DNA polymerase ζ generates tandem mutations in immunoglobulin variable regions

We have previously shown that Polη^−/− mice had a normal frequency of germinal center B cells and CSR (Martomo et al., 2005). In this study, we tested if Polη^−/− cells with one allele of pol ζ behaved differently than cells with two alleles. To confirm that pol ζ is down-regulated in the heterozygous cells, mRNA levels were measured by quantitative PCR from B cells. Primers were used that were located within the two deleted exons, 26 and 27, encoding Rev3 on the targeted allele to ensure that we did not detect any mRNA from potential alternatively spliced transcripts. Compared with Polζ^+/+ Polη^−/− cells, Rev3 transcripts were significantly decreased (25–30%) in Polζ^+/− Polη^−/− cells that were either resting (Fig. 1 A; P = 0.001) or stimulated in vivo with 4-hydroxy-3-nitrophenylacetyl conjugated to chicken γ globulin (NP₃₆-CGG) for 14 d (Fig. 1 B; P = 0.016). In contrast, there was no change in mRNA encoding Rev7, or in Rev1, a binding partner of pol ζ, showing that lower Rev3 levels did not affect these transcripts. There was also no significant change in the levels of pols ι and κ, indicating that they did not overcompensate for the reduction in Rev3.

Because complete loss of pol ζ in conditional knockout mice caused slower B cell growth (Schenten et al., 2009), we examined Polζ^+/− Polη^−/− cells after stimulation with LPS and IL-4 for cell proliferation by CFSE. As shown in Fig. 1 C, haploinsufficiency of pol ζ did not affect cell division after 3 d. Pol ζ is also recruited to sites of double-strand DNA breaks (Hirano and Sugimoto, 2006), which are key intermediates in CSR. To see if a deficiency of pol ζ cripples recombination, Polζ^+/+ Polη^−/− and Polζ^+/− Polη^−/− naive B cells were stimulated for 4 d to induce switching from IgM to different IgG isotypes. Compared with Polζ^+/+ Polη^−/− cells, Polζ^+/− Polη^−/− cells had a slight decrease in recombination to IgG1, IgG2a, and IgG2b, although the numbers were not significantly different (Fig. 1 D). Thus, in contrast to the conditional knockout mouse, which showed a 50% decrease in switching caused by defective nonhomologous end joining (Schenten et al., 2009), the residual pol ζ protein in the heterozygous cells was sufficient to permit CSR at normal levels.

To analyze the contribution of pol ζ to SHM, we immunized mice with NP₃₆-CGG, and isolated germinal center cells 14 d later. SHM was analyzed in the J_H4 intron downstream of rearranged V_HJ558 genes, and Polζ^+/− Polη^−/− clones had a similar mutational frequency and mutations per clone compared with Polζ^+/+ Polη^−/− clones (Fig. 2, A–C). Analysis of the types of mutations showed no difference between Polζ^+/+ Polη^−/− and Polζ^+/− Polη^−/− clones (Fig. 2 D).

The ascribed role of pol ζ is to extend mismatched primer termini (Lawrence and Maher, 2001) produced by other low-fidelity polymerases when copying past lesions, by adding the correct base (Johnson et al., 2000). However, recent evidence indicates that pol ζ, which has an error rate of ∼10⁻³, can also extend mismatches produced by itself or other polymerases with a second mutation, generating tandem double-base mutations (Zhong et al., 2006; Sakamoto et al., 2007). The second mutation occurs at a rate of 10⁻², indicating that pol ζ frequently inserts incorrect bases when confronted with mismatched termini. Therefore, we examined DNA sequences to see if we could find this signature of pol ζ (Fig. S1).

First, we looked at published wild-type sequences (Rada et al., 2002; Martomo et al., 2008; Schenten et al., 2009; Saribasak et al., 2011) to establish the location and spectra of tandem mutations. The aligned mutational data in 492 bp encoding the J_H4 intron in Fig. 3 A shows the distribution of 1,811 single mutations and their location relative to the 6 WGCW (W = A/T) hotspots in the sequence. As seen in Fig. 3 B, the distribution of 50 pairs of tandem mutations resembled the location of single mutations, and only 18% of them occurred in WGCW motifs.

Second, to see if the distribution of tandem pairs was different in the absence of the dominant pol η, we examined sequences from Polζ^+/+ Polη^−/− clones. Data were combined from approximately two-thirds immunized spleen cells (Fig. 2 A) and one-third Peyer’s patch cells generated in this study, because there was no difference in SHM between the two populations. The location of 33 pairs of tandem substitutions and 1,060 single mutations were similar, and just 21% tandems occurred in WGCW hotspots (Fig. 3, C and D). Furthermore, only 40% tandems were located in 27 of the less stringent DGYW motifs (D = A/G/T, Y = C/T; Rogozin and Diaz, 2004). Thus, the majority of tandem pairs in wild-type and pol η–deficient mice were found outside of the hotspots, suggesting that they did not arise from multiple events targeting only these motifs.

Third, we analyzed sequences from pol η–deficient mice with one or two alleles of pol ζ to quantify the contribution of pol ζ to the tandem signature. Adjacent mutations could be produced by polymerases in a single event or sequentially after cell division. Using a calculation based on frequency and random distribution of mutations (Winter et al., 1998), the expected number of tandems increases as the number of mutations per clone increases, indicating that they can occur in multiple events (Michael et al., 2002). Table 1 lists the observed and expected number of adjacent mutations in clones from spleen and Peyer’s patch cells. Notably, the majority of tandem mutations occurred in clones which had very few mutations (Table 1), further supporting the hypothesis that most tandems occur during a single event. The total observed values were compared with the total expected values and tested for significance using the Poisson distribution, as shown in Fig. 4 A. Tandem pairs were observed at a significantly higher frequency compared with the expected number in sequences from Polζ^+/+ Polη^−/− (P < 10⁻⁷), Polζ^+/− Polη^−/− (P < 10⁻²), and wild-type Polζ^+/+ Polη^+/+ (P < 10⁻⁶) clones.

The ratio of observed to expected values was then determined, and was higher in Polζ^+/+ Polη^−/− clones at 3.1 compared with Polζ^+/− Polη^−/− clones at 2.5 (Fig. 4 A). The increase in contiguous mutations from clones with two copies of pol ζ compared with clones with one copy suggests that pol ζ is important for introducing tandem mutations into the Ig loci. Furthermore, the ratio in wild-type mice expressing the low-fidelity polymerases was lower at 2.2 compared with pol η–deficient clones. Because both pol η (Matsuda et al., 2001) and pol ζ have the capability to extend mismatches in vitro, and therefore potentially introduce contiguous mutations, the observation that the ratio in Polζ^+/+ Polη^−/− clones was higher than in Polζ ^+/+ Polη ^+/+ clones suggests that most of these tandems are not introduced by pol η. Furthermore, the spectra of tandem mutations were similar to the single mutations (Fig. 3), which suggest that pol ζ can extend mismatched termini produced by other polymerases. In wild-type mice, pol ζ could extend mismatched termini produced by pols η (A/T bias), κ (A/T bias), Rev1 (C bias), and itself (no bias). In pol η–deficient mice, the overall single mutation spectra and the tandem spectra are decreased for A/T mutations because there are no pol η–generated mismatches. Thus, the first mutation has the signature of whatever polymerase is being used, and the second mutation produced by pol ζ will be random, because pol ζ does not have a strong substitution bias (Zhong et al., 2006).

To further confirm that pol ζ generated the tandem signature, sequences from Polζ^+/− and Polζ^−/− single cells from conditional knockout mice were analyzed (Schenten et al., 2009). Although there were fewer mutations in the single-cell analyses, the formula used to calculate the expected number of tandems takes into account the overall frequency of mutations per clone. The observed number of tandems was significantly higher than the expected number in Polζ^+/− cells (P = 0.03), but not in Polζ^−/− cells (P = 0.35). The Polζ^+/− cells had a ratio of observed to expected tandems of 2.1, whereas Polζ^−/− cells had a lower ratio of 1.4. Therefore, in the complete absence of pol ζ, the observed number of tandems was similar to the expected number, indicating that they arose by two separate events. This data, summarized in Fig. 4 B, supports the hypothesis that pol ζ generates tandem mutations during SHM by extending mismatches produced by other polymerases. Slippage of polymerases on DNA templates also has the potential to produce multiple substitutions (Wilson et al., 1998), but the frequency of insertions and deletions in V genes is low, suggesting that this is not a major mechanism.

There are two pathways that could allow synthesis by pol ζ during SHM. In the UNG pathway, the polymerase could synthesize across abasic sites introduced by UNG removal of uracils. In the MSH2–MSH6 pathway, pol ζ could synthesize in the gap produced by exonuclease 1. To determine which pathway is used by pol ζ, we analyzed previously published sequences of the J_H4 intron from Ung^−/− (Rada et al., 2002), Msh2^−/− (Frey et al., 1998), Msh6^−/− (with new sequences added for this study; Martomo et al., 2004), and Ung^−/− Msh2^−/− clones (Rada et al., 2004). The observed number of tandems was significantly higher than the expected number in Ung^−/− clones (P < 10⁻³), and the ratio of observed to expected tandems was 2.2, which was identical to the ratio seen in wild-type clones (Fig. 4, A and B). This indicates that pol ζ is not introducing tandem substitutions in the UNG pathway. However, disruption of the mismatch repair proteins MSH2 and MSH6 decreased the observed number of tandem pairs, so there was no difference between observed and expected values (P = 0.19). Mice deficient for both UNG and MSH2 also had no significant difference between observed and expected values (P = 0.12). Thus, in the absence of MSH2 and MSH6, the ratio of observed to expected tandems was lower at 1.4, and in the combined absence of MSH2 and UNG, the ratio was 1.2. These results indicate that tandem mutations are produced more frequently in the MSH2–MSH6 pathway. The notion that pol ζ might be used during mismatch repair is consistent with reports from yeast, where pol ζ synthesis is reduced in mismatch repair-deficient strains (Lehner and Jinks-Robertson, 2009). One potential way pol ζ could function during gap synthesis in SHM would be if another polymerase stalls after replicating over AID-induced uracils on the template strand to produce a A:U mismatch. Pol ζ could then be recruited to extend the mispair and generate a second mispair. Additionally, pol ζ could extend mismatched termini produced by other error-prone polymerases, i.e., pols η, κ, or Rev1, and generate a second mispair. Because pol ζ has no preferential mutational signature, both mechanisms will produce a spectra of tandems that matches the overall single mutation spectra as shown in Fig. 3.

A potential advantage of using pol ζ to introduce contiguous mutations could be to more efficiently change amino acids in V genes to produce higher affinity. Codons are degenerate in the third base, and if two adjacent bases are changed simultaneously, the chance for an amino acid substitution is greater than with a single substitution. Interestingly, tandem mutations are found at extraordinary levels in nurse shark Ig genes (Zhu and Hsu, 2010), suggesting that nurse sharks may use pol ζ more frequently than mice. Thus, our data support the participation of pol ζ in SHM, and suggest that it has an increased role in the absence of the dominant pol η protein.

Polζ^+/− mice on a mixed C57BL/6 X 129 background were obtained from J. Wittschieben and R. Wood (University of Texas, MD Anderson Cancer Center, Smithville, TX; Wittschieben et al., 2010) and bred to Polη^−/− mice on a C57BL/6 background (Martomo et al., 2005). We have previously reported that SHM is identical in both the C57BL/6 and 129 strains of mice (McDonald et al., 2003). Genotypes were confirmed by PCR of tail DNA using Rev3L⁺ and Rev3L⁻ primer sets (Wittschieben et al., 2010) and Polη primer sets (Martomo et al., 2008). Polζ^+/− Polη^−/− mice were then bred to each other, which generated Polζ^+/+ Polη^−/− mice and Polζ^+/− Polη^−/− mice. Littermate mice were used at 3–8 mo of age. All animal procedures were reviewed and approved by the Animal Care and Use Committee of the National Institute on Aging.

Resting splenic B cells were isolated using negative selection with anti-CD43 and anti-CD11b magnetic beads (Miltenyi Biotec). For in vivo activation, mice were immunized with 100 µg NP₃₆-CGG (Biosearch Technologies) for 14 d. Splenic B cells were then stained with FITC-conjugated anti-B220 (SouthernBiotech) and Alexa Fluor 647-labeled anti-GL7 (eBioscience), and B220⁺GL7⁺ cells were isolated by flow cytometry. RNA was isolated using RNeasy Mini kit (QIAGEN), and SuperScript III First-Strand Synthesis System for RT-PCR was used to prepare cDNA (Invitrogen). For quantitative real-time PCR, Power SYBR Green PCR Master Mix (Applied Biosystems) was used with the following primers: Rev3 forward, 5′-GCTCAATGCCCGTCAACTAGGACTAAA-3′ and reverse, 5′-CCAAGGTCTCTCTGGCTTTGTGAAC-3′; Rev7 forward, 5′-CTCGAAACAT­GGAGAAGATAC­AGGTCATCA-3′ and reverse, 5′-AAATGTCCGACGTCATGGTTTTTAGGG-3′; Rev1 forward, 5′-GGAAGAGAACGGAGAATGATGGCTGG-3′ and reverse, 5′-GTTAGCAGCATCTGTGCGGAACTG-3′; Polι forward, 5′-GCAGACGGTCTTACTACCTGAAAGTTGTC-3′ and reverse, 5′-CAGAACTTCGAGACGCTTGGCAG-3′; Polκ forward, 5′-CAGTTAAACA­ACCCAAAGAAA­GCTCGAGAAG-3′ and reverse, 5′-CAAAGTTGAGCTCTTTGTCTTCGTTCCTG-3′; and β-actin primers (Saribasak et al., 2011). For proliferation, B cells were treated with 5 µM CFSE (Invitrogen), incubated with LPS and IL-4 for 3 d, and analyzed by flow cytometry.

Resting splenic B cells were plated at a density of 500,000 cells/ml and stimulated with 5 µg/ml Escherichia coli LPS serotype 0111:B4 (Sigma-Aldrich) to induce switching to IgG3; LPS plus 5 ng/ml recombinant IL-4 (BD) to induce switching to IgG1; LPS plus 25 ng/ml IFN-γ (R&D Systems) for switching to IgG2a; and LPS plus 2 ng/ml TGF-β (R&D Systems) for IgG2b. After 4 d, the cells were stained with FITC-conjugated antibody to B220 along with phycoerythrin-conjugated antibodies to mouse IgG1, IgG2a, IgG2b, or IgG3 (SouthernBiotech) for analysis by flow cytometry.

Mice were immunized by intraperitoneal injection of 100 µg NP₃₆-CGG (Biosearch Technologies) in adjuvant complex (Sigma-Aldrich). On day 14, splenic B cells were stained with FITC anti-B220 and Alexa Fluor anti-GL7. In nonimmunized mice, Peyer’s patch B cells were also isolated and stained. B220⁺ GL7⁺ cells were isolated by flow cytometry, and DNA was prepared. The 492-bp intronic region downstream of J_H4 from rearranged V_HJ558 genes was amplified by nested PCR. The first round used forward primer J558 5′-AGCCTGACATCTGAGGAC-3′ and reverse primer JH2906 5′-GTGTTCCTTTGAAAGCTGGAC-3′, and the second round used forward primer J558Eco 5′-CCGGAATTCCTGACATCTGAGGACTCTGC-3′ and reverse primer JH2827Bam 5′-CGCGGATCCG­ATGCCTTTCTC­CCTTGACTC-3′. The amplified DNA was then digested with EcoRI and BamHI restriction enzymes, cloned into pBS-SK vector, and sequenced. Only clones with unique VDJ joins were scored.

Fig. S1 lists 101 sequences with tandem mutations in the J_H4 intron and shows where they occur.

We thank Richard Wood and John Wittschieben for Polζ ^+/− mice; Klaus Rajewsky for sequence data; R. Wersto, C. Morris, J. Scheers, C. Nguyen, and T. Wolf for flow cytometry analyses; and Ranjan Sen and Kimberly Zanotti for valuable comments.

This research was supported entirely by the Intramural Research Program of the National Institutes of Health, National Institute on Aging.

The authors declare no competing financial interests.