Different Mismatch Repair Deficiencies All Have the Same Effects on Somatic Hypermutation: Intact Primary Mechanism Accompanied by Secondary Modifications

Mlh1 and Pms2 knockout mice were provided by Dr. R.M. Liskay (Oregon Health Sciences University, Portland, OR). Heterozygote males and females were bred and offspring were genotyped using a PCR-based method as previously described 17,18. The mice were kept in a conventional mouse colony and are summarized in Table. For the study of splenic GC B cells, mice were immunized twice by intraperitoneal injection of 2 × 10⁸ SRBCs (ICN Biomedicals) on days 1 and 8 and killed 7 d after the second immunization.

Single cell suspensions from spleens or Peyer's patches were stained with PE-conjugated anti-B220 (CD45R) antibody (GIBCO BRL), FITC-conjugated PNA (Sigma Chemical Co.), and FITC-conjugated GL7 antibody (PharMingen). B220⁺/(PNA GL7)^high cells were isolated using FACStar^plus (Becton Dickinson). For the study of the Vk167/PEPS transgene in splenic GC B cells, the depletion of subsets of cells was carried out before FACS^® as described by Jacobs et al. 13 except for the following changes. T cells (Thy1.2⁺), B1 cells (CD5⁺), macrophages (Mac3⁺), and virgin B cells (IgM⁺, IgD⁺) were depleted from the splenic cells by the magnetic-activated cell sorting method. A single cell suspension was first incubated with biotinylated antibodies to Thy1.2, CD5, Mac3, sIgM, or sIgD (PharMingen) and then incubated with streptavidin-coated magnetic beads (Dynal). Cells bound by beads were depleted by a strong magnet and cells free of beads were used in GC B cell isolation as described above.

The synthesis of cDNA was carried out as suggested in the Superscript II kit (Stratagene) using oligo d(T)16 primer (PE Applied Biosystems). Amplification of the VH11 gene from the cDNA was carried out as previously described by Rogerson 23.

560-bp fragments of the Vk167/PEPS transgene containing EPS sequences were amplified with Pfu DNA polymerase (Stratagene) and 5′ primer Vk3B (5′ GTCAGTGGGGATATTGTGATAACC) and 3′ primer EKVK2 (5′ TCAACTGATAATGAGCCCTC). Both primers were phosphorylated in reactions containing 1 mM ATP and T4 polynucleotide kinase (New England Biolabs) before the PCR reaction. PCR conditions were 30 cycles of 94°C for 15 s, 62°C for 20 s, and 72°C for 30 s.

All PCR-generated fragments were cloned by blunt end ligation into the Srf1 site of pCRscript (Stratagene) or the Sma1 site of pKS(+). Ligation reaction mixtures contained T4 DNA ligase and either SrfI or SmaI (New England Biolabs) in order to minimize vector religation.

Analysis of the EPS in Vk167/PEPS transgene has been previously described 24. In brief, each plasmid clone with the 560-bp fragment insert of the transgene was amplified using primers Vk8B (5′ GTTTCAGCTCCAGCTTG) and Vk9B (5′ CTCCTCAGCTCCTGATC) (35 cycles of 94°C for 20 s, 60°C for 30 s, and 72°C for 30 s). The PCR product was digested with either EcoRV or PvuII (New England Biolabs) and separated on a nondenaturing 18% polyacrylamide/5% glycerol gel and visualized by ethidium bromide staining. Mutation at one of the restriction sites results in disappearance of smaller bands and/or appearance of larger bands (see Fig. 2 B).

Single strand conformation polymorphism (SSCP) analysis was carried out according to Orita et al. 25 with the following modifications in the protocol: plasmid DNA was amplified with Vk3B and EkVk2 primers in 10 μl reactions containing PFU DNA polymerase, 2 mM of each dNTP, and 1 μCi of [α-³²P]dCTP. PCR conditions were as described above. PCR products were digested with restriction endonuclease TaqαI (New England Biolabs) at 65°C, which results in two fragments of 300 and 260 bp, respectively. Digested PCR products were diluted with dilution solution (10 mM EDTA, 0.1% SDS) and stop solution (95% formamide, 20 mM NaOH, 20 mM EDTA, 0.15% bromophenol blue, and 0.15% xylene cyanol). The mixture was denatured by incubation at 98°C for 10 min and quick chill on ice. Resolution was carried out by electrophoresis through a 6% polyacrylamide/10% glycerol gel at 6 W for 17–19 h. The gel was dried and exposed to Kodak X OMAT film (Eastman-Kodak Co.).

Sequencing was carried out either manually using the Sequenase II Dideoxy Terminator Sequencing kit (United States Biochemical Corp.) or by University of Chicago Cancer Research Center DNA Sequencing Facility using ABI Prism model 377 sequencer (Applied Biosystems, Inc.). The sequencing primers T7, T3, or reverse primer were used. The reverse primer (5′ GGAAACAGCTATGACCATG 3′) is a primer just upstream of the T3 primer in the multiple cloning site of pKSII(+) or pCRSCRIPT.

Oligonucleotide primers (D4Mit42F and D4Mit42R) specific to simple sequence repeat (SSR) locus D4Mit42 were obtained from Research Genetics. Before the PCR reaction, the D4Mit42F primer was phosphorylated with T4 Polynucleotide Kinase (New England Biolabs) and [γ-³²P] ATP. PCR reaction contained 1 pmol each of labeled D4Mit42F primer and unlabeled D4Mit42R primer, 0.2 mM of each dNTP, 0.25 U of Taq polymerase, and 20–40 pg of DNA in a total vol of 10 μl. The PCR conditions were 32 cycles of 15 s at 94°C, 30 s at 55°C, and 30 s at 72°C. The products were separated on a 6% polyacrylamide denaturing sequencing gel and exposed to phosphor screen (Molecular Dynamics). The image was visualized by scanning with a Storm 860 scanner and analyzed by ImageQuant software (Molecular Dynamics).

Somatic hypermutation of an endogenous Ig gene was studied in knockout mice twice immunized with SRBCs. B220⁺PNA^hi B cells, which represent the GC B cell compartment, were isolated from spleens of Mlh1^−/− or Pms2^−/− mice. RNA was extracted from these cell populations and the VH11 gene, which is a member of the S107 family of heavy chain variable genes, was amplified by the reverse transcription PCR method and cloned for sequencing. In addition to the S107 family gene–specific primer, a primer specific to the γ constant region was used in amplification so that only the VH11 gene sequences from activated B cells that have switched to the γ constant region would be represented. Sequence analysis showed normal frequencies of somatic hypermutation in Pms2^−/− as well as in Mlh1^−/− mice (Fig. 1). The frequencies of somatic hypermutation in the clones analyzed were 0.7% in Pms2^−/− mice and 1% in Mlh1^−/− mice. Comparable mutation frequencies were previously reported for VHS107 genes in GC B lymphocytes of wild-type mice immunized in the same way 23. Incidentally, five out of six Mlh1^−/− clones sequenced were potentially related clones containing the same VDJ junction and sharing some identical mutations. As a control, the VH11 gene was amplified from a B220⁺PNA^lo population using VH11 and μ constant region–specific primers and no mutation was found in five sequenced clones (data not shown).

By analysis of VH11 gene clones, we have confirmed that the disruption of the Mlh1 gene or of the Pms2 gene does not have a significant effect on the frequency of Ig somatic mutation. To address the issue of altered mutation spectra without the complication of selection, we decided to pursue the analysis of a passenger transgene in the MMR-deficient strains of mice. The Vk167/PEPS transgene contains an insertion of a 108-bp artificial sequence within a rearranged κ light chain transgene (Fig. 2 A). The insertion, termed EPS, contains multiple EcoRV and PvuII sites that allow the detection of mutation by restriction fragment length variance. Of 108 bp in EPS, 76 bp lie in EcoRV or PvuII recognition sites. Mutation at any one of these positions results in the loss of the particular restriction site (Fig. 2 B). Previous study of one transgenic mouse line, PEPS4, which carries four copies of the Vk167/PEPS transgene, showed that this transgene can be targeted for Ig somatic hypermutation 24,26. Furthermore, the 108-bp EPS portion of the transgene showed a mutation frequency ∼10 times higher than the rest of the transgene sequences analyzed (Table) 24,26.

Mice heterozygous for Mlh1 or Pms2 disruption were bred with the PEPS4 line of transgenic mice. Mice homozygous for the disrupted allele and positive for the transgene were immunized twice with SRBCs and killed 7 d after the second immunization for harvest of the spleens. We enriched for activated GC B lymphocytes from splenic cells by depleting T cells, B1 cells, macrophages, and virgin B cells. Activated GC B lymphocytes were isolated from this enriched population by fluorescence-activated cell sorting. A part of the transgene was amplified with Pfu high fidelity DNA polymerase from the DNA of these GC B cells and cloned for analysis. Identification of mutations in the EPS was carried out as previously described. Approximately 1 out of 30 clones analyzed was identified to have mutations in one of the restriction sites. Subsequently, ∼560 bp of each transgene clone including EPS and flanking Ig gene sequences were sequenced. A total of 26 clones were sequenced from Mlh1^−/− mice and 15 clones from Pms2^−/− mice with total numbers of mutations identified being 41 and 27 respectively (Fig. 3 and Table). Mutation frequencies were 2.9 × 10⁻³ for Mlh1^−/− mice and 3.3 × 10⁻³ for Pms2^−/− mice. These are approximately twofold lower than 4.7 × 10⁻³, which is the frequency we observed in wild-type mice in a previous study (P < 0.0005 for Mlh1^−/− and Pms2^−/−) 26. We did not find increased numbers of tandem mutations in Pms2^−/− mice in contrast to the report by Winter et al. 14. The proportion of mutations from G and C nucleotides is significantly greater in Mlh1^−/− and Pms2^−/− mice than in wild-type mice. Increased mutations from G and C nucleotides have been reported in Msh2^−/− mice 13,20 but in analogous studies comparable alterations are not found in Pms2^−/− or Mlh1 mice 14,27. This discrepancy is unexplained. Our analysis of Mlh1^−/− and Pms2^−/− mice indicates that the increase in mutations from G and C nucleotides is the effect of the lack of MMR function in general and not the effect of the lack of functional mispair binding proteins such as Msh2/3/6.

Frey et al. 22 recently reported that Msh2 deficiency interferes with the accumulation of high numbers of mutations in Peyer's patch GC B cells. We wanted to ascertain whether the effect of disruption of the Mlh1 gene is similar in chronically stimulated B cells from Peyer's patches. 14-wk-old littermates, Mlh1^−/− and Mlh1^+/+, were used in the study. The mice were killed and Peyer's patches were dissected out from the intestines of these mice. GC B cells were then isolated by flow cytometry. Cloning and analysis of the EPS transgene were carried out as for the splenic GC B cells. In addition to restriction fragment length variance analysis, single strand conformation polymorphism (SSCP) analysis was carried out in identifying clones with somatic mutations. For wild-type mice, a total of 82 mutations from 24 clones were identified for a mutation frequency of 6.5 × 10⁻³ (Fig. 4 A). This mutation frequency represents an ∼38% increase from the mutation frequency found in the splenic GC B cells of wild-type mice. The total number of mutations found in Mlh1^−/− was 53 in 29 clones analyzed (Fig. 4 B). The mutation frequency was 3.5 × 10⁻³, which is not significantly different from that of the splenic GC B cells of Mlh1^−/− mice (2.9 × 10⁻³) but is twofold less than that of Peyer's patch cells of wild-type mice (P = 0.02) (Table). An increase in mutations from G and C nucleotides was again observed in the Peyer's patch B cells of Mlh1^−/− mice. We also noted that clones with high numbers of mutations were much reduced in Mlh1^−/− mice (Fig. 5). About 65% of mutated clones from Mlh1^−/− contained a single mutation, whereas ∼75% of the mutated clones from wild-type mice contained multiple mutations, up to 16 mutations per clone. In their analysis of Msh2^−/− mice, Frey et al. 22 observed a great increase in the microsatellite instability in Peyer's patch GC B cells (PNA^hi) compared with non-GC B cells from Peyer's patches (PNA^lo). We analyzed the stability of CA nucleotide repeat microsatellites at the SSR D4Mit42 locus in Mlh1^−/− mice, and observed that Peyer's patch GC B cells contained deletions or insertions in the D4Mit42 locus at a frequency of 11.5% (9 out of 78) (Fig. 6). Non-GC B cells showed a microsatellite instability at a three times lower frequency of 3.6% (3 out of 84). It seems that the accumulation of mutations in Peyer's patch cells is affected by Mlh1 disruption as well as by Msh2 disruption.

MMR-deficient mice, Msh2^−/−, Mlh1^−/−, or Pms2^−/−, have severe health problems relating to the gene disruption 17,18,19. In all of these, a high incidence of cancer and a dramatic increase in microsatellite instability is noted. Additionally, Mlh1 or Pms2 gene disruption leads to abnormalities in meiosis and thus to fertility problems. However, we did not find any gross defects in the immune function of Mlh1^−/− or Pms2^−/− mice that might compromise the somatic hypermutation process. The levels of serum Igs of various isotypes were normal in the Pms2⁻ and Mlh1⁻ mice (data not shown), thus isotype switching appeared to be normal. Also, when compared with the serum of wild-type mice immunized at the same time, Mlh1^−/− or Pms2^−/− mice showed no discernible defect in immune response to the heterologous antigens in SRBCs (data not shown). Early B cell development in bone marrow was studied by FACS^® analysis. We did not see any consistent difference in populations of B220^lo or B220^hi cells or in IgM/IgD double positive cells (data not shown). No significant differences in GC cells as assayed were noted; we found similar proportions of B220⁺PNA^hi cells in the spleens as well as the Peyer's patches of wild-type or MMR-deficient mice (data not shown). These observations are in contrast to the recent finding of defective immune responses in Msh2 knockout mice 21,28. It remains to be determined whether the differences are due to the type of MMR defect studied (mutS homologue versus mutL homologue), the immunization schedules, or perhaps background genes of the mouse strains studied.

As with wild-type mice, we found in the analysis of MMR-deficient mice that mutations from A were much more frequent than were mutations from T (Table and Table). This bias of A over T has been seen in most studies of Ig somatic hypermutation and suggests that there is a strand bias and a nucleotide preference (A or T) that are intrinsic to the mutation mechanism 3. From previous study of the Vk167/PEPS transgene, it was noted that the EPS fragment of the transgene that contains six PvuII sites and seven EcoRV sites was an order of magnitude more mutable than the rest of the VJ region 26. We have proposed that the hypermutability may be related to the highly stable RNA secondary structures predicted in the EPS region. The striking correlation between the RNA stem formation energy and the location of highly mutated sequences led us to suggest that the mutations could be directed by the secondary structure of nascent RNA transcripts and perhaps by the RNA polymerase pausing due to such RNA secondary structure 26. In MMR-deficient mice, we found that the hypermutability of the EPS is intact (Table and Table). Thus, the hypermutability of the EPS sequence in wild-type and MMR-deficient mice is most likely the consequence of the primary mechanism of the introduction of the mutations (see below). We also found that in Mlh1^−/− mice, just as in wild-type mice, there exists a mutation preference for certain nucleotides within the EcoRV or PvuII sites (Fig. 7). Not enough data from Pms2^−/− mice were accumulated for analogous consideration. Among the six nucleotides within the EcoRV sites, T and A nucleotides at positions 3 and 4, respectively, are much more frequently mutated than are the remaining four nucleotides. Similarly, within the PvuII sites, the G nucleotide at position 3 and the C nucleotide at position 4 are much more frequently mutated. These preferred nucleotides within the restriction sites correlate with the sequences of known hot spots as described by Smith et al. 5. The di- and trinucleotide hot spots listed by Smith et al. were derived from mutations in noncoding (therefore unselectable) regions of Ig genes and must reflect the targeting preference ingrained in the Ig mutation mechanism. Rada et al. 21 found that mutations were increasingly focused on the mutational hot spots in Msh2^−/− mice. The proportion of mutations found at the hot spots was elevated from 8.2% in wild-type mice to 25.1% in Msh2^−/− mice. Our analysis of Mlh1- or Pms2-deficient mice showed the persistence of mutation hot spots within the EPS but not a noticeable increase in hot spot focusing. The proportion of mutations found in the hypermutable EPS in wild-type mice was ∼40%, and the frequencies in Mlh1- and Pms2-deficient mice were 51 and 44%, respectively. The discrepancy with Rada et al. 21 could be due to the fact that the mutated bases in their hot spots were biased toward G and C, whereas ours were balanced for A, T, G, and C (Fig. 7). Perhaps GC-rich hot spots are treated differently from AT-rich hot spots in the primary mutation mechanism.

Thus, our analysis of Mlh1^−/− and Pms2^−/− mice shows that several signature properties of Ig somatic hypermutation remained unchanged in these MMR-deficient mice. A preference for targeting A over T (in the case of the nontranscribed strand of DNA) or of targeting T over A (in the case of the transcribed strand), a strand bias, mutational hot spots, and the hypermutability of the EPS are all observed in the analysis of the unselectable Vk167/PEPS transgene in the MMR-deficient backgrounds. The implication is that MMR proteins are not required for the primary step of introducing Ig somatic hypermutations, and also, that the absence of functional MMR does not alter the primary mechanistics of somatic hypermutation.

Recently, a number of labs reported that MMR deficiency in mice results in altered mutation spectra of Ig gene somatic hypermutation. One of the conflicting points among these reports is that even though both Msh2 and Pms2 are essential factors in MMR function, Msh2 and Pms2 deficiency reportedly carry different consequences for Ig hypermutation. Phung et al. 20 reported increased mutation from G and C nucleotides in Msh2^−/− mice, whereas an increase in tandem mutations was noted in Pms2^−/− mice 14. This has led to speculation that the presence or absence of certain functional mismatch binding complexes (Msh2/3 heterodimers or Msh2/6 heterodimers) might differentially affect the final outcome of the somatic hypermutation process 22,29. In addition, evidence from knockout mice suggests that Msh2, Mlh1, or Pms2 gene products are involved in more than postreplicative DNA MMR and perhaps each of these factors might be involved in some unique function. For example, Msh2-deficient mice are fertile, whereas Mlh1 deficiency leads to both male and female infertility and Pms2 deficiency leads only to male infertility 17,18. This may imply that Mlh1 and Pms2 perform a role, at least in meiosis, independent of Msh2 and perhaps independent of each other. However, we have found that disruption of the MutL homologues Mlh1 and Pms2 result in altered mutation spectra very similar to those seen in Msh2^−/− mice. There is a noticeable increase in mutations arising from G or C nucleotides in both Mlh1^−/− and Pms2^−/− mice. On the other hand, we did not observe an increase in tandem mutations in our analysis of Pms2^−/− mice. This new evidence suggests that the altered spectra of mutations, that is, the increase in mutations from G and C nucleotides, are due to the absence of DNA MMR function in general and not to the loss of any divergent function of any one of the MMR proteins. Another similarity between Msh2^−/− and Mlh1^−/− mice is the absence of highly mutated clones among Peyer's patch B cells. Frey et al. 22 suggested that the dramatic increase in microsatellite instability in Peyer's patch GC B cells observed in Msh2^−/− mice is responsible for the lack of the accumulation of mutations. Our observation of the increase in microsatellite instability in Peyer's patch GC B cells of Mlh1^−/− mice supports this proposal. GC B cells in Peyer's patches are highly proliferative and so the DNA damage (such as microsatellite instability) due to MMR deficiency could accumulate to a high degree, causing elimination of GC B cells after fewer rounds of mutation compared with wild-type mice. The effect of microsatellite instability and accumulative DNA damage is not as pronounced in splenic GC B cells. In spleen, GCs form in response to a recent antigenic stimulation and exist only for a short time (∼3 wk), whereas in Peyer's patches the presence of food and bacterial antigens lead to chronic stimulation of GC B cells 1.

During normal DNA replication, errors are corrected first by the proofreading activity of the DNA polymerase itself and then by postreplication DNA MMR. Postreplication MMR would not affect somatic mutation patterns unless the point mutations were introduced during S phase DNA replication. Recent evidence is strong that somatic mutation occurs during transcription 6,8. There is no real evidence that it occurs during the S phase of the cell cycle. However, MMR proteins also seem to be involved in transcription-coupled repair of UV-damaged DNA in association with the NER process in Escherichia coli and in humans 30,31. Mismatch-deficient human cell lines from hereditary nonpolyposis colorectal cancer patients lack the preferential repair of the transcribed strand of DNA 30. This implies that the MMR function may be present during interphase of the cell cycle and can function apart from the DNA replicative machinery. Several possible ways by which the MMR mechanism can contribute to the somatic hypermutation of Ig genes have been discussed previously. First, as Cascalho et al. 32 proposed, it can act to “fix” the mutations introduced in one of the strands by correcting the wrong strand of DNA. Mismatch repair deficiency, in this case, would lead to much reduced somatic hypermutation frequencies. However, several reports, including ours presented here, noted that no drastic decrease in mutation frequency exists in Msh2^−/−, Pms2^−/−, or Mlh1^−/− mice. On the other hand, MMR might normally be downregulated in the GC B lymphocytes so as to allow the mutations to go uncorrected. The fact that somatic hypermutation occurs in MMR-deficient mice is compatible with this proposal. However, the mutation spectrum should be unchanged and the mutation frequency should be increased rather than decreased in these knockout mice if MMR were simply decreased in the course of the normal Ig somatic hypermutation process. The altered mutation spectra in MMR-deficient mice were clearly observed by several labs using several different systems: Vλ1 gene sequences from λ1⁺ memory B cells of nitrophenyl-chicken γ-globulin–immunized mice 13, rearranged VkOX1 gene sequences from the spleen of oxazolone-immunized mice 20, the JC intronic region of rearranged heavy chain genes in Peyer's patch PNA^hi cells 22, and, in this study, an unselectable transgene with a hypermutable insert both in spleens of SRBC-immunized mice and in Peyer's patches of old mice. Furthermore, recent reports indicated that MMR activity, as measured in an in vitro assay, is normal in isolated human centroblast cells 33 that are equivalent to GC (B220⁺PNA^hi) B cells in mice. Thus, it is unlikely that downregulation of MMR is a normal mechanism to induce/enhance somatic hypermutation.

The fact that the mutation spectrum is altered in a very similar way in Msh2^−/−, Mlh1^−/−, and Pms2^−/− mice suggests that an active MMR complex is present in GC B lymphocytes of normal mice and that it influences the outcome of somatic hypermutation by preferentially correcting some of the mutations introduced.

We thank Michael Liskay for the gift of MMR-deficient mice; Julie Auger for help with flow cytometry; and Kevin Fuller and Nancy Michael for comments on the manuscript.

This work was supported by National Institutes of Health (NIH) grant GM38649. N. Kim was partially supported by NIH training grant GM07183. The DNA sequencing facility is supported by the University of Chicago Cancer Center support grant P30 (A14599).