Mitotic recombination events and single-base mutations induced by ultraviolet light in G1-arrested yeast cells

Analysis of genomic rearrangements induced by UV in the diploid YZ1.

The two haploid strains used to generate the diploid YZ1 differ by about 55,000 SNPs allowing us to detect mitotic recombination events and other genomic alterations throughout the genome at high resolution (7, 10). We performed whole-genome sequencing on 22 individual isolates each from YZ1 in which the haploid parent was treated with either 20 J/m² (YZ1-20) or 80 J/m² (YZ1-80). Before sequencing, we did single-colony purifications from each of the original colonies. The sequences for all 44 of these isolates are available online (SRA database, with the accession number PRJNA1249510). The numbers and rates of various types of genomic alterations in the YZ1-20 and YZ1-80 are summarized in Table 1. The coordinates of the LOH breakpoints and other relevant details of the genomic alterations are shown in Dataset S1-1 and S1-2.

As expected from our previous studies and those of others, UV treatment greatly elevated the rates of many types of genetic changes. The rates of most genomic alterations were elevated several hundred- to several thousand-fold relative to the rates observed spontaneously (23). In addition, the rates observed in the cells irradiated with 80 J/m² were 5- to 10-fold higher than found in cells irradiated with 20 J/m². Although LOH events can result from either large deletions or mitotic recombination events (gene conversions and crossovers), in yeast, most LOH events reflect mitotic recombination (10). Large deletions result in loss of one type of allelic SNP without duplication of the other, whereas, in yeast, most LOH events represent loss of one type of allelic SNP and duplication of the other type. This distinction can be made by determining DNA sequence coverage of the heterozygous SNPs. No LOH events, large deletions/duplications, or single-base mutations were found in isogenic diploids (YZ1-0) made by crossing two unirradiated strains (one isolate examined by whole-genome sequencing and three analyzed with SNP-specific microarrays (10). The chromosome locations of all recombination events observed with YZ1 are shown in SI Appendix, Fig. S1.

As shown in Fig. 1, both I-LOH and T-LOH are a consequence of the repair of DNA breaks using the unbroken homolog as a template. We and others showed that interstitial deletions (I-DEL) and duplications (I-DUP) are a consequence of crossovers between nonallelic repeats (such as the Ty1 transposons) located within a chromosome. Terminal duplications (T-DUP) and terminal deletions (T-DEL) are the result of recombination between nonallelic repeats located on nonhomologous chromosomes (23, 24). Extra chromosomes (trisomy) are likely a consequence of chromosome nondisjunction whereas monosomy could reflect either nondisjunction or failure to repair a DSB (25). Examples of various types of genomic alterations as assayed by sequence coverage are shown in Fig. 4.

In the construction of YZ1, KO171 (the W303-related haploid) was irradiated and KO124 (the YJM789-related haploid) was unirradiated. The most critical class of LOH events relevant to the possibility that the irradiated chromosome can stimulate recombination in the unirradiated chromosome is the simple I-LOH event. As shown in Fig. 1B, an event initiated on the irradiated chromosome would result in an I-LOH event in which sequences from the irradiated chromosome are lost and those of the unirradiated chromosome are duplicated. As shown in Table 1, among the 107 simple I-LOH events observed in YZ1, 103 were initiated on the irradiated W303-derived homolog and only 4 were initiated on the YJM789-derived homology. Thus, we conclude that the level of trans induction of events on the unirradiated chromosome by the irradiated chromosome is very low or absent, and the UV-induced stimulation of mitotic recombination is primarily a consequence of a cis induction by recombinogenic lesions on the irradiated chromosome.

The small number of I-LOH events that initiated on the unirradiated chromosome may represent a minor degree of trans stimulation or a low level of spontaneous mitotic recombination events. The predicted number of spontaneous mitotic recombination events among 44 isolates (based on the number expected from the analysis of ref. 23) is 44 ×ばつ 25 ×ばつ 3.3 ×ばつ 10⁻³ or 4. It should be emphasized that our findings do not necessarily contradict those of Fabre and Roman, since their study did not examine the ratio of cis to trans induction of recombination. Finally, the bias in the initiation events of the W303- and YJM789-derived homologs cannot be explained by a tendency for one homolog to be more recombination-prone than the other. In experiments in which a G1-arrested diploid with a closely related background to YZ1 was irradiated, simple I-LOH events were initiated at similar frequencies from both types of homologs; 0.53 from the W303-derived homolog and 0.47 from the YJM789-derived homolog (7).

The other classes of LOH events that are relevant to the issue of trans recombination are the I-DEL and I-DUP classes. As discussed above, these classes are usually the result of unequal intrachromatid or sister-chromatid recombination between nonallelic repeats (reviewed in ref. 26). Adding these two classes together, we found seven events that deleted or duplicated W303-derived sequences and no events affecting the YJM789-derived sequences (Table 1), supporting the conclusion that most or all of the UV-stimulated LOH events are initiated on the irradiated chromosome.

Complex events were not used to calculate whether the irradiated or unirradiated chromosome initiated the genomic alteration. Complex events have multiple transitions between heterozygous SNPs and LOH regions and/or adjacent W303- and YJM789-derived LOH regions. Because these patterns could be produced by events initiating on either homolog, they were not included in our analysis. The complex LOH events (approximately one-quarter as frequent as the simple LOH events) are depicted in Datasets S1–S5.

SI Appendix, Fig. S2 summarizes the molecular mechanisms that produce gene conversion events and reciprocal crossovers. Both types of recombination are likely initiated by double-strand DNA breaks, followed by invasion of one broken end into the unbroken homolog. Simple LOH events (gene conversions) are primarily the result of heteroduplex formation between homologous sequences, followed by DNA synthesis. The invading end is then displaced, resulting in formation of a heteroduplex (Synthesis-Dependent Strand Annealing); repair of mismatches within the heteroduplex results in LOH. In conversion events associated with crossover, a double-Holliday junction is formed with two regions of heteroduplex. Cleavage of the junctions can result in a crossover associated with a conversion (SI Appendix, Fig. S2). Complex patterns of LOH can result from "patchy" repair of mismatches within the heteroduplexes or formation of symmetric heteroduplexes by branch migration (7, 10).

Last, we observed a UV-induced increase in aneuploidy in YZ1 (Table 1). This observation needs confirmation for two reasons. The number of events is small (only one event in cells exposed to 20 J/m²). In addition, 5 of the 13 events in the cells exposed to 80 J/m² were observed in a single isolate.

Analysis of mutations induced by UV in YZ1.

In addition to gross genomic alterations, the sequencing experiments also allowed us to determine the rate of single-base alterations, small (less than 100 bp) in/dels, and complex mutations (more than one mutation within a 6 bp window) induced by UV (Table 1). Relative to the rate observed in unirradiated cells (23), mutations were greatly elevated (>500-fold), roughly proportional to the UV dose. All mutations observed in YZ1 are listed in Dataset S2, and the spectrum of single-base mutations is shown in SI Appendix, Fig. S3. As expected from previous studies (for example, refs. 17, 18, and 27), C->T/G->A, A->T/T->A, and A->G/T->C mutations were elevated relative to their spontaneous frequencies. The location of these mutations on the chromosomes is shown in SI Appendix, Fig. S4. The numbers of mutations per chromosome are approximately proportional to chromosome size (SI Appendix, Fig. S5) and there are no very strong mutation hotspots.

We also observed complex mutations in which two mutations were within 6 bp of each other (Dataset S2). 16 of these 28 complex events involved adjacent mutations. The three most common types were (NN representing any mutant bases): CC or GG to NN (8 events), CT or AG to NN (4 events), and GT or AC to NN (4 events). The first two classes were also the most common adjacent mutations observed by Vanderberg et al. (27). In summary, the UV-induced mutations observed in our study are very similar to those observed previously in other S. cerevisiae studies.

One final issue is whether UV irradiation of one haploid can induce mutations in an unirradiated haploid following mating. As discussed above, the diploids examined in our study are heterozygous for about 55,000 SNPs. In our experiments, the average length of the Illumina "reads" was about 150 bp. For 20 J/m², of 42 events, 40 were on the W303 chromosome and 2 were on the YJM789 chromosome. For the 80 J/m²UV dose, of the 241 SNP-linked events, 239 events were on the W303 chromosome and two were on the YJM789 chromosome. Considering all SNP-linked mutations, 279 (99%) were linked to a W303 SNP (the irradiated haploid) and 4 (1%) were linked to a YJM789 SNP (Dataset S2). In summary, as observed for other types of genomic arrangements described above, the trans effect of UV irradiation on the induction of single-base mutations in the unirradiated chromosome is very small or absent, consistent with previous studies in wild-type yeast strains (28).

System for the analysis of both daughter cells containing the products of UV-induced recombination.

The YZ1 isolates were colonies that were unselected for LOH events. We next used a different diploid YZ2 in which we could monitor reciprocal mitotic crossovers on chromosome IV as well as unselected LOH events. For these experiments, we used a system similar to that described by Sui et al. (23). We crossed an irradiated haploid (Wspo11, relevant genotype: MATa ade2-1) to the unirradiated haploid SY166 (relevant genotype: MATα ade2-1 IV1510386::SUP4-o) using the same protocol as described above for YZ1; the resulting diploid is called YZ2. YZ2, like YZ1 is heterozygous for 55,000 SNPs, allowing high-resolution mapping of LOH events. As shown in Fig. 1, a crossover between the heterozygous SUP4-o mutation and the centromere can result in a red/white sectored colony. The analysis of sectored colonies allows detailed conclusions about the mechanism of the recombination events that were induced by UV, since the products of both daughter cells can be examined. For example, we could determine important features of the recombination event that cannot be examined by the examination of nonsectored colonies including: 1. Whether the event was initiated by a DSB formed in the irradiated G1 cell or a DSB generated during the subsequent S-period (G2 DSB), and 2. Whether T-LOH events are reciprocal crossovers or a product of break-induced replication (BIR) (10). Importantly, we can also infer details of unselected recombination and mutation events in daughter cells. As described below, this procedure also allowed us to unambiguously identify two-strand mutations induced by UV.

Analysis of genomic rearrangements induced by UV in the diploid YZ2.

As shown in Fig. 1, red/white sectored colonies in YZ2 are the result of reciprocal mitotic crossovers between a SUP4-o marker located near the right end of chromosome IV and CEN4. The frequencies of sectors (equivalent to the frequency of sectored colonies per total colonies) were 3.1 ×ばつ 10⁻⁵/genome in an unirradiated isogenic derivative of YZ2 (10), and 3.1 ×ばつ 10⁻³/genome (20 J/m²) and 3.9 ×ばつ 10⁻³/genome (80 J/m²) for irradiated cells. In the irradiated cells, these rates were the average of rates calculated for three independent experiments in which at least 1,000 colonies were examined for each UV dose. At both doses, the rates of recombination were elevated by UV roughly 100-fold, similar to the effects of UV in unsectored colonies of YZ1 (Table 1). It is likely that the sector frequency was not proportional to UV dosage in these experiments because the diploid cell viability was reduced to about 4% by the 80 J/m² UV dose, resulting in a decreased probability of recovering the two daughter cells necessary to form a sectored colony. As discussed below, the numbers of LOH events in the sectored colonies were roughly proportional to the UV dose (Table 2).

The patterns of LOH in the sectored colonies of YZ2 are shown in Dataset S3 and are summarized in Table 2; coordinates for LOH events are given in Dataset S1. The types of events are similar to those that we have observed previously in the analysis of sectored colonies resulting from spontaneous or UV-induced events (7, 10). Since red/white sectored colonies are generated by reciprocal crossovers on chromosome IV, all nine of the sectored colonies analyzed had such crossovers; most of these were associated with an adjacent region of gene conversion. In addition, there were many unselected recombination events with gene conversions unassociated with crossing-over (Class 1 in Table 2). As observed for the YZ1 strain, in Class 1 events in which the chromosome with the recombinogenic lesion could be identified, 98% (182/186) were located on the irradiated W303 chromosome, confirming that trans effects of UV on recombination are very rare or absent.

Because the sectors allow the analysis of all recombinant chromatids, gene conversions and crossovers can be analyzed throughout the genome in addition to the selected event on chromosome IV. As observed for YZ1 (SI Appendix, Fig. S1), the recombination events in YZ2 were found on all of the chromosomes with no pronounced hotspots (SI Appendix, Fig. S6). One relevant issue is whether the recombination events in our study are initiated in G1 or S/G2. These possibilities can be distinguished by examining the patterns of gene conversion in sectored colonies. As shown in Fig. 1 B and C, events resulting from a DSB in S/G2 would produce a 3:1 pattern of conversion (outlined in dotted lines). In contrast, following replication of chromosome broken in G1, DNA repair would be expected to induce a 4:0 conversion pattern (Fig. 1 D–F). If we divide the gene conversion events that are unassociated with crossovers (Class 1A-1C and Dataset S3) into those that have a 4:0 region of conversion and those that do not, about two-thirds of the events (121 of 185) reflect DSBs formed in G1. The DNA lesions that produce 3:1 conversion events may represent single-stranded nicks formed in G1 that result in one broken chromatid following DNA replication (5). The source of the G1-induced DSBs is not clear. It is possible that these DSBs result from the excision of two closely apposed dimers on opposite strands of the duplex, consistent with the observation that the ratio of 4:0 to 3:1 events is higher in cells with greater doses of UV (7).

In addition to 4:0 and 3:1 conversion events, we found conversion events in which 4:0 regions were adjacent to 3:1 tracts (Class 1C). Most such events can be explained by the independent repair of two DSBs with conversion tracts of different sizes (Fig. 1E). One striking observation is the abundance of conversion events (Class 1B) in which the endpoints of the 4:0 events are identical within resolution of the inter-SNP distances; this class is about half as frequent as the hybrid 3:1/4:0 tracts (Table 2). This result suggests that either the invading ends of the two broken DNA molecules catalyze DNA synthesis tracts of the same lengths during independent SDSA events or that the conversion event occurs between one broken DNA molecule and the unbroken template in G1. These possibilities will be discussed further below. In addition, we found some conversion events (for example, Classes 1D and 2E, Dataset S3.2) in which both the W303- and YJM789-derived chromosomes were donors in gene conversion events. Such events have been observed previously in UV-treated diploids and may reflect formation of symmetric heteroduplexes or template switching (7).

Another feature of mitotic recombination that is revealed by analysis of sectors is whether the T-LOH events are reciprocal (as expected for a crossover) or nonreciprocal (as expected for break-induced recombination [BIR]). In BIR events, the centromere-distal portion of the chromosome is lost, and the centromere-proximal end invades the homolog and copies sequences by a conservative type of DNA replication (29). As shown in SI Appendix, Fig. S7A, a reciprocal crossover produces sectored colonies with reciprocal patterns of T-LOH. A BIR event resulting from repair of a single broken sister chromatid would be expected to produce one sector with a T-LOH event and one sector without LOH (SI Appendix, Fig. S7B). We also observed double BIR events in which the two sectors had nonreciprocal T-LOH events (SI Appendix, Fig. S7 C and D). There were 29 reciprocal crossovers (excluding the selected crossover on chromosome IV) and 15 BIR events (8 involving repair of a single DSB and 7 involving the repair of two DSBs). The greater fraction of reciprocal crossovers compared to BIR events was also observed for spontaneous events (30).

Analysis of mutations induced by UV in sectored colonies of the diploid YZ2: evidence for two-strand mutations.

We also examined UV-mutations induced in the sectored colonies of YZ2. The location and types of all mutations are given in Dataset S4. The very striking observation is that the red and white halves of individual colonies often had the same mutation. This observation argues that the mutation induced by UV in the irradiated haploid strain must have had the same mutation in both strands of the duplex, a two-strand mutation (TSM; Fig. 5). A mutation in only one of the two duplex strands (OSM) would result in a mutation in only one of the two sectors. Of 103 mutations observed in YZ2 in which the haploid parent was treated with 20 J/m², 15 were OSM and 88 were TSM (80 involved single mutations and 8 involved two adjacent mutations). In the 80 J/m² isolate, we detected 789 mutations: 652 were paired mutations (326 pairs) involving one base (TSM), 56 were two adjacent paired mutations (TSM), 78 were unpaired single mutations (OSM), and 3 were more complex mutations. Thus, of the mutations that were unambiguously classified as TSM or OSM for the 80 J/m² data, 90% were two-strand mutations, very similar to the percentage observed in the 20 J/m² sample (85%). Since one-strand mutations produce one sector with a mutation whereas two-strand mutants produce two mutant isolates (one in each sector), we conclude that the ratio of TSMs to OSMs is about 5:1. Mechanisms that may induce TSMs will be considered in Discussion.

We also determined the mutation spectrum for the YZ2 isolates. The pattern of substitutions is similar to that observed for YZ1 and different from that observed in wild-type cells (SI Appendix, Fig. S3). Significant elevations in the frequencies of C->T/G->A, A->T/T->A, and A->G/T->C were observed. We also examined separately the types of mutations in one-strand and two-strand mutations. For this analysis, the number of two-strand mutations was divided by two since each pair is a consequence of a single event. The types of mutation were broadly similar, although the C->T/G->A class was no longer significant for the one-strand mutations.

We also determined whether the UV-induced mutations were linked to SNPs on the irradiated W303-derived chromosome or the unirradiated YJM789-derived chromosome (Dataset S4). Almost all of the paired two-strand mutations were on the irradiated chromosomes (195 of 196 pairs), as well as 31 of 31 of the one-strand mutations, confirming the lack of a trans effect for UV-induced mutagenesis. As observed for mutations in YZ1, the YZ2 mutations (both OSM and TSM) were distributed over all chromosomes without very pronounced hotspots (SI Appendix, Fig. S8). Although one possible difference between the OSM and TSM mutations could reflect the timing of DNA replication, no significant effect relationship was observed between the location of OSM and TSM mutations and early- and late-firing origins of replication (SI Appendix, Fig. S8).