Schizosaccharomyces pombe MutSα and MutLα Maintain Stability of Tetra-Nucleotide Repeats and Msh3 of Hepta-Nucleotide Repeats

Defective mismatch repair (MMR) in humans is associated with colon cancer and instability of microsatellites, that is, DNA sequences with one or several nucleotides repeated. Key factors of eukaryotic MMR are the heterodimers MutSα (Msh2-Msh6), which recognizes base-base mismatches and unpaired nucleotides in DNA, and MutLα (Mlh1-Pms1), which facilitates downstream steps. In addition, MutSβ (Msh2-Msh3) recognizes DNA loops of various sizes, although our previous data and the data presented here suggest that Msh3 of Schizosaccharomyces pombe does not play a role in MMR. To test microsatellite stability in S. pombe and hence DNA loop repair, we have inserted tetra-, penta-, and hepta-nucleotide repeats in the ade6 gene and determined their Ade+ reversion rates and spectra in wild type and various mutants. Our data indicate that loops with four unpaired nucleotides in the nascent and the template strand are the upper limit of MutSα- and MutLα-mediated MMR in S. pombe. Stability of hepta-nucleotide repeats requires Msh3 and Exo1 in MMR-independent processes as well as the DNA repair proteins Rad50, Rad51, and Rad2FEN1. Most strikingly, mutation rates in the double mutants msh3 exo1 and msh3 rad51 were decreased when compared to respective single mutants, indicating that Msh3 prevents error prone processes carried out by Exo1 and Rad51. We conclude that Msh3 has no obvious function in MMR in S. pombe, but contributes to DNA repeat stability in MMR-independent processes.

. After excision beyond the mismatch, a replicative DNA polymerase fills the resulting gap and the remaining nick is ligated.
A defect in the human MMR genes MSH2 and MLH1 causes microsatellite instability and a predisposition to colon and other types of cancer (Lynch et al. 2009;Da Silva et al. 2016). Mutations in other MMR genes are rarely correlated with cancer, probably due to functional redundancy. In contrast to other types of repetitive elements, microsatellites are often situated in genes, which is a critical factor for tumor development and important for the choice of drugs for treatment of cancer patients. For example, deletions in a T 11 repeat in intron 4 of MRE11 causes aberrant splicing, and as a consequence, a truncated MRE11 protein (Giannini et al. 2002). This sensitizes cancer cells to combined treatment with camptothecin and thymidine (Rodriguez et al. 2008).
Recombination processes can alter tract lengths of repetitive DNA either by unequal crossover between repeats or through secondary structures formed between repeats in the same strand. The Mre11-Rad50-Nbs1 (MRN) complex has single-stranded 39-exonuclease and endonuclease activities as well as structural functions in recombination processes (Cejka 2015). After 59-39 resection of DNA double-strand breaks or ends by Exo1 or other 59-exonucleases, Rad51-dependent homologous recombination (HR) can occur by invasion of the 39ended single strand into a complimentary DNA molecule. Rad51independent single-strand annealing (SSA) can also occur between two repeats and leads to deletion of the intervening sequence. SSA requires the nucleotide excision repair factors Rad1-Rad10 of Saccharomyces cerevisiae (XPF-ERCC1 in human) and MutSb (Bhargava et al. 2016). FEN1 is a flap endonuclease with multiple roles in DNA metabolisms. FEN1 is involved in processing of Okazaki fragments during replication, in long-patch base excision repair and in other processes (Marti and Fleck 2004). S. cerevisiae, rad27 (a FEN1 homolog) mutants exhibit instability of mono-and dinucleotide repeats and generate duplications of sequences flanked by repeats (Johnson et al. 1995;Tishkoff et al. 1997;Kirchner et al. 2000). In addition, FEN1 has been implicated in trinucleotide repeat stability (Freudenreich et al. 1998;Liu and Wilson 2012) and repair of large loops with up to 216 unpaired nucleotides (Sommer et al. 2008).
The genome of fission yeast Schizosaccharomyces pombe encodes the MutS homologs Msh1, Msh2, Msh3, and Msh6, the MutL homologs Mlh1 and Pms1, and the exonuclease Exo1. Based on homology with S. cerevisiae Msh1, S. pombe Msh1 likely acts in MMR of mitochondrial DNA. S. pombe Msh2, Msh6, Mlh1, and Pms1 are indispensable for repair of base-base mismatches and small loops with one or two nucleotides (Schär et al. 1997;Rudolph et al. 1999;Mansour et al. 2001;Tornier et al. 2001;Marti et al. 2003). In contrast, Msh3 seems to have no, or a minor and rather MMR-independent, function in repair of base-base mismatches and small loops. Exo1 appears to be involved in MMR of base-base mismatches but has a rather MMR-independent function in repair of small loops (Rudolph et al. 1998;Mansour et al. 2001;Marti et al. 2003).
In the present study, we tested stability of tetra-, penta-, and heptanucleotide repeats in S. pombe. Our aim was to analyze whether stability of such repeats is dependent on MMR, and if so, whether MutSa, MutSb, or both are involved. In addition, we analyzed mlh1, and thus MutLa-deficient strains as well as exo1 mutants.

Determination of mutation rates and spectra
Mutation rates were determined by fluctuation tests as described (Mansour et al. 2001). In brief, seven tubes containing 2 ml YEL were each inoculated with a single small colony and incubated at 30°until cultures were grown to stationary phase. Appropriate dilutions were plated on YEA for cell titer determination and on MMA for selection of Ade + revertants. Colonies were counted after 5 d of growth at 30°, except for strains with rad50 or rad51 background, where colonies were counted after 6 d to compensate for their slow growth. Reversion rates were calculated from at least three independent fluctuation tests. Statistical significance was calculated with a two-tailed Student's t-test.
The nature of mutations was determined by sequencing of PCR products from genomic DNA using primers ade6_F 59-ATTAACACT GATGCCTTGGC and ade6_R 59-ACAGAGAACGTTTAGCGATC. In the case of ade6-(GACC) 7 DT, repeat tract changes were also analyzed by inspection of the color of Ade + revertants ( Figure 2A). The color was best determined when revertants were restreaked on YEA without supplemented adenine. The proportion of white and pink revertants was determined after 1-2 d of growth at 30°. Final averages and SDs were calculated from averages of at least three independent fluctuation tests, each with seven cultures, and up to 20 random revertants (where available) per culture. Figure 1 Schematic of the position and nature of the various repeats. In the ade6::ura4 disruption strain, 13 nucleotides have been deleted and replaced by the ura4 marker. This strain was transformed with DNA fragments to produce ade6 mutants with the indicated repeats (highlighted by blue arrows). Nucleotides within the repeats that differ from the wild-type sequence are shown in red. Integration of the repeats caused frameshifts with nearby located stop codons. The numbers of net inserted nucleotides and of the major deletion/insertion events that lead to Ade + reversions are given on the right. (+14) indicates that, although this event restores the open reading frame, it was not found among the 30 ade6-(ATCGTCC) 5 DT revertants sequenced (Table 6). In (GACC) 7 DT and (ATCGTCC) 5 DT, a T in a T 3 stretch immediately downstream of the repeats is deleted. This T deletion allows detecting different types of deletions and insertions within respective repeats in comparison to (GACC) 8 and (ATCGTCC) 5 , as indicated on the right.
Data availability S. pombe strains are available on request. The authors state that all data necessary for confirming the conclusions presented in the article are represented fully within the article.

Genetic assay for microsatellite instability in S. pombe
We have previously reported that in-frame nucleotide insertions and codon changes in a region around nucleotide 1397 of the ade6 gene [the ATG start codon is at 875 as defined by Szankasi et al. (1988)] in S. pombe did not disrupt or only slightly disrupted its functionality (Mansour et al. 2001;Marti et al. 2003). Here, we inserted tetra-, penta-, and hepta-nucleotide repeats in this part of ade6 in order to analyze microsatellite instability in wild type and MMR mutants (Figure 1). Insertion of the repeats caused frameshifts, which rendered cells auxotrophic for adenine due to nearby located stop codons. Such ade6 mutants required a supply of adenine for growth and turned red on the medium with a limited amount of adenine due to accumulation of a red pigment. Reversions of such strains to Ade + can occur by deletions or insertions of repeats, when these events restore the open reading frame. With (GACC) 7 DT, (CTGCC) 6 , and (ATCGTCC) 7 , insertions of one repeat unit and deletions of two repeat units are the major events detectable. The opposite is the case with (GACC) 8 and (ATCGTCC) 7 DT, where deletions of one repeat unit and insertions of two repeat units are the principal events that can be detected, although in the case of (ATCGTCC) 7 DT, we identified exclusive deletions as described below.
In the case of ade6-(GACC) 7 DT, deletions and insertion can be distinguished by the color of revertants ( Figure 2A). Deletions caused white Ade + , while insertions caused pink Ade + . The pink color likely reflects that ade6 is not fully functional and thus that the red pigment is produced in low quantities. Sequencing of 18 Ade + revertants of the (GACC) 7 DT repeat in the various genetic backgrounds revealed deletions and sequencing of 23 (GACC) 7 DT revertants revealed insertions (Table 1). All revertants with deletions were white and all revertants with insertions were pink. Thus, the occurrence of deletions and insertions in the (GACC) 7 DT repeat can be easily determined from a large number of revertants. Analyzing revertants of the other ade6 repeats did not allow a distinction by color, probably because Ade + originating from deletions and insertions were both not fully functional. In these cases, PCR products of independent Ade + revertants were subjected to sequencing to identify the number of repeats.
GACC tetra-nucleotide repeats were unstable in msh2, msh6, and mlh1 mutants We first analyzed stability of GACC tetra-nucleotide repeats in wild type, and in mutants deleted for either msh2, msh3, msh6, mlh1, or exo1. In wild type, ade6-(GACC) 8 and ade6-(GACC) 7 DT reverted to Ade + with 1.2 · 10 25 and 2.1 · 10 25 reversions per cell division, respectively ( Table 2). The ade6-(GACC) 8 reversion rates increased 39-44-fold in msh2, msh6, and mlh1 mutants and slightly decreased in msh3 and exo1 mutants. Similarly, ade6-(GACC) 7 DT reversion rates increased 17-22-fold in msh2, msh6, and mlh1 mutants and decreased 3-5-fold in msh3 and exo1 mutants. None of the differences between msh3, exo1, and wild type was statistically significant. DT repeat that lead to Ade + can be distinguished by their color. ade6-(GACC) 7 DT mutants form red colonies on medium with limited amount of adenine due to a defective ade6 gene. They can revert to Ade + by deletion of two or five repeat units, producing white Ade + , or by insertion of one or four repeat units, producing pink Ade + . (B) Percentage of deletions in the various strain backgrounds. Wild type, msh3, and exo1 mainly reverted to Ade + by deletions, while msh2, msh6, and mlh1 mutants mainly reverted by insertions. Significantly different to wild type: ÃÃ p , 0.01; ÃÃÃ p , 0.001. Shown are average values with SDs. Reversion spectra are also presented as pie charts, with the proportion of deletions and insertions indicated in white and pink, respectively.
Repeat tract changes in ade6-(GACC) 7 DT can be determined by colony color as described above and in Materials and Methods, and as illustrated in Figure 2A. In wild type, msh3, and exo1 strains, reversions mainly occurred by deletions ( Figure 2B). Most were deletions of two repeat units (Table 1). In contrast, msh2, msh6, and mlh1 mutants mainly reverted to Ade + by insertion of one repeat ( Figure 2B and Table 1). Sequencing of Ade + revertants of ade6-(GACC) 8 revealed that in wild type, all 10 had two repeat insertions (Table 1). In contrast, msh2 and msh6 reverted mainly by deletion of one repeat. The spectrum for msh3 was more heterogeneous, with some preference for insertions. Importantly, the occurrence of deletions vs. insertions was significantly different to wild type (Table 1).
Increased penta-nucleotide repeat stability in the msh3 mutant The (CTGCC) 6 repeat reverted in wild type with a rate of 3.4 · 10 25 to Ade + (Table 3). Rates were not significantly different in msh2, msh6, mlh1, and exo1 mutants. In contrast, the (CTGCC) 6 repeat appeared to be more stable in the msh3 mutant. It reverted mostly through gain of one repeat unit in all strain backgrounds, with the possible exception of mlh1 (Table 4). We conclude that MutSa cannot repair loops with five unpaired nucleotides in the (CTGCC) 6 context, and that Msh3 has some function in stability of the penta-nucleotide repeat.
Hepta-nucleotide repeat stability was not affected in msh2, msh6, and mlh1 and was slightly decreased in msh3 and exo1 mutants In wild type, the hepta-nucleotide repeats ade6-(ATCGTCC) 5 DT and ade6-(ATCGTCC) 5 reverted to Ade + at rates of 7.2 · 10 26 and 7.5 · 10 26 , respectively (Table 5). Rates were not significantly different in msh2, msh6, and mlh1 mutants, suggesting that loops with seven unpaired nucleotides are not repaired by MMR in S. pombe. In msh3 and exo1 mutants, reversion rates were increased 1.7-1.9-fold for ade6-(ATCGTCC) 5 DT (not significant) and 2.8-2.9-fold for ade6-(ATCGTCC) 5 (significant) ( Table 5). All sequenced ade6-(ATCGTCC) 5 DT revertants contained four hepta-nucleotide repeats and thus originated from deletion of one repeat ( Table 6). The fact that insertions were not found at all could be either due to a relative low sample size or that insertion of 14 nucleotides, leading to a total of 39 additional nucleotides in ade6 (Figure 1), does not render cells Ade + . ade6-(ATCGTCC) 5 reverted mainly by insertion of one repeat and less frequently by deletion of two repeats without any significant differences between the spectra of wild type and any of the mutants (Table 6).
Hepta-nucleotide repeat stability in rad2 FEN1 , rad50, and rad51 mutants The msh3 and exo1 mutants, but not the mutants msh2, msh6, and mlh1, exhibited increased instability of the (ATCGTCC) 5 repeat (Table  5). Since defective msh2 or mlh1 is generally considered to completely inactivate MMR, the (ATCGTCC) 5 repeat instability in msh3 and exo1 is not due to a defect in MMR. We therefore wanted to analyze the genetic context of msh3 and exo1 defects in microsatellite stability. To do this, we measured reversion rates of the hepta-nucleotide repeat in rad2 FEN1 , rad50, and rad51 mutants and in various double mutants (Table 7). FEN1 and HR have been implicated in repeat stability in S. cerevisiae (Johnson et al. 1995;Tishkoff et al. 1997;Freudenreich et al. 1998;Kirchner et al. 2000;Sundararajan et al. 2010). We found that the rad2 FEN1 , rad50, and rad51 single mutants had $2-fold increased reversion rates (Table 7). The msh3 exo1 double mutant reverted significantly less frequently to Ade + than either single mutant. Mutation rates n  remained about the same in msh3 rad50, msh3 rad2 FEN1 , and exo1 rad50 but decreased in the msh3 rad51 and exo1 rad51 double mutants when compared to respective single mutants (Table 7). Like wild type and the other single mutants, rad50 and rad51 strains mainly reverted to Ade + by insertion of one repeat unit (Table 6).
In the rad2 FEN1 mutant, 50% of the reversions were due to deletion of two repeats, which was not significantly different to wild type. A reduction of reversion rates in the msh3 rad51 and exo1 rad51 double mutants indicates that Exo1 and Rad51 act error prone on the (ATCGTCC) 5 repeat when msh3 is mutated.

DISCUSSION
MutSa and MutLa are essential for S. pombe MMR, which is limited to loops with up to four nucleotides Eukaryotic MMR is initiated by MutSa for repair of base-base mismatches and loops or by MutSb for repair of loops (Marti et al. 2002;Jiricny 2013). In S. cerevisiae, msh3 and msh6 mutants show little to moderate increases of mutation rates in mono-and dinucleotide repeats (Johnson et al. 1996;Marsischky et al. 1996;Greene and Jinks-Robertson 1997;Sia et al. 1997). On the other hand, such repeats are highly unstable in msh3 msh6 double mutants and within the range of the msh2 instability, indicating redundancy of MutSa and MutSb for small loops in this organism. In humans, MutSa is the major factor for recognition of base-base mismatches and loops, while MutSb rather serves as a backup (Drummond et al. 1997;Genschel et al. 1998;Marra et al. 1998). In S. pombe, we knew to this date that MMR is able to repair basebase mismatches and loops with up to two nucleotides (Schär et al. 1997;Rudolph et al. 1999;Mansour et al. 2001;Tornier et al. 2001;Marti et al. 2003). This requires MutSa and MutLa but not MutSb. In the present study, we expanded analysis of loop repair in S. pombe to four to seven unpaired nucleotides. Our aim was to determine the contributions of Msh2, Msh3, Msh6, Mlh1, and Exo1, and particularly the relative roles of MutSa and MutSb in stability of repeats with four or more iterated nucleotides in this model organism. The microsatellites tested were (GACC) 7 and (GACC) 8 tetra-, (CTGCC) 6 penta-, and (ATCGTCC) 5 hepta-nucleotide repeats (Figure 1). All such insertions caused a frameshift, rendering cells defective in ade6. Reversions to Ade + occurred when deletions or insertions of repeats restored the open reading frame.
Inactivated Msh2, Msh6, and Mlh1 caused instability of the tetranucleotide repeats, while defective Msh3 and Exo1 rather made the repeats slightly more stable, although this was not significantly different to wild type ( Table 2). The (GACC) 8 repeat reverted in wild type by insertions of eight nucleotides, while in msh2 and msh6 mutants, mainly deletions of four nucleotides occurred (Table 1). Thus, MutSa of S. pombe is capable to initiate MMR of loops with four unpaired nucleotides, whereas MutSb is not.
The assay with the (GACC) 7 DT repeat allowed distinguishing deletions from insertions by the color of Ade + revertants (Figure 2A). We found that wild type, msh3, and exo1 mainly reverted by eight-nucleotide deletions, and msh2, msh6, and mlh1 mainly by four-nucleotide insertions ( Figure 2B and Table 1). Thus, this assay also revealed that MutSa but not MutSb initiates MMR of loops with four nucleotides. In addition, Mlh1, and by extrapolation MutLa, is involved in removal of four-nucleotide loops. Since four-nucleotide deletions, detectable with (GACC) 8 , and four-nucleotide insertions, detectable with (GACC) 7 DT, were the predominant reversion events in msh2 and msh6 mutants, slippage of one repeat can occur in the template and in the nascent strand during replication, and both types of events are corrected by MMR mediated by MutSa and MutLa.
Mutation rates of the penta-nucleotide repeat (CTGCC) 6 in msh2, msh6, mlh1, and exo1 mutants were similar to that of wild type, but decreased in msh3 (Table 3). All strains preferentially reverted to Ade + by insertion of one repeat (Table 4). Inactivation of msh2, msh6, and mlh1 did not affect reversion rates or spectra of the hepta-nucleotide repeats (ATCGTCC) 5 DT and (ATCGTCC) 5 (Table 5 and Table 6). Like wild type, the mutants reverted by deletion of one repeat in (ATCGTCC) 5 DT and mainly by insertion of one repeat in (ATCGTCC) 5 . We conclude that loops in penta-and hepta-nucleotide repeats are not substrates of MutSa and MutLa in S. pombe.

Msh3 has an MMR-independent function in repeat stability
Our previous data showed that msh3 mutants had no significant defects in repair of base-base mismatches and of loops with one unpaired nucleotide in a T 6 repeat and in nonrepetitive DNA (Tornier et al. n  b Relative to wild type. c p-values were calculated by a two-tailed Student's t-test in comparison to wild type. Marti et al. 2003). msh3 mutations caused some instability of a (GT) 8 dinucleotide repeat, which was mostly evident by a reversion spectrum different to wild type (Mansour et al. 2001). Wild type reverted mainly by four-nucleotide insertions, whereas msh3 mainly reverted by two-nucleotide deletions in the (GT) 8 repeat. However, this was clearly less frequent than in msh2, msh6, and pms1 mutants (Mansour et al. 2001). In the present study, we also found that the spectrum for the (GACC) 8 repeat was different to wild type and to msh2 and msh6 (Table 1). Wild type exclusively reverted via eightnucleotide insertions (10 out of 10 revertants analyzed), and msh2 and msh6 mostly by four-nucleotide deletions. In contrast, four out of 11 revertants in msh3 background were due to four-nucleotide deletions, six originated from eight-nucleotide insertions, and one by an insertion of 20 nucleotides (Table 1). Thus, Msh3 appears to have a function in tetra-nucleotide repeat stability, which is different to Msh2 and Msh6.

2001;
In the case of the penta-nucleotide repeat, we found that the msh3 mutant showed a lower mutation rate than wild type (Table 3), indicating a role of Msh3 in supporting tract length changes in this repeat. Intriguingly, the msh3 mutant exhibited increased instability of the hepta-nucleotide repeats (Table 5). Thus, Msh3 has a function in maintaining stability of such repeats. Since msh2, msh6, and mlh1 did not show instability of hepta-nucleotide repeats, the Msh3 function appears to be MMR independent. We have found a genetic link to Rad50, Rad51, and Exo1, indicating that the Msh3 function is related to recombinational processes. Such a function is well known for S. cerevisiae MutSb, which participates in SSA where repeats flank a double-strand break (Sugawara et al. 1997;Chakraborty and Alani 2016). MutSb of S. pombe likely acts similarly, as both Msh2 and Msh3 have a function in the recombinational process of mating-type switching (Fleck et al. 1992;Rudolph et al. 1999). However, genetic data presented here and in our previous studies suggest that the functions of Msh3 in repeat stability and recombination is likely independent of Msh2 Mansour et al. 2001;Marti et al. 2003). The Msh2 independent role of S. pombe Msh3 in recombination may relate to that of bacterial MutS2 proteins, which act in recombination and antirecombination mechanisms but not in MMR (Pinto et al. 2005;Burby and Simmons 2017). MutS2 of Helicobacter pylori binds to DNA structures that resemble recombination intermediates and inhibits strand exchange in vitro (Pinto et al. 2005). In this regard, S. pombe Msh3 may be functionally similar, although structurally, it lacks the endonuclease domain of MutS2, and homology of its amino acid sequence clearly allocates it to the group of eukaryotic Msh3 proteins.
Structural studies with human MutS heterodimers showed that mismatch binding largely occurs by Msh3 or Msh6, while Msh2 has few contacts with the DNA backbone of correctly paired nucleotides in the vicinity (Warren et al. 2007;Gupta et al. 2011). The human Msh6 protein interacts directly with mismatched bases via a phenylalanine, which is conserved in eukaryotic Msh6 orthologs and bacterial MutS. In contrast, human Msh3 lacks this residue and instead interacts with phosphate groups of the unpaired nucleotides (Gupta et al. 2011). Work by Lee et al. (2007) demonstrated that deletion of the mismatch binding domain of S. cerevisiae Msh2 causes loss of MutSb2dependent MMR activity and revealed that the domain in Msh2 is required for general DNA binding, and in Msh3 for binding to DNA loops. In complex with Msh2, a chimeric Msh6 protein of S. cerevisiae containing the mismatch binding domain of Msh3 showed substrate specificity of Msh3, i.e., high affinity to loops with one to four unpaired nucleotides (Shell et al. 2007). The amino acid sequence within the mismatch binding domain of S. pombe Msh3 is very similar to that of human Msh3. However, in contrast to S. cerevisiae and human Msh3 and orthologs of other eukaryotes, S. pombe Msh3 lacks a canonical PIP box, which mediates interaction with PCNA. The PIP box of human Msh3 overlaps with the Mlh1 binding domain . Thus, it is also conceivable that amino acid residues required for interaction with Mlh1 are not present in S. pombe Msh3. It is currently not known whether S. pombe Msh3 can interact with PCNA or MutLa. If it does not, this may explain that it does not participate in MMR.

Role of MutSa and MutSb in MMR
The Escherichia coli homodimer MutS enables repair of base-base mismatches and loops with up to four nucleotides (Iyer et al. 2006).
n Repeat tract changes were determined from random independent revertants. a Distribution of deletions and insertions in mlh1 background significantly different to wild type (x 2 = 4.57; p = 0.033). Reversion spectra of all other mutants were not significantly different to wild type.
n Reconstituted MMR with S. cerevisiae proteins in vitro revealed that both MutSa and MutSb could initiate repair of base-base mismatches and of loops with one, two, or four nucleotides . Human MutSa binds to base-base mismatches and to loops with up to eight nucleotides, while MutSb allows repair of loops with two to about eight nucleotides (Genschel et al. 1998). Our genetic data imply that S. pombe Msh6 as part of the MutSa heterodimer is able to bind to loops with up to four unpaired nucleotides. Thus, the substrate spectrum of bacterial MutS and MutSa of S. pombe appears to be similar. During evolution, the spectrum had been extended to enable recognition of larger loops in humans. On the other hand, the substrate spectrum of MutSb considerably differs between species. Human MutSb supports repair of loops with two to eight nucleotides (Genschel et al. 1998), whereas S. cerevisiae MutSb is also involved in repair of some basebase mismatches besides loop repair (Harrington and Kolodner 2007), and S. pombe MutSb apparently does not have a function in MMR. In addition, some eukaryotes, such as Caenorhabditis elegans and Drosophila melanogaster do not have an Msh3 ortholog (Marti et al. 2002) and likely carry out MMR with MutSa and MutLa and no other MutS and MutL heterodimers, like S. pombe does. Harrington and Kolodner (2007) interpreted mutation spectra of base substitutions in S. cerevisiae msh3 mutants that were different to wild type as a role of MutSb in MMR of base-base mismatches. We observed differences of msh3 in mutation spectra for a (GT) 8 repeat (Mansour et al. 2001), a reduction of recombination events , and of reversion rates at the (CTGCC) 6 repeat (Table 3), an altered reversion spectrum for (GACC) 8 (Table 1) and repeat instability of (ATCGTCC) 5 (Table 5). We interpret these differences as phenotypes caused by loss of MMRindependent functions of Msh3.

Does S. pombe Exo1 have a function in MMR?
Exo1 of S. pombe was the first eukaryotic exonuclease to be identified as having a function in repair of mismatches (Szankasi and Smith 1995). Further studies with S. pombe showed that Exo1 contributes to MMR of base-base mismatches (Rudolph et al. 1998), modulates MMR of two-nucleotide loops in nonrepetitive DNA (Marti et al. 2003), and has an MMR-independent function in dinucleotide repeat stability (Mansour et al. 2001). In the present study, we did not found any evidence for a role of Exo1 in tetra-and penta-nucleotide repeat stability (Table 2 and Table 3). However, we observed that loss of Exo1 caused instability of heptanucleotide repeats, in contrast to the MMR mutants msh2, msh6, and mlh1 (Table 5). Exo1 also acts in recombination and doublestrand break repair (Fiorentini et al. 1997;Tsubouchi and Ogawa 2000;Kirkpatrick et al. 2000;Cejka 2015). Thus, a defect in a n Table 7 Reversion rates of (ATCGTCC) 5 repeats in msh3, exo1, and rad mutants b Relative to wild type. c p-values were calculated by a two-tailed Student's t-test in comparison to the indicated strains. WT, wild type; rad, rad2 FEN1 , rad50, or rad51 single mutants.
n Repeat tract changes were determined from randomly selected revertants of independent cultures. None of the reversion spectra of the mutants is significantly different to wild type. ND, not determined.
recombination mechanism might cause hepta-nucleotide instability of the S. pombe exo1 mutant rather than MMR deficiency, as discussed below. Although a nuclease is essential for removal of unpaired nucleotides during MMR, Exo1 does seem to be dispensable for MMR-mediated loop repair in S. pombe. This may be attributed to redundancy with other nucleases. MutLa of S. cerevisiae and humans has endonuclease activity, which is sufficient for completing MMR in the absence of Exo1 (Kadyrov et al. 2006(Kadyrov et al. , 2007Smith et al. 2013;Goellner et al. 2014Goellner et al. , 2015. Thus, it is also likely that Exo1 of S. pombe participates in MMR, but that MutLa and maybe other nucleases can replace its function. In fact, the amino acids required for MutLa nuclease activity are all highly conserved between eukaryotes, including S. pombe , supporting the idea that having endonuclease activity is a general feature of eukaryotic MutLa. In S. cerevisiae, exo1 mutants exhibit weak defects in MMR Amin et al. 2001;Smith et al. 2013;Goellner et al. 2014), likely because MutLa and Exo1 nuclease activities are largely redundant. exo1 deletion strains and pms1 strains with mutations causing endonuclease deficiency generally showed subtle increases of mutation rates, which strongly increased when both mutations were combined .
Rad51 and Exo1 are involved in error prone repair at (ATCGTCC) 5 repeats in msh3 mutants Stability of the hepta-nucleotide repeat (ATCGTCC) 5 was influenced by processes involving Msh3, Exo1, Rad2 FEN1 , Rad50, and Rad51. Deletions of any of the genes caused an $2-3-fold increase of reversion rates, which was predominantly due to expansions by one repeat unit and therefore by insertions in the nascent strand (Table 6 and Table 7). Rates were not further increased in the msh3 rad2 FEN1 , msh3 rad50, and exo1 rad50 double mutants. Instead, the msh3 exo1, msh3 rad51, and exo1 rad51 double mutants had lower rates than the respective single mutants. In S. cerevisiae, CAG trinucleotide repeats were unstable in rad51, rad52, and mre11 single mutants (Sundararajan et al. 2010). However, increased rates of repeat expansions in mre11 were largely suppressed by additional mutation of rad52. These data suggest that the MRN complex plays a role in maintaining repeat stability, and that downstream steps of HR in mre11, but not in wild-type background, can carry out error prone recombination at repeats (Sundararajan et al. 2010). In summary, the (ATCGTCC) 5 repeat analyzed in our study might be stabilized by Msh3 and slipped-out loops correctly processed by HR requiring Rad50, Exo1, and Rad51, thereby preventing aberrant events. In the absence of Msh3, the Exo1 and Rad51 proteins might carry out error prone processes, such as misalignment of repeats after strand resection catalyzed by Exo1 and during strand invasion mediated by Rad51.

Conclusions
We conclude from our studies that S. pombe Msh6, as part of MutSa, recognizes base-base mismatches and loops with one to four unpaired nucleotides, while Msh3 does not play a significant role in MMR, but rather maintains repeat stability independently of MMR. Consequently, S. pombe MMR cannot repair loops with five or more nucleotides, in contrast to human MMR (Genschel et al. 1998). Microsatellites with five or six iterated nucleotides are rare in S. pombe (hepta-nucleotide repeats were not analyzed) (Karaoglu et al. 2005), but are relatively abundant in the human genome (Lander et al. 2001). Thus, to ensure genome stability, humans require repair of larger loops that occur by strand slippage in microsatellites, while larger loops may be formed rarely in S. pombe microsatellites. It is therefore critical for humans, but not for S. pombe, to have an MMR system that can deal with larger loops.