Positive selection of yeast nonhomologous end-joining genes and a retrotransposon conflict hypothesis
-
Edited by Susan R. Wessler, University of Georgia, Athens, GA, and approved September 14, 2006 (received for review June 30, 2006)
Abstract
Transposable elements have clearly played a major role in shaping both the size and organization of eukaryotic genomes. However, the evolution of essential genes in core biological processes may also have been shaped by coevolution with these elements. This would be predicted to occur in instances where host proteins are either hijacked for use by mobile elements or recruited to defend against them. To detect such cases, we have used the Saccharomyces cerevisiae–Saccharomyces paradoxus sibling species pair to identify genes that have evolved under positive selection. We identify 72 such genes, which participate in a variety of biological processes but are enriched for genes involved in meiosis and DNA repair by nonhomologous end-joining (NHEJ). We confirm the signature of positive selection acting on NHEJ genes using orthologous sequences from all seven Saccharomyces sensu stricto species. Previous studies have found altered rates of Ty retrotransposition when these NHEJ genes are disrupted. We propose that the evolution of these repair proteins is likely to have been shaped by their interactions with Ty elements. Antagonistic pleiotropy, where critical genes like those involved in DNA repair are also subject to selective pressures imposed by mobile elements, could favor alleles that might be otherwise deleterious for their normal roles related to genome stability.
Mobile genetic elements such as transposons, retrotransposons, and viruses have had a profound effect on the organization of host genomes. Their activity has led to rapid increases in eukaryotic genome size and has contributed to the stratification of eukaryotic genomes into euchromatic and heterochromatic compartments. The presence of homologous mobile elements at multiple sites in the genome provides fodder for intragenomic recombination events, leading to rearrangements, duplications, and deletions within the genome. Finally, because many mobile elements encode transcriptional regulatory sequences, introduction of such elements into a new genomic location can potentially change the transcriptional profile of nearby host genes. Although some innovations in genes or gene expression may be selectively beneficial to the host genome, most mobile element insertions are likely to be deleterious (1).
Because mobile elements have the potential to reduce host fitness, it is useful to think of mobile elements and host genomes as being locked in long-term “genetic conflict.” Under this model, host genes are under constant selective pressure to limit the success of mobile elements, whereas mobile elements are under constant selective pressure to evade these limitations. The outcome of such a scenario is the rapid evolution of genes encoded by both the mobile element and the host. Two types of host genes are relevant to this conflict. First, many host genomes encode intracellular restriction factors, recently dubbed “intrinsic immunity” proteins, whose sole purpose is to limit mobile genetic elements such as viruses. We have characterized the rapid evolution of two such genes locked in conflict between primates and their retroviruses and retrotransposons (2, 3). A second class of host genes likely to be involved in genetic conflict encode proteins that are commandeered by mobile elements (e.g., transcriptional and translational machineries) to ensure their successful transposition. This second class is especially interesting because it implies that, in addition to the dramatic change they wreak on eukaryotic genome organization, mobile elements have also influenced the evolution of genes involved in fundamental cellular processes.
To identify cellular processes that may be subject to such conflict, we have undertaken a bioinformatic screen for yeast genes that have evolved under positive selection, one characteristic evolutionary signature of genetic conflict. Positively selected domains retain a higher proportion of DNA changes that alter the encoded amino acid (nonsynonymous) compared with the proportion of changes that leave the amino acid unchanged (synonymous) because of the presumed conflict-driven selective advantage associated with some nonsynonymous changes. At least two studies have attempted to uncover instances of positive selection in the genomes of Saccharomyces species, but these studies concluded that few or no genes had undergone positive selection (4, 5). Our screen reveals 72 yeast genes evolving under positive selection. These 72 genes do not represent a random collection, because there is a significant enrichment of genes involved in meiosis and DNA repair. Strikingly, four of the candidate genes encode proteins specifically involved in nonhomologous end-joining (NHEJ). Based on their mutant phenotypes, we propose that NHEJ genes are under positive selection to impede the integrative success of the Ty LTR-retrotransposons that populate the yeast genome. Genetic conflict involving NHEJ genes could have profound consequences for genome integrity in many organisms, making mutations that are subtly deleterious for NHEJ function nonetheless selectively favored because of their ability to combat mobile element insertions. Thus, mobile elements may influence genome evolution in ways that have been previously underappreciated.
Results
Detection of Gene Regions Subject to Positive Selection.
We undertook a bioinformatic screen for genes bearing the signatures of positive selection. Such signatures can be identified by analyzing DNA sequence for regions that have retained more nonsynonymous changes (normalized to the total possible = dN) than synonymous changes (normalized = dS). We used the completed genome sequences of the sibling species Saccharomyces paradoxus and Saccharomyces cerevisiae to do a pairwise analysis of dN and dS for protein coding genes. These species diverged ≈5 million years ago, and previous analysis of orthologous gene pairs between these two species yielded an average coding region divergence of 10% and an average dN/dS ratio of 0.11 (4). This previous study calculated dN/dS as whole-gene averages only and was able to detect few or no signatures of positive selection on this basis (see Table 2, which is published as supporting information on the PNAS web site). In contrast, we decided to analyze orthologous gene pairs in a sliding-window analysis, because one would expect such signatures to be concentrated to particular protein domains (3), a signal that may be missed in a whole-gene analysis. Indeed, genes with whole-gene dN/dS values in the range of 0.25 have previously been shown to contain robust signatures of positive selection when investigated in finer evolutionary detail (6). In a second study of selective pressures acting on the S. cerevisiae–S. paradoxus species pair (5), the authors did perform sliding-window analysis but concluded that all windows where dN/dS was >1 could be explained because dS in those windows was stochastically lower than a genomic average. To overcome this obstacle, we have chosen to assess the significance in our sliding-window analysis with simulations based on observed dS for that coding region.
We initially focused on 435 genes with whole-gene dN/dS values >0.25 in the S. cerevisiae–S. paradoxus species pair (4). Orthologous gene pairs were aligned and dN/dS was calculated for 100-bp windows, with a 50-bp slide between windows, by using the K-Estimator software package (7). For windows showing dN/dS > 1, a confidence value was generated by using simulations performed in the K-Estimator package. This program generates a null distribution of dN from replicates in which dN is simulated to be equal to the value estimated for dS (under neutral evolution dN = dS), using parameters empirically derived from the DNA region being analyzed (dS, number of codons analyzed, transition:transversion ratios, and G+C content at third positions). Thus, each window with dN/dS > 1 was tested against a null hypothesis of neutral evolution, and only windows with a 95% or greater confidence value (P < 0.05) were considered positive hits.
An example of a sliding-window analysis that uncovered a window where dN/dS is significantly greater than one in the NUP53 gene (P = 0.004) is shown in Fig. 1 A. In this 100-bp region, the two sequences contained 13 replacement changes and only 2 synonymous changes. This ratio of 13R:2S is comparable to the strongest evidence of positive selection seen in all orthologous genes between human and chimpanzee, where the ratio for the whole PRM1 protamine gene is 13R:1S (8). We find 72 such gene pairs containing statistically significant windows of positive selection (Fig. 1 B). We believe that this analysis represents a majority of the yeast genes with statistically significant (P < 0.05) domains of positive selection, because the probability of finding additional genes declines steadily with decreasing whole-gene dN/dS (Fig. 1 C).
A dN/dS sliding-window strategy identifies yeast genes evolving under positive selection. (A) As an example, a window in the NUP53 gene rejects the neutral expectation of dN/dS = 1 with high significance (P = 0.004). This 100-bp window has 13 nonsynonymous (or replacement, R) changes and 2 synonymous (S) changes. (B) The GO Slim Mapping tool (Saccharomyces Genome Database web site) was used to assign 72 yeast genes found to have at least one significant window of positive selection to major, high-level categories in the “biological process” GO. Many of the 72 genes are listed in more than one major process category, as would be expected. (C) The incidence of significant windows of positive selection is plotted versus whole-gene dN/dS values (as determined in ref. 4). Upon reanalysis, several genes had whole-gene dN/dS values lower than previously reported, probably because of additional sequence information that is now available, so the 0.10–0.19 bin represents only 33 genes.
A number of false positives are expected because of problems of multiple testing intrinsic to any sliding-window approach. An important validation is whether certain categories of genes are enriched for instances of elevated dN/dS, as we might expect if a biological process was under genetic conflict, but not expected for local, stochastic fluctuations in dS, as proposed previously (5). We therefore assessed whether the 72 candidate genes represent a random sampling of biological processes in the yeast genome. We used the Saccharomyces Genome Database GO Term Finder tool (9) to categorize these 72 genes by annotated biological processes. Because many genes have multiple gene ontology (GO) process terms assigned to them, these 72 genes grouped into 77 GO process terms, many of which are parents or daughters of other terms. A summary of some of the major GO terms is presented in Table 1, and the complete list is included in Table 3, which is published as supporting information on the PNAS web site. A binomial P value was obtained for the probability of observing the number of positively selected genes in each family, based on the percentage of genes in these categories over the whole yeast genome (9). The process list is then ranked in descending order of significance. Under this criterion, positively selected genes were significantly overrepresented in 52 of the 77 categories, suggesting a highly nonrandom distribution of candidate genes and adding validation to the sliding-window approach. Although this statistical assessment is independent from the sliding-window statistical analysis, this test also has a potential multiple testing caveat, because 77 different P-values are obtained. To attempt to account for the fact that multiple comparisons were performed, we used the Bonferroni correction (9). Despite the fact that this correction is unsuitably stringent because of the nonindependence of categories tested (because GO categories are hierarchical, and some categories are subcategories of others, 77 independent tests were not truly performed), two categories were still at or close to significance: meiosis (P = 0.032) and DNA double-strand break (DSB) repair (P = 0.089) (Table 1). Intriguingly, the four genes in the DSB repair category (XRS2, SAE2, NEJ1, and POL4) are all specifically involved in NHEJ.
GO Term Finder sorting of yeast genes under positive selection
Signatures of Positive Selection in NHEJ Genes.
NHEJ and homologous recombination are the two cellular pathways for repairing DSBs in DNA. Because DSBs presumably look the same in all organisms, it is surprising that 4 of the 10 yeast NHEJ genes show a signature of positive selection (Fig. 2 A). Sliding-window analyses of all S. cerevisiae–S. paradoxus NHEJ orthologs are shown (Fig. 2 B). Although only four genes have significant windows in this species pair (XRS2, POL4, SAE2, and NEJ1), LIF1 also has a nonsignificant peak of dN/dS > 1. These evolutionary signatures are not the result of S. cerevisiae adaptation to the laboratory (10), because dN/dS comparisons with a “wild” S. cerevisiae isolate, YJM789, yielded essentially identical results (data not shown).
NHEJ genes are subject to positive selection. (A) The process of DSB repair via NHEJ is schematized in three steps: end-binding, bridging, and DNA synthesis/ligation. Proteins that play key roles in each of these steps are indicated. Those in bold (Xrs2, Sae2, Nej1, and Pol4) were found to be positively selected in the sliding-window dN/dS analysis. (B) The sliding-window analyses of all S. cerevisiae–S. paradoxus NHEJ genes are shown. Regions where dN/dS is >1 are potential sites of positive selection. Windows that significantly reject neutrality are indicated under the gene name in base pairs, labeled with a P value and the actual number of replacement (R) and synonymous (S) DNA changes. (C) All 10 NHEJ genes from the genomes of seven Saccharomyces sensu stricto species (S. cerevisiae, S. cariocanus, S. paradoxus, S. mikatae, S. kudriavzevii, S. bayanus, and S. pastorianus) were analyzed. S. cariocanus is denoted with a dashed line in the cladogram above the table. The second through fourth columns list results from sliding-window analyses (100-bp window, 20-bp step size) on orthologous gene sets. +, presence of a significant window in cer-par comparison from B; −, no statistically significant (P < 0.05) window of dN/dS >1 was detected. All other cells indicate the presence of a significant window and indicate the dN/dS value of that 100-bp window followed by the P value in parentheses. The final two columns report results of PAML analysis, where we compared the likelihoods obtained by modeling codon evolution under neutral evolution (M7) versus positive selection (M8) using a multiple alignment of all seven sensu stricto orthologs for each gene (see Materials and Methods). By evaluating twice the difference of the log-likelihoods (2*ΔlnL) between M7 and M8, under a χ2 distribution with 2 df, P values for positive selection were obtained. ns, not significant (P > 0.05). S. cariocanus sequence was not included in the PAML analysis of YKU70 and YKU80.
To determine why NHEJ genes may be subject to positive selection, we searched the literature for the phenotypes of strains bearing disruptions of these genes. In fact, deletion or disruption of the S. cerevisiae YKU70, YKU80, RAD50, MRE11, XRS2, and SAE2 NHEJ genes have one consistent phenotype: they all affect levels of Ty1 retrotransposition (11–13). Yeast Ty retrotransposons are related to vertebrate retroviruses and share a common lifecycle (14, 15). One of the most abundant transcripts in yeast, Ty1 mRNA comprises 0.8% of the total cellular RNA, yet successful retrotransposition is rare (16, 17). This suggests that the yeast cell has active systems for limiting the propagation of Ty elements (18).
If positive selection of NHEJ genes is due to genetic conflict with Ty retrotransposons, one might expect similar selective pressures to be acting on NHEJ genes in other species representing the Saccharomyces sensu stricto group, because Ty elements predate the differentiation of this clade (19). We sequenced orthologs of all NHEJ genes from Saccharomyces cariocanus and Saccharomyces pastorianus, and gathered database sequences for the other sensu stricto species (in a few cases, Saccharomyces kudriavzevii and Saccharomyces bayanus had to be resequenced). To analyze these additional sequences, we first performed two more pairwise comparisons of Saccharomyces species, Saccharomyces mikatae–S. kudriavzevii and S. bayanus–S. pastorianus (Fig. 2 C). We find that several of the NHEJ genes show evidence of positive selection in these pairwise comparisons. These include genes found as hits in our earlier analysis (XRS2, NEJ1, and POL4) as well as genes that were not hits in the S. cerevisiae–S. paradoxus analysis (MRE11 and LIG4). All five of these genes are involved in the latter two stages of NHEJ, specifically end-bridging and DNA synthesis/ligation. When found in two independent pairwise comparisons, windows of positive selection were typically seen in the same region of the gene (data not shown). RAD50, MRE11, and XRS2 also play a role in homologous recombination, but no instances of positive selection were found in a similar analysis of the five homologous recombination-specific genes RAD51, RAD52, RAD54, RAD55, and RAD59 (data not shown).
Statistical power to detect positive selection is maximized when interspecies divergence can be compared with intraspecific polymorphism (20, 21). However, diverse S. cerevisiae and S. paradoxus isolates show very modest levels of polymorphism (22, 23). We sequenced four NHEJ genes from a panel of eight strains of S. paradoxus and found only 4–16 polymorphic sites per gene, an insufficient number to make any well supported conclusions about positive selection (data not shown). An alternative approach for testing positive selection is to sequence multiple closely related species and evaluate positive selection on a codon-by-codon basis (24). Using multiple alignments of all seven sensu stricto orthologs for each gene, we ran maximum-likelihood simulations with PAML (Fig. 2 C). These simulations analyze whether specific codons have undergone repeated rounds of positive selection over the evolution of the Saccharomyces species, and the significance of this finding is summarized with a P value. We find strong support (P < 0.05) for positively selected codons in two of the NHEJ genes (XRS2 and NEJ1). For XRS2, 1% of codons had an average dN/dS of 3.5, whereas 8% of codons in NEJ1 had an average dN/dS of 2.1 (see Fig. 4, which is published as supporting information on the PNAS web site).
Discussion
We have undertaken an evolutionary screen for genes with the signatures of positive selection in Saccharomyces species. Two biological processes, meiosis and NHEJ, show the most striking signatures. It is unclear what drives the rapid evolution of meiosis genes in budding yeast. Because yeast have a symmetric meiosis (all four meiotic products survive), they are unlikely to be subject to centromere drive, which can drive rapid evolution of meiosis genes in plants and animals (25). This rapid evolution could be due to genetic conflict with the selfish, intracellular 2-μm plasmid for transmission during meiosis (26) or a process akin to sexual selection in plants and animals. Genetic screens for Ty reproductive success have not identified many meiotic genes (12, 13, 27, 28). This reflects an inherit limitation of these screens, because most retrotransposition assays were performed in mitotically dividing cells. There is some evidence that Ty elements (and retrotransposons in general) preferentially transpose during meiosis (29), probably because there are a large number of DSBs that are created for meiotic recombination. It will therefore also be important to evaluate host factors that affect Ty retrotransposition during meiosis to fully evaluate the causes of rapidly evolving meiosis genes.
The rapid evolution of NHEJ proteins seems counterintuitive because these proteins repair DNA DSBs that should look similar in all organisms. We find that positive selection has shaped NHEJ genes across the entire Saccharomyces genus and propose that this positive selection is being driven by antagonistic interactions with Ty retrotransposons attempting to insert into the yeast genome. Several genetic screens have used gene deletion or disruption sets to identify yeast genes that regulate Ty retrotransposition (12, 13, 27, 28). Although the lists of genes identified by these screens differ, they nonetheless suggest that core biological processes such as chromatin, transcription, translation, vesicular trafficking, nuclear transport, and DNA maintenance are involved in Ty regulation (28).
We propose two models for the positive selection of NHEJ genes (schematized in Fig. 3), one in which NHEJ proteins play a defensive role against Ty elements, and one in which they are on the offensive (acting as intrinsic immunity). Although retrotransposon integration is partially mediated by the integrase (IN) protein that they encode, host DNA repair factors are presumably required for the final repair of gaps left at the ends of retrotransposon insertions (30–32). Ty1 as well as hepatitis B virus have been shown to exploit artificially created DSBs that result in integration events that are apparently mediated by NHEJ (33–35). Based on these observations, the “defensive” model proposes that Ty preintegration complexes (PICs) (consisting of cDNA and Integrase proteins) interact with host NHEJ proteins as a way of localizing to exploitable DSBs that occur naturally. These Ty elements would therefore localize preferentially to breaks where NHEJ enzymes are currently available to facilitate DNA repair after their integration. Alternately, recruiting NHEJ proteins to the PIC might protect Ty cDNA from DNA degrading activities (36). In these “defensive” scenarios, NHEJ gene variants that can avoid recognition by the Ty integration complex will have a selectable advantage, while Ty elements will be selected to increase this affinity.
Two hypothetical models for genetic conflict between Ty retrotransposons and host NHEJ proteins. Ty elements are transcribed, and this RNA is exported to the cytoplasm where it is translated. Genomic RNA also serves as a template for reverse transcription, after which a double-stranded DNA intermediate protected by integrase (IN) proteins is formed in the cytoplasm (together known as the preintegration complex, PIC). The formation of virus-like particles and their disassembly in the cytoplasm is not shown. After entry of the PIC into the nucleus, the Ty elements may (upper arrow) seek out DSBs by virtue of recognizing NHEJ proteins. In this “defensive” model, NHEJ proteins would be predicted to evolve “away” from Ty PIC recognition to maximize host fitness. Under an alternative “offensive” model (lower arrow), entering PICs encounter NHEJ proteins, which circularize PICs into 2-LTR circles that are dead-end products. In this model, the NHEJ proteins are evolving “toward” recognition of Ty PICs.
A second possibility is that NHEJ proteins actively recognize PICs and join the DNA ends to create 2-LTR circles, thereby rendering them incompetent for integration (Fig. 3). 2-LTR circles are so commonly observed during retroviral infections that they are used as a molecular marker for nuclear entry of the PIC. In fact, NHEJ has already been linked to the circularization of HIV-1 (30) and hepatitis B cDNA (37). It is easier to imagine free cDNA ends as substrates for NHEJ than the integration complex, because retrotransposon integration does not create true DSBs. Under this “offensive” model, NHEJ genes would evolve under positive selection to increase recognition of Ty elements, while Ty elements would be selected to evade this interaction.
Experimental evidence provides support for both of these models. Disruption of SAE2, MRE11, or RAD50 increases Ty1 retrotransposition levels (13, 38), lending support to the “offensive” model for NHEJ. However, disruption of KU70, KU80, or XRS2 leads to decreased levels of Ty1 retrotransposition, supporting the “defensive” model (11, 12). This might suggest that NHEJ protein subcomplexes interact differently with Ty elements, or that their depletion may have both direct and indirect effects on controlling Ty retrotransposition. For instance, impairment of NHEJ genes can lead to compromised telomeres, invoking the telomere checkpoint process (13), which may in turn control rates of Ty retrotransposition (39). In addition, compromised telomeres may themselves lead to the checkpoint-induced enhanced recruitment of NHEJ proteins to the ends of chromosomes, making them unavailable for Ty cDNAs for either defensive or offensive purposes.
If either or both of these hypotheses are true, why are only some NHEJ genes found to have the signatures of positive selection? We feel that this outcome is entirely expected for two reasons. First, Ty elements and the NHEJ complex may only need to physically interact through one or a few NHEJ proteins. Positive selection is predicted to act most significantly on proteins that make direct physical contact with the proteins of competing entities (3), while genetic experiments might reveal that all NHEJ genes have an effect of regulating transposition. Second, changes in certain NHEJ genes may be even more deleterious than the effects of retrotransposition. Thus, even when there are direct interactions with Ty elements, we might only expect a subset of those genes to allow the rapid changes in their amino acid sequence. Further biochemical and genetic experiments should clarify the exact nature of the relationship between host NHEJ machinery and Ty retrotransposons.
Alternate models for the positive selection of NHEJ genes can also be considered. For instance, yeast species living in different environments may experience unique sources of DNA damage that yield different profiles of DNA break “structures.” This scenario could certainly drive evolution but is unlikely to drive positive selection of the relevant genes, because this requires recurrent rounds of natural selection to keep step with continually changing conditions. This would require multiple successive changes in the profile of DNA break structures encountered by yeast in the wild. Furthermore, the observation that positive selection of NHEJ genes is observed across the Saccharomyces genus, whose species are often found in sympatry, argues against this model.
NHEJ is critical for genome stability in many organisms including humans. However, positive selection for mutations conveying more successful defense against retroelements may have led to the fixation of alleles that are slightly deficient for the core process of NHEJ itself. This scenario of dueling selective pressures (antagonistic pleiotropy) is especially intriguing in light of the finding that the human cancer susceptibility gene BRCA1 plays a role in DNA repair and NHEJ, yet has been subject to positive selection in recent primate evolution (40). In addition, Cernunnos, a human homolog of NEJ1, also contains signatures of positive selection (41, 42). Our computational screen and evolutionary hypothesis thus provides a good framework to study the unusual and unexpected evolution of DNA repair proteins and its functional consequences in a process fundamentally important to genome stability and evolution in all eukaryotes.
Materials and Methods
Obtaining Sequences from Yeast NHEJ Genes.
Orthologous gene sequences (4, 43) were obtained from the Saccharomyces Genome Database (www.yeastgenome.org). In addition, we sequenced YKU70, YKU80, LIF1, SAE2, RAD50, MRE11, XRS2, LIG4, NEJ1, and POL4 from S. pastorianus and S. cariocanus genomic DNA (except YKU70 and YKU80, which were not sequenced from S. cariocanus). S. kudriavzevii RAD50, MRE11, and LIG4, as well as S. bayanus LIF1 and LIG4, were also resequenced. PCR from genomic DNA was performed with PCR SuperMix High Fidelity (Invitrogen), and PCR products were sequenced directly. These sequences have been deposited in the GenBank database (accession numbers DQ889943–DQ889967), and primer sequences are available upon request. S. bayanus and S. pastorianus strains were a gift from Dan Gottschling (Fred Hutchinson Cancer Research Center), and S. kudriavzevii strain 1802 was a gift from Ed Louis (University of Leicester, Leicester, U.K.).
Evolutionary Analyses.
Of the 435 genes with the highest whole-gene dN/dS as assessed in a previous study (4), 17 genes were not analyzed because of unclear orthology, and another 23 were excluded because of the presence of stop codons in the alignment. For each orthologous gene pair, the S. cerevisiae and S. paradoxus coding region DNA sequences were aligned by using CLUSTAL_X (44). dN and dS for 100-bp sliding-window comparisons, as well as their confidence values, were calculated by using the K-Estimator software package (7). Orthologous sequence pairs for NHEJ genes of other species were analyzed similarly.
Using CLUSTAL_X (44), we created multiple alignments for each NHEJ gene from the seven sensu stricto species. Maximum-likelihood analysis was performed with codeml in the PAML 3.14.1 software package (45). To detect selection, multiple alignments were fitted to the NSsites models M7 (codon values of dN/dS fit to a beta distribution, dN/dS > 1 disallowed) and M8 (similar to model 7 but with an extra category of dN/dS > 1 allowed) assuming the f61 model of codon frequencies. Simulations were run with multiple starting values for dN/dS. Likelihood ratio tests were performed to assess whether permitting codons to evolve under positive selection gives a significantly better fit to the data.
GO Analysis.
GO Slim Mapping and Term Finder (9) analyses were performed on the Saccharomyces Genome Database web site as described there.
Acknowledgments
We thank Ben Wiggins for help with polymorphism analysis and Rachel Brem, Nels Elde, Michael Emerman, Scott Goeke, Dan Gottschling, Shari Kaiser, Julie Kerns, and Danielle Vermaak for their comments on the manuscript. This work was supported by Fred Hutchinson Cancer Research Center/University of Washington Cancer Consortium Pilot Grant 5 P30 CA015704-31 (to Leland Hartwell) and by a Searle Scholar Award (to H.S.M.). S.L.S. was supported by National Institutes of Health Training Grant 5 T32 CA09657 and National Research Service Award 1 F32 GM074299. H.S.M. is a Sloan Fellow in Computational and Evolutionary Molecular Biology.
Footnotes
- †To whom correspondence should be addressed at: 1100 Fairview Avenue North, A1-162, Seattle, WA 98109. E-mail: hsmalik{at}fhcrc.org
-
Author contributions: S.L.S. and H.S.M. designed research; S.L.S. performed research; S.L.S. and H.S.M. analyzed data; and S.L.S. and H.S.M. wrote the paper.
-
The authors declare no conflict of interest.
-
This article is a PNAS direct submission.
-
Data deposition: The sequences reported in this paper have been deposited in the GenBank database (accession nos. DQ889943–DQ889967).
- Abbreviations:
- GO,
- gene ontology;
- NHEJ,
- nonhomologous end-joining;
- PIC,
- preintegration complex;
- DSB,
- double-strand break.
- © 2006 by The National Academy of Sciences of the USA