Polζ ablation in B cells impairs the germinal center reaction, class switch recombination, DNA break repair, and genome stability

This was achieved by flanking exon 2 of the Rev3 gene by loxP sites, to allow its deletion by Cre-mediated recombination (Fig. 1, A and B). Exon 2 is the second coding exon of Rev3, and messenger RNA splicing from exon 1 to exon 3 results in a frameshift mutation. Some Rev3 transcripts incorporate an additional exon downstream of exon 1. RNA splicing from this exon to exon 3 also results in a frameshift mutation (43). Two targeted embryonic stem (ES) cell clones were injected into blastocysts. Mice carrying the floxed Rev3 allele (subsequently called Polζ^f) were viable and born at Mendelian ratios. The correct targeting of the Rev3 locus was confirmed by a nested PCR assay on tail DNA from Polζ^f/+ mice and wild-type controls. The PCR relied on a forward primer located outside the targeting vector to ensure the integration of the latter into the Rev3 locus, and a reverse primer located 3′ of the downstream loxP site. The amplicons of both the wild-type and floxed alleles migrated at the expected size of ∼5.2 kb (unpublished data). The PCR products were then gel purified and used as templates for two separate PCRs that amplified across the first and second loxP site, respectively. Each PCR amplified only one band for wild-type mice. In Polζ^f/+ mice, however, an additional band appeared in both cases whose larger size was consistent with the presence of an inserted loxP site (Fig. 1 C). Mice carrying the Polζ^f allele were intercrossed with CD21-cre or deleter mice to generate Polζ^f/Δ/CD21-cre compound mutants, in which Polζ function was specifically inactivated in mature B cells (44, 45). The deletion of exon 2 in the Polζ^Δ allele was confirmed by gene amplification and subsequent sequencing of the PCR product (unpublished data).

No defects in B cell development were observed in Polζ^f/Δ/CD21-cre mice (unpublished data). Likewise, the number and proportions of resting B cells in the periphery was indistinguishable from those in control mice (unpublished data). We then characterized the antibody response of Polζ^f/Δ/CD21-cre mice against the T cell–dependent antigen NP-CG (3-hydroxy-nitrophenyl acetyl coupled to chicken γ-globulin). We injected Polζ^f/Δ/CD21-cre mice and control animals with 100 µg NP-CG in alum intraperitoneally and measured the serum titers of the NP-specific IgG1 and Igλ antibodies at days 7 and 14 after immunization. At both time points, the response of Polζ^f/Δ/CD21-cre mice was reduced by 50% compared with that of control mice (unpublished data). Polζ deficiency may affect cell viability of highly proliferating cells such as GC B cells. We therefore compared the proportions of CD19⁺Fas⁺PNA^high GC B cells in spleen, mesenteric lymph nodes, and Peyer's patches of Polζ^f/Δ/CD21-cre and control mice that had been immunized with 100 µg NP-CG in alum 14 d earlier. The GC B cell compartment of Polζ^f/Δ/CD21-cre mice was 2–3-fold smaller than that in control animals (Fig. 2 A), suggesting an impairment of cell proliferation or cell viability in the absence of Polζ. To test whether this resulted in an accumulation of B cells that had escaped Cre-mediated inactivation of Polζ, we isolated CD19⁺Fas⁺PNA^high GC B cells and CD19⁺Fas⁻PNA^low naive B cells from Polζ^f/Δ/CD21-cre mice and control mice 14 d after immunization by flow cytometry and determined the deletion efficiency of the Polζ^f allele in these cells by PCR. Competitive amplification of the Polζ^f and Polζ^Δ alleles and comparison to a standard curve with known ratios of the two alleles showed that ∼90% of the PCR products in the CD19⁺Fas⁻PNA^low B cell fraction and 80% in the CD19⁺Fas⁺PNA^high B cell fraction were derived from the Polζ^Δ allele in Polζ^f/Δ/CD21-cre mice (Fig. 2 B). This corresponds to a deletion efficiency of 80% in naive and 60% in GC B cells, as the mice were heterozygous for the Polζ^Δ allele. Thus, there was a moderate enrichment of cells that failed to inactivate Polζ in the GCs.

To examine the impact of Polζ deficiency on SHM, we isolated naive and GC B cells from Polζ^f/Δ/CD21-cre and control mice that had been immunized with NP-CG 14 d earlier. As in the mutants, a substantial fraction of these cells had escaped Cre-mediated recombination. An analysis of VDJ joints amplified from populations of such cells is a priori compromised. Indeed, when we analyzed the accumulation of somatic mutations by bulk PCR using a primer pair that anneals in the framework region 3 of most V_HJ558 gene segments and in the intron downstream of the J_H4 gene segment (46), we found only a moderate reduction of the frequency of SHM in the mutants compared with the controls (Fig. S1 A, available). 76 and 79% of the mutations in Polζ^f/Δ/CD21-cre mice occurred at A-T basepairs, compared with 56 and 65% in the controls. There was no difference in the ratio of transitions to transversions between mutants and controls (Fig. S1 B).

To approach the impact of Polζ deficiency on SHM at a higher level of resolution, we analyzed the accumulation of somatic mutations in Polζ^f/Δ/CD21-cre mice at the level of single GC B cells, which we simultaneously genotyped for the inactivation of the Polζ gene. This approach offered an additional advantage in that it is not sensitive to a different rate of PCR errors across the individual samples. We isolated single GC B cells by flow cytometry and performed a PCR using a primer pair that anneals in the framework region 3 of most V_HJ558 gene segments and in the intron downstream of the J_H4 gene segment together with primer pairs that allow the genotyping of the Polζ^f and Polζ^Δ alleles. The amplification efficiency of both the Polζ^f and the Polζ^Δ allele in sorted GC Polζ^f/Δ B cells was 80%. We successfully amplified the rearranged J_H gene segments in 66% of Polζ^f/Δ GC B cells from mice without the CD21-cre allele, in 42% of Polζ^f/Δ GC B cells from Polζ^f/Δ/CD21-cre mice, and in 66% of Polζ^Δ/Δ GC B cells from Polζ^f/Δ/CD21-cre mice. These differences in amplification efficiencies likely resulted from different extents of contamination of the sorted cells by non-B cells. The PCR products corresponding to the four introns downstream of the rearranged J_H gene segments were sequenced and analyzed for the presence of somatic mutations. GC B cells that had ablated Polζ mutated their Ig genes with a frequency of 0.19% (Fig. 3 A). In contrast, the mutation frequency in GC B cells that had retained the Polζ^f allele was 0.68%. Moreover, almost 50% of all sequences derived from Polζ^Δ/Δ cells contained only one mutation, whereas close to 80% of the sequences derived from Polζ^f/Δ cells had multiple mutations. Likewise, 38% of the sequences derived from Polζ^Δ/Δ cells remained unmutated.

To compare the pattern of somatic mutations in mutants and controls, we first established the mutation patterns in the individual introns downstream of each rearranged J_H gene segment. Subsequently, we normalized the mutation patterns to the base composition of the respective intron and then combined the mutation patterns of all four introns. There was no significant difference in the mutation pattern between Polζ^f/Δ and Polζ^Δ/Δ GC B cells. In Polζ^Δ/Δ cells, 62% of all mutations accumulated at A-T basepairs, which compared with 61% for Polζ^f/Δ cells (Fig. 3, B and C). Likewise, the ratio of transitions to transversions was not altered. We conclude from these data that Polζ deficiency has no impact on the pattern of SHM and that the differences in the A-T mutations observed in the bulk analysis must have been a result of differences in the fidelity of gene amplification between the experimental groups, a possibility which is excluded in the single cell analysis. With respect to the frequency of SHM, two interpretations offer themselves to explain the results: Polζ deficiency either directly impacts the SHM mechanism or it results in a rapid arrest of cellular proliferation preventing further SHM. The absence of a change of mutation pattern favors the latter interpretation.

To study a possible effect of Polζ deficiency on CSR, splenic B cells from Polζ^f/Δ/CD21-cre and control mice were stimulated with LPS in the presence or absence of IL-4, and the percentages of class-switched cells were measured by flow cytometry 3 or 4 d later. Compared with wild-type B cells, B cells from the mutant mice exhibited 2–3-fold reduced frequencies of cells switched to the expression of IgG1 or IgG3, respectively (Fig. 4 A and Table I). As CSR and cell proliferation are linked (47), we labeled Polζ^f/Δ/CD21-cre B cells and control cells with CFSE and measured their ability to divide. Proliferation of Polζ^f/Δ/CD21-cre B cells was reduced compared with controls (Fig. 4 A). Moreover, cell viability of Polζ^f/Δ/CD21-cre B cells dropped significantly 3 d after activation (unpublished data). Defective proliferation and decreased viability could explain the reduced frequency of Polζ-deficient cells switched to the expression of IgG1 and IgG3. However, the defect in CSR did not appear to be linked to the reduced ability of Polζ^f/Δ/CD21-cre B cells to proliferate. The percentages of class-switched cells were also 2–3-fold reduced when cells having undergone a similar number of divisions were analyzed for CSR (Fig. 4 A, gate indicated by red bracket). The latter result was independent of the location or width of the gate.

Germline transcription of the S regions is required for the initiation of CSR. Hence, we asked whether differences of germline transcription could account for the impairment of CSR in Polζ^f/Δ/CD21-cre B cells. We stimulated splenic B cells with either LPS + IL-4 or LPS alone for 2 d and then amplified germline transcripts of Sμ, Sγ1, and Sγ3 by RT-PCR. Polζ^f/Δ/CD21-cre and control B cells produced similar levels of Sμ, Sγ1, and Sγ3 transcripts (Fig. 4 B). To test whether Polζ is required for the repair phase of CSR, we amplified and sequenced Sμ-Sγ3 junctions from mutant and control cells, as it is known that the inactivation of some genes involved in DNA repair, such as MSH2, PMS2, or Polη, can lead to changes in the nature of the microhomologies or mutations at the junctions between the Sμ and Sγ3 or other S regions in class-switched cells (15, 19, 48, 49). PCR products between 500 and 800 nt were selected for analysis, subcloned, and sequenced, excluding clonally related cells. We then analyzed the sequences by BLAST alignment to the germline Sμ and Sγ3 sequences and counted the number of nucleotides at the S junctions that were identical to both the Sμ and Sγ3 region. The majority (53%) of the sequences derived from Polζ^+/Δ/CD21-cre B cells displayed little (1–2 nt) or no microhomology at the S junctions, whereas a smaller fraction (21%) of sequences exhibited a longer stretch (3–7 nt) of nucleotide overlap (Fig. 4 C). The remaining fraction (25%) had insertions of 1 nt or more at the S junctions. Similar to the control cells, the majority of sequences (56%) derived from Polζ^f/Δ/CD21-cre B cells contained no or only small microhomologies. In contrast, the frequency of sequences containing microhomologies of 3–7 nt was increased (34%), whereas the frequency of sequences containing insertions was reduced (10%). Insertions of 2 nt or longer were absent in sequences from the mutant B cells in contrast with control sequences. The changes in the nature of the Sμ–Sγ3 junctions, albeit modest, are consistent with a possible requirement for Polζ in the efficient processing of free DNA ends during CSR.

We also analyzed the switch junctions for somatic point mutations, which are known to occur in these regions through the recruitment of AID in the initiation of CSR (13, 50). In cells initiating but not completing CSR in the course of proliferation, such mutations are introduced into Sµ but are rarely found in Sγ1 or Sγ3. It is believed that this explains the higher load of mutations in the Sµ parts of Sμ–Sγ1 or –Sγ3 switch junctions (24, 49, 50). Comparing the sequences of Sμ–Sγ3 junctions within a 200-bp window, we found that in wild-type cells the mutation frequencies in the Sµ portions of the junctions were indeed higher than in Sγ3 portions (0.53 and 0.27%, respectively), whereas in the mutants they were similarly low (0.21 and 0.22%, respectively; Table S1, available). The limited number of mutations did not allow a meaningful comparison of the mutation pattern. These results indicate that somatic mutations were equally introduced in mutant and control cells into switch regions undergoing switch recombination but that because of impaired proliferation and survival, the mutant cells had less opportunity to accumulate mutations in Sμ before switching.

Polζ had been previously shown to be required for genome stability in the DT40 chicken B cell line and mouse embryonic fibroblasts (51, 52). As genomic instability might explain the defective proliferation and CSR in LPS-activated Polζ-deficient B cells, as well as the counterselection of Polζ-deficient GC B cells, we stimulated B cells from two Polζ^f/Δ/CD21-cre and two control (Polζ^f/+/CD21-cre) mice with LPS + IL-4 or LPS alone in vitro, prepared metaphase spreads 3 d later, stained the chromosomes with DAPI, and analyzed the metaphases by microscopy (Fig. 5 A). Chromosomal abnormalities were observed in 3% of the metaphases from B cells of the control mice in contrast with 39% of the metaphases from B cells of Polζ^f/Δ/CD21-cre mice. We saw approximately equal frequencies of broken chromosomes and chromatids versus fused chromosomes in the affected metaphases of Polζ-deficient B cells. The occurrence of broken chromosomes and broken chromatids (which are present in mitotic cells as a result of unrepaired DSBs introduced before or after replication, respectively) suggested a DNA repair defect throughout the cell cycle (Fig. 5 A, Table S2 [available], and see Fig. 7). Spectral karyotyping (SKY) confirmed the general chromosomal instability detected in Polζ-deficient B cells (Fig. 5 B). Consistent with the previous analysis, only a modest number (5%) of metaphases of B cells from control mice contained irregular chromosomes, whereas metaphases containing broken chromosomes, chromosome fusions, and translocations were much more frequent (23%) in B cells of Polζ^f/Δ/CD21-cre mice stimulated in vitro with LPS + IL-4 for 3 d. Even though Polζ-deficient metaphases were present with a broken chromosome 12, which carries the IgH locus in the mouse and where CSR occurs, other chromosomes (namely chromosomes 1, 3, and 11) were affected at a similar frequency. However, the sensitivity of either DAPI staining or SKY in the detection of CSR-specific breaks is limited, as the IgH locus is located in the subtelomeric region of chromosome 12, and such techniques may fail to distinguish an intact chromosome 12 from one lacking a small telomeric fragment as a result of a break within the IgH locus. We therefore proceeded to analyze CSR-specific chromosome breaks by fluorescence in situ hybridization (FISH) using IgH locus-specific probes.

In these experiments, we induced CSR in B cells of Polζ^f/Δ/CD21-cre mice and of control mice by stimulation with an α-CD40 antibody in the presence of IL-4. This treatment efficiently induces CSR to IgG1 and allows comparison to stimulation with α-CD40 mAb alone, which activates the cells by itself and, in contrast with LPS, does not induce CSR. On day 4 after stimulation, metaphases were analyzed by FISH. Using a telomere-specific probe, we confirmed the increased general genome instability in B cells of Polζ^f/Δ/CD21-cre mice in that we observed chromosomal breaks in 53 versus 13% of all metaphases of B cells from Polζ^f/Δ/CD21-cre and control (Polζ^f/+/CD21-cre) mice, respectively (Fig. 6 and Table I). The chromosomal breaks in B cells from both Polζ^f/Δ/CD21-cre and control (Polζ^f/+/CD21-cre) mice consisted of approximately equal proportions of chromatid and chromosome breaks (Fig. 7 and Table S2). The results in Table S2 also raise the possibility that Polζ haploinsufficiency affects genome stability in cells activated to undergo CSR, but more data would be required to firmly establish this point.

To assay for breaks or translocations within the IgH locus, we used a FISH assay in which the first probe recognizes sequences just upstream of the IgH V_H domain on the telomeric region of chromosome 12 (5′ Igh probe) and the second probe sequences immediately downstream of the C_H locus (3′ Igh probe) (26). This assay revealed chromosomal breaks within the IgH locus, visualized as separated 3′ and 5′ IgH signals, in 8.3% of metaphases from α-CD40 mAb + IL-4–activated B cells from Polζ^f/Δ/CD21-cre mice. This is in contrast with only 2% of metaphases containing such breaks in activated control B cells (Fig. 6 and Table I). The breaks led to IgH locus-specific translocations in 28% of the affected cells from Polζ^f/Δ/CD21-cre mice and in 43% of affected control B cells (unpublished data). These findings suggest that Polζ is required for repair of a subset of IgH locus breaks that occur during CSR.

To ensure that the enhancement of IgH breaks in activated Polζ^f/Δ/CD21-cre B cells is indeed caused by CSR, we also stimulated B cells from Polζ^f/Δ/CD21-cre and control mice with an α-CD40 or α-RP105 mAb alone because such treatments induce proliferation but not CSR (53). Although B cells of Polζ^f/Δ/CD21-cre mice again accumulated chromosomal aberrations at a higher level than the control cells, IgH locus-specific breaks were not detected in either cell type under these conditions (Fig. 6, B and C; and Table I). We conclude that the increased frequency of chromosomal breaks in the IgH loci in B cells from Polζ^f/Δ/CD21-cre mice after α-CD40 mAb + IL-4 stimulation is a result of defective repair of CSR-induced DSBs.

Ubiquitous Polζ deficiency in the mouse is embryonic lethal (39–41). We have overcome this lethality by generating a conditional allele of Rev3 and used this allele to render mature resting CD21-expressing B cells deficient of Polζ. Using this system, we show that Polζ is not essential for B cell maintenance but is required for proper B cell proliferation, presumably because of its role in the maintenance of genome stability. The genomic instability and proliferation defect of Polζ-deficient B cells observed in the present study is reminiscent of the phenotype of Polζ-deficient chicken DT40 B cells, which also accumulate genomic aberrations and are impaired in their ability to proliferate (52). Genomic instability was also observed in Polζ-deficient mouse embryonic fibroblasts, which barely divide and require inactivation of p53 to proliferate (51). It has been speculated that Polζ may prevent the collapse of replication forks that have stalled at DNA lesions and may facilitate postreplication DNA repair via homologous recombination or translesion synthesis and mismatch extension (42, 51). However, the role of Polζ in the maintenance of genome stability has not been determined with precision. Many components of the NHEJ or DNA damage response machinery do not affect proliferation and SHM (24, 28, 54). B cells deficient of the classical NHEJ factors Xrcc4 and ligase 4 harbor predominantly chromosome breaks, which are indicative of failed DSB repair in the G1 phase of the cell cycle (23). In contrast, we found high frequencies of both chromosome and chromatid breaks in the Polζ-deficient B cells, indicating pre- and postreplication repair defects, respectively. It is thus also conceivable that the impaired proliferation and reduced cell viability of Polζ-deficient B cells are a result of a function of Polζ in NHEJ, translesion synthesis, and homologous recombination.

Genomic instability and impaired proliferative capacity of Polζ-deficient B cells are likely the main reasons for the impaired GC and IgG1 antibody response in our mutant animals, in which Polζ was selectively ablated in mature B cells. GCs were reduced in size, the frequency of somatic mutations introduced into V region genes was reduced, and CSR was impaired. We also observed counterselection of cells that had undergone deletion of the Rev3 gene in GCs. Analyzing the extent and pattern of SHM in Polζ-proficient and -deficient GC B cells by single cell PCR, we found a >3.5-fold reduction of the frequency of somatic mutations in the mutant cells but no change in the mutational pattern. Given that Polη has been identified as the sole polymerase responsible for mutations at A-T basepairs (55), a contribution of Polζ to SHM should have resulted in a change of the mutational pattern in Polζ-deficient B cells. In the absence of such a change, our results are consistent with the interpretation that the impairment of SHM by Polζ deficiency is caused by the impaired cell viability and consequent counterselection of the mutant cells in the GC reaction.

In two earlier papers a direct involvement of Polζ in the mechanism of SHM had been suggested (37, 38). In one approach, Diaz et al. (37) constitutively expressed antisense RNA against Polζ in transgenic mice and observed compromised SHM in the absence of a change of mutational pattern, which is similar to the present data. Although these results were difficult to interpret because of the non–B cell-autonomous expression of the transgenic antisense RNA and possible genomic instability also in that system, a study of SHM in a Burkitt lymphoma cell line again demonstrated a reduced frequency of somatic mutation upon transfection of the cells with Rev3-specific antisense oligonucleotides, without an effect on the mutation pattern (37, 38). Rev3 expression was only partially inhibited in these experiments, and cell proliferation was not detectably affected. It remains to be seen to which extent these latter experiments reflect the requirements for SHM in GC B cells in vivo, where a direct participation of Polζ in its control clearly remains tentative.

In contrast with SHM, where our results suggest indirect effects of Polζ deficiency, our data indicate a direct involvement of Polζ in CSR. We observed a reduction in CSR efficiency of Polζ-deficient B cells compared with control B cells even when we compared cells that had gone through the same number of cell divisions. Together with the observation that germline transcription of the S regions remains unaffected in Polζ^f/Δ/CD21-cre mice, this suggested an involvement of Polζ in the repair phase of CSR, a notion which was directly supported by the increased frequency of IgH locus-specific DNA breaks in Polζ-deficient B cells undergoing CSR in response to anti-CD40 stimulation in the presence of IL-4. Our evidence clearly indicates that these breaks were caused by a defective CSR process because they were absent in Polζ-deficient B cells activated with α-CD40 or α-RP105 mAbs alone, an activation protocol which induces proliferation but not CSR. Among 1,200 chromosomes analyzed by telomere FISH from individual α-CD40 + IL-4 cultures, ∼3% of the chromosomes from Polζ-deficient B cells showed abnormalities including chromatid breaks, chromosome breaks, and chromosome fusions (Fig. 7, Table I, and Table S2). However, telomere FISH does not distinguish between CSR-specific and -unspecific chromosomal aberrations. To approach this problem, we determined the frequency of such CSR-specific breaks in the activated Polζ-deficient cells within the IgH locus on chromosome 12, where CSR occurs. Based on the genome instability detected by telomere FISH, we estimated that if chromosome 12 was targeted randomly, three breaks in this chromosome should be found among 50 metaphases and these breaks should be randomly distributed along the chromosome. In contrast with the latter prediction, we detected three to five IgH locus-specific chromosome breaks in 50 metaphases (Table I). Considering that the IgH locus covers only ∼3% of chromosome 12, this implies that IgH locus-specific breaks in Polζ-deficient B cells stimulated to undergo CSR are the result of a targeted rather than a random process. In further support of this notion, we failed to observe chromatid breaks at the IgH locus in metaphases of Polζ-deficient cells (in contrast with their frequent occurrence at other loci), suggesting that, specifically within the IgH locus, Polζ functions to repair a subset of DSBs restricted to prereplicative stages of the cell cycle (G1) precisely when AID is known to target S regions (13, 56). In principle, an increased frequency of IgH locus-specific DNA breaks could also result from defective V(D)J recombination. This has been recently observed in ATM-deficient B cells, which accumulate such breaks in an AID-independent and RAG-dependent manner (53). Because the CD21-cre allele causes the deletion of the floxed Polζ allele in mature but not developing B cells (44) and no IgH locus-specific DNA breaks were detected in Polζ-deficient B cells stimulated with α-CD40 or α-RP105 mAbs, such a possibility seems to be excluded in the present context.

CSR involves the generation of staggered DSBs during the G1 phase of the cell cycle, which need to be processed to create blunt DNA ends that serve as substrates for the alternative and classical NHEJ machinery (13, 23, 56–58). This processing requires the activity of exonucleases and DNA polymerases, with the former being provided by exonuclease 1 among perhaps other enzymes (20). The present results suggest that Polζ catalyzes fill-in reactions in the course of CSR-associated DSB repair by NHEJ. This is in line with observations in Saccharomyces cerevisiae that ATR homologue Mec1-dependent phosphorylation, which plays a major role in DNA damage checkpoint responses, promotes the association of Polζ/Rev1 with DSBs (59, 60). Our observation that CSR is impaired in the absence of Polζ, yet not completely abolished, points to the participation of additional polymerases in the repair phase of CSR. These could include Polμ and Polλ, both of which have been suggested to operate in classical NHEJ (61, 62).

In normal B cells undergoing CSR, strict dosing of AID and efficient repair by NHEJ prevents aberrant rejoining of IgH locus-specific DSBs to DSBs within other loci. Dysregulation of AID activity or deficiency of specific repair factors has been shown to increase the frequency of oncogenic IgH-myc translocations in in vitro cultured primary B cells undergoing CSR (27). In this context, we report frequent chromosomal translocations involving the IgH locus (and multiple uncharacterized partners) in primary Polζ-deficient B cells undergoing CSR in vitro. We also observed a slight trend toward increased genomic instability in Polζ^f/+/CD21-cre heterozygous B cells compared with wild-type B cells (Table S2). Thus, genetic or epigenetic alterations that result in Polζ (haplo) insufficiency might act in vivo to increase the frequency of oncogenic IgH locus translocations.

A 25-kb clone comprising 4 exons of rev3 was obtained by screening a P1 library generated from C57BL/6 genomic DNA. The clone was used to construct a targeting vector containing a loxP-flanked exon 2. The culture and transfection of C57BL/6-derived Bruce4 ES cells has been described previously. Homologous recombination was determined by Southern hybridization of probe A with EcoRI-digested genomic DNA or of probe B with BamHI-digested genomic DNA. The latter assay was also used to ensure the integration of the third loxP site. After homologous recombination, the neomycin resistance gene was removed in vitro by transfection of ES cells with the Cre recombinase-expressing plasmid pIC-Cre. Two independently targeted ES cell clones were injected into CB.20-derived blastocysts, which were implanted into pseudo-pregnant CB.20 foster mice to generate chimeric mice. Both clones transmitted the targeted allele into the germline. The correct targeting was confirmed by a nested PCR. The first PCR round used the primers polz-1 [5′-GTCCGTCTGCCATCAAGTCTAGCTTCAG-3′] and polz-2 [5′-GGAAGCCCTGGATTTCATCACGCCATTG-3′], which anneal 5′ of the region of homology and 3′ of the second loxP site, respectively (Fig. 1 B). Both the Polζ⁺ and Polζ^f allele yielded a PCR product of ∼5.2 kb. This was followed by an amplification step of the gel-purified PCR products using primers that flank each loxP site (first loxP site: primers polz-3 [5′-GTCAGCTGGTCAGAGTGAGGGATAC-3′] and polz-4 [5′-TAACCAGCACACTACCACTACCTGGG-3′]; second loxP site: primers polz-5 [5′-CCTCCTCAGTAGTAGTAGGACCTTTGGAGG-3′] and polz-6 [5′-CGTATCTGCCCACTCTCTGTCATTCTCTC-3′]). The primers polz-1 and polz-2 were also used for the amplification of the Polζ^Δ allele from tail DNA of Polζ^+/Δ mice. Parts of the PCR product (3.2 kb) were sequenced using the primers polz-1, polz-3, and polz-6 to confirm the integration of the targeting vector into the Rev3 locus and the excision of exon 2.

Polζ^f/+ mice were intercrossed with deleter mice to obtain mice carrying a nonfunctional Polζ allele (Polζ^Δ/+ mice) (45). Subsequent intercrosses between Polζ^f/+ mice, Polζ^Δ/+ mice, and CD21-cre mice generated Polζ^f/Δ/CD21-cre mice (44). The mice were kept on a C57BL/6 background. All mouse experiments were approved by the Institutional Animal Care and Use Committee of Harvard University and by the Immune Disease Institute.

Single-cell suspensions prepared from lymphoid organs were stained with the following monoclonal antibodies conjugated to FITC, PE, CyC, APC, or biotin: anti-CD3 (145-2C11), anti-CD4 (L3T4), anti-CD8 (53–6.7), anti-CD19 (ID3), anti-CD21, anti-CD23 (B3B4), anti-CD43 (S7), and anti-Fas (Jo2; BD). Monoclonal antibodies against IgM (R33-2412) and B220 (RA3-6B2) were prepared and conjugated in our laboratory. PNA conjugates were PNA-FITC or PNA-biotin (Vector Laboratories). Biotin conjugates were visualized with CyC or APC-streptavidin (BD). All samples were acquired on a FACSCalibur (BD) and analyzed using CellQuest software (BD).

Mice were immunized by intraperitoneal injection with 100 µg of alum-precipitated NP₃₆-CG ((4-hydroxy-3-nitrophenyl)acetyl coupled to chicken γ-globulin; Biosearch Technologies, Inc.). NP-specific Ig serum titers were monitored every 7 d by bleeding of the mice from the tail vein and subsequent ELISA.

CD19⁺PNA^lowFas^low naive and CD19⁺PNA^highFas^high GC B cells were sorted by FACS, subsequently lysed in 50 µl of 10-mM Tris, ph 7.5, plus 250 µg/ml proteinase K for 3 h at 56°C, and heat inactivated for 15 min at 80°C. Serial dilutions of 20,000 cell equivalents were used for the amplification of the Polζ^f and Polζ^Δ alleles in a competitive PCR (30 s and 94°C denaturation; 30 s and 67°C annealing; and 1 min and 72°C elongation, 30 cycles) using the forward primers Seq1 (5′-AAAGGTCAGCTGGTCAGAGTGAGGGA-3′) and Seq7 (5′-ATGTAAAAGAATACAGTGGAAGTGAGG-3′), which anneal 5′ and within the floxed region, respectively, and the common reverse primer Seq2, 5′-GGGAAGCCCTGGATTTCATCACG-3′, which anneals 3′ of the floxed region. The ratios of the Polζ^f- and Polζ^Δ-derived PCR fragments were compared with a standard curve of known ratios to estimate the deletion efficiency.

Single GC B cells (CD19⁺PNA^highFas^high) were sorted by flow cytometry into 96-well plates and analyzed individually using a nested PCR for the genotyping of the cells and a seminested PCR for the amplification of DNA fragments corresponding to the introns downstream of the rearranged V_HD_HJ_H elements. After cell lysis with proteinase K in PCR buffer, the PCR was performed amplifying DNA fragments corresponding to the Polζ^f and Polζ^Δ using the primers zeta1 (5′-ACCAAAAGAAAGGTCAGCTGGTCAGAGTGAGGG-3′), zeta3 (5′-TTTTAGACTGGGTGCTCAGGAAGGGTCTCATTGAC-3′), and zeta6 (5′-TGCTCTTGTACTCAGAGACTGGCTCACGGAGC-3′) and DNA fragments containing the introns downstream of the rearranged V_HD_HJ_H elements using the primers J558Fr3 (5′-CAGCCTGACATCTGAGGACTCTGC-3′) and JH4-int2 (5′-GCTCCACCAGACCTCTCTAGACAGC-3′), which anneal in the framework 3 region of most V_HJ558 elements and the intron downstream of J_H4, respectively (46). The annealing temperature was five cycles at 62°C followed by five cycles at 60°C and 30 cycles at 58°C. The second PCR reactions for the amplification of the Polζ alleles and the IgH introns were performed separately using 5 µl of the initial PCR reaction mix. The deletion of the Polζ^f allele was determined with the primers zeta4 (5′-AAGAATACAGTGGAAGTGAGGGAGCAGGATCCG-3′) and zeta5 (5′-CCTATGGGAAGCCCTGGATTTCATCACGC-3′) detecting the Polζ^f allele and the primers zeta5 and zeta2 (5′-TGCAAGGTACTGACGAGCAACATGAGAAATATGGG-3′) detecting the Polζ^Δ allele. The IgH introns were amplified with the primers J558Fr3 and JH4Hint-1 (5′-TGCCTTTCTCCCTTGACTCAATCAC-3′). The annealing temperature of the secondary PCR reactions was 58°C. DNA products corresponding to the JH introns were directly sequenced with the primers J558Fr3 and JH4Hint-1. To compare the frequency of somatic mutations, a 500-bp stretch downstream of the rearranged J_H gene segments was analyzed for the accumulation of somatic mutations. Cells carrying only deleted Polζ alleles (Polζ^Δ/Δ) were grouped and compared with cells from the same mouse that retained a floxed Polζ allele (Polζ^f/Δ). To compare the pattern of somatic mutations, the mutations were analyzed in each J_H intron and subsequently normalized according to the base composition of the individual J_H intronic regions analyzed (intronic region J_H1 [A, 0.97; C, 1.147; G, 1.04; T, 0.87], intronic region J_H2 [A, 1.15; C, 0.94; G, 1.06; T: 0.88]; intronic region J_H3 [A, 1.08; C, 1.28; G, 0.83; T, 0.91], and intronic region J_H4 [A, 0.94; C, 1.56; G, 0.92; T, 0.82]).

The analysis of SHM in GC B cells by bulk PCR has been published previously (32, 33). Splenic GC B cells were sorted into a naive fraction (CD19⁺PNA^lowFas^low) and a GC fraction (CD19⁺PNA^highFas^high). PCR fragments corresponding to the intron downstream of J_H4 were obtained from DNA of purified cell populations using the primers J558Fr3 and JHCHint-1 (46). PCR fragments containing VDJ rearrangements involving a J_H4 element were selected for further analysis, subcloned, and sequenced. A stretch of 500 bp of intron sequence immediately downstream of the J_H4 element was analyzed for somatic mutations.

B cells were purified from splenic single cell suspensions by MACS depletion using anti-CD43 microbeads (Miltenyi Biotech). The cells were stimulated with either 20 µg/ml LPS alone, 20 µg/ml LPS and 25 ng/ml IL-4 (R&D Systems), or 1 µg/ml anti-CD40 mAb (clone HM40-3; BD) and 25 ng/ml IL-4. Cells were cultured for 4–5 d. The percentage of class-switched cells was determined between days 3 and 5 by flow cytometry. Proliferation was determined by labeling of the cells before stimulation with 2.5 mM CFSE in PBS and measuring the number of cell divisions on day 4 by flow cytometry. For the analysis of the switch junctions, DNA was isolated from the cells on day 3 and amplified by a nested PCR reaction using the primers Cμ1.1 and Cγ1.1, followed by Cμ1.2 and Cγ1.2. PCR products with a length of 500–800 nt were subcloned and subsequently sequenced. Clonally related sequences were excluded from the analysis.

MACS-purified B cells were stimulated with 20 µg/ml LPS and 25 ng/ml IL-4 and cultured for 3 d, at which 50 ng/ml colcemid (KaryoMAX Solution; Invitrogen) was added to the medium for 2 h. The preparation of metaphases has been described elsewhere. In brief, the cells were swelled in 0.4% potassium chloride solution for 10 min at 37°C, incubated four times in 3:1 methanol/acetic acid solution, dropped onto microscopy slides, and subsequently passed through hot steam and air dried. The metaphases were either covered with DAPI-containing mounting medium (Vectashield; Vector Laboratories) or further processed for SKY using the SkyPaint DNA kit (Applied Spectral Imaging) according to the manufacturer's instructions. The analyses were performed on a microscope (Eclipse; Nikon) using a black and white 12-bit charge-coupled device camera (Applied Spectral Imaging) and a 63× objective lens. The cytogenetic analysis by FISH has been published elsewhere (26). In brief, the cells were stimulated with 1 µg/ml of an α-CD40 mAb (clone HM1-3; BD), 2.5 µg/ml of an α-RP105 mAb (anti-CD180; clone RP/14; eBioscience), or an α-CD40 mAb and 25 ng/ml IL-4 for 4 d. Subsequently, metaphases were prepared as described and hybridized with the fluorescence-labeled BAC207 and BAC199 for the detection of the 5′ and 3′ ends of the IgH locus, respectively. The telomeres were stained with an Cy3-conjugated (TTAGGG)₃ PNA probe (Applied Biosystems).

Fig. S1 contains the analysis of SHM in GC B cells of Polζ^f/Δ/CD21-cre mice and control mice by bulk PCR. Table S1 shows the frequency of mutations in Sµ and Sγ3 regions near Sμ–Sγ3 junctions in Polζ-deficient and control B cells, and Table S2 describes the types of genomic abnormalities observed in stimulated Polζ-deficient and control B cells.

We thank Angela Egert, Anthony Monti, Margaret Curnutte, and Victoria Dreier for expert mouse work. We also appreciate the helpful advice and comments of Sergei Koralov, Stefano Casola, Marc Schmidt-Supprian, John Manis, and Daniel Stetson.

This work was supported by the Deutsche Forschungsgemeinschaft (SFB 243) and the National Institutes of Health (grants AI054636 and CA092625 to K. Rajewsky and AI31541 and CA092625 to F.W. Alt). F.W. Alt is an Investigator of the Howard Hughes Medical Institute.

The authors have no conflicting financial interests.