The ATPase activity of MLH1 is required to orchestrate DNA double-strand breaks and end processing during class switch recombination

A highly conserved GxG sequence motif (Fig. 1 A) is located in the loops defining the boundary of the ATP binding pocket of bacterial MutL (Dutta and Inouye, 2000). The mammalian MLH1 ATP binding pocket behaves similarly in a crystal structure resolved at 2.5 Å (Protein Data Bank accession no. 3NA3; Fig. 1 A). Accordingly, the GxG motif is in a polypeptide stretch devoid of secondary structures, suggesting that a G67R mutation should not significantly affect the overall integrity of the ATPase or the adjacent MutS interaction domains. However, G67 appears to be in close proximity to the adenine base of the ATP molecule (Fig. 1 A), suggesting that a G67R mutation could greatly alter the bulkiness and charge of the ATP pocket, thereby hindering the accommodation of the ATP molecule.

ATP binding and/or hydrolysis by the ATPase of MLH1 is believed to induce large conformational changes (Sacho et al., 2008; Johnson et al., 2010) capable of mediating an MMR signaling cascade by recruiting and activating downstream effector proteins after MutSα binding (Ban et al., 1999; Dutta and Inouye, 2000; Spampinato and Modrich, 2000; Tomer et al., 2002; Chahwan et al., 2011). Although MLH1 and its ATPase activity are required for efficient DNA MMR in vitro and in vivo (Räschle et al., 2002; Tomer et al., 2002; Jiricny, 2006; Avdievich et al., 2008), there has been some uncertainty about the role of MLH1 in the resolution of the AID-generated mismatches during SHM. In fact, the frequency and characteristics of mutations in MLH1-deficient mice were previously shown to be altered in an artificial exogenous transgene but normal when an endogenous Ig variable gene was studied in knockout mice (Kim et al., 1999; Phung et al., 1999). To determine whether the continued presence of the protein but the loss of the ATPase activity of MLH1 had any effect on SHM, three WT and three Mlh1^GR/GR mice were immunized with NP₃₆-CGG to promote SHM in the rearranged V_H186.2 gene. We found no significant differences in the overall mutation frequencies between WT and Mlh1^GR/GR mice (P = 0.4252; Fig. 1 B), and the distribution of mutations was similar (P ≥ 0.0576; Fig. 1 C). Although there was a slight tendency for more G:C mutations in mutant mice, the number of transversions or transitions (P ≥ 0.2652), the different types of substitutions (P ≥ 0.0576), and the cumulative spectrum of mutations from G, C, A, or T sites (P ≥ 0.2183) were also comparable between WT and mutant mice (Fig. 1, B and D). Furthermore, the overall mutation frequencies of the hotspots for AID (WRC/GYW) and Polη (WA/TW), which are major hallmarks of the SHM process, were similar in both groups (P ≥ 0.5287; Fig. 1 B). Finally, the strand bias signatures previously reported in antigen-selected V(D)Js (Steele, 2009; Roa et al., 2010; van Oers et al., 2010), which show WA>TW and GYW>WRC biased frequencies, were similarly evident in both WT and Mlh1^GR/GR mice (Fig. 1, B and D). To determine whether selection pressures on the V_H186.2 coding region concealed hypermutation differences, we compiled unique mutations within the intronic Jh2-Jh4 region of splenic B cells from unimmunized mice. As expected, Jh2-Jh4 sequences yielded fewer mutations than V_H186.2 sequences (Fig. 1 E), but the overall mutation frequencies and the relative spectrum of mutations in both scenarios were comparable between the mutant and WT mice (Fig. 1, E and F). Fig. 1 F also shows a difference in G:C/A:T ratios between V_H186.2 and Jh2-Jh4 regions, an observation which is not completely understood yet but has been connected before to intrinsic differences between antigen and nonantigen selected regions (Steele, 2009).

Other MMR factors, including MutSα, the exonuclease EXO1, the ubiquitylated form of PCNA, and error-prone polymerases such as Polη, have been clearly shown to behave as error-prone factors at the Ig locus, modifying the spectrum of substitutions and introducing mutations at A:T sites within the V (variable) region (Chahwan et al., 2011). However, our data here on the catalytic mutant MLH1 expand previous findings that show little or no effect of MLH1 (Kim et al., 1999; Phung et al., 1999) or PMS2 deficiency (Péron et al., 2008; van Oers et al., 2010) in SHM of B cells. The reason why the heterodimer MLH1–PMS2 collaborates with MSH2/MSH6/PCNA/EXO1/Polη outside the V region and is essential in global MMR but seems to be irrelevant for SHM remains unknown. It is likely that there is some mechanism to prevent MutLα from being recruited by the MutSα complex or becoming active at the V region so as to prevent it signaling for additional downstream elements and reduce the likelihood of DSBs.

In contrast to the lack of a significant effect of the loss of MLH1 and of its ATPase activity on SHM, Mlh1^−/− mice have a substantial decrease in CSR (Schrader et al., 1999; Stavnezer et al., 2010; Eccleston et al., 2011). Because this decrease in CSR could be caused by the absence of the MLH1–PMS2 complex, including its scaffolding functions, or by the loss of the ATPase activity of MLH1, we examined whether the selective inactivation of the ATPase activity of MLH1 would affect CSR. We purified splenic B cells from mutant and WT mice and stimulated them with LPS or with LPS + IL4 to induce ex vivo switching from IgM to IgG3 and IgG1, respectively. After normalization of the efficiencies of switching to WT levels, Mlh1^GR/GR mice showed an ∼70% reduction in CSR to all studied isotypes (P < 0.001; Fig. 2, A and B). No appreciable differences in proliferation that might delay CSR were detected between WT and Mlh1^GR/GR B cells, as assessed by CFSE dilution histograms (Fig. 2 C).

A similar phenotype has been shown in the absence of PMS2, an essential partner of MLH1 during CSR and postreplicative MMR (Schrader et al., 1999; Ehrenstein et al., 2001; Péron et al., 2008; van Oers et al., 2010). These reduced efficiencies in CSR have been associated with impaired generation and resolution through NHEJ of DSBs, both of which are critical for CSR (Stavnezer et al., 2010). However, little is known about the functional domains of MLH1 and the role, if any, of its distinct catalytic activities during CSR. Based on the deficiency in CSR exhibited here by Mlh1^GR/GR B cells, it became plausible to explore the hypothesis that the ATPase activity of MLH1 might play a role in the generation and/or resolution of the DSBs produced during CSR.

Because MLH1–PMS2 ATPase activities are thought to be required for the activation of the MutL endonuclease function (Pillon et al., 2010; Polosina and Cupples, 2010) leading to DSBs in SRs (van Oers et al., 2010), we determined whether DSB generation during B cell activation was affected in Mlh1^GR/GR mice. To examine whether there was a specific decrease of DSBs at SRs, chromatin immunoprecipitation (ChIP) experiments using anti-53BP1 antibodies were conducted on activated B cells from WT, Mlh1^−/−, and Mlh1^GR/GR mice. 53BP1 has been shown to be essential for CSR and the mediation of DNA end joining (Ramachandran et al., 2010; Bothmer et al., 2011), and therefore the recruitment of 53BP1 to the targeted SRs serves as a reporter for the presence of DSBs. As expected, 53BP1 protein was enriched approximately two- to fourfold on the Sμ and Sγ1 region in LPS + IL4–stimulated WT B cells compared with the respective Mlh1^−/− or Mlh1^GR/GR (Fig. 3 A). We did not find a significant increase on the Sγ3 region, which switches infrequently after LPS + IL4 treatment (Stavnezer et al., 2010), suggesting that the observed 53BP1 enrichment was specific. Furthermore, we used Western blots to examine the levels of H2AX that undergo rapid phosphorylation (γ-H2AX) in response to the generation of DSBs in the stimulated B cells (Rogakou et al., 1999). As expected, all cells stimulated with LPS or with LPS + IL4 had considerably higher γ-H2AX signal compared with nonstimulated cells (Fig. 3 B). However, the increase in the overall amount of γ-H2AX was much lower in activated Mlh1^GR/GR splenic B cells than in WT cells (Fig. 3 B), also suggesting a smaller accumulation of DSBs in the mutant cells.

Collectively, these data suggest that there is a decrease in the formation of DSBs in the relevant SRs in the MLH1-ATPase mutant B cells that is comparable with that found in mice lacking MLH1. This is consistent with reports that MMR factors are upstream of IgH DSB formation (Schrader et al., 2007; Péron et al., 2008). Our observations further suggest that at least some of the DSBs generated during CSR are orchestrated by MLH1 through its ATPase activity. One possibility is that the ATPase-dependent conformational changes of MLH1 are important to activate PMS2 endonuclease (Pillon et al., 2010; Polosina and Cupples, 2010), allowing MutLα to introduce additional nicks into the SRs and thus facilitating the generation of DSBs in an attempt to repair the recurrent AID-induced mismatches. In fact, this is consistent with our recent report that the endonuclease activity of PMS2 is a major function of the MutLα complex during CSR (van Oers et al., 2010).

The lack of MMR function and the reduction of DSBs have been attributed to delayed recombination efficiencies and differences in the length of microhomologies detected at the switch junctions (Lieber, 2010; Stavnezer et al., 2010). The use of long microhomologies during CSR has been interpreted to suggest that the alternate pathway of NHEJ (alt-NHEJ) is being used, and so the analysis of S-S junctions has been proven to be very informative of the role of end-joining factors (Yan et al., 2007; Kotnis et al., 2009). To determine the nature of the DSB repair pathway or pathways during CSR in activated Mlh1^GR/GR B cells, we analyzed recombined switch junctions. We activated control and Mlh1^GR/GR splenic B cells with LPS for 96 h and then amplified and sequenced Sμ-Sγ3 junctions. The distribution of breakpoints within the SRs was similar when distances up to 1 kb from the edges of the Sμ-Sγ3 amplicon were explored (Fig. 4 A). In contrast, comparison of the actual recombination sites between the Sμ and Sγ3 regions from mutant B cells to those of controls revealed profound differences. WT sequences contained ∼20% direct (blunt) junctions and only a few insertions, with most of the remainder being short microhomology junctions (1.64 bp mean; Fig. 4, B and C). In contrast, only 5% of direct junctions and also few insertions were found in Mlh1^GR/GR B cells, which showed more junctions with longer microhomologies (5.45 bp mean) than the WT cells (P < 0.0001; Fig. 4, B and C).

Our finding that sequence microhomology at Sμ-Sγ3 junctions is greater in Mlh1^GR/GR switching B cells than in WT suggests a deficiency in canonical NHEJ in MLH1 mutant B cells and provides additional information regarding the role of MLH1 during the DNA end processing of SRs. The use of long microhomologies is unusual in canonical NHEJ at CSR junctions and suggests the participation of alt-NHEJ (Yan et al., 2007; Kotnis et al., 2009). This has been suggested to be the result of additional end processing of DSBs or the impaired recruitment of NHEJ proteins in the absence of MLH1–PMS2 proteins (Stavnezer et al., 2010). In fact, DNA helicases and other DNA repair factors have been shown to be part of the MutLα interactome (Cannavo et al., 2007) and may require the ATPase activity of MutLα for recruitment to the complex. More specifically, these interactome studies have identified EXO1 and DNA-PKcs as partners of MLH1, both of which have been shown to function during CSR in B cells (Stavnezer et al., 2010). EXO1 is not only an exonuclease member of the MMR complex but has also been shown to be involved in DNA end resection downstream of a DSB, whereas DNA-PKcs is a kinase protein integral for the mediation of DSB repair using the canonical NHEJ pathway (Kotnis et al., 2009; Lieber, 2010; Chahwan et al., 2011). During MMR, MutLα plays a role in the termination of EXO1-mediated strand excision by RPA (Genschel and Modrich, 2009). It is possible that the defect in the ATPase activity of MutLα interferes with this activity also during DNA end processing of SRs, which might lead to longer excision tracts and increased use of microhomologies during NHEJ.

Recombinant junctions usually accumulate mutations at the donor Sμ and acceptor Sγ regions (Stavnezer et al., 2010). To test whether the switching deficiencies observed in Mlh1^GR/GR mice were accompanied by any alteration in SR mutations, we sequenced Sμ-Sγ3 junctions from LPS-activated mutant B cells. The comparison with WT sequences showed a significant increase in the overall frequency of mutations accumulated at the Sμ-Sγ3 junctions of Mlh1^GR/GR B cells (P < 0.01; Fig. 4 D). This approximately twofold increase affected all types of mutations and hotspots in Mlh1^GR/GR B cells (Fig. 4 D), but it did not affect the distribution of mutations around the recombination breaks, which was comparable in both mutant and WT sequences (Fig. 4 E).

The fact that Mlh1^GR/GR B cells accumulate higher frequencies of mutations at the SRs seems consistent with the notion that deficiencies in MLH1 compromise the genome integrity of the cell and the efficiency of MMR mechanisms (Kunkel and Erie, 2005; Iyer et al., 2006; Jiricny, 2006). Because the frequency of switching was reduced to ∼30% that of WT cells (Fig. 2), it is possible that the defect in MLH1 slowed the recombination processes in most SRs and led to futile cycles of repair, which would allow the accumulation of additional AID hits and error-prone repair in a continuous attempt to progress through CSR.

Supported by all of these arguments, we therefore propose that the ATPase domain of MLH1 appears to be functional both upstream by promoting the generation of DSBs and downstream of the AID-generated mismatches by recruiting NHEJ factors and/or limiting EXO1-mediated resection. Our results cannot distinguish between these two latter alternatives but strongly suggest that the orchestration of the end-repair mechanisms downstream of the DSB require not only the scaffolding functions of MLH1 and PMS2 but also the ATPase activity of MLH1. It should be noted that the loss of the endonuclease activity of PMS2 did not affect the preferential use of canonical NHEJ (van Oers et al., 2010), whereas alt-NHEJ with its longer microhomologies appears to be favored in the Mlh1^GR/GR B cells studied here. Therefore, the MLH1GR mutation does not only result in the loss of activation of the latent endonuclease of PMS2 but also must result in the loss of additional functions of MLH1.

It is becoming clearer that MMR factors can have separate scaffolding and catalytic functions, which could explain the distinct roles of these proteins during MMR processes. In this sense, mouse models representing separation of function mutations have proven to be useful in exploring the multiple roles of MMR protein complexes during error-free DNA repair, error-prone SHM, or DSB-mediated CSR (Chahwan et al., 2011). The role of the endonuclease activity of PMS2 during CSR was separated from the scaffolding functions of MutLα in the Pms2^EK/EK mice (van Oers et al., 2010). Here, we suggest that MLH1 has ATPase-dependent functions that influence both the generation and the NHEJ repair of DSBs at the SRs. In contrast, other ATPase-independent functions of MLH1, such as those involved in the apoptotic response to DNA damage, are retained in the mutant MLH1^GR/GR but not in the null mice (Avdievich et al., 2008). Therefore, some MLH1 functions are independent of its ATPase activity and seem to be irrelevant to CSR because they are not sufficient to enable moderate or normal efficiencies of switching in Mlh1^GR/GR mice.

The Mlh1^−/− mice were generated previously by deleting exon 2 of the Mlh1 gene (Edelmann et al., 1996). The Mlh1^G67R/G67R mice were generated more recently and carry a missense mutation within exon 2 of the Mlh1 gene (Avdievich et al., 2008). The G67R mutation slightly reduced the levels of the mutant MLH1 protein, but this effect was very modest and did not affect the interaction of MLH1 with its MutLα partner PMS2 (Avdievich et al., 2008). All mouse experiments were approved by the Albert Einstein College of Medicine Animal Use Committee.

To analyze the V_H186.2 region, three 8-wk-old Mlh1^G67R/G67R mice along with three WT littermates were immunized intraperitoneally with 200 µg NP₃₆-CGG (Biosearch Technologies) in alum (Thermo Fisher Scientific). To analyze the Jh2-Jh4 region, two aged mice (>8 mo) per genotype were used. In both cases, splenic B cells were purified to analyze SHM as previously described (Roa et al., 2010). Alignment and analysis of unique mutations were performed using SeqMan (DNASTAR) and SHMTool (MacCarthy et al., 2009). As consensus sequences, GenBank/EMBL/DDBJ no. J00530.1 (nucleotides 224–496) for V_H186.2, GenBank no. NT_039551.8 (nucleotides 25620771–25619943) for Jh2-Jh4, and GenBank no. NT_039551.8 (nucleotides 25521839–25521801) for the rearranged IgG1 constant region were used. The presence of significant A>>T and WA>>TW bias in the WT sequences was used as a quality assessment of mutation datasets (Steele, 2009).

After depletion of T cells by complement-mediated lysis, splenic B cells were obtained from six WT and seven Mlh1^G67R/G67R 8-wk-old mice. Purified B cells were plated at a concentration of 0.5 × 10⁶ cells/ml and stimulated with either 50 µg/ml LPS (Sigma-Aldrich) or 50 µg/ml LPS and 50 ng/ml rIL-4 (R&D Systems) for 4 d. Simultaneously, purified splenic B cells were stained with 10 µM CFSE (Invitrogen) in PBS/0.1% BSA at a final concentration of 10⁷ cells/ml. At various time points after stimulation, cells were harvested and fixed with 1% paraformaldehyde in PBS to evaluate cellular proliferation. Surface IgM, IgG1, and IgG3 and intracellular CFSE were analyzed with a FACSCalibur (BD) and FlowJo software (Tree Star).

Genomic DNA was prepared from splenic B cells cultured for 4 d with 50 µg/ml LPS to induce specific switching from IgM to IgG3. Junctional Sμ-Sγ3 DNA was amplified using high-fidelity PfuTurbo DNA polymerase (Agilent Technologies) as previously described (van Oers et al., 2010). Alignment and analysis of sequenced switch junctions were performed using BLAST2seq (NCBI) with the low-complexity filter disabled. Sequences from GenBank accession no. NC_000078.6 were used as SR consensus sequences. Microhomology was defined as matched successive nucleotides that cannot be uniquely assigned to either germline Sμ or Sγ3 consensus sequences.

For Western blot analysis, 1.0 × 10⁶ cells were lysed in Laemmli sample buffer followed by DNA shearing using a syringe. After boiling for 5 min, samples were loaded on 4–20% polyacrylamide gels followed by immunoblotting according to the manufacturer’s specification as described previously (Ramachandran et al., 2010). ChIP was performed according to published protocols (Kuang et al., 2009). In brief, sonicated chromatin from 2.0 × 10⁶ purified splenic B cells was preincubated for 2 h at 4°C with Protein A–agarose beads (Roche) to avoid nonspecific binding. After centrifugation at maximum speed at 4°C for 5 min, the supernatant was incubated overnight at 4°C with 2 µg 53BP1 antisera (Novus Biologicals). The mixture was then centrifuged as in the preceding step, and chromatin–antibody complexes were collected by incubation with Protein A–agarose beads for 90 min at 4°C. Washes and isolation of DNA, as well as quantitative PCR conditions and primers are described previously (Kuang et al., 2009).

This work was supported by Postdoctoral Fellowship EX-2006-0732 from the Spanish Ministry of Education and Science (to S. Roa) and National Institutes of Health grants R01CA72649 and R01CA102705 (to M.D. Scharff) and R01CA76329 and R01CA93484 (to W. Edelmann). M.D. Scharff is supported by the Harry Eagle Chair provided by the National Women’s Division of the Albert Einstein College of Medicine.

The authors have no conflicting financial interests.