ATM influences germinal center integrity

Abstract

The DNA damage response protein ATM has long been known to influence class switch recombination (CSR) in ex vivo cultured B cells. However, an assessment of B cell-intrinsic requirement of ATM in humoral responses in vivo was confounded by the fact that its germline deletion affects T cell function, and B:T cell interactions are critical for in vivo immune responses. Here we demonstrate that B cell-specific deletion of ATM in mice leads to reduction in germinal center (GC) frequency and size in response to immunization. We find that loss of ATM induces apoptosis of GC B cells, likely due to unresolved DNA lesions in cells attempting to undergo CSR. Accordingly, suboptimal GC responses in ATM-deficient animals are characterized by decreased titers of class switched antibodies and decreased rates of somatic hypermutation. These results unmask the critical B cell intrinsic role of ATM in maintaining an optimal GC response following immunization.

INTRODUCTION

Upon engaging antigens, mature B cells undergo immunoglobulin heavy chain (Igh) class switch recombination (CSR) and somatic hypermutation (SHM) (1, 2). During CSR, AID-instigated generation of DNA double strand breaks (DSBs) at the Igh locus leads to the replacement of the default Cμ constant region (Ch) exons with an alternate set of Ch exons (Cγ, Cε, Cα) so that the B cell switches from expressing IgM to one expressing a secondary antibody isotype (IgG, IgE or IgA) (1, 2). SHM proceeds through AID-induced mutations at the variable region exons of both Igh and light chain (Igκ, Igλ) genes and occurs primarily in the context of highly specialized, transient microanatomical structures called germinal centers (GCs) that develop in the follicles of secondary lymphoid organs (3). GCs comprise a complex array of cell types including B cells, T-follicular (Tfh) cells, follicular and conventional dendritic cells, and macrophages. Within GCs, B cells undergo iterative cycles of proliferation and SHM, followed by Tfh-mediated selection of higher-affinity B cells, leading to affinity maturation of the B cell receptor (3). While CSR can occur in the GC and in the extra-follicular regions, SHM occurs almost exclusively in the GC (3).

The serine-threonine kinase ATM (ataxia telangiectasia mutated) is a critical regulator of the DNA damage response (DDR) (4). ATM is recruited to and activated by DSBs to phosphorylate several key proteins that initiate activation of the DNA damage checkpoint leading to cell cycle arrest and DNA repair (4). Mutations in ATM cause ataxia telangiectasia (A-T) in humans, a neurodegenerative disorder associated with immunodeficiency, genomic instability and a marked predisposition to lymphoid malignancy (4). Work from many laboratories over the past decade demonstrates that B cells from both A-T patients and ATM-deficient mice have profound CSR defects (5–10). However, many of these studies were limited to studying ATM within the context of ex vivo activated splenic B cells. Investigating the B cell intrinsic function of ATM in vivo has proven challenging, largely due the known impact of ATM deficiency on T cell development, and the integral role T cells play in orchestrating humoral immune responses, (10, 11). Because GCs requires close collaboration of B and T cells (3), the B cell intrinsic requirement of ATM in vivo can only be assessed following B cell specific deletion of ATM. Here we demonstrate that ablation of ATM in B cells impairs GC B cell responses, and decreases class switched antibodies and frequency of SHM following immunization. We observe an increase in GC B cell apoptosis, perhaps due to a failure to resolve DSBs, leading to a disruption of germinal center integrity.

MATERIALS AND METHODS

Mice

Atm^fl/fl, Atm^KD/fl, and CD21-cre mice were previously characterized (12–14). Whenever possible, littermates and mice of both sexes were analyzed. Animals were housed in a pathogen-free facility approved by the Institutional Animal Care and Use Committee of MSKCC.

Cell culture, flow cytometry and sorting

For ex vivo assays, naïve splenic B cells were purified by CD43 negative selection (Miltenyi Biotec), cultured at a density of 10⁶ per ml, and stimulated with either LPS (30 μg/ml, Sigma), LPS (30 μg/ml) + IL-4 (12.5ng/ml, R&D Systems), or LPS (10 μg/ml) + TGF-β (2ng/ml, R&D Systems) + anti-IgD dextran conjugates (300 ng/ml, Fina BioSolutions) for CSR to IgG3, IgG1, or IgA, respectively. For proliferation assays, naïve splenic B cells were labeled with 5 μM CellTrace Violet (CTV, Thermo Fisher Scientific) according to manufacturer’s protocol, activated with LPS, LPS+IL-4 or LPS+TGFβ+ anti-IgD dextran and dye dilution tracked from d0 to d4. Antibodies for flow cytometry were as follows: B220 (BV510, FITC, PerCPCy5.5; clone RA3–6B2), IgA (PE; clone mA-6E1), IgG1 (BV510, APC; clone X56), IgG3 (FITC; clone R40–82), CD69 (PE/Cy7; clone H1.2F3), CD86 (AF700; clone GL-1), MHC Class II I-Ab (eFLuor450; clone AF6–120.2), Fas (BV510; clone Jo2), GL7 (FITC, PerCP-eF710; clone GL7) and Zombie Red fixable viability dye; all were purchased from eBioscience, BD, and BioLegend. Samples were analyzed on an LSR II flow cytometer (BD). Cell sorting was carried out in a FACSAria cell sorter (BD). All data analysis was performed using FlowJo software (version 9.9; Tree Star).

Immunofluorescence

Spleen samples embedded in optimal cutting tissue reagent (Sakura) were sliced to sections of 6 to 10μm thickness and stained with appropriate primary and secondary antibodies. Primary antibodies: IgD (Southern Biotech), GL7 (clone GL7; eBioscience), B220 (clone RA3–6B2; abcam), Ki-67 (abcam) and cleaved caspase-3 (clone D175; Cell Signaling). Secondary reagents: Alexa Fluor 488-conjugated anti-rat pAb, Alexa Fluor 546-conjugated anti-rat pAb, and cyanine 5-conjugated streptavidin (Jackson ImmunoResearch Laboratories). Nuclei were visualized with DAPI (Boehringer Mannheim), and sections mounted with FluorSave (Calbiochem). Manufacturer instructions (TACS2 TdT-Fluor, Trevigen) were followed for in situ apoptosis detection. Slides were scanned with Pannoramic Flash (3DHistech, Hungary) using 20x/0.8NA objective, and regions of interest were drawn manually using CaseViewer (3DHistech, Hungary) and exported into TIFF files. Raw unedited images were then analyzed using ImageJ/FIJI where the area of the interest was automatically measured. Signals of interest were thresholded and the area and count were measured. Scoring of desired area was done manually with randomly shuffled pictures to reduce bias.

Immunization and ELISA

Mice (8–10 weeks old) were immunized i.p. with 1×10⁹ packed sheep red blood cells (SRBC, Innovative Research) and boosted with 1×10⁹ SRBC after 10 days. Spleen and Peyer’s patch samples were harvested at d7 or d14 after booster immunization. For NP-CGG immunization, 8- to 10-week-old mice were injected i.p. with 50μg of NP (30)-CGG (Biosearch Technologies) precipitated in 10% alum and boosted with 50 μg NP-CGG on day 14. Serum samples at d0, d14 and d21 post-immunization were analyzed by ELISA. Antibodies and reagents for ELISA have been described elsewhere (15). Endpoint titers were calculated by a one-phase exponential decay curve using Prism7 software (GraphPad).

Somatic Hypermutation

Genomic DNA from sorted GC B cells was amplified using primers targeting framework region 3 (FR3) of the V_HJ558 gene and a sequence located at the 3’ end of the Igh intronic enhancer (Fwd: 5’-GGAATTCGCCTGACATCTGAGGACTCT-3’ and Rev: 5’-GACTAGTCCTCTCCAGTTTCGGCTGAA-3’, respectively) (16) with Phusion High fidelity DNA polymerase (NEB). The ~1.2kb product was cloned using the Zero Blunt TOPO cloning kit (Invitrogen). For each genotype, 40–50 individual clones were sequenced with the internal primer JH4: 5’-CCATACACATACTTCTGTGTTCC-3’. Mutation frequencies were calculated by dividing the total number of unique mutations by the total number of bases sequenced. Sequence alignment was performed using GenBank NC_000078 as consensus sequence.

RESULTS AND DISCUSSION

Loss of ATM impairs CSR ex vivo

We employed two different genetic models of ATM deficiency. In Atm^fl/fl:Cd21-cre mice (referred to as ATM-KO), CD21-cre-mediated deletion of the floxed ATM PIKK domain generates ATM-null mature B cells (12, 14) with no detectable ATM protein (Fig. 1A). In Atm^KD/fl:Cd21-cre (ATM-KD) mice, deletion of the floxed allele generates mature B cells in which the ATM^KD allele encodes a kinase-dead ATM protein as the only source of ATM (13) (Fig. 1A). Atm^fl/fl and Atm^fl/wt:Cd21-cre (referred to as ATM-Het) served as controls for ATM-KO and ATM-KD, respectively. When splenic B cells were activated ex vivo, both ATM-KO and ATM-KD B cells were markedly impaired in their ability to undergo CSR to IgG1, IgG3 and IgA, compared to controls (Fig. 1B). ATM deficiency did not significantly alter B cell proliferation (Fig. 1C, Supplemental Fig. 1A), apoptosis (emental Fig. 1B) or expression of activation markers CD86, CD69 and MHC-II (Fig. 1D). The CSR defect in ATM-KO and ATM-KD B cells is thus consistent with impaired CSR observed in B cells from mice with germline deletion of ATM (5–10).

Figure 1.

ATM deficiency impairs CSR in vitro. (A) Immunoblot for ATM and TRAPP (loading control) in splenic B cells derived from mice of indicated genotypes. Immunoblot is representative of 3 independent experiments. (B) Splenic B cells were stimulated for 72h in culture with LPS+IL-4, LPS or LPS+TGF-β+αIgD, and CSR to IgG1, IgG3 and IgA, respectively, was assessed by flow cytometry. Data represents two independent experiments (mean ± SD). (C) Splenic B cells were labeled with CTV, stimulated with LPS+IL4 and analyzed for CTV dilution at 72h post-stimulation. Data is representative of two independent experiments. (D) Cell-surface expression of the activation markers CD86, CD69 and MHC-II in B cells from unstimulated (control) or stimulated in culture for 72h with LPS+IL4. * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001 (unpaired t-Student with post-hoc Mann-Whitney test).

ATM influences the maintenance of GC B cells

To assess the impact of ATM-deficiency during an in vivo immune response, mice were immunized with sheep red blood cells (SRBC), a complex polyepitope antigen that elicits a T cell-dependent GC response (15). We observed that the frequency of GC B cells (GL7⁺ Fas⁺, gated on B220⁺) was reduced in ATM-KO and ATM-KD mice compared to controls at both d7 and d14 post-immunization (Fig. 2A, 2B, 2D), with a more pronounced and statistically significant reduction observed at d14 (Fig. 2D). Amongst the GC B cells, the frequency of IgG1⁺ B cells was also reduced in both ATM-KO and ATM-KD mice (Fig 2C, E), indicating a CSR defect in vivo. To further explore the reduction in GC frequency, we immunostained splenic sections from SRBC-immunized mice (Fig. 2F) and observed a marked and significant reduction in both the GC area (Fig. 2G) and GC density (Fig. 2H) in ATM-deficient mice relative to control animals. Thus, ATM loss impairs the formation and/or maintenance of GC B cells.

Figure 2.

ATM influences germinal center (GC) responses. Mice were immunized with SRBC, boosted at day 10 and analyzed at d7 and d14 after immunization. (A) Representative flow cytometric analysis of splenic B cells from mice of indicated genotypes at d14 after SRBC immunization. (B, D) Frequency of GC (Fas⁺GL7⁺) cells at d7 and d14 post-immunization in the splenic B220⁺ cell population. (C, E) Frequency of IgG1⁺ cells amongst the GC population. Data (mean ± SD) was compiled from Atm^fl/fl (n=6), ATM-Het (n=6), ATM-KO (n=6) and ATM-KD mice (n=5) for day 7 and from Atm^fl/fl (n=9), ATM-Het (n=5), ATM-KO (n=8) and ATM-KD mice (n=5) for day 14. (F) Representative immunofluorescence image of GC. Spleen sections were stained for IgD (green), GL-7 (red), B220 (grey) and the nuclei counterstained with DAPI (blue). Dashed white boundary indicates the area in which GC B cells were quantified. T, T-cell zone. (G) Quantification of GC B cell area. (H) Germinal center density. Data (G, H) represents mean ± SD compiled from Atm^fl/fl (n=3), ATM-Het (n=3), ATM-KO (n=3) and ATM-KD (n=3) mice at day 14 after SRBC immunization, with at least 100 GCs quantified per genotype. White solid bars represent 100μm scale bar. *, p < 0.05; **, p < 0.01; ***, p < 0.001. (unpaired t-Student with post-hoc Mann-Whitney test).

We further dissected the GC phenotype upon immunization by quantifying the levels of apoptosis in situ. Mouse spleens from SRBC immunized mice were immunostained to detect apoptosis by TUNEL (Fig. 3A, 3B) or generation of cleaved caspase-3 (Fig 3C, D). We observed significantly increased levels of TUNEL-positive and active caspase-3-positive cells in both ATM mutants relative to control mice. Thus, the increased level of apoptosis in ATM deficient GC B cells coincides with the diminished numbers of GC B cells observed after immunization with T-dependent antigens, indicating a failure to maintain a successful GC response.

Figure 3.

ATM-deficient GC B cells undergo increased apoptosis. (A) Representative immunofluorescence of spleen sections, IgD (green), Ki-67 (red), TUNEL (blue). (B) Quantification of active in situ TUNEL signal within GC (dashed white lines). (C) Representative immunofluorescence of IgD (green), cleaved caspase-3 (red), B220 (white) and DAPI (blue). (D) Quantification of cleaved caspase-3 signal within GC (white dashed line). Data represents mean ± SD compiled from Atm^fl/fl (n=3), ATM-Het (n=3), ATM-KO (n=3) and ATM-KD mice (n=3) 14 days after SRBC immunization, with at least 200 GCs quantified per genotype. Dashed white boundary indicates GC where quantification of signal was assessed. White solid bars represent 100μm scale bar. *, p < 0.05; ****, p < 0.0001 (unpaired t-Student with post-hoc Mann-Whitney test)..

ATM loss impairs generation of high-affinity switched antibodies

To assess how ATM deficiency impacts affinity maturation, we immunized mice with the T cell-dependent antigen 4-hydroxy-3-nitrophenylacetyl hapten conjugated to chicken globulin (NP-CGG) (15). Serum levels of total IgM were comparable across genotypes at every time point analyzed, while levels of IgG1 and IgG2b were markedly reduced in ATM-KO and ATM-KD mice (Fig. 4A). Additionally, while the titers of NP-specific IgM antibodies were similar between ATM-deficient and control mice (Fig. 4B), there was a significant reduction in both high-affinity (NP8) and broad-affinity (NP30) NP-specific IgG1 antibodies at d21 (Fig. 4C). Thus, reduced GC numbers and density stemming from the loss of ATM leads to impaired production of antigen-specific switched antibodies. The ratio of the high-affinity to low-affinity IgG1 levels was not altered in the ATM-KO or ATM-KD mice compared to controls (Fig. 4D and 4E), suggesting that amongst the cells that have managed to undergo CSR, affinity maturation was not grossly impaired.

Figure 4.

Somatic hypermutation is impaired in ATM deficient mice. Mice were immunized with NP-CGG, boosted at day 10 and analyzed at indicated time points after immunization. (A) Quantification of serum antibody levels in Atm^fl/fl (n=8), ATM-Het (n=5), ATM-KO (n=7) and ATM-KD mice (n=4) before immunization (d0) and at d14 and d21 post-immunization (mean and SD). (B-C) Antibody titers for IgM (B) and IgG1 (C) specific for NP(8) and NP(30). (D, E) Affinity maturation for IgM (D) and IgG1 (E) was analyzed by calculating the ratio of anti-NP(8)/NP(30). Data was compiled from Atm^fl/fl (n=4), ATM-Het (n=3), ATM-KO (n=4) and ATM-KD (n=3) mice. (F-G) SHM of J_H4 intron. Total pooled GC B cells (F), IgM⁺ GC B cells (G) or IgA⁺ GC B cells from Peyer’s patch (G) were sorted and mutation frequency in the J_H4 intron was assessed. Pie charts depict the proportions of clones with the indicated number of mutations. The total number of clones analyzed is shown at the center of each pie. Total mutation frequency is indicated. Data compiled from Atm^fl/fl and ATM-KO (n=3); ATM-Het and ATM-KD (n=2). *, p < 0.05; **, p < 0.01 (unpaired t-Student with post-hoc Mann-Whitney test).

To directly assess the role of ATM in SHM, we sequenced the J_H4 intron, a region frequently mutated in Peyer’s patch B cells independent of a specific immunogen (16, 17). When the total pool of GC B cells from Peyer’s patches was analyzed, the frequency of SHM at J_H4 was slightly reduced in ATM-deficient cells compared to control mice, even though the difference was not statistically significant, and no marked alteration in SHM spectrum was observed (Fig. 4F, Supplemental Fig. 2A). Likewise, when we sorted and analyzed IgA⁺ GC B cells from Peyer’s patches, a reduction in SHM frequency was observed in ATM-deficient B cells (Fig. 4G, Supplemental Fig. 2B); however, the difference did not reach statistical significance, and no alterations in mutation spectrum was observed, suggesting that B cells that have managed to class switch can undergo normal SHM or that there is a strong selection for mutated IgA⁺ cells in the Peyer’s patches. On the other hand, when IgM⁺ GC B cells were examined, a marked and significant reduction in the frequency of SHM was observed in ATM-deficient mice relative to controls (Fig. 4G). The spectrum of mutations in IgM⁺ GC B cells from ATM-deficient mice was also altered (Supplemental Fig. 2B), however the number of mutated clones in ATM KD mice was too low to reach a firm conclusion regarding mutation spectrum. Overall, these results support the notion that IgM⁺ GC B cells attempting to undergo CSR are unable to resolve AID-induced DSBs, leading to increased cell death, which culminates in smaller GCs and impaired antigen-specific immune responses.

It is now firmly established that ATM co-ordinates the recruitment of a multitude of factors, including H2AX, 53BP1, MDC1 and Rif-1, to resolve DSBs (18), and the CSR defect in vitro and in vivo in ATM-deficient B cells reflects this DSB resolving activity. It is worth noting that in vitro, a failure to resolve DSBs in B cells undergoing CSR does not lead to either proliferation defect or increased cell apoptosis, potentially because DNA damage checkpoints are not activated. In contrast, our in vivo results suggest that a failure to resolve DSBs in GC B cells leads to apoptosis and a profound loss of GC integrity, a surprising result given that GC B cells are programmed to survive, despite DNA damage, through BCL6-mediated repression of p53 (19). These observations raise the possibility that ATM, in addition to acting as an enforcer of DNA repair, also participates in dampening the DNA damage response to provide a window for B cell survival in the face of ongoing DNA damage. These dual but opposing roles of ATM are not without precedent. Previous work has demonstrated the existence of a positive feedback loop in which ATM participates in both the generation and in the repair of DSBs during CSR (9). Further studies are required to determine if ATM indeed has a role in dampening the DNA damage response to promote GC integrity.

Key Points.

• Loss of ATM leads to apoptosis of germinal center B cells

• Deletion of ATM in B cells reduces germinal center frequency and size

• ATM loss leads to decreased titers of class switched antibodies after immunization