Key Points
HR supports germinal center B cell survival and Ab affinity maturation.
HR is active at the hypermutating Ig locus and influences the SHM mutation pattern.
The magnitude of A:T mutagenesis during SHM of the Ig V region increases over time.
Visual Abstract
Abstract
During somatic hypermutation (SHM) of Ig genes in germinal center B cells, lesions introduced by activation-induced cytidine deaminase are processed by multiple error-prone repair pathways. Although error-free repair by homologous recombination (HR) is crucial to prevent excessive DNA strand breakage at activation-induced cytidine deaminase off-target genes, its role at the hypermutating Ig locus in the germinal center is unexplored. Using B cell–specific inactivation of the critical HR factor Brca2, we detected decreased proliferation, survival, and thereby class switching of ex vivo–activated B cells. Intriguingly, an HR defect allowed for a germinal center reaction and affinity maturation in vivo, albeit at reduced amounts. Analysis of SHM revealed that a certain fraction of DNA lesions at C:G bp was indeed repaired in an error-free manner via Brca2 instead of being processed by error-prone translesion polymerases. By applying a novel pseudo-time in silico analysis of mutational processes, we found that the activity of A:T mutagenesis during SHM increased during a germinal center reaction, but this was in part defective in Brca2-deficient mice. These mutation pattern changes in Brca2-deficient B cells were mostly specific for the Ig V region, suggesting a local or time-dependent need for recombination repair to survive high rates of SHM and especially A:T mutagenesis.
Introduction
Immunoglobulin diversification in germinal center B cells is based on a unique combination of targeted introduction of DNA lesions by activation-induced cytidine deaminase (AID), followed by mostly error-prone processing via several of the major DNA repair pathways of the cell (1). At the same time, damage introduced by AID into several non-Ig genes may be repaired in an error-free manner, ensuring genome maintenance (2). The molecular basis of this phenomenon is quite elusive, but locus-specific malfunction of a major error-free repair pathway, such as homologous recombination (HR), may be involved.
Processing of the U:G lesion caused by AID-mediated cytosine deamination may lead to one of five following possible outcomes (1, 3): 1) during class switch recombination of the Ig C region, processing to double-strand breaks leads to constant module deletion and use of downstream modules to produce Abs with variant effector functions; 2) during Ig gene conversion occurring in several farm animals, but not in mice and humans, lesions in the Ig V region trigger HR-based sequence transfers from upstream pseudogenes (4); 3) during somatic hypermutation (SHM) of Ig genes (in, e.g., humans and mice), lesions can be processed by a) replication over the uracil, leading to transitions at C:G, b) translesion synthesis (TLS) over abasic sites formed by uracil excision, allowing also for C:G transversions, or c) nonconservative mismatch repair of the U:G pair, triggering A:T mutagenesis via Polη (1); 4) misprocessing of the lesions may lead to chromosomal translocations (5); and finally, 5) base excision repair (BER) may restore the original sequence (6). The pathways governing the choice between these options are largely enigmatic to date. Most importantly, the question of why error-free processing (e.g., by BER) does not occur at the normal high efficacy in Ig genes is a mystery.
In mammalian cells, lesions that fail to be repaired by BER before entrance into S phase can then be processed by HR (7). In fact, in this case the HR pathway becomes essential for the prevention of replication collapse, genetic instability, or cell death, as double-strand breaks formed upon encounter of a single-strand break by the replication fork can only be processed in an error-free manner by HR (8). A reminiscent scenario has been shown for AID-induced DNA double-strand breaks occurring in G1 in non-Ig genes (off-target sites) of activated B cells, which require HR for resolution in S/G2 (9, 10). Unless completely resolved in the G1 phase of the cell cycle, remaining AID-induced lesions in the Ig genes should thus in part also be processed by HR. However, the profound toxicity shown before for HR defects in AID-expressing cells (10) may raise concerns that the investigation of HR function during SHM in mouse cells, which requires a functional germinal center response, might be a risky endeavor.
Gene-targeting studies in the chicken DT40 cell system, which is an established model for Ig gene conversion (4), have led to an intriguing concept of the relationship between HR and SHM. In particular, inactivation of HR leads to abolishment of Ig gene conversion and a concomitant switch to SHM based on TLS (11). We have previously shown that checkpoint signaling via Chk1 decreases SHM in chicken and human B cells, and our data in DT40 cells imply a role of HR regulation via Chk1 in this phenomenon (12). Conceptual application of this model to the mammalian situation is limited until a role for HR in Ig gene SHM is shown in vivo.
To clarify this issue, we generated a conditional knockout of the early recombination factor Brca2 in mouse B cells (13, 14). Brca2 is crucial for the loading of Rad51 onto resected DNA to promote HR at double-strand breaks or stalled replication forks (15). We show that ex vivo–activated Brca2-deficient B cells proliferate less and die more and are impaired in class switch recombination, although to a milder extent than observed before in Xrcc2-deficient cells (10). Strikingly, we found that Brca2-deficient murine B cells may survive AID expression and form germinal centers where SHM and affinity maturation do occur. Although the overall mutation load was lower in Brca2-deficient cells, pattern analysis showed a significant relative increase in C:G transversion mutations but surprisingly also a drop in relative A:T mutagenesis. In silico analysis of changes in mutation patterns upon mutation accumulation indicated that the relative probability of A:T mutator activity increases over the course of the germinal center reaction in normal but not Brca2-deficient germinal center B cells. This increase of A:T mutagenesis seemed to be specific for the IgH V region and to pose a serious threat to genomic stability that is normally counterbalanced by HR.
Materials and Methods
Mice
Brca2fl/+ mice on a C57BL/6 background were obtained from P.O. Frappart with permission of Dr. J. Jonkers (13). Mb1-cre mice on a C57BL/6 background were obtained from Dr. C. Kosan with permission of M. Reth (14). Brca2fl/flMb1cre/+ and Brca2+/+Mb1cre/+ littermates were generated from intercrosses of Brca2fl/+Mb1+/+ and Brca2fl/+Mb1cre/+ mice. Genotyping was performed with the primers and conditions indicated in Supplemental Table I. Mice were bred in specific pathogen-free conditions, and all animal experiments were approved by the Thüringer Landesamt für Verbraucherschutz. For the immunization experiments, five mice per genotype per addressed question were chosen to meet the requirements of the principle for reduction of animal numbers while at the same time minimizing variations caused by individual immunizations. No randomization or blinding was done.
B cell isolation and stimulation
B cells were isolated from splenic single-cell suspensions of 8–16-wk-old male and female mice using MACS depletion with α-CD43 beads (Miltenyi Biotec). A total of 2 × 105
Flow cytometric analyses
Immunization, germinal center cell sorting, and analysis of JH4 and pre-Sμ mutagenesis
Eight- to twelve-wk-old male and female mice were immunized by peritoneal injection of 100 μg of alum-precipitated nitrophenylacetyl (NP) chicken γ globulin (NP-CGG; Biosearch Technologies). Blood samples for ELISA analyses were taken from the cheek 7 and 10 or 7 and 14 d after injection. Splenic B cells were sorted for germinal center cells (B220+, PNAhigh, CD95+) in a FACSAria (BD Biosciences) 14 d after immunization, followed by genomic DNA isolation and amplification of the IgH JH4 intron (using primers specific for VH186.2 and JH4) or the pre-Sμ region. All primers and PCR programs are listed in Supplemental Table I. Gel-purified PCR fragments were excised and cloned into pGEM-T (Promega). Individual colonies were picked and sequenced by the Sanger method using the JH4 reverse or the Sμ reverse primer. The so-obtained sequences had to be inverted to analyze the actual coding strand. Sequence alignments were done in Geneious (Biomatters), and mutational analyses were performed using the SHMTool (http://shmtool.montefiore.org/cgi-bin/p1) (16). Whenever clonally identical mutated sequences were found, all but one were excluded from the analysis to rule out the influence of highly abundant B cell clones on the mutational pattern.
Analysis of non-Ig mutagenesis
Twenty-seven– to thirty-three–wk-old mice were sacrificed and a single-cell suspension of Peyer patches was prepared, which was then sorted for germinal center and nongerminal center cells as described above. Amplification of the Cd83 gene was done from genomic DNA, and the Cd83 forward primer was used for Sanger sequencing. The resulting sequences were analyzed as described for JH4 and pre-Sμ.
ELISA
IgG1 titers of activated B cell culture supernatants were determined by coating plates with α-IgG1 (BD Pharmingen) and detecting with α-IgG1–biotin (BD Pharmingen) using an O-phenylenediamine substrate (Sigma) and purified IgG1 (BD Pharmingen) dilutions as standard. Absorbance was measured at 492 nm after stopping the reaction with 3 N HCl. Analyses were done in duplicates with three different sample dilutions each.
For measuring the amount of NP-binding IgG1 Abs in mouse sera, plates were coated with NP3 or NP15 (Biocat), and pooled serum from immunized mice was used as standard. Serum samples were serially diluted starting with 1:100 (day 7), 1:400 (standard), and 1:500 (day 14). Detection was performed as described above.
Statistical and bioinformatics analysis
For mutation pattern analyses, p values were calculated using the two-sided χ2 test for 2 × 2 contingency tables. When the prerequisites for χ2 were not given, a two-sided Fisher exact test was used instead. For all other analyses, p values were calculated using a two-sided Student t test with *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 after checking for homo- or heteroscedasticity by an F-test. Bar graphs show the mean with error bars depicting the SD; n refers to the number of mice per genotype.
The algorithm used for pseudo-timing of mutation accumulation was implemented in Python. Starting from the fasta file assembly required for the SHMTool, it generates bins of sequences with k = 1, 2, 3, 4, etc. mutations, and subsequently analyzes the mutational pattern in these sequences as the total number of each type of base change in the respective bin, divided by the total amount of bases in the bin for which such a change would be possible (frequency of original base in the consensus sequence times number of sequences in the bin). Visualization of the results was achieved using the Matplotlib library.
The relative mutation frequency was denoted φ, and for a given number of mutations we calculated the posterior probability density function (PDF) of φ by Bayes theorem as follows:(1)Where D represents the data (i.e., sequences with k mutations), and the likelihood function is written as follows:
(2)In Eq. 2,
denotes the number of mutations of type M from sequence i originating from either
or
Brca2BKO, and
is the number of sequences with exactly k mutations. The number of mutations follows a binomial distribution governed by the relative mutation frequency as follows:
(3)We chose a flat, uninformative prior,
(4)which does not bias the value of φ.
To quantify changes in the relative mutation frequency as a function of the number of overall mutations, we performed linear fits to values drawn from the posterior PDF using Monte Carlo simulation (17, 18). Each fit is parametrized by the slope (aj) and intercept (bj). Values of φ were drawn 1000 times and 95% confidence intervals (CIs) for the differences of slope and intercept between Ctrl and Brca2BKO were calculated. Differences in intercept reveal whether the basal relative mutation frequencies were different between Ctrl and Brca2BKO, whereas differences in slope indicate differences in acceleration or retardation of the respective mutation.
Data and computer code availability
The JH4 sequences presented in this article have been submitted to GenBank (https://www.ncbi.nlm.nih.gov/genbank/) under accession numbers MN098330 to MN098764. The Python code for the pseudo-time analysis and the underlying JH4 alignments are available at https://github.com/noxxi/bioinfo-pseudotiming-somatic-hypermut. The code for the Bayesian inference pipeline is available at https://github.com/applied-systems-biology/Bayesian-Analysis-SHM.
Results
We inactivated HR specifically in B cells by crossing the previously described Brca2fl/fl mice (13) to Mb1-cre mice (14) and used Brca2fl/flMb1cre/+ (henceforth termed Brca2BKO) and Brca2fl/flMb1+/+ or Brca2+/+Mb1cre/+ (henceforth termed Ctrl) mice for further analyses. As expected for the Mb1-cre driver (14), B cell–specific deletion of Brca2 occurred at close to 100% efficiency (Fig. 1A). Despite the reported essential function of HR for normal B cell development (19), FACS analyses revealed that splenic T versus B cell ratios as well as relative numbers of marginal zone versus follicular cells were not significantly altered in Brca2BKO as compared with Ctrl mice (Fig. 1B). We thus conclude that Brca2BKO mice are suitable for the analysis of the influence of HR on late B cell maturation events.
Normal splenic B cell composition in Brca2BKO mice. (A) Schematic representation of the modified Brca2 locus with PCR primers used for the determination of deletion efficiency in purified B cells. LoxP sites flank exon 11 (approximately 4800 bp), encompassing the BRC repeats responsible for Rad51 binding. Arrows mark the PCR products indicative of wildtype, floxed, or deleted alleles. (B) Representative FACS plots and percentages of splenic B and T cells and marginal zone versus follicular B cells in Brca2BKO and Ctrl mice. Cells were pregated on alive (DAPI-negative) single cells. Mean and SD determined for five and three mice of each genotype are given, two-sided Student t test. p > 0.5.
Stimulation of isolated Brca2BKO and Ctrl splenic B cells with α-CD40/IL-4 to induce AID expression and initiate class switch recombination from IgM to IgG1 revealed a substantial defect of Brca2BKO B cells in survival, leading to an increasing loss of living cells of up to 50% until day 4 after stimulation (Fig. 2A). This was congruent with a previous study (10) although less pronounced. The survival defect was accompanied by decreased proliferation of the cells stimulated with α-CD40/IL-4 (Fig. 2B). CSR is known to be dependent on proliferation. Accordingly, the appearance of class-switched surface IgG1-positive cells was reduced by half in Brca2BKO B cells (Fig. 2C), as was IgG1 secretion into the culture supernatant measured by α-IgG1 ELISA (Fig. 2D). As CSR is mechanistically based on nonhomologous end-joining processes, in which Brca2 is not known to play a role, we presumed that the lack of class-switched cells was due to the observed proliferation and survival problems. Stimulating with α-CD40 only, and thereby not inducing proliferation, resulted in significant but milder survival defects in Brca2BKO B cells (Supplemental Fig. 1A, 1B). A mitigated reduction of class-switched Brca2BKO B cells was observed when comparing populations with the same number of cell divisions (Supplemental Fig. 1C, 1D). Brca2 deficiency thus leads to reduced class switching, supposedly by impairing proliferation and survival of ex vivo–activated B cells, confirming the cytotoxic phenotype of AID expression in HR-deficient cells (10).
Reduced survival, proliferation, and class switching in Brca2BKO B cells. (A) Survival of B cells isolated from spleens upon stimulation with α-CD40 and IL-4 as determined by FACS analysis on day 0, day 3, and day 4 (d 0/3/4). Living cells were defined as TO-PRO–negative signals of total events. Mean with SD of five mice per genotype is shown. Significance was calculated using a two-sided Student t test. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. (B) Proliferation of purified CFSE-stained B cells stimulated with α-CD40 and IL-4 measured by FACS on day 1, 3, and 4. Representative experiment with n = 2 mice per genotype. (C) Percentage of cells with surface IgG1 expression among alive cells at the indicated days after stimulation measured by FACS. Cells were pregated on single cells. Mean and SD of seven mice per genotype are shown. (D) IgG1 secretion of stimulated B cells, as determined via α-IgG1 ELISA in culture supernatants at day 4. Shown is the mean and SD of five mice per genotype, analyzed in duplicates.
We next asked whether Brca2BKO mice are capable of forming normal germinal centers and conducting successful affinity maturation despite survival and proliferation defects. Therefore, we immunized Brca2BKO and Ctrl mice with the model Ag NP-CGG and analyzed the frequency of germinal center cells as well as the distribution into light and dark zone at day 10 and 14 postimmunization by FACS. PNAhighCD95+ germinal center cells could clearly be observed in the Brca2BKO mice with approximately the same frequency as in Ctrl mice at day 10 (Fig. 3A, 3C) but with reduced frequency at day 14 (Fig. 3B, 3D, no statistical significance because of one outlier). This implied ongoing depletion of HR-deficient germinal center cells over the time course of the germinal center reaction. Comparing the amount of centroblasts in the dark zone or of centrocytes in the light zone, indicative of cells undergoing high rates of proliferation and mutagenesis or selection and differentiation, respectively (20), revealed a normal distribution of germinal center cell subsets in Brca2BKO compared with Ctrl mice (Fig. 3A–D). To assess the output of the germinal center reaction, the accumulation of NP-specific Abs in blood was analyzed at day 14 (Fig. 3E). Brca2BKO mice showed a strongly reduced amount of overall (α-NP15) and of high-affinity (α-NP3) IgG1 Abs compared with Ctrl mice. Despite the differences between individual immunization experiments, the overall decrease in serum NP-specific IgG1 Abs approaches significance. However, the presence of NP3-binding high-affinity Abs in Brca2BKO mice implied that the germinal center response, albeit impaired, is in principle functional.
Germinal center formation and affinity maturation in Brca2BKO mice. (A and B) Representative FACS plots of B220+CD95+PNAhigh germinal center cells, CXCR4highCD86low dark zone (DZ), and CXCR4lowCD86high light zone (LZ) cells in spleens of Ctrl and Brca2BKO mice 10 (A) and 14 (B) d after immunization with NP-CGG. Cells are pregated on alive (DAPI-negative) single cells. (C and D) Percentages of germinal center or light and dark zone cells in immunized Ctrl and Brca2BKO mice and nonimmunized (n.i.) Ctrl mice on day 10 (C) and day 14 (D) after immunization. n = 5 (day 10), n = 8 (day 14, among which n = 3 for the DZ/LZ staining). Significance was calculated using a two-sided Student t test. ***p < 0.001. (E) Serum titers of IgG1 Abs specific for NP 14 d after immunization from five mice shown in (D). Binding to NP15 detects most NP-specific Abs, whereas binding to NP3 indicates high affinity. The values represent arbitrary units relative to a pooled serum control of NP-immunized mice, which was defined as 500 U. Each dot represents the mean of technical replicates derived from serially diluted serum of one mouse. There was no detection of NP-specific Abs in n.i. mice in any of the experiments. Significance was calculated using a two-sided Student t test. (F) Frequency of the W33L exchange at day 14 after immunization indicative of affinity maturation in the NP-immunized mice shown in (D). Data are based on a total of 34 and 53 V186.2 sequences of Ctrl and Brca2BKO mice, respectively; each dot represents data acquired from one mouse.
To evaluate the ability of Brca2BKO mice to undergo the full affinity maturation program more clearly, we sorted PNAhighCD95+ germinal center cells and PNAlowCD95− nongerminal center cells at day 14 postimmunization from five Brca2BKO and five Ctrl mice and assessed the selected V186.2 gene implied in the NP response for the telltale W33L mutation indicative of increased Ab affinity. Intriguingly, the frequency of this mutation in germinal center cells of Brca2BKO mice reached the level of some of the Ctrl mice (Fig. 3F), implying that the decrease seen in serum high-affinity IgG1 Abs (Fig. 3E) may be due to survival defects during class switching rather than the actual hypermutation and selection processes.
To directly analyze the role of HR in somatic hypermutation, we performed sequence analysis of the nonselected JH4 intron flanking the V186.2 gene rearrangement used in the NP response from splenic germinal center cells (Fig. 4). Previous studies in DT40 cells with a defect in HR or HR-activating pathways showed an increase of mutagenesis or point mutations (11, 12, 21, 22). However, the overall mutation load in the Brca2BKO B cells was lower than in Ctrl mice with a reduction of the mutation frequency of ∼50% (Fig. 4A, 4B). This coincided with a lower accumulation of highly mutated sequences in Brca2BKO mice (Fig. 4A, 4C). Because the Brca2 defect led to an increasing loss of cells over time (Figs. 2A, 3C, 3D), the more pronounced reduction in the mutation frequency in the JH4 intron may be attributed to the loss of highly mutated clones at later time points of the germinal center reaction.
Altered SHM in Brca2BKO mice. (A) Mutated sequences detected in the Jh4 intron in germinal center B cells isolated at day 14 post NP-CGG immunization from five mice of each genotype. Numbers of analyzed sequences are given in the middle of the pie charts; the fractions show the portion of sequences with 0, 1, 2, etc. mutations. The mutation frequency was determined as mutated bases divided by all analyzed bases among all sequences (or among mutated sequences only, shown in parentheses). Significance was calculated comparing the total number of mutated versus unmutated bases of all mice per genotype with a two-sided χ2 test. ***p ≤ 0.001. (B) Absolute numbers of nonunique base substitutions in the sequences analyzed in (A), compiled for all mice of one genotype. (C) Amount of JH4 intronic sequences with a certain number of mutations. (D) Relative frequency of transition versus transversion mutations at C:G and mutations at C:G versus at A:T among a total of 578 (Ctrl) and 350 (Brca2BKO) mutations. Significance was calculated using a two-sided χ2 test comparing absolute numbers of the indicated types of mutations [shown in (B)] between Ctrl and Brca2BKO (i.e., comparing TS at C:G versus TV at C:G, or comparing TS at C:G versus all other mutations). *p ≤ 0.05, **p ≤ 0.01. No correction for base composition of the JH4 intron was applied for this figure.
Strikingly, Brca2BKO mice also showed significant differences in the mutation pattern of JH4 sequences compared with Ctrl mice (Fig. 4B, 4D, Supplemental Fig. 2A). To get an idea about changes in the different mutator systems described in the model of Neuberger and Di Noia (1), we looked more closely at transitions and transversions at C:G and mutations at A:T (Fig. 4D). At C:G residues, Brca2BKO mice displayed a relative increase in transversion mutations (Fig. 4D), indicating that TLS over the abasic site generated by uracil excision (1) is proportionally increased. Concomitantly and unexpectedly, relative mutagenesis at A:T residues was significantly decreased. This difference also remains significant when the data of the five mice per genotype were not compiled (Supplemental Fig. 2B).
An increase in transversions at C:G, which are mostly caused by activity of the TLS polymerase Rev1 (23) upon HR inactivation, would be consistent with observations in the DT40 system (11, 12). This implies that a certain fraction of DNA lesions is in fact repaired in an error-free manner via HR, thereby preventing the processing of these lesions via TLS.
At the same time, the observed changes in A:T mutagenesis, which is mediated by noncanonical MMR and TLS Polymerase η (24), were unexpected and could, in fact, not be studied in the DT40 system before. Considering the lower amount of highly mutated JH4 sequences in Brca2BKO cells (Fig. 4A, 4C), we wondered whether this might be linked to the decreased A:T mutagenesis. We first compared the Rev1 and Polη hallmark mutations in lowly and highly mutated JH4 sequences (Fig. 5A). In Brca2BKO cells, the relative frequency of G>C and C>G mutations, indicative of Rev1 activity, is higher in both cases, implying that a preference for inserting transversions instead of transitions at C:G occurs throughout the process. Intriguingly, in case of the Polη hallmark mutation A>G (or T>C on the other strand), a very different picture emerged: Brca2BKO cells with lowly mutated sequences have as many or even more mutations caused by Polη than the Ctrl, but in highly mutated sequences, the relative amount of these mutations decreased significantly compared with the Ctrl.
In silico staging and mutation classification in Brca2BKO mice. (A) Relative frequency of Rev1 (G>C, C>G) and Polη (A>G, T>C) hallmark mutations as percentage of a total of 76 (Ctrl) or 122 (Brca2BKO) mutations found in lowly mutated sequences (one and two mutations per sequence) and as percentage of a total of 140 (Ctrl) or 111 (Brca2BKO) mutations found in highly mutated JH4 sequences (four and five mutations per sequence) compared with all observed base changes; significance was calculated using the two-sided χ2 or Fisher exact test comparing the absolute number of a certain base change to all other base changes among (e.g.) lowly mutated sequences. *p ≤ 0.05. (B) In silico pseudo-timing analysis of Rev1- and Polη-mediated mutational events in Ctrl and Brca2BKO mice. Mutation pattern of sequences containing k = 1–5 mutations are shown. The y-axis depicts mutation frequencies of the indicated base substitution, corrected for the abundance of the original base within the respective sequences. (C) The posterior PDF of the relative mutation frequency (φ) of Rev1 (G>C, C>G) and Polη (A>G, T>C) hallmark mutations in pseudo-time (x-axis) for Ctrl and Brca2BKO as density maps. The lines show the median fit based on 1000 Monte Carlo simulations. For full details on significant differences between Ctrl and Brca2BKO see Table I.
Because this finding suggested a differential evolution of mutation accumulation, we decided to stage mutational processes in the germinal center reaction on the basis of the sequence data obtained. B cells entering the germinal center undergo multiple cycles of proliferation, mutation, and selection with an overall mutation rate of ∼1 mutation per V gene sequence per cell division (25, 26). B cells with sequences harboring more mutations have thus undergone more “productive cycles” of mutation insertion, and the history of mutational events can therefore be traced by sequentially evaluating mutation patterns in sequences carrying k = 1, 2, 3, 4 or more mutations, establishing an in silico pseudo-timing of mutation accumulation.
Using a custom Python algorithm, we visualized mutation frequencies for each possible base change depending on their appearance in sequences with certain numbers of mutations (k) (Fig. 5B, see Supplemental Fig. 3A for every base change). Most types of base changes showed a relatively steady mutation accumulation in sequences from both Ctrl and Brca2BKO B cells, whereas the situation for the Polη hallmark mutation A>G (e.g.) was strikingly different. Although a steep increase was observed in Ctrl sequences, in case of Brca2BKO, these base changes were inserted with similar efficiency during early productive cycles but then failed to further accumulate.
To more clearly evaluate this idea, the posterior PDF of the relative mutation frequency, denoted by Φ, was calculated using Bayesian methods (27, 28) (Fig. 5C). The relative mutation frequency Φ describes the probability to observe a certain base change in sequences with k = 1, 2, 3, 4, 5 mutations, thus now depicting the relative activity of the respective mutator system at any given step in our pseudo-timing approach, rather than the result (as in Fig. 5B). Complete data from this simulation are shown in Supplemental Fig. 3B, whereas Fig. 5C focuses on the Rev1 and Polη hallmark mutations. This analysis revealed that for most base changes/mutator systems, the activity is relatively constant throughout the process of mutation accumulation. For Polη-mediated A>G mutations, an increase in relative activity is seen at later pseudo-timepoints in Ctrl samples, whereas for Brca2BKO cells, the activity remains constant at the initial level throughout the process. To determine whether there are significant differences in the mutation accumulation curves of Ctrl and Brca2BKO clones, we calculated 95% CIs of the slope and intercept of linear fits to the relative mutation frequency in pseudo-time using Monte Carlo simulations based on the posterior PDFs (Table I). No significant difference in initial relative mutation rate was detected between Ctrl and Brca2BKO (see values for differences in intercept [bj] in Table I). For Polη-mediated A>G mutations, in contrast, the fitted slope is significantly steeper in Ctrl than Brca2BKO (see values for differences in slope [aj] in Table I).
To tackle the question why Brca2BKO mice fail to increase the A:T mutator activity at a certain (pseudo-)time point, we decided to look at AID-induced mutagenesis in other gene loci: the pre-Sμ region and non-Ig genes targeted by AID like Cd83. Class switch recombination and thereby mutagenesis of the Sμ region is considered to be an early process in the germinal center reaction and even starts before the germinal center has fully matured (29, 30), whereas SHM, leading to V region mutagenesis, supposedly reaches its maximum when the dark zone is established where proliferation and AID and Polη expression are at height (20, 31, 32). The mutation frequency in the pre-Sμ region of Brca2BKO splenic germinal center cells was reduced by only ∼22%, and the lack of highly mutated clones was not as evident as among the JH4 sequences (Fig. 6A, 6D). In contrast to the JH4 intron, no significant changes were detected in the mutation pattern of the pre-Sμ region (Fig. 6B), with A:T mutagenesis being generally lower than in JH4 sequences (Figs. 4D, 6B). To analyze the Cd83 gene, which is a known off-target of AID (2), we isolated genomic DNA from Peyer patch germinal center cells of unimmunized Ctrl and Brca2BKO mice. In this study, Brca2BKO cells had a similar to somewhat higher (not significant) mutation frequency compared with Ctrl cells (Fig. 6C, 6D). The overall very low mutation frequency in non-Ig genes, however, did not allow a mutation pattern analysis because only 36 (Ctrl) and 51 (Brca2BKO) nonunique mutations were found in the Cd83 sequences out of which 10 (Ctrl) and 8 (Brca2BKO) were mutations at A:T.
Analysis of pre-Sμ and Cd83 sequences in Brca2BKO mice. (A) Mutation frequency of pre-Sμ sequences 14 d after NP-CGG immunization of four mice per genotype among all sequences (and among mutated sequences only, shown in parentheses), *p ≤ 0.05 (two-sided χ2 test of mutated versus unmutated bases of all Ctrl and all Brca2BKO mice). Numbers of analyzed sequences are given in the middle of the pie charts; the fractions show the portion of sequences with 0, 1, 2, etc. mutations. (B) Relative mutation frequency as percentage of a total of 193 (Ctrl) and 151 (Brca2BKO) detected mutations among all pre-Sμ sequences. No significances were detected in a two-sided χ2 test. (C) Mutation frequency of Cd83 sequences from isolated Peyer patch germinal center B cells analyzed as in (A). Mice were 27–33 wk old, three mice per genotype. (D) Comparison of mutation frequencies (among all sequences) in the JH4 intron, the pre-Sμ region, and the Cd83 gene for each mouse tested. Significance was calculated using a two-sided Student t test. **p ≤ 0.01.
We may thus conclude that a deficiency in Brca2 does not affect the mechanism of A:T mutagenesis as such, but rather interferes with the accelerated manifestation of A:T mutations in more highly mutated JH4 sequences. This suggests that HR-deficient cells cannot survive once A:T mutagenesis has reached a certain threshold. The pre-Sμ region, in contrast, showed an overall lower A:T mutagenesis than the JH4 intron (Figs. 4D, 6B). Thus, the critical threshold might not yet be exceeded during Sμ mutagenesis. Likewise, non-Ig mutagenesis with even lower A:T mutagenesis did not show any of the effects observed in JH4 sequences, suggesting that the increase of the A:T mutator activity and the subsequent challenges for genomic stability are specific for the V region and later phases of the germinal center reaction.
Discussion
In the current study, we identified Brca2floxMb1cre mice as a model system to study the role of HR on the germinal center reaction. B cells lacking Brca2 showed a profound survival and proliferation defect in ex vivo experiments, leading to a reduced amount of class-switched cells. We observed a functional germinal center reaction in Brca2BKO mice but with reduced numbers of germinal center cells as well as affinity-matured Abs. Importantly, we showed that error-free repair by HR is active at the hypermutating V region of the Ig locus. Upon its inactivation, B cells showed a higher propensity to process some C:G lesions by error-prone TLS, leading to a relative increase in C:G transversions. Simultaneously, HR was not able to prevent the vast majority of AID-induced point mutations but is, on the contrary, indispensable for the manifestation of highly mutated B cell clones and a supposedly V region–specific increase in A:T mutagenesis.
The lesions introduced by AID, cytidine deaminations processed to abasic sites, and single-strand breaks, are among the most frequent spontaneous changes in our genome, occurring between 100 and 10,000 times per day in each cell of our body. The repair pathways dealing with them, BER and HR, are accordingly robust, and their interplay is well described (7); lesions that are left unprocessed by BER will become targets of HR in replicating cells. Together with a previous study on XRCC1 (6), our analysis allows for the conclusion that both standard repair pathways for the critical lesions introduced by AID are active at hypermutating Ig loci in vivo yet fail to complete error-free processing. Although this may be due to a sheer cellular damage overload, such a scenario is unlikely, as it has been shown directly that the repair defect is locus-specific (2). Accordingly, so must be the deregulating mechanism, whatever its nature.
Landmark studies for the chicken B cell line DT40 (11, 33, 34) have shown that the repair process taking over after HR inactivation is TLS, leading to transversions at C:G. Although our study now also shows for murine B cells a relative increase in C:G transversions, the effect is not as substantial as in DT40. Chicken cells normally diversify their Ig genes by gene conversion, a process taking place in the chicken’s bursa (35) using HR in cis and presumably in G1 (36), whereas the error-free HR pathway we have assessed is restricted to S/G2 in mammalian cells when a sister chromatid is available. This, in contrast, confirms that SHM via TLS over abasic sites partially takes place in S/G2, where it competes with HR for the repair of these lesions, a hypothesis that has also been suggested by Sharbeen et al. (37).
In the DT40 system, A:T mutagenesis is essentially absent (38). In mice and humans, A:T mutations relying mostly on Polη can be observed (24, 39). Brca2 has been reported to interact with Polη during repair synthesis by HR (40), so direct effects of the Brca2BKO on A:T mutagenesis would be possible, provided the A:T mutator were dependent on the HR machinery or replication. This has not been reported so far, although canonical mismatch repair is tightly associated with replication (41). A:T mutagenesis during SHM is instead believed to happen in the G1 phase only (42). The results of our in silico analysis also imply that HR is not required for A:T mutagenesis as such, but rather for the efficient manifestation of A:T mutations in highly mutated sequences. These sequences represent cells that have performed more productive cycles or have advanced further in the germinal center reaction. We surmise that this is an indirect and secondary effect due to a higher susceptibility to cell death of those cells acquiring their single mutational event per cycle via A:T mutagenesis. Notably, A:T mutagenesis requires strand breaks and by exonuclease activity creates stretches of ssDNA (43), whereas C:G mutagenesis does not. We hypothesize that during the first productive cycles, AID-induced DNA damage is still on a tolerable level for HR-deficient cells, but when the frequency of DNA lesions processed by the A:T mutator increases, the cells reach a nontolerable threshold and die. This implies that the mechanism of A:T mutagenesis is challenging for genomic stability, presumably by leaving unprocessed DNA single-strand breaks and/or stalling the replication fork. Interestingly, mice with a defect in the replication fork-stabilizing Fanconi anemia pathway showed a reduction in Polη-mediated A:T mutagenesis and a lack of highly mutated sequences, reminiscent of what we observed in Brca2BKO mice (44).
Our initial analysis of lowly and highly mutated sequences as well as the pseudo-timing analysis revealed that the occurrence of problems in A:T mutagenesis in BRCA2BKO is delayed. Whether this reflects the timing of high Bcl-6 induction in mutating B cells (45, 46) which would interfere with checkpoint responses (47, 48) or rather the onset of dark zone formation, of which Polη is a signature gene (32), and where proliferation and AID expression is highest (31), will need to be investigated with tools enabling the tracing of single mutator system activities in B cells throughout the germinal center response. Further investigations into the timing of mutational events and potential subsequent pattern differences would also be of interest for the identification of germinal center exit time points of (e.g.) memory B cells.
As a general notion, based on our finding of differential evolution of mutagenesis throughout the germinal center response, we would like to point out the importance of considering proliferation or survival defects when interpreting SHM data, in particular concerning A:T mutagenesis. We also support the idea of exploiting the profound problems of HR-defective cells caused by AID-activity (10) for the treatment of AID-expressing cancer types, as it has already been approached (49), and would like to contribute as an addition that specifically the A:T mutagenesis machinery may play an important role in AID-induced toxicity under these conditions.
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We thank J. Jonkers, M. Reth, and E. Hobeika for providing transgenic mouse models, the staff of the animal facility of the Faculty of Biological Sciences for expert caretaking, and the FACS facility of the Fritz Lipmann Institute for cell sorting. We thank all members of the Jungnickel laboratory and C. Kosan for critical reading of the manuscript, members of the Kosan laboratory for sharing expertise and reagents, and A. Schmidt, J. Efremova, L. Giebeler, A. Müller, and P. Gruebner for expert technical assistance.
Footnotes
This work was supported by grants from the Deutsche Forschungsgemeinschaft (JU2690/4-1, JU2690/1-2) and the Jena School of Microbial Communication (all to B.J.), CRC/TR124 FungiNet Projects B4 (to M.T.F.) and C4 (to B.J.), and the Deutsche Krebshilfe (Grant 70112155 to B.J.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
- AID
- activation-induced cytidine deaminase
- BER
- base excision repair
- CI
- confidence interval
- HR
- homologous recombination
- NP
- nitrophenylacetyl
- NP-CGG
- NP chicken γ globulin
- probability density function
- TLS
- translesion synthesis.
- Received April 29, 2019.
- Accepted July 4, 2019.
- Copyright © 2019 by The American Association of Immunologists, Inc.