Evidence that MutSβ repairs indels generated by mispair initiated template slippage

Abstract

This study examines the specificity of MutSβ dependent DNA mismatch repair (MMR) at the URA3 reporter gene in Saccharomyces cerevisiae. Addition and deletion mutations containing multiple base changes are observed at elevated rates in msh3Δ strains. This is consistent with a role for MutSβ in repairing multibase mismatches, such as those important in development of certain diseases and possibly including those made during Okazaki fragment maturation. Mutations containing single C-G base pair deletions are also observed at elevated rates in msh3Δ strains. These mutations are observed at still higher rates in msh3Δ strains with mutator variants of the three major replicases, implicating a defect in insertion/deletion repair during nuclear DNA replication. These deletions are observed at only a subset of detectable locations in URA3 and they occur at higher rates in one of two orientations of URA3 relative to the nearest replication origin. This subset of sites has a strong local DNA sequence context bias suggesting that the deletions are initiated by transient base•base mismatches which lead to template strand slippage, a type of transient initiator mutagenesis (TIM). We conclude that these mispair-initiated template slippage events require MMR via MutSβ before the next round of DNA synthesis to prevent deletion of an unpaired cytosine in the template strand.

Introduction

The fidelity of nuclear DNA replication in eukaryotes is enhanced by repair of mismatches made during replication that escape exonucleolytic proofreading. This DNA mismatch repair (MMR) is initiated when mismatches made by DNA polymerases (Pols) α, δ and/or ε are bound by either of two heterodimers (recently described in [1]). One is MutSα consisting of Msh2-Msh6. This heterodimer binds and contributes to the repair of base-base mismatches that cause base substitution mutations, and loops of one to several unpaired bases that cause insertion/deletion (indel) mutations. The other is MutSβ, consisting of Msh2-Msh3. This heterodimer has limited ability to repair base•base mismatches, but it efficiently repairs one or multiple unpaired bases that can also result in indel mutations [2–4]. MSH3 defects are associated with microsatellite instability (MSI) in human tumors, where MSI represents unrepaired indels in direct repeat tracts [5, 6]. Biallelic germline MSH3 mutations are linked to a hereditary MSI tumor syndrome [7, 8]. As of the date of writing, at least 411 non-silent human germline MSH3 variants have been identified as pathogenic or likely pathogenic (ClinVar database [9]). Over three quarters of those (n = 310) were annotated with either Familial adenomatous polyposis 4 (n = 235), Hereditary cancer-predisposing syndrome (n = 215), and/or Endometrial carcinoma (n = 98).

Substantial evidence (considered in [10]) supports the hypothesis that the four-subunit, exonuclease-deficient Pol α-primase/polymerase initiates nuclear DNA replication by synthesizing a short RNA primer and then extending it with about two helical turns of DNA. This priming is thought to occur during synthesis of Okazaki fragments (OFs) at replication origins and during lagging strand replication. At origins, Pol δ is suggested to rapidly extend the RNA-DNA primer until a collison-release handoff to Pol ε, which then continuously synthesizes the bulk of the leading strand [11–14]. Pol δ also extends the RNA-DNA primers during the bulk of lagging strand replication and it is thought to participate in synthesis of both DNA strands during termination of replication [14]. DNA ligase 1 completes replication by sealing the nicks between OFs to form a continuous lagging strand.

Given the complexity of replication enzymology, here we examine the role of one of the mismatch recognition complexes, MutSβ. To do this, we examine ten Saccharomyces cerevisiae strains, two with and eight without the MSH3 gene. We compare URA3 reporter gene mutation rates in wild type strains to those in the msh3Δ single-mutant strains and in three double mutant strains. The double mutants also contain mutator variants of nuclear DNA replicases: DNA Pols α, δ or ε. The mutator variants, Pol α-L868M, Pol δ-L612M, and Pol ε-M644G, respectively, have catalytic subunits encoded by pol1-L868M, pol2-M644G and pol3-L612M. The three variant replicases retain normal or near normal replicative capacity but have elevated spontaneous mutation rates and biased mutagenesis that suggests which polymerase is responsible for generating specific replication errors in vivo (reviewed in [10]). MMR efficiency is determined by comparing specific mutation rates in MMR-proficient yeast strains to rates in strains lacking MutSβ-dependent MMR due to loss of MSH3. The results reveal that certain deletion mismatches are inefficiently repaired in the absence of MutSβ-dependent MMR. We consider several ideas to explain the results, including the hypothesis that MutSβ participates in repairing indel mismatches made by mispair-induced template slippage (MITS), a form of transient initiator mutagenesis (TIM) [15].

Experimental Procedures

Strain construction.

Strains were isogenic derivatives of strain Δ|(−2)|-7B-YUNI300 (MATa CAN1 his7-2 leu2::kanMX ura3Δ trp1-289 ade2-1 lys2-ΔGG2899-2900) [1, 16]. The URA3 reporter was introduced through deletion-replacement as previously described [16]. Briefly, the URA3 reporter was introduced near the efficient origin ARS306, in either orientation 1 or orientation 2 by transformation of a PCR product containing URA3 and its endogenous promoter flanked by sequence targeting the AGP1 locus. Haploid lines with variant polymerases (pol1-L868M, pol2-M644G and pol3-L612M) were created previously [17–19]. Briefly, a plasmid containing the appropriate active site mutation was introduced via the integration-excision method [20]. Haploid strains were transformed with plasmid YEpHO [21], causing mating type switching and the formation of homozygous diploids. Diploid cells that lost the YEpHO plasmid were selected. The msh3Δ strains were constructed through deletion-replacement, as previously described [22]. Diploids cells were transformed with a PCR product containing the nourseothricin resistance cassette (NAT-R) amplified from pAG25 [23] using primers containing 50 nucleotides homologous to the intergenic regions upstream and downstream of the MSH3 open reading frame. Resulting heterozygous MSH3/msh3Δ diploids were verified by PCR and by resistance phenotype. Haploid msh3Δ samples for mutation rate and spectra collections were independently generated by sporulating MSH3/msh3Δ diploids and dissecting the resulting tetrads, followed by NAT-R selection of the resulting spore colonies.

Mutation rates and spectra.

Spontaneous mutation rates were determined via fluctuation analysis, as previously described [24]. Briefly, for each strain, spore colonies from independent tetrad dissections were inoculated into 5 ml YPDA rich liquid media with 100 μg/ml supplemental adenine. The cultures were incubated at 30 °C until saturation (approximately three days). Cultures were diluted and plated on complete (non-selective synthetic) media and on selective media containing 5-fluoroorotic acid, incubated for 5 days, and then colonies counted. Based on small pilot experiments, dilutions were adjusted to target colony counts of 50-500 per plate. Mutation rates were calculated from colony counts using Drake’s formula [25]. Mutation spectra were constructed via Sanger DNA sequencing of URA3 PCR amplicons from independently derived 5-FOA^R colonies. Mutations were called from sequencing chromatograms via DNAStar (Lasergene) with default settings and manual verification. Ambiguous chromatograms were excluded from the analysis.

Results and Discussion

Spontaneous mutation rates in the URA3 reporter gene in S. cerevisiae were measured in msh3Δ strains expressing wild type DNA polymerases or in combination with mutations in each of the three major B-family replicases, pol1-L868M for Pol α, pol2-M644G for Pol ε and pol3-L612M for Pol δ [17–19]. The URA3 reporter gene was placed in one of two orientations at the AGP1 locus, near an early-firing replication origin on chromosome III, ARS306. In Orientations 1 and 2 (OR1 and OR2) the coding strand is primarily replicated as either the nascent leading strand or nascent lagging strand, respectively. In both orientations, the overall mutation rates are lower in the msh3Δ strains (Table 1) than previously observed rates in msh2Δ [26, 27] and msh6Δ strains [28, 29], consistent with the predominant role for MutSα in MMR that stabilizes the yeast nuclear DNA genome. However, when independent 5-FOA-resistant (5-FOA^R) mutants from the msh3Δ strains were examined by sequencing the URA3 reporter gene (Figures 1, 2, 3, and 4), they primarily contained increased rates for two specific types of indel mutations: additions or deletions of two or more base pairs and a specific subset of all possible single-base G•C deletions. These indel mutations were distributed throughout the URA3 open reading frame (OR1 spectrum above the coding strand in red, OR2 spectrum below in blue). From the results, we calculated mutation rates for these two types of mutations and compared them to rates in the wild type strains and to rates in msh6Δ strains that only lack MutSα (Msh2•Msh6). The results lead us to propose two specific mechanisms for their production.

Table 1.

URA3 mutation rate and specificity in msh3Δ strains.

Strain	URA3 orientation	OR1	95% c.i.	OR2	95% c.i.
Wild type	Mutants sequenced	211		257
	Fluctuation replicates	72		82
	Mutation rate (×10−7)	0.18	(0.14-0.23)	0.21	(0.16-0.27)
	1 bp G•C deletions	0.0026	(1x)	0.0025	(1x)
	≥ 2 bp mutations	0.0034	(1x)	0.009	(1x)
msh3Δ	Mutants sequenced	168		174
	Fluctuation replicates	42		40
	Mutation rate (×10−7)	0.39	(0.1-0.14)	0.44	(0.31-0.62)
	1 bp G•C deletions	0.058	(22x) (1x)	0.17	(68x) (1x)
	≥ 2 bp indels	0.052	(15x) (1x)	0.046	(5x) (1x)
msh3Δ pol1-L868M	Mutants sequenced	208		208
	Fluctuation replicates	34		35
	Mutation rate (×10−7)	1.9	(1.3-2.6)	4	(3.1-5.0)
	1 bp G•C deletions	0.35	(6x)	1.7	(10x)
	≥ 2 bp indels	0.12	(2x)	0.34	(7x)
msh3Δ pol2-M644G	Mutants sequenced	122		125
	Fluctuation replicates	35		34
	Mutation rate (×10−7)	1.6	(1.1-2.5)	2.9	(2.2-3.9)
	1 bp G•C deletions	0.35	(6x)	1.62	(10x)
	≥ 2bp indels	0.091	(2x)	0.12	(3x)
msh3Δ pol3-L612M	Mutants sequenced	150		211
	Fluctuation replicates	32		26
	Mutation rate (×10−7)	3.5	(2.1-5.9)	16	(11-23)
	1 bp G•C deletions	1.8	(31x)	14	(85x)
	≥ 2 bp indels	0.14	(3x)	0.23	(5x)

Figure 1. URA3 mutation spectra in single mutant msh3Δ strains.

(A) A diagram of URA3 orientations relative to replisome direction and the nearest origin (ARS306). DNA strands containing the coding sequence are colored to match. (B) Spectra in msh3Δ strains with wild type DNA polymerases. The URA3 coding sequence is shown with every 10^th base indicated by a grey dot. Base substitutions (single letters), single base deletions (open triangles), short deletions (lines), and single base or short multi-base additions (closed triangles) are depicted above (OR1, red) or below (OR2, blue).

Figure 2. URA3 mutation spectra in the pol1-L868M msh3Δ strains.

As per Figure 1, but for the pol1-L868M msh3Δ strains. (A) A diagram of URA3 orientations relative to replisome direction and the nearest origin (ARS306). DNA strands containing the coding sequence are colored to match. (B) Spectra in msh3Δ strains with wild type DNA polymerases. The URA3 coding sequence is shown with every 10^th base indicated by a grey dot. Base substitutions (single letters), single base deletions (open triangles), short deletions (lines), and single base or short multi-base additions (closed triangles) are depicted above (OR1, red) or below (OR2, blue).

Figure 3. URA3 mutation spectra in the pol2-M644G msh3Δ strains.

As per Figure 1, but for the pol2-M644G msh3Δ strains. (A) A diagram of URA3 orientations relative to replisome direction and the nearest origin (ARS306). DNA strands containing the coding sequence are colored to match. (B) Spectra in msh3Δ strains with wild type DNA polymerases. The URA3 coding sequence is shown with every 10^th base indicated by a grey dot. Base substitutions (single letters), single base deletions (open triangles), short deletions (lines), and single base or short multi-base additions (closed triangles) are depicted above (OR1, red) or below (OR2, blue).

Figure 4. URA3 mutation spectra in the pol3-L612M msh3Δ strains.

As per Figure 1, but for the pol3-L612M msh3Δ strains. (A) A diagram of URA3 orientations relative to replisome direction and the nearest origin (ARS306). DNA strands containing the coding sequence are colored to match. (B) Spectra in msh3Δ strains with wild type DNA polymerases. The URA3 coding sequence is shown with every 10^th base indicated by a grey dot. Base substitutions (single letters), single base deletions (open triangles), short deletions (lines), and single base or short multi-base additions (closed triangles) are depicted above (OR1, red) or below (OR2, blue).

MutSβ is involved in repairing mismatches containing two or more bases.

Compared to published results for msh6Δ strains [26, 27], the spectra of mutations in the msh3Δ strains (Figures 1–4) reveal two major differences. One is the high proportion of multi-base insertions and deletions in the msh3Δ strains, which were rare in msh6Δ strains (10 of 3,824 total mutations in 16 spectra [26–29]. In contrast, in the msh3Δ OR1 and OR2 strains, multi-base indel rates were 15-fold and 5-fold higher, respectively, than the rates in the corresponding wild-type yeast strains (Table 1, ratios shown in black parentheticals). The enzymological origins of such mismatches could theoretically be any of several DNA transactions that occur in a yeast cell. However, the rates in double mutant strains are 2-7-fold higher than in the msh3Δ single mutant strains (Table 1, ratios shown in red parentheticals). These increases are consistent with the hypothesis that MutSβ-dependent MMR corrects replication errors consisting of multiple unpaired bases in either the template or the primer strand, thereby preventing deletions or additions, respectively. One potential source of these muti-base indels is Okazaki fragment maturation (OFM), a possibility that can be examined in future studies.

Included among the multi-base indels observed in the msh3Δ strains are indels located at four of the eleven triplet repeat sequences present in the URA3 reporter gene. This implies either that indels do not occur at the other triplet repeat sites, or that non-frameshifting indels are tolerated by orotidine-5′-phosphate decarboxylase, thereby avoiding a 5-FOA^R phenotype. Two triplet repeats stand out, with multiple observed indels at the (ATT)₃ repeat at nucleotide positions 688 through 696 and at the (TTG)₂ repeat at positions 695-700. In the msh3Δ strain (Fig. 1), one unit of the (ATT)₃ sequence was deleted in six and eight independent 5-FOA^R mutants in OR1 and OR2, respectively. Moreover, one new copy of the ATT sequence was independently added twice for each orientation. Similarly at the adjacent (TTG)₂, one copy of the TTG unit was deleted and three were added in different 5-FOA^R mutants. Thus, the collective mutation rates of triplet-repeat indels at these two sites are 2.3 and 3.0 ×10⁻⁹ in OR1 and OR2, respectively. These rates are ≥27-fold and 19-fold higher than those in our previous study of mutagenesis in wild type yeast (Table 1). These increases strongly imply that MutSβ is normally involved in the repair of triplet repeat indels, as previously suggested [2–4]. Moreover, because yeast Msh3 is conserved in humans, this interpretation further suggests the possibility that a reduction in Msh3 function could have a role in the origins of several human diseases associated with expansion of triplet repeat sequences. Indeed, trinucleotide repeat expansions underlie at least twenty severe neuromuscular and neurodegenerative diseases [30, 31].

Are these indels due to replication errors that are not repaired, or is another mechanism responsible for generating the misaligned substrates? We cannot say for certain, but trinucleotide indel rates are higher in both URA3 orientations and in strains harboring mutator variants of the three replicases (Table 1; 1.7 to 5.1-fold). Because Pol δ is involved in many DNA transactions, the responsible intermediates could be made during repair, recombination or some other more specialized DNA synthesis reaction. However, increased rates with the other two more specialized mutator replicases suggest that the trinucleotide indels are formed in the absence of Msh3 during the nuclear activity common to all three, DNA replication.

MutSβ preferentially repairs of a subset of G•C base pair deletions.

The second and predominant class of mutations observed in the msh3Δ spectra are single-base deletions that were not previously observed in the msh6Δ strain [29]. These are deletions of single G•C base pairs at specific locations (Figure 1–4). Overall, and usually in both orientations of URA3, the rates for these deletions are higher in the single-mutant msh3Δ strain than in a wild-type strain (Table 1). They are higher still in the double mutant strains harboring mutations in Pols α, δ and ε (Table 1). The latter observation implies that these rate increases reflect replication errors made during nuclear DNA replication by any of the three replicases.

G•C deletions were observed at many sites but concentrated in a few. There are 346 guanines and cytosines in URA3 where single-base deletions would shift the reading frame and presumably cause a 5-FOA^R phenotype. Among them, 204 are non-iterated G•C base pairs, 88 are in 44 runs of two base pairs, and 54 are in 18 runs of three, for a total of 310 locations. G•C deletions in the msh3Δ strains are concentrated in only 40 of these locations, with over 95% located at 13 sites and over 83% located at just five runs of three bp (positions 321, 604, 612, 639 and 767). This non-uniform distribution could reflect locations where Msh3 contributes to repair of mismatches made during OFM. Alternatively, or in addition, they could occur at locations where DNA damage arises, and/or they could depend on flanking DNA sequence contexts. Because 93% of the deletions were in runs of two or three base pairs, DNA strand slippage during replication may be involved in creating the mismatches repaired by MutSβ. The most prominent G•C deletion at a non-iterated base pair was at position 115 in the msh3Δ pol1-L868M mutant (Fig. 2).

A possible mechanism to explain deletions of the G•C base pairs in the msh3Δ strains.

The unique specificity of the G•C deletions observed here in msh3Δ strains is consistent with mutations resulting from a base•base mismatch made during replication. The model for this mechanism differs from our previous studies that described the rates and genomic locations at which each of the three major yeast nuclear replicases commit several types of replication errors, e.g. base•base mismatches [32–34] and the incorporation of ribonucleotides [10, 12, 14, 35]. In those earlier studies, differences between Pols α and δ as compared to Pol ε supported the hypothesis that Pol ε preferentially replicates the leading DNA strand while Pols α and δ preferentially replicate the lagging DNA strand (reviewed in [36]). Those observations are distinct for each of the three replicases, and importantly, they differ from what is observed in the msh3Δ strains. Here, deletions of G•C base pairs are preferentially generated at higher rates in one particular orientation, and a similar bias is generally observed for all three variant replicases (Figs. 1–4 and 5d). Moreover, deletions are concentrated in G•C runs and within a particular sequence context. Bases immediately 5′ to G runs containing deletions (Fig. 5b–c) are As far less frequently than are bases 5’ to all G runs (Fig. 5a). Positions 3′ to deletion-bearing runs are enriched in Ts (compare to Fig. 5e–f to 5g). Likewise, positions immediately 5′ to C runs with deletions are enriched in As (compare Fig. 5i–j to 5h) and bases 3′ are almost always Gs (compare Fig. 5. Panels L-M to panel N). In the context of the URA3 coding strand, the consensus sequences for G and C deletion targets (Figure 5) are thus 5′-YG_nT-3’ and 5′-AC_nG-3’, respectively, where Y represents pyrimidines, C or T. Accounting for less diversity in C run flanks due to lower counts (G₂ = 45 vs. C₂ = 17; G₃ = 14 vs. C₃ = 4), these two consensus sequences are complementary (Figure 5).

Figure 5. Rates and sequence contexts of G•C deletions in msh3Δ strains.

“Runs” are defined as strings of identical consecutive DNA bases, with non-iterated bases considered runs of one base. URA3 was inserted into S. cerevisiae chromosome 3 at the AGP1 locus, near efficient replication origin ARS306, in two orientations. In orientation 1 (OR1), the coding sequence shown here is replicated by DNA Pol ε as the nascent leading strand and the reverse complement is replicated by Pols α and δ as the nascent lagging strand in > 90% of replication events [10, 35]. In orientation 2 (OR2), the coding and noncoding strands are the nascent lagging and leading strands, respectively. All base identities are drawn from the URA3 coding sequence. Data for A, T, G, and C bases are color coded amber, teal, green, and crimson, respectively. (A) The fraction of each possible base immediately 5′ to G runs in URA3. The definition of G runs precludes Gs in flanking positions. (B) 5′ flanking base fractions for G runs containing deletions in each msh3Δ strain and each URA3 orientation. Each deletion is independently considered, so sites with higher rates will have more influence. (C) The mean fractions in panel B, averaged by URA3 orientation. All data sets are given the same weight, regardless of observed mutation count or calculated mutation rate. (D) Rates of 5-FOA resistance acquisition via G deletions in URA3. (E-G) As per panels C, B, and A, respectively, but for the base immediately 3’ to G runs. (H-N) As per panels A-G, but for C runs.

From this sequence specificity, we suggest that many of the G•C base pair deletions in the msh3Δ strains may be explained by a mechanism common to all three replicases that involves transient intermediate errors. Transient misalignment during DNA synthesis was proposed many years ago [37] and used to explain mutational hotspots generated during DNA synthesis in vitro by several DNA polymerases [38–42], including one of the three replicases studied here, yeast Pol α [43]. We suggested four different mechanisms, collectively dubbed transient initiator mutagenesis (TIM), to explain many mutations arising during nuclear DNA replication in S. cerevisiae catalyzed by Pol δ and Pol ε [15]. One such mechanism, Mismatch-Initiated Template Slippage (MITS), involves incorporation of a base•base mismatch at the last position in a mononucleotide run. We posited that after the base•base mismatch is made but before continued DNA synthesis, this mismatch rearranges to form an indel mismatch followed by correct base pairs at the 3′-OH terminus to allow further correct incorporation (Fig. 6).

Figure 6. Mispair-Initiated Template Slippage model for deleting G•C base pairs.

The proposed MITS [15] mechanism for single-base G and C deletions in msh3Δ strains using two examples of deletion-bearing loci. Sequence pairs represent complementary DNA strands with the primer strand above and the template below. Replication proceeds from left to right by extension of the primer strand. Bases identified in the consensus sequences in Figure 5 use the same color scheme: A = amber; T = teal; G = green; and C = crimson. Other bases (complementary strand, bases beyond immediate flanks) are colored black. Mispaired bases are surrounded by grey boxes.

As an example, consider the 11-fold preference in the rate of G•C deletions at positions 604-606 in OR1 observed the msh3Δ pol3-L612M strain (Fig. 4). In this orientation, a template CCC run is flanked by an adenine that is present at the next template position for correct incorporation. Here Pol δ is thought to synthesize the G-containing mononucleotide run using the CCC run on the template strand (Fig. 6A). In this circumstance, we hypothesize that Pol δ mistakenly inserts a thymine opposite the final template cytosine in the run (nucleotide 604) to create a C•dTMP mismatch (top of Fig. 6A). Because mismatches are generally more difficult to extend than correct base pairs, this mismatch is likely to create a barrier to further elongation. Before continued synthesis from a mismatch, strand slippage forms a template-primer containing an extrahelical C in the template strand that is followed by two correct C•G base pairs. The mis-inserted T now correctly pairs to the A at the next position in the template strand. This rearrangement provides three correct base pairs to be present between the unpaired cytosine and the next correct incorporation, favoring formation of the indel mismatch. Thus, in absence of MutSβ-initiated MMR of the template strand containing the unpaired cytosine, the new strand is the template for a G•C deletion in the next round of replication. The same model can also explain the hotspot for G•C deletions at position 321-323, a run of three G residues flanked by a 3′-template T (Figure 6B). The G•C deletions observed at this location are preferentially observed in OR2, so MITS can explain the G•C deletions in the same fashion as those observed at positions 604-606. We further speculate that many of the other G•C deletions observed in the msh3Δ strains can also be explained by this model, using different combinations of an initiating base-base mismatch and the local sequence, including both the length of a run and the identity of the flanking base pair.

Previously observed strand biases in mutation rates led to the hypothesis that Pol ε is the major leading strand replicase while Pols α and δ are the predominant lagging strand replicases. Here, however, the MITS model suggests that, in the absence of Msh3, the bulk of G•C deletions occur when the G run is in the nascent lagging strand. Given higher rates in the pol2M644G msh3Δ strain, this appears to hold true even when the responsible polymerase is thought to work primarily on the leading strand. Nonetheless, the extent of the bias does differ among the three replicases, as do the locations for loss of a G•C base pair. The latter is best exemplified by the hotspot for loss of a non-iterated G•C base pair in the msh3Δ pol1-L868M strain (Figure 2). Although we do not yet understand these differences, several possible explanations can be considered. The differences could be related to the nature of the initiating base•base mismatch. Consistent with the MITS model in Figure 6, our previous work demonstrated that misinsertion of dTMP opposite template guanine is one of three mismatches preferentially made at elevated rates by all three variant replicases across the yeast genome [33]. Moreover, a second study showed that among the four deoxynucleotide triphosphates present in the S. cerevisiae strain used in our studies, dTTP is normally present at the highest concentration [26]. Each of these two parameters may be relevant to the patterns of mutagenesis observed here. It is also possible that the relative strengths of local base stacking interactions are important, because pyrimidine bases are known to be more easily unstacked than purine bases. We plan to test these ideas in the future by examining the pattern of G•C deletion mutations observed in a URA3 reporter gene whose primary DNA sequence has been modified. We have also begun studies to examine mutation specificity in msh3Δ and msh6Δ strains using whole genome sequencing. Because the yeast genome consists of 6,000,000 base pairs rather than the 804 base pairs in URA3, the whole genome approach may provide insights into other roles for MutSβ and MutSα in specifically determining the four different transient initiator mutagenesis (TIM) models [15] and more generally, the stability of the whole eukaryotic nuclear genome. Such studies could be informative regarding associations between mutations in Msh3 and/or Msh6 with several human diseases (see Introduction for references).