MSH2 ATPase Domain Mutation Affects CTG•CAG Repeat Instability in Transgenic Mice

Myotonic dystrophy type 1 (DM1) is associated with one of the most highly unstable CTG•CAG repeat expansions. The formation of further repeat expansions in transgenic mice carrying expanded CTG•CAG tracts requires the mismatch repair (MMR) proteins MSH2 and MSH3, forming the MutSβ complex. It has been proposed that binding of MutSβ to CAG hairpins blocks its ATPase activity compromising hairpin repair, thereby causing expansions. This would suggest that binding, but not ATP hydrolysis, by MutSβ is critical for trinucleotide expansions. However, it is unknown if the MSH2 ATPase activity is dispensible for instability. To get insight into the mechanism by which MSH2 generates trinucleotide expansions, we crossed DM1 transgenic mice carrying a highly unstable >(CTG)300 repeat tract with mice carrying the G674A mutation in the MSH2 ATPase domain. This mutation impairs MSH2 ATPase activity and ablates base–base MMR, but does not affect the ability of MSH2 (associated with MSH6) to bind DNA mismatches. We found that the ATPase domain mutation of MSH2 strongly affects the formation of CTG expansions and leads instead to transmitted contractions, similar to a Msh2-null or Msh3-null deficiency. While a decrease in MSH2 protein level was observed in tissues from Msh2G674 mice, the dramatic reduction of expansions suggests that the expansion-biased trinucleotide repeat instability requires a functional MSH2 ATPase domain and probably a functional MMR system.


Introduction
Myotonic dystrophy type 1 (DM1) is a neuromuscular disease characterized by highly variable clinical manifestations including skeletal muscle, cardiorespiratory and brain defects and endocrine abnormalities [1]. DM1 results from the expansion of an unstable CTG trinucleotide repeat in the 39-untranslated region of the dystrophia myotonica protein kinase (DMPK) gene [2][3][4]. In normal individuals, the CTG repeat at the DM1 locus varies from 5 to 37 repeats and remains stable over generations. Tract lengths of .50 CTG repeats are dramatically unstable in successive generations, with a strong bias towards expansions, often reaching thousands of CTGs in the most severe form of the disease [1]. The level of CTG repeat instability depends on the sex and repeat length of the transmitting parent [5,6]. The size of the expanded repeat inversely correlates with the age at onset and is positively correlated with the severity of symptoms providing the molecular basis for anticipation [7]. In addition to intergenerational instability, the CTG repeat is also somatically unstable in DM1 patients. Somatic instability, which is age-dependent, biased towards expansions of CTG repeat length, presents variable expansion patterns between tissues, and is probably associated with the progression of disease symptoms in DM1 patients [8]. A better understanding of the molecular mechanisms underlying trinucleotide repeat instability is fundamental for the development of therapies aimed at controlling trinucleotide repeat instability. Such therapeutic approaches would hopefully reduce the disease severity and progression of the symptoms.
Extensive analysis of repeat instability in various model systems in vitro and in vivo has revealed crucial insights into the role of ciselements, DNA replication and genome-maintenance repair [9][10][11][12]. Numerous transgenic mouse models and their derived cell lines revealed that the repeat expansion rate does not always correlate with cell division rate, suggesting the involvement of various replication-independent mechanisms [13,14]. Studies have reported that several mismatch repair (MMR) proteins are involved in expansion-biased trinucleotide instability in transgenic mouse models carrying expanded CTGNCAG repeat tracts. Accumulation of CTGNCAG expansions in mitotic, non-mitotic and germline tissues clearly depends on the MMR proteins MSH2, MSH3 and to a lesser extent on PMS2 (postmeiotic segregation homologue 2) [15][16][17][18][19][20]. Precisely how MutS homologues MSH2 and MSH3 mediate CTGNCAG repeat expansions has not yet been clearly determined.
MMR proteins are required to maintain genomic integrity in prokaryotes and eukaryotes, by correcting single mismatches and short unpaired regions, such as small insertions and deletions [21,22]. In eukaryotes, three proteins are involved in mismatch recognition, MSH2, MSH3 and MSH6. The three proteins form two heterodimers MutSa (MSH2-MSH6) and MutSb (MSH2-MSH3). MutSa is thought to be involved primarily in the recognition and repair of base-base mismatches and small insertion/deletion loops. MutSb acts preferentially on insertion/ deletion loops up to 12 nucleotides in length [23]. MMR is tightly linked with the mismatch-binding-dependent ATPase activities of the MSH complexes but the precise roles of these activities are still in debate. Upon binding to the mismatch, the MSH complex exchanges a bound ADP for an ATP, which permits a change in the proteins conformation and the signaling of downstream repair events [21,22,24,25]. The MSH2, MSH3, and PMS2 mismatch repair proteins are also involved in other DNA repair pathways such as single-strand annealing and homologous recombination, anti-recombination, DNA damage signaling, apoptosis, as well as site-specific mutagenesis during immunoglobin somatic hypermutation and class switch recombination [26][27][28][29]. While the MSH2 ATPase activity is known to be required for some of these non-MMR functions, the specific roles of ATP-binding and ATPhydrolysis remain to be elucidated. Different scenarios have been suggested to explain the role of MMR proteins in trinucleotide repeat instability. It has been shown that CTGNCAG repeat tracts can form hairpin structures and/or slipped-strand structures in vitro [30,31]. These structures are suspected to disturb DNA metabolism in vivo, leading to trinucleotide repeat instability [9]. An initial study reported that in vitro MSH2 can recognize and bind to slipped-strand (S-DNA) structures formed by expanded CTGNCAG repeat tracts [32]. More recently, a detailed study of MutSb binding in vitro to long CAG hairpin structures was reported [18], as well as an in vitro assay to process slipped-DNAs [33].
Three models explaining the mechanism through which the MMR proteins MSH2 and MSH3 act to generate CTGNCAG expansions have been put forth. The first model proposes that the role of the MMR proteins MSH2-MSH3 (MutSb) in trinucleotide repeat expansions is through protecting CAG loops from repair by binding the loops [18]. This model was based upon the reduced ability of the human MutSb complex to hydrolyze ATP when bound to hairpins formed by either (CAG) 13 or (CTG) 13 ; suggesting that the protein complex would be stuck on the loop, unable to translocate along the helix, and unable to signal or interact with downstream repair proteins. In this manner binding of MutSb to CAG or CTG hairpins would abort their repair, thereby allowing the loops to be incorporated into the genome as expansions. Thus, in this model MutSb binding but not ATP hydrolysis would be critical for expansions.
The second model posits that the MutSb complex is not required for CTG or CAG slip-out processing or protecting them from repair, but rather MutSb may be involved in expansion prior to the processing of the slip-outs [33]. An in vitro repair assay using slipped-DNAs with slip-outs of either (CAG) 20 or (CTG) 20 revealed that these can be correctly repaired, escape repair, or be repaired in an error-prone manner (only partially excising the slip-outs). All of these processes occurred in human cell extracts deficient in either MSH2 or MSH3. These authors suggested that rather than acting in the processing or protecting CTG or CAG slip-outs from repair, MutSb may be involved in forming the slipouts. It is unknown if such an activity would require the endogenous ATPase activity of the protein.
The third model suggests that the mismatch repair process itself is involved in trinucleotide repeat expansions [17]. This study demonstrated that the MutL MMR homologue PMS2 is also involved in CTGNCAG repeat somatic expansions in mice. Some repeat expansions are still formed in PMS2-deficient mice suggesting a partial role of PMS2 in trinucleotide repeat instability [17]. The authors proposed that the MMR activity is required to generate somatic expansions implying that the role of MMR proteins would not be limited to the binding of MutSb complex to the A-A mismatches in CAG hairpins and/or to the MutSbmediated formation of slipped-strand structures. In this model the ATPase activity of the MutSb would be critical to expansions.
In an attempt to distinguish between the different hypotheses and to clarify the role of the ATPase activity of MSH2 in the mechanism of CTGNCAG expansion, we crossed Msh2 ATPasedefective mice (Msh2 G674A mice) with mice carrying a large unstable CTG expansion (DM300-328) [27,34,35]. The DM300-328 mice carrying .300 CTG repeats in the context of the human genomic DM1 locus of over 45 kb display very high levels of intergenerational and somatic instability similar to that observed in DM1 patients [36]. The Msh2-mutant mice carry a missense mutation of a glycine-to-alanine within the Walker ''type A'' motif GXXXXGKS/T of the MSH2 ATPase domain [27,34] . This ATPase domain in MSH2 and in many proteins with ATPase function, is known to coordinate the phosphate groups of ATP [34]. Mismatch binding experiments using Msh2 G674A/G674A ES cell extracts demonstrated that the mutant MSH2 G674A -MSH6 complex retained normal mismatch binding activity. However, the complex was resistant to ATP-mediated mismatch release and had lost its capacity to signal mismatch repair resulting in MMRdeficiency similar to that observed in Msh2 2/2 null mutant extracts. Moreover, the homozygous ATPase mutation caused a mutator phenotype, increased genome-wide mononucleotide and dinucleotide instability in the genome of mutant mice similar to that observed in Msh2 2/2 mice. However, in contrast to the Msh2null allele, the Msh2 G674A mutation did not significantly affect the cellular response to DNA damage-inducing agents, indicating that normal ATP processing with subsequent repair is not essential for the apoptosis signalling function of MSH2 [34]. Therefore, Msh2 G674A mutant mice appeared to be an excellent model to determine if MSH2 ATPase activity is required for CTGNCAG repeat expansions or if MSH2 binding to the trinucleotide repeats

Author Summary
Myotonic dystrophy type 1 is a neuromuscular disease characterized by highly variable clinical manifestations, including muscular and neuropsychological symptoms. DM1 results from the dramatic expansion of an unstable CTG repeat in the DMPK gene. Longer CTG repeats cause a more severe form of the disease and an earlier age of onset. The DNA mismatch repair proteins MSH2 and MSH3 are known to be major players in the formation of trinucleotide expansions. Nevertheless, the mode of action of these proteins remains elusive. In order to get further insight into the role of MSH2 in the formation of CTG expansions, we used a mouse model carrying a mutation in the conserved ATPase domain of Msh2. This mutation affects the function of this domain and alters the DNA repair mismatch activity. After breeding of these mice with mice carrying highly unstable CTG repeats, we found that the ATPase domain mutation of MSH2 strongly affects the formation of CTG expansions. Our findings show that expansion-biased trinucleotide repeat instability requires a functional MSH2 ATPase domain and support the hypothesis, according to which a functional MMR activity is required to generate expansions.
is sufficient to generate expansions in transgenic mice. We assessed intergenerational repeat instability over successive generations and somatic mosaicism in CTG repeat containing mice deficient for MSH2 ATPase activity. The Msh2 G674A/G674A mice showed a dramatic decrease in CTG repeat expansions and an increase in contractions; a pattern paralleling that observed in Msh2 2/2 transgenic mice. Our data indicate that the ATPase-defective Msh2 G674A mutation affects intergenerational and somatic instability of CTGNCAG repeats in transgenic mice.

Results
To assess the consequence of the G674A Msh2 missense mutation in the generation of expanded trinucleotide repeats, we crossed ATPase-defective Msh2 mutant mice [27,34] with DM300-328 mice carrying the human DMPK gene with .300 CTG repeats, which show high levels of intergenerational and somatic instability that is biased towards expansions and agedependent [36]. We analyzed the length of the CTG repeat inherited from males and females, after weaning, in tail DNA from mice with different Msh2 genotypes, as previously described [36]. The CTG instability in the various transmissions are reported in Table 1 and Figure 1. In Msh2 +/+ transgenic mice, we observed a high level of intergenerational instability biased towards expansions (100% and 62.3% expansions in offspring from male and female transmissions, respectively). In the Msh2 G674A/+ transmissions, the frequency of expansions significantly decreased for both male and female transmissions (from 100% to 75% and from 62.3% to 37.9%, respectively). For both paternal and maternal Msh2 G674A/G674A transmissions, the frequency of expansions dramatically decreased (down to 12.1% and 0% in male and female transmissions, respectively). At the same time, the frequencies of CTG contractions increased in Msh2 G674A/+ and Msh2 G674A/G674A transgenic mice. The distributions of the magnitudes of intergenerational CTG repeat length changes were significantly different between Msh2 +/+ and Msh2 G674A/+ and between Msh2 +/+ and Msh2 G674A/G674A transmissions; p,0.005 and p,0.0001, respectively, both for male and female transmissions (assessed by the Mann-Whitney test). Interestingly, much larger contractions are observed in Msh2 G674A/G674A female transmissions (with a mean of 241.9 CTG units). The sizes of contractions were positively correlated with the age of the transmitting female (r 2 = 0.718, p,0.0001).
In summary, these data support the idea that the G674A mutation of one or both Msh2 alleles significantly decreased the frequency of repeat expansions in paternal and maternal transmissions, suggesting that a functional MSH2 ATPase domain is required to generate expansions. Furthermore, these data indicate that the ATPase-mutant MSH2 is sufficient to lead to a preferential formation of CTG contractions that also arise in the Msh2-null and Msh3-null DM1 mice [19,20]. Together these results support a requirement of a functional MSH2 ATPase domain for the expansion-biased parent to offspring CTG transmissions, and that this defect is sufficient to cause the formation of transmitted CTG contractions, which also arise in the absence of either MSH2 of MSH3 [19,20].
Both MSH2 and MSH3 proteins are required for CTG expansions in somatic tissues [19,20,37], yet the requirement of the MSH2 ATPase domain is unknown. Thus, we analysed the somatic CTG instability in several tissues collected from 8-monthold transgenic mice with different Msh2 genotypes. Wild-type mice show inter-and intra-tissue CTG length mosaicism biased towards expansions, pronounced in pancreas and germinal tissues, as previously observed [19,36] (Figure 2 and data not shown for female mice). A striking change in somatic instability was not readily observed by standard PCR for the Msh2 G674A/+ mice. In contrast, expansions were clearly not detected in Msh2 G674A/G674A tissues and contractions were easily observed in testis and sperm, but not in somatic tissues. Somatic instability was assessed with greater sensitivity using small-pool PCR (SP-PCR). Tissues assessed by SP-PCR include ovaries, testis, brain and cerebellum; tissues known to display a range of instability levels in repairproficient mice. The inherited size in each mouse is represented by the CTG repeat length in blood, liver or in tail at weaning, in which somatic instability is minimal throughout the animal's life [19]. In Msh2 G674A/+ mice, the degree of somatic mosaicism seems to be unchanged in all tissues analysed, and where present showed an expansion bias ( Figure 3). In contrast, the degree of somatic CTG mosaicism was strikingly decreased in all tissues tested from the Msh2 G674A/G674A mice. Neither expansions nor high levels of contractions were evident by either standard PCR or SP-PCR for the ATPase mutant mice, except in brain, where very small expansions could be detected. The absence of expansions in Msh2 G674A/G674A mice parallels their loss in Msh2 2/2 mice. However, the absence of contractions in somatic tissues of Msh2 G674A/G674A mice contrasts with their detection in tissues of Msh2 2/2 mice [19]. These results show that the homozygous Msh2 G674A mutation reduces the formation of CTGNCAG expansions in somatic tissues, while mutation of a single allele has no significant effect. These results also indicate that the Msh2 G674A/G674A mutation does not lead to high levels of CTGNCAG contractions in somatic tissues, which contrasts with their formation in tissues of Msh2 2/2 mice [19]. We have shown that CTG instability is dramatically decreased in Msh2 G674A/G674A , supporting the idea that the G674A mutation in MSH2 ATPase domain affects CTG repeat instability. Lin et al. demonstrated that this mutant ATPase domain alters the MMR pathway in Msh2 G674A/G674A mice without affecting the mismatch binding activity [34]. The stability of MSH2, MSH6 and MSH3 proteins depends on the presence of each partner involved in the formation of the MutSa and MutSb complexes [38,39]. In embryonic fibroblasts from Msh2 G674A/G674A mice, expression levels of MSH2 and MSH6 proteins were normal, suggesting that the ATPase mutation did not alter the stability of these proteins [34]. We assessed the amount of MSH2 protein in germinal and cerebral tissues from 3 to 4-month-old transgenic mice with different Msh2 genotypes, by western blot. We observed a decrease of MSH2 protein in Msh2 G674A/G674A germinal tissues (to 13% in ovaries, to 50% in testis, p,0.001, t-test), cerebellum (to 19% p,0.001, t-test) and brain (to 62%, p = 0.005, t-test) (Figure 4). In mice heterozygous for the G674A mutation, MSH2 protein levels were also decreased only in germinal tissues (to 37% and 80% in ovaries and testis respectively, p,0.001, t-test). No detectable variation in protein levels was observed in brain and cerebellum of heterozygous mice. To ensure that the observed decrease of MSH2 protein was not due to a decreased sensitivity/avidity of the mutant MSH2 to the anti-MSH2 antibody, we also assessed the levels of MSH3 and MSH6 proteins and found these to be decreased in tissues of the Msh2 ATPase-mutant mice with a pattern similar to the decrease of MSH2 ( Figure 5).
To determine if the mutation decreases MSH2 expression at the RNA level, we measured transcript levels in Msh2 G674A/G674A and wild-type tissues by qRT-PCR. Our analysis showed that the quantity of Msh2 G674A transcripts is unchanged in homozygous mutant mice compared to wild-type mice (data not shown). Therefore, our results suggest that the missense mutation in MSH2 ATPase domain probably destabilizes MSH2 protein in adult mice.

Discussion
Since the discovery of dynamic mutations, which are responsible for more than 20 neurological and/or neuromuscular diseases, many studies have focused on understanding the molecular mechanism of instability. Various data have clearly revealed that MSH2, MSH3, MSH6 and PMS2 are required for CTGNCAG expansions in mouse model systems [15][16][17][18][19][20]. However, the mode of action of these proteins in the generation of trinucleotide expansions remains unclear. It has been reported that MSH2-MSH3 ATPase activity was reduced upon binding to CAG hairpins, this was interpreted as impairing the repair of the CAG hairpin permitting its integration as an expansion [18]. In contrast, MutSb was not required for either the correct or protected repair of slipped-CTGNCAG repeats, suggesting that MutSb may be required to form these DNA structures with an unknown involvement of ATP hydrolysis [33]. Nevertheless, the role of the MutL homologue PMS2 in the  formation of somatic expansions suggests that MMR pathway could be involved in the mechanism of instability [17]. In our study, we have shown that the G674A mutation in MSH2 ATPase domain strongly affects the formation of intergenerational expansions. Msh2 G674A/G674A transgenic mice showed a strong bias towards contractions compared to Msh2 +/+ transgenic mice which showed a strong bias towards expansions in their offspring. In the homozygous Msh2 G674A/G674A mice, we observed only 12.1% of expansions for paternal transmissions and no expansions for maternal transmissions (versus 100% and 62.3% in both male and female wild-type transmissions, respectively). These results suggested that a functional MSH2 ATPase domain is necessary to generate intergenerational expansions. These results also suggest that the defective ATPase domain is sufficient to yield a contraction bias similar to the Msh2null mice. In testis, where 89.3% of seminiferous tubules are germinal cells [40], we observed a decrease of MSH2 levels of ,50% in the Msh2 G674A/G674A mice. This coupled with the severe reduction of expansions upon male Msh2 G674A/G674A transmissions (from 100% for Msh2 +/+ down to 12% for the mutant mice) compared to the negligible effect for Msh2 2/+ mice (92% , Table S1 and [19]) strongly argue that a functional ATPase activity is required for expansions. Then, the deficiency of the ATPase domain in conjunction with decreased protein levels are responsible for the decrease of expansion frequency. Furthermore, it demonstrates that the apoptosis signaling function of MSH2, described to be non affected by the G674A mutation [34], is not involved in the mechanism generating CTG repeat expansions.
We also observed that the G674A mutation of one Msh2 allele was sufficient to decrease the frequency of expansions in both male and female transmissions (from 100% to 75% expansions and from 62.3% to 37.9% for male and female transmissions, respectively). Such a decrease in the frequency of expansions was not observed for male transmissions in heterozygous Msh2 2/+ mice which displayed 92% expansions similar to Msh2 +/+ mice; indicating that in this context a single functional Msh2 allele was sufficient to drive the same level of expansions as two functional alleles [19]. In contrast, our results show that a functional MSH2 protein in the presence of an ATPase mutant MSH2 G674A protein was not sufficient to drive the same levels of expansions; but yielded reduced expansion levels for paternal and maternal transmissions -further supporting a requirement of the ATPase domain. To this degree our data suggest that the MSH2 G674A mutant protein could act as a dominant-negative mutant, effectively blocking the ability of the wild-type MSH2 protein from driving CTGNCAG expansions. This purported dominant negative function would be specific to CTGNCAG expansions, as a dominant-negative mutator effect was not observed for either mono-or dinucleotide microsatellites in the heterozygous Msh2 G674A/+ mice [34]. In concordance with our observed CTGNCAG effect, the S. cerevisiae yMsh2 protein with the G693A (equivalent to the mammalian G674A mutation) or G693D mutated ATPase domains cause a dominant mutator phenotype [41,42]. The mutant yMsh2 G693A -yMsh6 complex binds to DNA mismatches and is not released in the presence of ATP. This could block access of wild-type proteins and subsequently inhibit normal repair [43]. In the case of CTGNCAG repeats, the MSH2 G674A mutant protein could partially block the access of the wild-type MSH2-MSH3 complex to unpaired CAG repeats, therefore protecting it from being processed to expansions. There is no data concerning the ability of the MSH2 G674A -MSH3 complex to bind trinucleotide repeats. The analogous mutation in the conserved Walker domain of the prokaryote MutS protein (K620M) results in a weak mismatch binding activity and in a decrease in ATP binding and hydrolysis [24,44]. However, in the case of MutS homoduplex in bacteria, both subunits of the complex are affected by the mutation. In eukaryotes, MSH2 partners (MSH3 or MSH6) retain their capacity to bind and hydrolyse ATP and it has been shown that MSH6 binds ATP with higher affinity and hydrolyses ATP faster than MSH2 [45]. Mispair binding is regulated by mismatchstimulated ADP to ATP exchange inducing conformational changes to the N-terminus mispair binding domain [24,46]. Recent data have shown that human MutSa and MutSb have different ADP binding and ATPase activities in vitro that could explain the preferential processing of base-base and insertion/ deletion mispairs by these two complexes [25]. In the mice used in this study, the MSH2 G674A -MSH6 complex is able to bind to mismatches but the ATP-mediated disassociation from the substrate containing a G/T mismatch is reduced [34]. It is not known if the binding of MSH2 G674A -MSH3 is affected by the mutation. However, the possible dominant-negative function of MSH2 G674A on CTGNCAG expansions suggest that MSH2 G674A -MSH3 may still be able to bind to trinucleotide repeat expansions.
In Msh2 G674A/G674A mice, we observed a dramatic decrease of somatic CTG expansions together with a decrease of MSH2 protein levels in adult somatic tissues, such as brain and cerebellum. Because the stability of MSH2, MSH3 and MSH6 depends on the ability of these proteins to form heterodimers, the levels of MSH3 and MSH6 were also found decreased in mutant tissues. However, MSH3 and MSH6 were clearly detected in homozygous mutant mice showing that the mutant MSH2 G674A protein was still able to form the MutSb and MutSa complexes. While this reduction varied between tissues, a role of the ATPase domain in somatic instability seemed to be implicated. The dramatic loss of CTG instability in Msh2 G674A/G674A mouse brains, which show as much as 62% of MSH2 protein strongly suggests that a functional ATPase domain is required for the formation of somatic expansions. The decrease of protein levels in the germinal and somatic tissues of Msh2 G674A/G674A mice was unexpected as no decrease was observed in MEF cells from these mice [34]. Furthermore, Ollila et al recently observed that the production of human recombinant MSH2 G674A -MSH6 complex was expressed in amount similar to the wild-type hMutSa suggesting that the G674A mutation did not affect protein stability [47]. However, a lowered interaction of MSH2 G674A with MSH6 was reported in this study. Interestingly, the difference in hMutSa expression levels between various cell lines was shown to result from cell-dependent differences in the degradation rate of both proteins forming the complex, by the ubiquitin-proteasome pathway [48]. Furthermore, the protein kinase PKCf is able to phosphorylate hMutSa and regulates MSH2 and MSH6 protein stability and levels by inhibiting the ubiquitin-proteasome dependent degradation of these proteins [49]. It is unknown if the stability of MSH3 is regulated by the same pathway but the regulation of MSH2 stability has a direct impact on the regulation of MutSb. It is possible that the Msh2 G674A mutation affects the sensitivity of MSH2 to the ubiquitin-proteasome degradation pathways possibly by altering its ability to be phosphorylated. In this manner a tissuespecificity in the efficiency of the ubiquitin-proteasome degradation pathways would account for the different decrease of MSH2 G674A protein that we observed in the various mouse tissues. Interestingly, we found a putative PKC phosphorylation site at threonine 677; proximal to the G674A mutation, in the human and mouse MSH2 proteins using the NetPhosK 1.0 server (output score 0.89 using the Evolutionary Stable Sites filter) [50]. While tissue-specific regulation of MSH2 protein levels is beyond the scope of the present study, it is clear that the altered CTG instability we observe with the Msh2 G674A mutation indicates that an ATPase-competent MSH2 protein is necessary to drive CTGNCAG expansions and that in its absence contractions predominate in parent-to-offspring transmissions.
It is noteworthy, that the levels of CTGNCAG contractions in either Msh2 G674A/G674A , or Msh3-null mice is considerably greater in parent-offspring transmissions than in their somatic tissues ( [20] and this study). The formation of high levels of contractions of expanded CTGNCAG repeats transmitted from parent-to-offspring depends upon the homozygous loss of Msh2 or Msh3, or a homozygous Msh2 G674A/G674A mutation, and occur to some degree in a heterozygous Msh2 G674A/+ mutant. One interpretation is that the contractions would normally have been prevented by the presence of MSH3 and MSH2 with a functional ATPase domain. It would be of interest to learn if the germline-specific MSH2-MSH3:MLH3-MLH1/PMS2 complexes, detected at centromeric and Y chromosome repeats and proposed to prevent deleterious crossover events at repetitive DNAs [28], also form at expanded CTGNCAG transgenes in germ cells of transgenic mice. They may be linked to transmitted repeat contractions.
In conclusion, our finding that a functional MSH2 ATPase domain is required for CTGNCAG expansion mutations makes this the third mutagenic process for which its activity is required. The ATPase activity of MSH2 has been previously demonstrated to be required for both somatic hypermutation and class switch recombination at immunoglobin genes [26,27]. Our data demonstrate that the Msh2 G674A mutation affects CTGNCAG repeat instability in transgenic mice and almost completely abolishes the formation of intergenerational and somatic expansions and leads to a high level of contractions in parental transmissions. The G674A mutation also decreases the levels of MSH2 protein in a tissue-specific manner. The decrease of MSH2 protein cannot account totally for the dramatic decrease of expansions. Therefore, CTGNCAG repeat expansions are likely affected by cumulative effects of its defective ATPase domain and reduced MSH2 protein levels. If we assume that the binding of MSH2 G674A -MSH3 is not affected by the mutation, our data would suggest that the binding of MutS complexes to the trinucleotide repeats is not sufficient to drive instability towards expansion and that a fully functional MSH2 ATPase domain is required. This does not support the hypothesis that a decrease of ATP hydrolysis by MSH2-MSH3 bound to trinucleotide repeats and aborted repair of DNA hairpin loops could cause expansions. On the contrary, our data sustain the hypothesis according to which a functional MMR activity is required to generate expansions.

CTG Repeat Analysis
Intergenerational instability was investigated by comparison of the CTG repeat lengths in transgenic parent and offspring. DNA was extracted from tail collected at weaning by isopropanol precipitation. To determine the CTG repeat length in transgenic mice, 15 ng of tail DNA samples were amplified in 25 ml reaction using 0.4 mM 101 and 102 primers [36], 16 Custom master mix (Thermo Scientific) and 0.04U Thermoperfect Taq polymerase (Integro BV). The following cycling conditions were used: 5 min at 96uC; 45 s at 96uC, 30 s at 68uC and 3 min at 72uC (30 cycles); 1 min at 68uC and 10 min at 72uC (1 cycle). Electrophoresis of PCR products was performed using a previously described method [52]. The amplified product (from 2 ml to 5 ml) was mixed with DNA-loading dye and subjected to electrophoresis in 0.7-0.8% agarose gels at 300 V for 30 min and then 160 V overnight at 4uC. After electrophoresis, the gel was incubated in depurinating solution (0.25 M HCl) for 10 min, denaturing solution (0.5 M NaOH, 1.5 M NaCl) for 30 min and neutralization solution (1.5 M NaCl, 0.5 M Tris-HCl, pH 7.5) for 30 min. DNA was blotted onto positive nylon membrane (MP Biomedicals), UV-crosslinked (UV Stratalinker 2400, Stratagene) and hybridised with a double-strand CTG DNA probe radiolabelled with a-32 P-dCTP using Amersham ready-to-go TM DNA labelling beads (-dCTG) (GE Healthcare). CTG repeat lengths were compared with a 250 bp DNA ladder loaded on the same gel and analysed by Quantity One-1-D Analysis Software (Bio-Rad Laboratories). Somatic instability was analysed in various tissues collected from 8-month-old mice. DNA was phenol/chloroform-extracted from tissues [14] and sperm DNA was extracted according to Seznec et al. [36]. CTG repeat instability was determined as described above.

Small-Pool PCR (SP-PCR) Analysis
DNA extraction, SP-PCR amplifications and PCR product electrophoretic analyses were performed using the methods previously described [52]. DNA samples were digested with HindIII and SP-PCR was performed with DM-C and DM-BR primers [52]. SP-PCR products from mice with different Msh2 genotypes and tail/blood or tail/liver PCR products were loaded on the same 0.7% agarose gel to compare instability between Msh2 +/+ , Msh2 G674A/+ and Msh2 G674A/G674A mice.

Protein Sample Preparation and Western Blotting
The different tissues were collected from 3 to 4-month-old-mice with different Msh2 genotypes and pooled for protein extraction (three mice for each genotype). Proteins were extracted by mechanical homogenisation in lysis buffer (0.125 M Tris-HCl pH 6.8, 4% SDS, 10% glycerol) containing complete Mini 76 protease inhibitor cocktail (Roche) and 1 mM PMSF (Sigma).
Western blotting was reproduced at least three times for each tissue to make semi-quantitative analysis. Densitometric analysis of protein levels was performed using Quantity One-1-D Analysis Software (Bio-Rad Laboratories) using non-saturated exposures.

Statistical Analysis
Statistical analyses for instability distributions were performed with StatView software using the Mann-Whitney test (SAS Institute Inc.). The statistical analysis of the levels of MSH2 protein was performed with StatView software using t-test. The detected differences in total transmissions and in MSH2 protein levels were considered statistically significant only if p,0.05.