEC family deaminases in yeast where it largely depends on uracil excision, which generates an abasic site for strand breakage. Localized kataegis can also be nucleated by an I-SceI-induced break. Genome-wide patterns of APOBEC3-catalyzed deamination in yeast reveal APOBEC3B and 3A as the deaminases whose mutational signatures are most similar to those of breast cancer kataegic mutations. Together with expression and functional assays, the results implicate APOBEC3B/A in breast cancer hypermutation and give insight into the mechanism of kataegis. https://doi.org/10.7554/eLife.00534.001 eLife digest The genomes of cancer cells contain mutations that are not present in normal cells. Some of these prevent cells from repairing their DNA, while others give rise to tumours by causing cells to multiply uncontrollably. Moreover, some of the mutations in breast cancer cells occur in clusters—a phenomenon known as kataegis (from the Greek for 'thunderstorm'). Kataegic mutations occur almost exclusively at a cytosine preceded by a thymine. This suggests that a family of proteins called AID/APOBEC enzymes—which remove amine groups from cytosines—may be involved in generating these mutations. In this study, Taylor et al. confirm this possibility by showing that expressing individual members of the AID/APOBEC family of enzymes in yeast cells increases the mutation frequency and induces kataegis. The kataegis triggered by the AID/APOBEC enzymes could be localised through the introduction of double-stranded breaks into the DNA: Taylor et al. suggest that this might happen because repairing the breaks exposes single-stranded DNA, which the AID/APOBEC enzymes then act upon. By comparing the mutations induced in the yeast cells with those observed in breast cancer cells, Taylor et al. identified APOBEC3B as the enzyme most likely to be responsible for kataegis in breast cancer (with APOBEC3A also a strong candidate in some cancers). Moreover, they showed that APOBEC3B was highly expressed in breast cancer cell lines, and that APOBEC3B and APOBEC3A can also cause DNA damage in human cells. Taken together, the findings provide key insights into the mechanism by which kataegis arises, and identify two proteins likely to contribute to the mutations seen in breast cancer. Further work is now required to determine whether these enzymes also give rise to mutations in other forms of cancer. https://doi.org/10.7554/eLife.00534.002 Introduction Whole genome sequencing of 21 breast cancers recently revealed the presence in more than half the cancers of a novel form of localised hypermutation (termed kataegis) in which clusters of multiple same-strand mutations at C:G pairs are focused on multikilobase-long genomic regions with adjacent mutations within each cluster separated by one to several hundred base pairs (Nik-Zainal et al., 2012). Although the mechanism underlying kataegis is unknown, the fact that the mutations occurred nearly exclusively at C residues preceded by a 5′-T suggested that AID/APOBEC cytidine-DNA deaminases might possibly be involved in the process since these enzymes are sensitive to the 5′-flanking sequence context (Conticello et al., 2007; Nik-Zainal et al., 2012). Members of the AID/APOBEC family of enzymes (reviewed in Conticello et al., 2007) deaminate cytosine in the context of a single-stranded polynucleotide substrate, and function in adaptive and innate immunity. AID acts on C residues in the DNA of the genomic immunoglobulin loci in activated lymphocytes to trigger antibody gene diversification whereas APOBEC3 family members, of which there are seven in humans, act on C residues in the DNA of viral replication intermediates (usually in the cytoplasm) as part of a host restriction pathway. Off-target deamination by AID results in nucleotide substitutions and genomic rearrangements in B lymphocyte tumours, some of which are implicated in oncogenesis (reviewed by Gazumyan et al., 2012). Although AID is the only member of the AID/APOBEC family known to act physiologically on endogenous nuclear DNA, it is possible that other members of the AID/APOBEC family might occasionally get access to the nucleus and cause cancer-associated genomic damage or mutation (Harris et al., 2002; Beale et al., 2004; Vartanian et al., 2008; Stenglein et al., 2010; Landry et al., 2011; Nik-Zainal et al., 2012; Nowarski et al., 2012). Here we have asked whether we could recapitulate cancer-like kataegis by expression of different AID/APOBEC enzymes in budding yeast and if so, use the tractability of yeast to gain insight into the kataegic process. The results provide insight not only into the mechanism of kataegis but also provide a strong pointer to the identity of the deaminases responsible for the kataegis observed in breast cancers. Results Recapitulating kataegis in yeast with AID/APOBEC deaminases Several members of the AID/APOBEC family members were expressed in yeast and all were found to give a significant increase in the mutation frequency as judged by the yield of colonies resistant to canavanine (CanR) (Figure 1—figure supplement 1). Genome sequencing, however, revealed that most CanR colonies had typically accumulated less than 10 point mutations (>98% at C:G pairs) during the period of AID/APOBEC induction and clonal expansion (Figure 1A). A hyperactive mutant of AID (AID*; Wang et al., 2009) gave a significantly higher mutation load (median of 25 mutations per genome). We therefore initially focused on the mutations in AID*-transformants (1078 mutations in total, of which all except 14 were at C:G pairs) to see if there were signs of kataegis. Figure 1 with 1 supplement see all Download asset Open asset AID/APOBEC-induced kataegis in yeast. (A) The total number of mutations detected in canavanine-resistant (CanR) AID/APOBEC yeast transformants, median frequency indicated. (B) Rainfall plot of genome-wide intermutational distances (IMD) in individual CanR AID* transformants. Clone identifier is indicated along the top, mutations shown as dots with the y-axis giving the distance to the next downstream mutation on the same chromosome. For each clone, dots are ordered sequentially along the genome. Dot colours represent: cluster mutations (at C, red; G, black), unclustered mutations (at C, pink; G, grey). Mutations at A:T (14 out of 1078 total), single mutations on individual chromosomes (15% of the database), the most downstream mutation in each chromosome and transformants without any multiply mutated chromosomes (5/40) are not depicted. Supplementary file 1B contains the location of all identified mutations. (C) Observed distribution of IMDs in the AID* dataset compared to a simulation assuming mutations are randomly scattered throughout the genome. (D) IMD plots of APOBEC3A/B/G*-expressing yeast transformants (28/78 transformants harboured no multiply mutated chromosome and are not depicted). (E) Detailed view of AID* mutation clusters. Each line represents an individual cluster with the clone identifier (grey box), number of mutations in the cluster and total mutations in the clone indicated. Mutations are coloured as in (B), a horizontal line indicates mutations that have coalesced, * indicates clusters localising within 10 kb of CAN1. All clusters containing ≥5 mutations are depicted. (F) Mutation clusters identified in yeast APOBEC3 transformants. https://doi.org/10.7554/eLife.00534.003 The distances between neighbouring mutations in the AID* yeast transformants are displayed as rainfall plots (Figure 1B). While the median overall intermutational distance (IMD) is 13 kb, it is apparent that rather than the mutations being scattered randomly over the genome, mutation distribution is bimodal (Figure 1C). Dividing the mutations into two groups using k-means cluster analysis (Hartigan and Wong, 1979) reveals that one group exhibit a median IMD of 156 kb with a distribution of distances that is as expected for a set of individual mutations randomly scattered over the yeast genome as judged by Monte Carlo simulation (Figure 1C). We designate these as singlet mutations: they account for 52% of the total mutations. The remaining 48% of the mutations are much more closely spaced than would be expected on a random basis. We designate these as proximal mutations, which are separated from each other by a median IMD of only 727 bp with >99% of them being within 8.5 kb of their closest neighbour. The overwhelming majority of the proximal mutations in the AID* transformants do not occur as isolated mutational pairs but, rather, are found in clusters. Thus, if we define proximal mutations as a pair of mutations that are located <8.5 kb apart (a distance that excludes 99% of the singlet mutations) and define a cluster as a stretch of DNA containing ≥5 proximal mutations, we find that 75% of the AID*-induced proximal mutations are actually parts of clusters. These clusters typically extend over 6–15 kb (with the full range detected being 1.8–30 kb) and contain anything up to 26 mutations (Figure 1E). This clustering is far in excess of anything that would be expected on a random basis. The level of mutation clustering observed with AID* is such that more than one-third of the transformants analysed (16/40) contain at least one mutatio