Activation-induced cytidine deaminase (AID) is induced in B cells during an immune response and is essential for both class-switch recombination (CSR) and somatic hypermutation of Ab genes. The C-terminal 10 aa of AID are required for CSR but not for somatic hypermutation, although their role in CSR is unknown. Using retroviral transduction into mouse splenic B cells, we show that the C terminus is not required for switch (S) region double-strand breaks (DSBs) and therefore functions downstream of DSBs. Using chromatin immunoprecipitation, we show that AID binds cooperatively with UNG and the mismatch repair proteins Msh2-Msh6 to Ig Sμ and Sγ3 regions, and this depends on the C terminus and the deaminase activity of AID. We also show that mismatch repair does not contribute to the efficiency of CSR in the absence of the AID C terminus. Although it has been demonstrated that both UNG and Msh2-Msh6 are important for introduction of S region DSBs, our data suggest that the ability of AID to recruit these proteins is important for DSB resolution, perhaps by directing the S region DSBs toward accurate and efficient CSR via nonhomologous end joining.
Activation-induced cytidine deaminase (AID) initiates both class-switch recombination (CSR) and somatic hypermutation (SHM) in B cells during an immune response. AID deaminates deoxycytidines (dCs) to deoxyuridines (dUs), initiating DNA repair processes that lead to DNA breaks in Ig switch (S) regions and to SHM in Ig V region genes (1, 2). The C terminus of AID is required for CSR but not for SHM (3, 4). However, the role of the AID C terminus in CSR is unknown.
The structure of AID is well conserved; mouse and human AID have 92% peptide sequence identity (5). AID has a domain required for SHM but not for CSR between aa 13 and 23, a catalytic (deaminase) domain between aa 56 and 94, and a domain required for CSR but not SHM between aa 182 and 198 (3, 4, 6, 7). It has a robust chromosome region maintenance-1–dependent nuclear export signal also located at the C terminus between aa 190 and 198 (8–10). Nuclear localization of AID appears to be tightly regulated, because AID predominantly resides in the cytoplasm, and nuclear AID undergoes rapid ubiquitin-mediated proteasomal degradation (11). The half-life of nuclear AID is three times shorter than that of cytoplasmic AID. However, nuclear export is not required for CSR, as mouse AIDF198A is not exported from nuclei, and yet CSR and SHM are only modestly reduced in cells expressing this mutant (9). Also, many AID mutants with substitution mutations in the C terminus that retain a functional nuclear export signal are unable to potentiate CSR (12); this has been suggested to be due to interaction with an unidentified factor(s) (13).
During CSR, the dUs in S regions generated by AID are converted into double-strand breaks (DSBs) in the donor (Sμ) and acceptor Sx regions (1). The dU bases are excised by UNG, and deficiency in UNG causes a dramatic reduction in S region DSBs and in CSR (14–16). Apurinic/apyrimidinic endonucleases (APE) 1 and 2 have been shown to be important for S region DSBs during CSR (17); these enzymes can nick the abasic sites generated by UNG activity. When the nicked sites on opposite DNA strands are sufficiently near, they can form DSBs, which can be recombined with DSBs in another S region. Recombination occurs predominantly by nonhomologous end-joining (NHEJ), resulting in CSR (18).
If the single-strand breaks (SSBs) created by AID–UNG–APE activities are too far apart to spontaneously form a DSB, mismatch repair (MMR) proteins can process the SSBs into DSBs during CSR (19–21). The Msh2-Msh6 heterodimer binds G:U mismatches (22) and recruits Mlh1-Pms2 heterodimers (23). Subsequently, exonuclease 1 (24) loads at SSBs generated by APE1/2 activity and most likely excises past the U:G mismatch until reaching an SSB on the other DNA strand, thus creating a DSB with staggered ends. DNA polymerase can then fill in the staggered ends to generate a blunt-ended DSB or a DSB with a small overhang that can be recombined by NHEJ (20).
Mouse B cells expressing ΔAID (C-terminal 10 aa deleted; i.e., AIDΔ189–198) have greatly reduced CSR (∼10% of wild-type [WT]), but have normal levels of AID-induced mutations in Ig Sμ regions (3). Thus, AID targeting to the Sμ region is retained in this mutant. Furthermore, DSBs are detected in both donor and acceptor S regions in activated aid−/− CH12F3-2 B lymphoma cells expressing this mutant (25). In fact, the DSBs appeared to be somewhat increased relative to cells expressing full-length AID, and there was also a 10-fold increase in c-myc-IgH translocations. It has also been shown that the in vitro deaminase activity of ΔAID is at least as high as full-length AID (26, 27). Taken together, these data indicate that the C terminus of AID is not needed for introduction of dUs and DSBs in S regions. To explain these data, it is possible that the C terminus of AID regulates the timing of DSB creation and repair and/or interacts with proteins that direct S region DSBs toward accurate NHEJ and CSR.
Although the mutation frequency in the Sμ region is unchanged in ΔAID-expressing cells, the frequency of transition mutations relative to transversions at G:C bp within Sμ is increased by 2-fold compared with that of WT AID (3). This is similar to the effect on V region SHM of treating cells with an UNG inhibitor (28) and suggests that UNG activity might be overwhelmed or otherwise compromised in ΔAID-expressing cells. We hypothesized that the AID C terminus might be involved in recruiting UNG and other proteins that work together to convert dUs to DSBs and that these interactions are important for directing DSBs toward proper CSR.
To begin to address these possibilities, in this study we asked if AID interacts with UNG via its C terminus. We find that UNG interacts with full-length AID but not ΔAID, and this interaction is only detected in association with Ig S regions. Because MMR proteins also recognize dU in DNA, in the form of U:G mismatches, we also asked if MMR proteins interact with AID via its C terminus in association with S regions. Indeed, we find that Msh2-Msh6 interacts with AID when associated with S regions and that this depends on the AID C terminus. Finally, we show that the interactions between AID and UNG and between AID and MMR proteins increase the stability of binding of AID, UNG, and Msh2-Msh6 to S regions and that in cells expressing ΔAID, MMR does not contribute to CSR.
Materials and Methods
All mouse strains were extensively (≥8 generations) backcrossed to C75BL/6. AID-deficient mice were obtained from T. Honjo (Kyoto University, Kyoto, Japan) (29). Msh2-deficient mice were obtained from T. Mak (University of Toronto, Toronto, ON, Canada) (30). UNG-deficient mice were obtained from D. Barnes and T. Lindahl (London Research Institute, London, U.K.) (31). Msh6-deficient mice were obtained from W. Edelmann (32). Double-knockout mice were bred by mating heterozygous mice. Mice were housed in the Institutional Animal Care and Use Committee-approved specific pathogen-free facility at the University of Massachusetts Medical School; these mice were used according to the guidelines from University of Massachusetts Medical School Animal Care and Use committee. For each experiment, splenic B cells were isolated from littermates.
Primary Abs were for: estrogen receptor (ER) (sc-8002, lots G2307 and D0110; Santa Cruz Biotechnology); UNG (rabbit anti-peptide aa 280–295 from mouse UNG; prepared by Anaspec); MSH2 (sc-494), MSH6 (sc-10798), and GAPDH (sc-25778). For immunoblotting, secondary Abs used were goat anti-rabbit (sc-2004) or donkey anti-mouse–HRP (sc-2020) (all from Santa Cruz Biotechnology).
pMX-PIE-AID-FLAG-ER-IRES-GFP-puro and pMX-PIE-ΔAID-FLAG-ER-IRES-GFP-puro (3) were received from Drs. V. Barretto and M. Nussenzweig (The Rockefeller University, New York, NY). To create the control retrovirus pMX-PIE-ER-IRES-GFP, the ER gene from retroviral (RV)-AID-ER was PCR amplified with a forward primer, 5′-GCCGGATCCCGCCATGTCTGCTGGAGACATGAGAGCT-3′, and a reverse primer, 5′-GGCCTCGAGGAGCTCAAGCTGTGGCAGGGAAACC-3′. The PCR product was digested with BamHI and XhoI and inserted into BamHI- and XhoI-digested pMX-PIE vector containing an internal ribosome entry site (IRES), GFP, and puromycin genes downstream of the XhoI site. The construct was sequenced, and we confirmed proper expression of the RV-ER construct by Western blotting. To create the AIDH56R/E58Q-ER mutant, the AID gene was subcloned into Bluescript (Stratagene), mutated using Quik-Change (Stratagene), and then reinserted into pMX-PIE retroviral plasmid.
B cell purification and cultures
Mouse splenic B cells were isolated as previously described by T cell depletion with Ab and complement (33). B cells were cultured at 105 cells/ml in 5% CO2. LPS (25 μg/ml), anti-IgD dextran (α–δ-dex) (0.3 ng/ml), and IL-4 (20 ng/ml) were used to induce class switch recombination to IgG1; LPS and α–δ-dex were used to induce IgG3 switching. Human BAFF/BLyS (50 ng/ml; Human Genome Sciences) was included in all cultures.
To analyze cell-cycle progression of cells expressing AID-ER, ΔAID-ER, or ER, aid−/− mouse splenic B cells were transduced with retroviruses after culture for 48 h. Cells were harvested 1 d later, and flow cytometry was used to determine GFP expression levels. Cells were stained with propidium iodide and analyzed by flow cytometry to study the cell-cycle progression. For this experiment, we used cultures in which >95% of cells expressed GFP.
Cell proliferation assay
Proliferation of aid−/− B cells expressing AID-ER, ΔAID-ER, or ER was assayed by staining freshly isolated aid−/− mouse splenic B cells with PKH26 (Sigma-Aldrich), as prescribed in the product manual. Cells were then cultured for 48 h and infected with retroviruses expressing AID-ER, ΔAID-ER, or ER. PKH26 fluorescence was assayed by live-cell FACS 2 d postinfection.
Production of retroviruses in Phoenix-E cells
We seeded 3 × 106 Phoenix-E cells per 10-cm plate with 10 ml complete RPMI 1640 medium with heat-inactivated serum 24 h before transfection with the retroviral constructs. To transfect the retroviral constructs, Fugene 6 (Roche) was brought to room temperature and mixed well before use. A total of 27 μl Fugene 6 was added to 600 μl serum-free RPMI 1640 at room temperature. Fugene 6 was pipetted directly into the medium without contacting the walls of the plastic tube, mixed, and incubated at room temperature for 5 min. Six micrograms retroviral plasmid construct and 3 μg pCL-Eco plasmid (expressing RV-Gag, Pol, and Env proteins) were added, mixed, and incubated for 45 min at room temperature. The mixture was then added dropwise to a 10-cm plate containing Phoenix-E cells. The plate was swirled to ensure even distribution. The plate was incubated at 37°C for 24 h and moved to a 32°C incubator, which resulted in a higher titer of virus production. Transfection efficiency of Phoenix cells was checked with FACS for GFP. Cultures that have ∼65–90% GFP+ cells yielded good to excellent virus titers. Virus supernatant was collected 2 d after transfection and replaced with 10 ml fresh complete RPMI 1640 with heat-inactivated serum. The supernatant was then centrifuged at 1200 rpm for 5 min. Virus was harvested two more times every 48 h thereafter. The supernatant was used immediately or snap-frozen in liquid nitrogen before storing at −80°C. Because the virus titer decreases at each freeze-thaw cycle, appropriate aliquots were made before they were frozen.
Retroviral infection of B cells
A total of 4 ml B cells at 105/ml were cultured for 2 d in 6-well plates or 2 ml in 24-well plates in IgG1 or IgG3 CSR conditions. Each culture was split into two cultures on day 2. The culture plates were centrifuged at 1200 rpm for 5 min after the split to remove the supernatant without removing the cells. The cells on the plate were resuspended in the retrovirus mixture containing: virus supernatant (1 ml), RPMI 1640 (1 ml), heat-inactivated FBS (30 μl), LPS (25 μg/ml), IL-4 (20 ng/ml) (not included in IgG3 switching conditions), BLyS (50 ng/ml), anti–δ-dex (0.3 ng/ml), 4-hydroxytamoxifen (4-OHT; 1 μM; Sigma-Aldrich) (3), and polybrene (10 μg/ml) (3). After 45 min at room temperature in the mixture, the cells were resuspended in the mixture and the culture plates centrifuged at 2000 rpm (Allegra-12R centrifuge; Beckman Coulter) for 45 min at 20°C. The plates were returned to 37°C, 5% CO2, for 24 h.
CSR assay and FACS analysis
To study CSR in cells expressing AID-ER, ΔAID-ER, or ER, we cultured 2 ml B cells (105/ml) for 2 d in IgG1 or IgG3 switching conditions described above and infected these cells with retrovirus to express AID-ER, ΔAID-ER, or ER. Cells were treated with 1 μM OHT at time of infection. Cells were harvested 24 h postinfection and subjected to FACS staining and analysis as described previously (33), except that cells were not fixed to detect enhanced GFP (EGFP) and were stained with 7-aminoactinomycin D (BD Biosciences) to allow us to gate out dead and dying cells. Statistical difference in CSR between different cultures was determined by a two-tailed t test.
Chromatin immunoprecipitation (ChIP) assays were performed as described previously (34). Live cells were isolated by flotation on Lympholyte M (Cedarlane Labs, Toronto, ON, Canada) 24 h after retroviral transduction. After recovery and washing twice, 2 × 107 live cells were resuspended in PBS and cross-linked with formaldehyde at a final concentration of 1% for 5 min at 37°C. Cross-linking was stopped by adjustment to 125 mM glycine, incubating 5 min at room temperature. Cross-linked cells were sonicated on ice for 200 s. Sonication was performed in 10-s bursts; the sample was incubated for 15 s in ice between each bursts. A total of 106 cell equivalents were used per immunoprecipitation (IP), and 105 cells were used for the input samples. To correct for this, ChIP results were divided by 10. Significance was calculated by a two-tailed t
In vitro transcription translation
Equal volume of GST or GST-UNG was added to each tube containing AID-ER, ΔAID-ER, and ER in dilution buffer containing: PBS with complete protease inhibitor mixture (Roche), 200 μM PMSF, 200 μM sodium orthovanadate, 50 μM NaF, 2 μM β-glycerophosphate, 50 μM ZnCl2, and 200 μg/ml ethidium bromide. The mixtures were incubated at 4°C for 12 h and then incubated with glutathione beads for another 2 h. The beads were spun down and then washed three times with wash buffer. The wash buffer was made by adding 200 mM NaCl and 0.05% Nonidet P-40 to the dilution buffer without ethidium bromide. The proteins were then eluted with elution buffer after the third wash. The eluates were then subjected to Western blotting.
Genomic DNA preparation and linker ligation-mediated PCR
After culture for 24 h after RV infection, viable cells were isolated by flotation on Lympholyte M, cells were imbedded in low-melt agarose plugs, DNA was isolated, and ligation-mediated PCR (LM-PCR) performed as described (16), with slight modifications. For linker ligation, 50 μl 1× ligase buffer was added to the 50 μl plugs, which were then heated to 65°C to melt the agarose. A total of 20 μl DNA (∼200,000 cell equivalents) was added to 2 μl T4 DNA Ligase (2 Weiss units; MBI Fermentas, Hanover, MD), 10 μl double-stranded annealed linker in 1× ligase buffer, 3 μl 10× ligase buffer, and 30 μl distilled H2O and incubated overnight at 18°C. Linker was prepared by annealing 5 nM each LMPCR.1 (5′-GCGGTGACCCGGGAGATCTGAATTC-3′) and LMPCR.2 (5′-GAATTCAGATC-3′) in 300 μl 1× ligase buffer, which results in a double-stranded oligonucleotide with a 14-nt single-stranded overhang that can only ligate unidirectionally. Ligated DNA samples were heated at 70°C for 10 min, diluted five times in distilled H2O, and then assayed for gapdh DNA by PCR to adjust DNA input prior to LM-PCR. The primers 5′ Sμ (5′-GCAGAAAATTTAGATAAAATGGATACCTCAGTGG-3′), Sg3-AP (5′-AACATTTCCAGGGACCCCGGAGGAG-3′), and CmuL2 (5′-CTGCGAGAGCCCCCTGTCTGATAAG-3′) were used in conjunction with linker primer (LMPCR.1) to amplify DNA breaks in Sμ, Sγ3, or Cμ, respectively. Three-fold dilutions of input DNA (0.5, 1.5, and 4.5 μl for Sμ/Cμ LM-PCR; 1.5, 4.5, and 13.5 μl for Sγ3 LM-PCR) were amplified by HotStar Taq (Qiagen) using a touchdown PCR program (28 cycles after touchdown for Sμ LM-PCR and 35 cycles after touchdown for Sγ3 LM-PCR). PCR products were electrophoresed on 1.25% agarose gels and blotted onto nylon membranes (GeneScreen Plus; PerkinElmer, Waltham, MA). Blots were hybridized with an Sμ-specific oligonucleotide probe (μ probe 5′: 5′-AGGGACCCAGGCTAAGAAGGCAAT-3′) for 5′ Sμ LM-PCR or Sγ3-LP (5′-GGACCCCGGAGGAGTTTCCATGATCCTGGG-3′ or Cμ: 5′-TGGCCATGGGCTGCCTAGCCCGGGACTTCCTG-3′) that had been end-labeled with [γ32
Expression of ΔAID-ER or AID-ER does not alter cell-cycle progression or cell proliferation
To test the role of the AID C terminus in CSR, we expressed full-length AID or truncated AID lacking the C-terminal 10 aa by retroviral transduction in activated aid−/− mouse splenic B cells. The retroviral constructs encode full-length AID or ΔAID (Δ189–198) fused to a FLAG tag and the ER (AID-ER and ΔAID-ER, respectively) (3) (Fig. 1A). The fusion protein is followed by an IRES and a gene encoding EGFP. This entire construct is flanked by long terminal repeats. We also prepared a control vector that expresses the ER protein but no AID.
The three retroviral constructs (AID-ER, ΔAID-ER, or ER) were transduced into aid−/− splenic B cells 2 d after activation with LPS, BLyS, and anti-IgD dextran ± IL-4 to induce CSR to IgG1 or IgG3, respectively. The cultures were treated with tamoxifen (4-OHT) at the time of retroviral transduction, which induces nuclear localization of ER-tagged proteins, to allow AID to reach its target. One day later, transduction efficiency was assayed by GFP expression (Fig. 1B). In this study, we only include experiments in which the three different retroviruses yielded similar GFP expression. Consistent with previous reports, full-length AID-ER induces ∼10-fold more IgG1 and IgG3 CSR than does ΔAID-ER in aid−/− splenic B cells, as shown by flow cytometry analyses (Fig. 1F). In the absence of OHT treatment, AID-ER does not induce CSR (data not shown). The expression levels of AID-ER and ΔAID-ER proteins in cells with similar GFP expression are quite similar (Fig. 1C), and the proteins are present in both nucleus and cytoplasm (Fig. 1G). Previously, it was reported that ΔAID-ER is not expressed as well as AID-ER in splenic B cells (3). Perhaps this is because in those experiments, cells were cultured for 3 d after viral transduction, whereas we assay our experiments 1 d after transduction. We chose this time because there is adequate CSR and less cell death on day 1 than on subsequent days.
Because CSR is cell division linked (35, 36), it is important to examine whether the cell-cycle profiles and cell proliferation rates in cells expressing ΔAID-ER are similar to that of cells expressing AID-ER. We stained cells expressing AID-ER, ΔAID-ER, or ER with propidium iodide on day 1 after transduction and examined the cell cycle profile of each using flow cytometry. AID-ER or ΔAID-ER expression has no effect on cell-cycle progression of aid−/− B cells (Fig. 1D). To examine cell proliferation, B cells were stained with PKH26 prior to activation and transduction with RV-AID-ER, RV-ΔAID-ER, or RV-ER. The rate of PKH26 dilution by cell division was assayed by flow cytometry and indicates that cells transduced with the three different RV constructs proliferate similarly (Fig. 1E).
ΔAID-ER induces S region DSBs in aid−/− B cells
S region DSBs are an essential intermediate in CSR and are dependent upon AID, UNG, and MMR (15, 16, 21, 37–39). We asked whether we could confirm the previous report that S region DSBs are present in cells expressing ΔAID (25). We used LM-PCR to analyze Sμ and Sγ3 DSBs in aid−/− splenic B cells expressing AID-ER, ΔAID-ER, or ER. As expected, nontransduced WT B cells showed abundant DSBs in Sμ and Sγ3 regions, whereas aid−/− B cells, either nontransduced or transduced with the control RV-ER, showed very few S region DSBs (Fig. 2A, 2B). We also performed LM-PCR on GFP-negative cells FACS-sorted from RV-infected populations and found very few S region DSBs (Fig. 2A, 2C). However, aid−/− cells expressing either AID-ER or ΔAID-ER, but not ER alone, contain abundant blunt DSBs in Sμ and Sγ3 region DNA, but very few DSBs in the Cμ gene. The GAPDH bands shown below the blots in Fig. 2 are an internal PCR control for template loading. The GFP expression among the different cell populations used for LM-PCR was similar in the experiment shown in Fig. 2A, ranging from 69–79%, suggesting that the expression of AID and ΔAID was similar. We also assayed staggered DSBs in RV-transduced cells by treating the genomic DNAs with T4 DNA polymerase prior to linker ligation. As shown in Fig. 2B, the number of DSBs in T4-Pol–treated DNA from cells expressing AID and ΔAID appear similar. This experiment was performed three times with similar results.
To determine whether there is a difference between AID- and ΔAID-induced Sμ and Sγ3 DSBs in cells expressing low versus high amounts of AID and ΔAID, we sorted cells according to levels of GFP expression. Supplemental Fig. 1 shows the profile of GFP expression and gates used for this sort. As we found when unsorted cells were analyzed, the numbers of Sμ and Sγ3 DSBs detected in cells expressing high amounts of AID or ΔAID (GFP-high) were similar (Fig. 2C, Supplemental Figs. 2, 3). In GFP-low cells, ΔAID appears to induce fewer S region DSBs than does full-length AID. Perhaps, in cells expressing lower amounts of AID, the conversion of dUs to DSBs is less efficient in ΔAID-expressing cells than in cells expressing full-length AID, despite the fact that ΔAID has high deaminase activity. This reduction in DSBs cannot account for the severe reduction in CSR in ΔAID-expressing cells, because abundant breaks are found in unsorted ΔAID-expressing cells (Fig. 2A), and switching is reduced 10-fold.
One possible role of the C terminus might be to promote interaction between Sμ and a downstream acceptor S region, perhaps thereby allowing AID to attack the acceptor S region. However, we found that the AID C terminus is not required for introduction of DSBs into Sγ3 in B cells induced to switch to IgG3, as our LM-PCR data indicate that cells expressing full-length AID and ΔAID have similar numbers of Sγ3 DSBs (Fig. 2, Supplemental Fig. 3).
To determine whether loss of the C terminus alters the sites at which DSB are found, we cloned and sequenced the LM-PCR products. The sites of DSBs in Sμ of AID-ER– or ΔAID-ER–expressing aid−/− B cells are indistinguishable from each other (Table I) and from those in WT B cells expressing endogenous AID (17), but different from those few DSBs cloned from aid−/− cells (16). As true for Sμ DSBs instigated by endogenous AID, ≥80% of the DSBs occur at G:C bp and preferentially at AID (WRC/GYW and AGCT) target hotspots. These data suggest that deletion of the AID C terminus does not prevent AID from generating DSBs in S regions or alter the sequence specificity of Sμ DSBs.
AID and UNG bind cooperatively to Sμ DNA, and their binding depends on the AID C terminus
As it is clear that ΔAID mutates Sμ and induces DSBs, we expected ΔAID-ER to be detected at the Sμ region in vivo using ChIP assays. Surprisingly, the ChIP data show decreased association of ΔAID-ER with Sμ DNA compared with AID-ER in aid−/− splenic B cells induced to switch to IgG3 (Fig. 3A). As anti-ER Ab can precipitate AID-ER, ΔAID-ER, and ER proteins equally well from total cell extracts (Fig. 3B), we conclude that binding of AID-ER to the Sμ region that is stable enough to be detected by ChIP depends upon its C terminus. Supplemental Fig. 4 shows a gel analysis of the anti-ER ChIP at Sμ. The association of full-length AID with Sμ is only observed if cells are treated with OHT to induce the ER-tagged proteins to move to nuclei (data not shown). Because ΔAID induces the same frequency of mutations at Sμ as does full-length AID (3), it is clear that the ability of AID to stably associate with Sμ is not important for its ability to deaminate Sμ, although it might be important for CSR.
Cells expressing ΔAID-ER show an increased proportion of C to T and G to A transition mutations in the 5′ Sμ region relative to C to A and C to G or G to C and G to T transversion mutations, similar to the phenotype of ung−/− B cells (3). This suggests that AID might associate with UNG via the AID C terminus and help recruit UNG to S regions. We asked, therefore, if association of endogenous UNG with Sμ is compromised in cells expressing C-terminal–deleted AID. Indeed, ChIP with anti-UNG Ab demonstrated that whereas UNG associates with Sμ in aid−/− cells expressing AID-ER, it is found only at background levels in cells expressing ΔAID-ER (Fig. 3D). As UNG is required for S region DSBs (16), it is clear that it must associate with S regions in ΔAID-expressing cells, although apparently this interaction is not stable enough to be detected by ChIP under our conditions.
Our finding that binding of UNG to Sμ in vivo depends on the AID C terminus prompted us to ask if the preferential binding of full-length AID to Sμ likewise depends upon UNG. We expressed the three retroviruses in aid−/−ung−/− splenic B cells and performed ChIP with anti-ER Ab. Our results demonstrate that in cells lacking UNG, full-length AID-ER does not associate better than does ΔAID-ER with Sμ (Fig. 3A). Fig. 3C shows that AID and ΔAID can be detected in both whole-cell and nuclear extracts from aid−/−ung−/− cells. AID-ER, ΔAID-ER, and UNG are not found at the Cμ region in these cells, which indicates that the association is specific for Sμ (Fig. 3A, 3D). We conclude that the ability of full-length AID but not ΔAID to associate stably with Sμ depends upon the presence of UNG. The inability of UNG to bind Sμ in cells expressing ΔAID is not due to a lack of dUs, as AID lacking the C terminus is a highly active cytidine deaminase (26, 27). Also, this result is not an artifact due to possible differences in numbers of Sμ SSBs or DSBs, which theoretically could affect the ability to amplify Sμ DNA in the ChIP, because the quantitative PCR results in these experiments showed that there were equal amounts of amplifiable input DNA in the WT, ΔAID, and ER samples (data not shown). To ensure that not all proteins bind better to the Sμ region in AID-expressing cells than in ΔAID-expressing cells, we performed ChIP for RNA polymerase II. RNA PolII binds Sμ similarly in cells expressing each of the three RV constructs (Fig. 3E). Taken together, these data suggest that AID and UNG bind cooperatively to the Sμ region in B cells undergoing CSR and that this cooperative binding depends on the AID C terminus. Despite the fact that in cells expressing ΔAID, neither UNG nor ΔAID bind stably enough to be detected in our ChIP assays, it is clear that ΔAID and UNG have enzymatic activity at the S region in cells expressing ΔAID due to the presence of DSBs (Fig. 2) (25). These results suggest that the binding detected by ChIP might be important for a process downstream of DSB formation.
AID and Msh2-Msh6 bind cooperatively with Sμ DNA dependent upon the AID C terminus
The MMR heterodimer Msh2-Msh6 binds to U:G mismatches that result from AID activity (22). Interestingly, msh6−/− mice have altered targeting of AID, as the AID-induced mutations in V regions are more focused at WRC hotspots in these mice than in WT mice (40, 41). We asked if endogenous Msh2-Msh6 can be detected at S regions expressing RV-AID and, if so, whether association of endogenous Msh2-Msh6 with Sμ in vivo depends upon the C terminus of AID. ChIP for Msh2 and for Msh6 in aid−/− B cells demonstrated that both Msh2 and Msh6 associate with Sμ in cells expressing full-length AID-ER, but not significantly above background when expressing ΔAID-ER or ER (Fig. 4A). Binding of Msh2 and Msh6 is not detected at the Cμ region (Fig. 4B), in agreement with our finding that AID-ER binds Sμ and not Cμ.
We next asked if binding of full-length AID to Sμ in vivo might depend upon these MMR proteins, in addition to UNG, by expressing the RV constructs in aid−/−msh2−/− and in aid−/−msh6−/− cells. Indeed, we found that in the absence of Msh2 or Msh6, AID-ER does not bind better than ΔAID-ER to the Sμ region, and the binding was at background levels (Fig. 4C). This is not due to differential expression of the AID proteins, as the amounts of AID-ER and ΔAID-ER are similar in cytoplasmic and nuclear extracts of B cells from both aid−/−msh2−/− and aid−/−msh6−/− mice (Fig. 4D). Together with the results showing reduced binding of Msh2-Msh6 to Sμ in cells expressing ΔAID-ER, these data suggest that AID binds cooperatively with Msh2-Msh6 to the Sμ region and that this cooperative binding depends on the C terminus of AID. We hypothesize that cooperative binding of AID with UNG and Msh2-Msh6 stabilizes the binding of AID to Sμ, thus allowing the binding to be detected by ChIP. This increased stability does not appear necessary, however, for the deamination activity of AID or for S region DSBs. As UNG is essential for DSBs (16), and MMR proteins contribute to formation of Sμ DSBs (20, 21), it is clear that these proteins must be binding and acting at S region in ΔAID-expressing cells, but perhaps binding only transiently. We conclude that a complex of AID, UNG, and Msh2-Msh6 that is stable enough to be detected by ChIP requires the AID C terminus and that this complex might be required for CSR.
Mutations of the catalytic domain of AID prevent cooperative binding of AID, UNG, and Msh2 to Sμ and Sγ3 in ChIP assays
We have been unable to detect direct binding between AID and UNG or between AID and Msh2 (data not shown). We hypothesized that detectable binding of UNG and Msh2-Msh6 to Sμ depends on the presence of their dU substrates and therefore that the deaminase activity of AID would be required to detect cooperative binding of AID, UNG, and Msh2 to Sμ. To test this hypothesis, we transduced aid−/− B cells with RV-AIDH56R/E58Q (i.e., full-length AID with two mutations in the catalytic domain resulting in complete absence of deaminase activity) (42). ChIP assays demonstrated that RV-AIDH56R/E58Q, UNG, or Msh2 did not detectably bind Sμ in cells expressing this mutant AID, although in these experiments, AID, UNG, and Msh2 bind Sμ in cells expressing full-length WT AID (Fig. 5A–C). Thus, the cooperative binding of UNG and Msh2-Msh6 with AID depends upon the presence of dUs in DNA.
As AID also acts on acceptor S regions, we asked whether the cooperative binding of these three proteins would also occur at Sγ3 and would also depend upon the deaminase activity of AID. Indeed, as shown in Fig. 5A–C, AID, UNG, and Msh2 bind to Sγ3 regions in cells transduced with WT AID but not in cells transduced with RV-AIDH56R/E58Q, nor in cells transduced with ER alone. A Western blot of total cell extracts shows that the mutant AID was expressed as well as WT AID (Fig. 5D). Thus, we conclude that AID cooperatively binds with UNG and Msh2-Msh6 to both donor S and acceptor S regions dependent upon the presence of dUs in DNA. Our results support the hypothesis that in addition to deamination of dCs, leading to DSBs, AID has a role in CSR subsequent to generation of DSBs, and this role results in stabilization of the binding of UNG and MMR specifically to S regions.
MMR deficiency has no effect on CSR in cells expressing ΔAID
CSR is reduced by 50–75% in MMR-deficient B cells (1, 33, 40, 43–45). As we found that cooperative binding between AID and the Msh2-Msh6 heterodimer at Sμ depends upon the AID C terminus, we next asked if the C terminus of AID is important for the ability of MMR to increase CSR efficiency. We hypothesized that recruitment of Msh2-Msh6 by AID is important for the function of MMR during CSR and that in cells expressing ΔAID-ER, deficiencies in Msh2 or Msh6 proteins will not reduce CSR efficiency relative to MMR-sufficient cells. To test this hypothesis, we examined CSR to IgG1 and IgG3 in aid−/−msh2−/− and aid−/−msh6−/− B cells transduced with the three RV constructs (Fig. 6). In cells expressing full-length AID, CSR was reduced by 45–70% by Msh2 or Msh6 deficiency, consistent with results in MMR-deficient cells expressing endogenous AID. These results demonstrate that despite being overexpressed, AID-ER behaves similarly to endogenous AID, as optimal CSR depends upon MMR.
Importantly, although switching to IgG1 in aid−/− B cells expressing ΔAID-ER is reduced by ∼90% compared with that of cells expressing AID-ER, it is not further reduced by deficiencies in Msh2 (Fig. 6A, 6B, upper panels) or Msh6 (Fig. 6C, 6D, upper panels). Similarly, switching to IgG3 in ΔAID-ER-expressing aid−/− B cells is reduced by ∼90% relative to AID-ER-expressing aid−/− B cells, but not further reduced in aid−/−msh2−/− B cells (Fig. 6A, 6B, lower panels) or in aid−/−msh6−/− B cells (Fig. 6C, 6D, lower panels). As ΔAID is an active deaminase and induces Sμ mutations, this is not due to the lack of U:G mismatch substrates for Msh2-Msh6 in Sμ. These data indicate that Msh2 and Msh6 do not increase CSR in cells in which AID lacks the C terminus.
In conclusion, these data reveal that the C terminus of AID is important for recruitment of UNG and Msh2-Msh6 to Ig S regions and that this recruitment is important for the function of Msh2-Msh6 in CSR. Although both UNG and MMR are important for DSB formation during CSR, our results indicate that the AID C terminus is not required for Sμ or Sγ3 DSBs, suggesting that UNG and MMR can contribute to DSB formation without binding stably. These data clearly indicate that the block in CSR caused by lack of the AID C terminus occurs downstream of S region DSB formation, although it is possible that the efficiency of DSB formation is reduced. We conclude that a complex of AID, UNG, and Msh2-Msh6 that is stable enough to be detected by ChIP requires the AID C terminus and that this complex is important for CSR.
We have demonstrated that AID binds cooperatively with both UNG and Msh2-Msh6 to the Ig Sμ region in B cells induced to switch to IgG3 and that the cooperative binding requires the C-terminal 10 aa of AID. Full-length AID was not detected at Sμ by ChIP in cells lacking UNG, Msh2, or Msh6, and thus, all three components are required for complex formation. The interactions might be direct or indirect. Consistent with the ChIP results, we found that the MMR proteins Msh2 and Msh6 do not contribute to CSR in cells expressing ΔAID, indicating that the ability of AID to recruit MMR proteins to Sμ is important for the role of MMR during CSR. The associations between AID and UNG and between AID and Msh2-Msh6 are detected when AID is localized on S regions; we have not been able to detect binding in the absence of DNA or at the Cμ region. The binding to both Sμ and Sγ3 are only detected when AID is competent for deaminase activity, suggesting that the cooperative binding between AID and UNG and Msh2-Msh6 occurs after AID has deaminated dCs. However, as not all dUs are likely to be excised by UNG and converted to SSBs, it is likely that at any one moment in cells undergoing CSR, S regions contain a mixture of dUs, abasic sites, and SSBs (1). Thus, S regions could have substrates for UNG and MMR binding, despite also having SSBs and DSBs.
It is puzzling that there are about as many Sμ DSBs in cells expressing ΔAID as in cells expressing full-length AID, despite lower recruitment of AID, UNG, and MMR. Considerable evidence indicates that UNG is essential for S region DSBs (16) and that MMR is important for conversion of SSBs to DSBs in S regions during CSR. In the absence of MMR, DSBs appear to only form from SSBs that are near on opposite strands (19–21, 46, 47). Sμ region DSBs have also been observed using LM-PCR in patients with the autosomal recessive form of hyper-IgM syndrome (HIGM-2) with C-terminal deletions of AID (48). Also, CH12F3-2 lymphoma B cells expressing C-terminal–deleted AID have been shown to have as many Sμ region DSBs as cells expressing full-length AID (25). Therefore, it seems clear that in the absence of the AID C terminus, both UNG and MMR can still access dU and U:G mismatches resulting from AID activity. One possibility to explain the numerous DSBs detected in cells expressing ΔAID is that ΔAID might have higher catalytic activity than full-length AID (27), and the increased dUs might require fewer UNG and MMR molecules to induce DSBs. However, we favor the alternative possibility that DSBs are induced with lower efficiency in cells expressing ΔAID, but are also repaired or recombined inefficiently and thus accumulate. This appears consistent with the fact that GFP-low cells expressing ΔAID have fewer DSBs than GFP-low cells expressing full-length AID. We hypothesize that in GFP-high cells, this repair problem would be exacerbated due to overwhelming the repair/recombination mechanisms because of the high AID levels, thus explaining the similar numbers of DSBs observed in ΔAID- and AID-expressing GFP-high and in total GFP+ cells. Inefficient repair/recombination is also consistent with the increased translocations found in B cells expressing RV-ΔAID (25), as unresolved DSBs increase the risk of DNA translocations, carcinogenesis, and cell death.
Our results suggest that the C terminus of AID promotes the formation of a protein complex that not only generates S region DNA breaks, but also promotes their repair in a manner that leads to productive CSR. UNG and MMR proteins might help recruit other proteins involved in subsequent repair steps during CSR. For example, Msh2 and Msh6 have been found in a large complex of DNA repair proteins in HeLa cells, termed BASC, which also includes Mre11-Rad50-Nbs1. Mre11-Rad50-Nbs1 binds DSBs induced by AID and is known to be important for repair of DSBs and for CSR (1, 49–52). Another possibility is suggested by findings in overexpression experiments that AID binds DNA-dependent protein kinase, catalytic subunit (DNA-PKcs), dependent upon the AID C terminus (53), and that Msh2-Msh6 binds to Ku70-Ku80 (54). As Ku and DNA-PKcs are important for directing DNA breaks toward NHEJ, it is possible that in addition to its role in converting SSBs to DSBs, MMR associated with the AID C terminus is involved in recruiting DNA-PK–Ku proteins to sites of AID-induced DSBs. Chromosomal translocations have been shown to be mediated by microhomology-mediated end joining (55), an alternative form of end-joining in which recombination is mediated by short microhomologies between the two recombining sequences. In addition, B cells with mutations in DNA damage response or repair genes that are involved in CSR, including mutations in MMR or UNG genes, often show increased lengths of junctional microhomology (46, 56–59). Most importantly, Sμ-Sα junctions from two human HIGM-2 patients expressing AID with C-terminal deletions show highly significantly increased lengths of junctional microhomology (59). These results suggest that the AID C terminus, MMR and UNG are involved in the recombination step during CSR and perhaps direct S region DSBs toward NHEJ.
CSR and V-D-J recombination both involve NHEJ and occur during the G1 phase of the cell cycle. Recent studies have shown that a mutant of RAG-2 (T490A), which prevents degradation of RAG-2 during S phase, results in aberrant V-D-J recombination and increased chromosomal translocations, indicating the importance of cell cycle control for accurate recombination (60, 61). AID-dependent Sμ DSBs are only observed during G1 phase in normal splenic B cells undergoing CSR, indicating that most DSBs are introduced, recombined, and repaired during G1 phase (21, 49). It is possible that the interaction of AID with UNG and MMR might be important to recruit these proteins to S regions during G1 phase, because UNG and MMR proteins are primarily associated with S phase activity. MMR is known to associate with the replication machinery and to correct nucleotides misincorporated during S phase due to DNA polymerase errors (23, 62). UNG is important for removal of dU incorporated instead of deoxythymidine during replication, and in both primary fibroblasts and HeLa cells, UNG is expressed at much higher levels during S phase than during G1 or G2 phases (63, 64). However, in splenic B cells induced to undergo CSR, UNG expression is slightly higher in G1 phase than in S phase. UNG is active in G1 phase B cell extracts and is clearly important for excision of dUs due to deamination by AID (14, 16, 21). It is possible that if UNG and MMR are not specifically recruited to Sμ during G1 phase by full-length AID, these proteins might only be available at S regions during S phase, thus creating DSBs in S phase. During DNA replication, DSBs cause stalling of DNA polymerase and collapse of replication forks, which are then sometimes repaired by break-induced replication, which is highly inaccurate (65). This could lead to translocations and greatly reduced CSR. Alternatively, inefficient repair of DSBs in G1 phase cells expressing ΔAID could cause DSBs to be retained during S and G2 phases and thus lead to chromosomal rearrangements.
The binding of UNG and MMR to AID might be indirect and occur via other proteins. AID has been shown to interact with several proteins (66). Protein kinase A, which phosphorylates AID at S38, is specifically localized at S regions during CSR and might participate in a protein kinase A–RPA–AID complex (67, 68). Phosphorylation of AID at S38 is required for interaction with RPA and for CSR (69, 70), but it is possible that the AID C terminus is also involved. Interestingly, RPA32 has been shown to interact with UNG by a yeast two-hybrid assay (71). Also, a recent publication reported that the C terminus of AID interacts with the 14-3-3 adaptor protein complex and that this interaction is important for recruitment of AID to S regions, as 14-3-3 binds specifically to AGCT motifs (72). Deficiency in 14-3-3 proteins results in a 50% decrease in CSR. However, because ΔAID-ER mutant induces as many Sμ mutations as AID-ER, the recruitment of AID by 14-3-3 is not necessary for the ability of AID to deaminate Sμ. Also, we find that the preferential occurrence of Sμ DSBs at AGCT hotspots is identical in cells expressing both AID and ΔAID. It is possible, however, that 14-3-3, when bound to AID, is important for recruiting UNG and MMR and other repair proteins to Sμ.
The authors have no financial conflicts of interest.
We thank V. Barreto and M. Nussenzweig (The Rockefeller University, New York, NY) for the pMX-PIE-AID-FLAG-ER-IRES-GFP-puro and pMX-PIE-ΔAID-FLAG-ER-IRES-GFP-puro plasmids. We also thank K. Rauch for preparing the AIDH56R/E58Q-ER mutant, our colleagues at the University of Massachusetts Medical School and R. Gerstein, R.T. Woodland, N. Rhind, M. Volkert, and J.E.J. Guikema at the Academic Medical Center, University of Amsterdam for helpful suggestions, and the University of Massachusetts Medical School Flow Cytometry Core Facility for excellent technical assistance.
This work was supported by National Institutes of Health Grants R01 AI023283 and R21 AI88578 (to J.S.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
- activation-induced cytidine deaminase
- apurinic/apyrimidinic endonuclease
- chromatin immunoprecipitation
- class-switch recombination
- DNA-dependent protein kinase, catalytic subunit
- double-strand break
- enhanced GFP
- estrogen receptor
- HIGM syndrome
- hyper IgM syndrome
- internal ribosome entry site
- ligation-mediated PCR
- mismatch repair
- nonhomologous end-joining
- somatic hypermutation
- single-strand break
- Received May 16, 2011.
- Accepted June 21, 2011.
- Copyright © 2011 by The American Association of Immunologists, Inc.