MutLα suppresses error-prone DNA mismatch repair and preferentially protects noncoding DNA from mutations

The DNA mismatch repair (MMR) system promotes genome stability and protects humans from certain types of cancer. Its primary function is the correction of DNA polymerase errors. MutLα is an important eukaryotic MMR factor. We have examined the contributions of MutLα to maintaining genome stability. We show here that loss of MutLα in yeast increases the genome-wide mutation rate by ∼130-fold and generates a genome-wide mutation spectrum that consists of small indels and base substitutions. We also show that loss of yeast MutLα leads to error-prone MMR that produces T > C base substitutions in 5′-ATA-3′ sequences. In agreement with this finding, our examination of human whole-genome DNA sequencing data has revealed that loss of MutLα in induced pluripotent stem cells triggers error-prone MMR that leads to the formation of T > C mutations in 5′-NTN-3′ sequences. Our further analysis has shown that MutLα-independent MMR plays a role in suppressing base substitutions in N3 homopolymeric runs. In addition, we describe that MutLα preferentially protects noncoding DNA from mutations. Our study defines the contributions of MutLα-dependent and independent mechanisms to genome-wide MMR.

DNA damage and replication errors constantly challenge the integrity of genetic information.To deal with DNA damage and replication errors, the cell evolved DNA repair.The key DNA repair pathways that have been identified are direct repair by photoreactivation, homology-directed double-strand break repair, base excision repair, mismatch repair, nonhomologous end joining, nucleotide excision repair, repair of DNA interstrand crosslinks, and ribonucleotide excision repair (1)(2)(3)(4)(5)(6)(7).If left unrepaired, DNA damage and replication errors cause mutations.Mutations are the molecular basis of many diseases including cancer.In addition to being detrimental, mutations are sometimes beneficial for the organism.For example, mutations that are produced in B cells during somatic hypermutation of immunoglobulin genes lead to antibody diversification, an essential step in the immune response (8).
The DNA mismatch repair (MMR) system is a major DNA repair system that has been conserved from bacteria to humans (1,9).It promotes genetic stability by strand-specific removal of base-base mismatches and small insertion/deletion loops that are formed during DNA replication and strand exchange in homologous recombination (10,11).The MMR system also stabilizes genetic information by suppressing homeologous recombination and initiating apoptosis in response to irreparable DNA damage caused by several anticancer drugs (12,13).Although MMR mostly occurs in an error-free manner, it is known to be error-prone at certain genomic loci (14).The error-prone MMR causes triplet repeat expansions and the formation of base substitution mutations at A:T base pairs of immunoglobulin genes (15)(16)(17).
Studies in the bacterium Escherichia coli led to the discovery and elucidation of the strand-specific MMR that is directed to the daughter strand by transient absence of DNA methylation at GATC sites (18)(19)(20)(21).The MMR reaction in E. coli is initiated by recognition of a mismatch by MutS.After recognizing a mismatch, MutS recruits MutL to the heteroduplex DNA in an ATP-dependent reaction.The ternary MutS-MutL-heteroduplex complex activates MutH endonuclease to incise the daughter strand at a hemimethylated GATC site.A nick produced by MutH serves as a loading site for helicase II that unwinds a stretch of the mismatch-containing daughter strand in a MutS-and MutL-dependent reaction (22).The unwound strand is hydrolyzed by an ssDNA exonuclease (ExoI, ExoVII, ExoX, or RecJ).After the excision step, the generated DNA gap is filled in by the action of the DNA polymerase (Pol III) holoenzyme and the nick is sealed by DNA ligase.
Unlike MMR in E. coli and its closely related bacteria, MMR in many other bacteria involves MutL proteins that function as endonucleases (23)(24)(25)(26).Studies of the mechanisms of bacterial MutL endonuclease-dependent MMR revealed that some of them rely on MutL endonuclease-b clamp interactions and others do not (24,27).
The finding that inactivation of the MMR system by deletion of the MSH2 gene increases the spontaneous genomewide mutation rate in S. cerevisiae by 100-fold emphasizes the critical role of this DNA repair system in maintaining whole-genome stability (37).Eukaryotic MMR proceeds via a series of coordinated events that include mismatch recognition, incision of the daughter strand near the mismatch, and mismatch removal (9,26,38).Biochemical studies have resulted in the reconstitution of MutLa-dependent and independent MMR pathways from purified components (39)(40)(41).A critical step in MutLa-dependent MMR pathways is a mismatch-, MutSa-, proliferating cell nuclear antigen-, replication factor C-, and ATP-dependent incision of the discontinuous strand by the endonuclease activity of MutLa (23).A MutLa-generated strand break located 5 0 to the mismatch is utilized by MutSa-activated EXO1 to excise the mismatch in a 5'→3 0 hydrolytic reaction modulated by replication protein A (23,42).The generated gap is filled in by the DNA Pol d or Pol ε holoenzyme (39,43,44).When EXO1 is not available, a 5 0 strand break created by the activated MutLa endonuclease is utilized by the DNA Pol d holoenzyme to remove the mismatch in a strand-displacement DNA synthesis reaction (40,44) enhanced by the nuclease activity of DNA2 (41).
In this study, we investigated MutLa-dependent and independent mechanisms of MMR at a whole-genome level in S. cerevisiae.We show that MutLa plays a major role in genomewide MMR in yeast, preferentially protects noncoding DNA from mutations, and suppresses error-prone MMR in yeast and human cells.We also show that a yeast MutLa-independent mechanism contributes to genome-wide MMR.

MutLa plays a major role in yeast genome-wide MMR
Yeast MutLa (Mlh1-Pms1 heterodimer) is required for MMR events that take place in the CAN1, his7-2, and lys2 mutation reporters (45).However, the involvement of MutLa in MMR events that occur across the whole yeast genome has not been defined.Furthermore, it has not been understood whether a MutLa-independent MMR mechanism promotes genetic stability at a whole-genome level.We initiated this study to examine the contributions of MutLa-dependent and independent mechanisms to genome-wide MMR in S. cerevisiae.We first passaged multiple isolates of diploid pms1D, msh2D, and WT strains for 900 generations/isolate to allow them to accumulate de novo mutations.We then performed whole-genome sequencing and identified de novo mutations in each of the passaged isolates by removing variants that were present in the isolate at generation 0. The identified de novo mutations that are henceforth referred to as mutations were pooled according to the genotype to generate the mutation spectra (Table S1).
To determine the contribution of MutLa to genome-wide MMR, we calculated spontaneous genome-wide mutation rates in the WT, pms1D, and msh2D strains (Fig. 1  Mutation rate (x 10 ) Figure 1.MutLa is required for the majority of MMR events in Saccharomyces cerevisiae.Total rates of genome-wide spontaneous mutations (A) and rates of genome-wide spontaneous deletions (B), base substitutions (C), and insertions (D) in the pms1D, msh2D, and WT strains.The mutation rates were calculated as described in Experimental procedures and are presented as means ± SD.The data in A to D were obtained using independent biological replicates (n wt = 15, n pms1D = 9, and n msh2D = 8).The numbers above the bars are relative mutation rates.The p values were calculated using the Mann-Whitney U two-tailed test (GraphPad Prism 6 software, https://www.graphpad.com/features).Deletions were not observed in the WT strain (B).MMR, mismatch repair.
Table S2).As shown in Figure 1A, the genome-wide mutation rate in the pms1D strain was 130 times higher than that in the WT strain.Importantly, the genome-wide mutation rates in the pms1D and msh2D strains did not significantly differ from each other (Fig. 1A).Further calculations showed that the rates of deletions in the pms1D and msh2D strains were indistinguishable from each other (Fig. 1B).Likewise, no significant difference between the rates of base substitutions, or between the rates of insertions, in the two mutant strains was detected.Together, these data indicate that yeast MutLa is required for most of genome-wide MMR events.
Trinucleotide signatures of base substitutions in the pms1D and msh2D strains Analyses of mutational signatures reveal essential information about the nature of mutational processes and DNA repair mechanisms (37,46,47).We extracted trinucleotide signatures of base substitution mutations from the pms1D and msh2D strains (Fig. 2).The majority of mutations in both signatures are C > A, C > T, and T > C base substitutions.Each of the mutational signatures has three large peaks.Two of the peaks are for C > T base substitutions at the 5 0 -GCA-3 0 and 5 0 -ACA-3 0 sequences, and the third peak is for C > A base substitutions at the 5 0 -CCT-3 0 sequences.The C > T base substitutions at 5 0 -GCA-3 0 and 5 0 -ACA-3 0 sequences and C > A base substitutions at 5 0 -CCT-3 0 sequences account for 23% and 27% of all base substitutions in the pms1D and msh2D strains, respectively.Overall, the two mutational signatures display significant similarities to each other.The observed similarities between the two mutational signatures support the view that MutLa plays a key role in yeast genome-wide MMR.

MutLa suppresses error-prone MMR in yeast and human cells
It has been unknown whether an MMR factor suppresses an error-prone MMR mechanism that acts at a specific locus or throughout a genome.Our analysis of the trinucleotide base substitution signatures of the pms1D and msh2D strains showed that there was a significant difference between the two signatures (Fig. 2).Specifically, the base substitution signature of the pms1D strain contains a peak for T > C substitutions at 5 0 -ATA-3 0 sequences which is significantly larger than the corresponding peak present in the mutational signature of the msh2D strain (p = 0.0147) (Figs. 2 and 3).This finding suggests that yeast MutLa suppresses a genome-wide error-prone MMR mechanism that generates T > C base substitutions at 5 0 -ATA-3 0 sequences.
To study whether MutLa suppresses error-prone MMR in higher eukaryotes, we analyzed recently published mutation data that were obtained by whole-genome sequencing of human DMSH2 and DPMS2 induced pluripotent stem cells (iPSCs) (48).The results of our analysis showed that the rate of T > C base substitutions is significantly higher in DPMS2 iPSCs that lack MutLa (MLH1-PMS2 heterodimer) than DMSH2 iPSCs (Fig. 3C).In contrast, the total rate of base substitutions in DPMS2 iPSCs is significantly lower than the total rate of base substitutions in DMSH2 iPSCs (48).We also  observed that one of the peaks of T > C substitutions in the trinucleotide mutational signature of DPMS2 iPSCs is at 5 0 -ATA-3 0 sequences (48) and that the rate of T > C substitutions at 5 0 -ATA-3 0 sequences in the DPMS2 iPSCs is higher than the rate of T > C substitutions in the identical sequences in the DMSH2 iPSCs (Fig. 3D).Collectively, these data demonstrate that MutLa suppresses error-prone MMR in yeast cells and human iPSCs on the genome-wide level.

MutLa preferentially protects non-coding DNA from mutations
Previous research revealed that microsatellite instability, a hallmark of MMR deficiency, in colorectal and endometrial cancer genomes preferentially occurs in noncoding DNA (49).We next analyzed how mutations were distributed between coding and noncoding DNAs in the WT and pms1D strains.In agreement with a previous study (47), we observed that mutations in the WT strain did not show a preference for accumulation in coding or noncoding DNA (Fig. 4A).In contrast, mutations in the pms1D strain preferentially accumulated in noncoding DNA (Fig. 4A).In fact, there was an 8-fold preference for accumulation of mutations in noncoding DNA in the pms1D strain (Fig. 4B).Further analysis showed that in the MutLa-lacking strain the preferences for accumulation of deletions and insertions in noncoding DNA were 10-fold and 8-fold, respectively (Fig. 4, C and D).These findings show that MutLa preferentially protects noncoding DNA from mutations.
A previous study revealed that Msh2 is necessary for preferential protection of noncoding DNA from deletions (37).In line with that study, we determined that Msh2 also preferentially protected noncoding DNA from insertions (Fig. 4D).

MutLa-independent MMR at a whole-genome level
Biochemical studies with purified proteins revealed a MutLa-independent mechanism that corrects mismatches in vitro (39,42).However, it has remained unclear whether MutLa-independent MMR functions in vivo.To study whether MutLa-independent MMR acts at a whole-genome level, we took advantage of WebLogo (50) to visualize patterns in which T > C, C > T, and C > A mutations accumulated in the pms1D and msh2D strains.We determined that the sequence pattern in which T > C substitutions accumulated in the pms1D strain was different from that in the msh2D strain (Fig. 5A).In contrast, the sequence patterns in which C > T or C > A substitutions amassed in the pms1D and msh2D strains were similar (Fig. 5, B and C).A subsequent analysis showed that the most frequent nucleotide that was immediately downstream from A 11 runs in which 1-bp deletions accumulated was a G in pms1D cells and a T in msh2D cells (Fig. 6).These analyses indicate that genome-wide MutLa-independent MMR contributes to the suppression of T > C base substitutions and 1-bp deletions in A 11 runs.
Base substitutions in the pms1D and msh2D strains frequently occurred in homopolymeric runs (Fig. 7A).We next analyzed the involvement of MutLa-independent MMR in the suppression of base substitutions in homopolymeric runs.C > T mutations are the most common base substitutions in the mutation spectra of the pms1D and msh2D strains (Table S1 and Fig. 2).Our initial analysis showed that C > T base substitutions in homopolymeric N 3 runs accumulated at a higher rate in the msh2D than pms1D strain (Fig. 7B).A following analysis revealed that the total rate of base substitutions in N 3 runs was higher in the msh2D strain than the pms1D strain (Fig. 7C).Therefore, MutLa-independent MMR is more important for the suppression of base substitutions in homopolymeric N 3 runs than MutLa-dependent MMR. and eukaryotes lack MutH endonuclease and its homologs.In these organisms, MutL homologs act as endonucleases in MMR (23-26, 45, 51, 52).MutLa is the founding member of the MutL endonuclease family (23,45).MutLa functions downstream from the mismatch recognition step that is accomplished by MutSa or MutSb (23,53).In this study, we investigated the involvement of MutLa in genome-wide MMR in S. cerevisiae.We have found that loss of MutLa caused by deletion of the PMS1 gene elevates the genome-wide mutation rate to a level that is present in the msh2D strain (Fig. 1A).Because MMR does not function in the absence of MSH2, the finding that the genome-wide mutation rates in the msh2D and pms1D strains do not significantly differ from each other (Fig. 1) demonstrates that MutLa is required for most of MMR events occurring across the yeast nuclear genome.Unlike MutLa of S. cerevisiae, MutLa of human iPSCs is only required for 55% MMR events (48).

Discussion
Mutational signatures have been instrumental in understanding mutational processes operating in human cancers as well as replication fidelity and DNA repair mechanisms (46,47,54).We extracted and examined the trinucleotide signatures of base substitution mutations from the pms1D and msh2D strains (Fig. 2).As described below, these two mutational signatures display significant similarities to SBS44, a mutational signature of human MMR deficiency that was extracted from cancers (54).First, the majority of base substitutions in all three mutational signatures are C > T, T > C, and C > A alterations.Second, the three most prominent peaks that are present in the mutational signatures of PMS1 and MSH2 deficiency (Fig. 2) are also present in SBS44.These similarities provide strong evidence that DNA Pols in yeast and human cells produce similar base-base mismatches in the same sequence contexts.
Error-prone MMR occurs in variable regions of immunoglobulin genes and certain triplet repeat loci (8,14,15).Because MMR includes the step of DNA resynthesis it is not a surprise that MMR can be error-prone.MutLa has been best known for providing the endonuclease function for error-free MMR.We found evidence that the loss of yeast MutLa triggers error-prone MMR that results in the formation of T > C mutations at 5 0 -ATA-3 0 sequences (Fig. 3, A and B).In line with this evidence, our analysis of the human whole-genome mutational data has demonstrated that human MutLa deficiency in the iPSCs also leads to error-prone MMR that triggers T > C mutations at 5 0 -ATA-3 0 sequences (Fig. 3D).In addition, we have determined that error-prone MMR caused by the loss of human MutLa in the iPSCs produces T > C mutations at ten other trinucleotide sequences (Table S3).Thus, not only does MutLa play a major role in the error-free correction of DNA Pol errors but it also suppresses errorprone MMR.The finding that MutLa deficiency results in error-prone MMR in yeast and human cells improves our understanding of the contribution of MutLa to the correction of DNA replication errors.The loss of yeast MutLa triggers error-prone MMR that results in T > C substitutions at 5 0 -ATA-3 0 sequences, whereas the loss of human MutLa in the iPSCs causes error-prone MMR that gives rise to T > C substitutions at 5 0 -ATA-3 0 (Fig. 3D) and ten other trinucleotide sequences (Table S3).One interpretation of this observation is that one error-prone MMR mechanism functions in yeast pms1D cells and two or more error-prone MMR mechanisms operate in human DPMS2 iPSCs.Another interpretation of this observation is that the properties and/or nuclear concentrations of the involved error-prone DNA Pol(s) are different in the yeast and human cells.These two interpretations are not mutually exclusive.The mechanism(s) of error-prone MMR occurring in response to the absence of MutLa is not known.Because a low-fidelity DNA Pol, Pol h, and EXO1 are involved in errorprone MMR in somatic hypermutation of immunoglobulin genes (8,15), it is possible that these two enzymes also contribute to error-prone MMR that is triggered by the loss of MutLa.It would be important to define this errorprone mechanism(s) using genetic and biochemical approaches.
Noncoding DNA contains gene promoters and other critical regulatory elements.In S. cerevisiae, many promoters include homopolymeric dA:dT sequences that are required for normal levels of transcription (55,56).Importantly, the reduction of the size of promoter homopolymeric dA:dT sequences significantly decreases the level of transcription (55).In agreement with this finding, transcription of yeast HIS3 was stimulated 3-fold after incorporation of a 17-bp poly(dA:dT) sequence in its promoter region (55).Therefore, it is beneficial for the organism to maintain the stability of noncoding DNA.
A previous study revealed that Msh2 is required for preferential protection of noncoding DNA from deletions (37).We show here that MutLa preferentially protects noncoding DNA from deletions and insertions but not from base substitutions (Fig. 4, F-H).Our data analysis indicates that loss of MutLa causes a 10-fold bias for the accumulation of deletions and insertions in noncoding relative to coding DNA (Fig. 4, C and D).Deletion of MSH2 leads to the formation of the same biases (Fig. 4, C and D).Of importance is the observation that in the pms1D and msh2D strains the mutation rates in noncoding DNA are 7 times higher than the mutation rates in coding DNA (Fig. 4E), despite the fact that the size of yeast noncoding DNA is 3 times smaller than that of yeast coding DNA (57).The increased rates of mutations in noncoding DNA of the pms1D and msh2D strains are due to the elevated rates of deletions and insertions, but not base substitutions (Fig. 4, F-H).The following observations that were obtained in previous studies explain the preferential protection of noncoding DNA from indels by MMR.First, relative to coding DNA, noncoding DNA has a significantly larger number of longer homopolymeric sequences (58), which, when clustered, create mutation hot spots (59).Second, longer homopolymeric runs are much more unstable than the shorter ones (60).Third, MMR is much more efficient in correcting insertion/deletion loops in longer than shorter homopolymeric runs (60).The more efficient removal of indel loops at longer poly(dA:dT) runs by the MMR system is probably a result of reduced nucleosome density at these homopolymeric runs (61).Biochemical studies reconstituted a human MutLa-independent MMR pathway that corrects mismatches on 5 0 -nicked DNA (39,42).This MutLa-independent pathway relies on the mismatch recognition factor MutSa and the 5'→3 0 exonuclease activity of EXO1 to excise mismatches in vitro.However, the contribution of this pathway to MMR in eukaryotic organisms in vivo had remained unclear.We investigated whether yeast MutLa-independent MMR occurred throughout the nuclear genome (Figs.5-7).We observed that the sequence pattern for T > C mutations in the pms1D cells is different from the sequence pattern for the same mutations in the msh2D cells (Fig. 5A).Likewise, the sequence pattern for 1-bp deletions in homopolymeric A 11 runs of pms1D cells differs from the sequence pattern for 1-bp deletions in homopolymeric A 11 runs of msh2D cells (Fig. 6).Furthermore, we determined that in N 3 homopolymeric runs the rates for C > T and other base substitutions in the msh2D cells is significantly higher than the rates for the same genetic alterations in the pms1D cells.Collectively, these data indicate that MutLa-independent MMR contributes to the maintenance of whole-genome stability in S. cerevisiae.Assessment of the mutation rates in human DMSH2 and DPMS2 iPSCs supports the view that a MutLa-independent process plays a role in genome-wide MMR (48).
In summary, we show here that MutLa plays a major role in genome-wide MMR, suppresses error-prone MMR, and preferentially protects noncoding DNA from mutations.We also show that MutLa-independent MMR functions in vivo.

Experimental procedures S. cerevisiae strains
Yeast WT haploid strains that were used in this study are isogenic BY4741 (MATa his3D1 leu2D0 met15D0 ura3D0) and BY4742 (MATa his3D1 leu2D0 lys2D0 ura3D0), both of which are derivatives of S288C.PMS1 and MSH2 gene deletions in the haploid WT strains were generated by lithium/ PEG-based transformations of PCR-amplified gene replacement cassettes.The gene deletions were confirmed by PCRs.Homozygous diploid yeast strains lacking PMS1 (FKY2291, FKY2292, FKY2293, and FKY2294) or MSH2 (FKY1982, FKY1983, and FKY1984) were constructed by crossing haploid BY4741 and BY4742 strains carrying pms1D or msh2D.WT diploid strains FKY1719, FKY1720, and FKY1721 are isogenic to the homozygous diploid pms1D or msh2D strains and were described previously (47).

Mutation accumulation, library preparation, and genome sequencing
To amass spontaneous mutations in the diploid yeast pms1D, msh2D, and WT strains mutation accumulation experiments were performed.In these mutation accumulation experiments, we utilized 30 single-cell bottlenecks to passage fifteen WT, nine pms1D, and eight msh2D isolates for 900 generations at 30 C on solid yeast peptone dextrose medium supplemented with 60 mg/l adenine and 63 mg/l uracil (47).A representative single colony was randomly selected for the next bottleneck.Glycerol stocks of the passaged isolates that were at generations 0 and 900 were prepared and stored at −80 o C.
The glycerol stocks were used to prepare patches of the multiple isolates of the passaged yeast pms1D, msh2D, and WT strains on solid yeast peptone dextrose medium supplemented with 60 mg/l adenine and 63 mg/l uracil at 30 C for 20 to 24 h.Genomic DNAs from the fresh patches were isolated using a MasterPure DNA purification kit (LGC Bioresearch Technologies).Four hundred nanograms genomic DNA of each sample was used to construct whole-genome DNA libraries with an NEBNext Ultra II FS DNA Library prep kit (NEB) and NEBNext Multiplex Oligos for Illumina (NEB).DNA fragments were size-selected for an average insert size of 450 bp.The libraries were analyzed using a TapeStation system (Agilent).The 151-bp 2D paired-end sequencing was performed using a NovaSeq 6000 sequencing system (Illumina) with an S4 PE 2 × 150 flow cell in XP mode.

Mutation spectra and calculation of mutation rates
The initial binary base calls sequencing files were demultiplexed and converted to FASTQ files using bcl2fastq Conversion Software, v. 2.20.0 (Illumina, https://support.illumina.com/sequencing/sequencing_software/bcl2fastq-conversionsoftware.html).The obtained paired-end sequencing data were imported into CLC Genomics Workbench (Qiagen) and aligned to the S. cerevisiae S288C reference genome.Mutations were called as described previously (47).Variants present within repetitive elements were not called if they could not be uniquely mapped.Mutations that were present in an isolate at generation 0 were removed from the list of mutations that were present in the same isolate at generation 900.The mutation spectra were generated by pooling mutations accumulated in the multiple isolates of the WT, pms1D, and msh2D strains according to the genotype.
Mutation rates (m) were calculated as previously described (47) using the following equation: m = N i / gen / N g , where N i is the number of mutations of type i, N g is the size of the diploid genome in which the variants were called (22,983,805 bp), and gen is the total number of generations for all isolates of the genotype.

Mutational signatures and sequence logos
The trinucleotide signatures of base substitution mutations (46) and the sequence logos were generated as previously described (47).Briefly, the position of each mutation in the genome and the 5 0 and 3 0 flanking sequences were determined using the S. cerevisiae S288C reference genome sequence and the CLC Genomics Workbench (Qiagen).The mutated trinucleotide sequences were sorted into the 96 different classes (46) using the Excel Data Filter Tool (Microsoft).WebLogo 3 (https://weblogo.threeplusone.com/create.cgi)was used to prepare sequence logos (50).

Figure 2 .
Figure 2. Similarities between trinucleotide signatures of base substitution mutations in MutLa-deficient and Msh2-deficient strains.The trinucleotide mutational signatures were extracted from sequenced nuclear genomes of the pms1D (A) and msh2D (B) strains as described in Experimental procedures.All base substitution mutations present in the mutation spectra of the pms1D and msh2D strains (n pms1D = 734 and n msh2D = 741) were used to generate the trinucleotide mutational signatures.

Figure 3 .
Figure 3. MutLa suppresses error-prone MMR in Saccharomyces cerevisiae and human iPSCs.A and B, rates of spontaneous genome-wide T > C base substitutions in 5 0 -ATA-3 0 (A) and 5 0 -NTN-3' (B) sequences in the yeast pms1D and msh2D strains.The Mann-Whitney U two-tailed test (GraphPad Prism 6 software) was used to determine the p values.C and D, genomewide rates of T > C substitutions in 5 0 -NTN-3' (C) and 5 0 -ATA-3' (D) sequences in human DPMS2 and DMSH2 iPSCs.The genome-wide mutation rates were calculated as the number of base substitutions per number of cell divisions and are shown as means ± SD.The p values were determined by unpaired t test (GraphPad Prism 6 software).The genome-wide mutations and the mutated sequences in iPSCs are from a previous study that utilized independent biological replicates (n DPMS2 = 4; n DMSH2 = 3) (48).iPSC, induced pluripotent stem cells; MMR, mismatch repair.

Figure 4 .
Figure 4. MutLa preferentially protects noncoding DNA from mutations.A, distribution of genome-wide mutations in coding and noncoding nuclear DNA of the pms1D, msh2D, and WT strains.B, genome-wide mutations in coding and noncoding DNA of the pms1D and msh2D strains.The numbers above the bars are ratios of the observed to expected mutations.C and D, genome-wide deletions (C) and insertions (D) in coding and noncoding DNA of the pms1D and msh2D strains.The numbers above the bars are ratios of the observed to expected mutations.E, genome-wide mutation rates in coding and noncoding DNA of the WT, pms1D, and msh2D strains.The numbers above the bars are relative mutation rates.F-H, genome-wide rates of deletions (F), insertions (G), and base substitutions (H) in coding and noncoding DNA of the WT, pms1D, and msh2D strains.Data that are shown in A-H were obtained using independent biological replicates (n wt = 15, n pms1D = 9, and n msh2D = 8).The p values were determined by the Mann-Whitney U two-tailed test (GraphPad Prism 6 software).

Figure 5 .Figure 6 .
Figure 5. MutLa-independent MMR contributes to the suppression of T > C mutations.Sequence patterns for T > C (A), C > T (B), and C > A (C) base substitutions in the pms1D and msh2D strains.The sequence logos were generated using WebLogo 3 as described in Experimental procedures.To produce the sequence logos, 213 (A), 316 (B), and 144(C) mutated sequences in the pms1D strain and 177 (A), 352 (B), and 154(C) mutated sequences in the msh2D strain were analyzed.MMR, mismatch repair.

Figure 7 .
Figure 7. MutLa-independent MMR plays a role in the suppression of base substitutions in N 3 homopolymeric runs.A, distribution of base substitution mutations in homopolymeric sequences of different length.B, rates of C > T transitions in 3-nt homopolymeric runs of the Pms1-and Msh2deficient strains.C, rates of base substitutions in 3-nt homopolymeric runs of the pms1D and msh2D strains.The mutation rates in B and C were calculated as described in the Experimental procedures and are presented as means ± SD.The Mann-Whitney U two-tailed test (GraphPad Prism 6 software) was used to compute the p values.MMR, mismatch repair.